Compléments_couches_denses_corrigé.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/hocine-drioueche/OpenclassroomsProject/blob/main/Compl%C3%A9ments_couches_denses_corrig%C3%A9.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ljUdbmTomESG"
      },
      "source": [
        "<center><img src=\"https://assets-datascientest.s3-eu-west-1.amazonaws.com/train/logo_datascientest.png\" style=\"height:150px;center\"></center>\n",
        "\n",
        "<hr style=\"border-width:2px;border-color:#75DFC1\">\n",
        "<center><h1>Deep-Learning - Modules complémentaires</h1></center>\n",
        "<center><h2>Compléments sur le dataset MNIST</h2></center>\n",
        "<hr style=\"border-width:2px;border-color:#75DFC1\">\n",
        "\n",
        "\n",
        "Le but de cet exercice est d'approfondir nos connaissances sur l'entraînement d'un modèle en deep learning.\n",
        "\n",
        " Nous allons travailler avec la base de données **```MNIST```**. Cette base contient des images de chiffres manuscrits que nous tenterons de classifier.\n",
        "\n",
        "> Nous aborderons les points suivants :\n",
        ">> I - [Préparation du dataset](#preparation)\n",
        ">>\n",
        ">>\n",
        ">> II - [Influence de la taille des batchs](#batch_size)\n",
        ">>> A - [Pour nombre d'epoch fixé](#fixed_epochs)\n",
        ">>>\n",
        ">>>\n",
        ">>> B - [Pour un nombre d'itérations fixé](#fixed_iterations)\n",
        ">>\n",
        ">>\n",
        ">> III - [Illustration du surentraînement](#overfitting)\n",
        ">>\n",
        ">>\n",
        ">> IV - [Influence du nombre d'images dans le jeu d'entraînement](#train_size)\n",
        ">>\n",
        ">>\n",
        ">> V - [Influence de l'architecture du réseau de neurones](#architecture)\n",
        ">>> A - [Architecture très basique VS architecture plus complexes](#dense)\n",
        ">>>\n",
        ">>>\n",
        ">>> B - [Introduction aux réseaux de neurones convolutionnels](#CNN)\n",
        "\n",
        "- (a) Exécuter la cellule ci-dessous pour importer les modules nécessaires à l'exercice ainsi que le dataset ```MNIST```."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "P-dWe34ii2J9",
        "outputId": "58690b43-86ae-47d0-85b9-3fa0ca3fc2e5"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11493376/11490434 [==============================] - 0s 0us/step\n",
            "11501568/11490434 [==============================] - 0s 0us/step\n"
          ]
        }
      ],
      "source": [
        "import numpy as np\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "from matplotlib import cm\n",
        "%matplotlib inline\n",
        "\n",
        "from keras.datasets.mnist import load_data\n",
        "from keras.utils import np_utils\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Flatten\n",
        "from keras.layers.convolutional import Conv2D, MaxPooling2D\n",
        "\n",
        "import time\n",
        "\n",
        "(X_train, y_train), (X_test, y_test) = load_data()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1xQq2gwazUE6"
      },
      "source": [
        "- (b) Afficher, dans une grille de figures, 6 images tirées aléatoirement de l'échantillon ```X_train```, avec en titre les labels correspondants de ```y_train```.\n",
        "> - Cet affichage a déjà été fait dans le premier notebook du cours."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 264
        },
        "id": "E39tYMX7FxqN",
        "outputId": "bd7691d2-edd0-4c2c-a203-8b58354eac96"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 6 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "j = 1\n",
        "for i in np.random.choice(np.arange(0, len(y_train)), size = 6):\n",
        "    img = X_train[i]\n",
        "    plt.subplot(2, 3, j)\n",
        "    j = j + 1\n",
        "    plt.axis('off')\n",
        "    plt.imshow(img, cmap=cm.binary, interpolation='None')\n",
        "    plt.title('Label: ' + str(y_train[i]))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1yxBTeHulVUw"
      },
      "source": [
        "# <a name=\"preparation\"></a> I - Préparation du dataset\n",
        "\n",
        "- (a) Changer la forme des images de ```X_train``` et de ```X_test``` en des vecteurs de tailles **28*28** à l'aide de la méthode ```reshape```."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "B0VGyYJrj-HA",
        "outputId": "d22a2120-882f-44c8-fc4d-fa2d9302ce30"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Shape of X: (60000, 784)\n",
            "Shape of y: (60000,)\n"
          ]
        }
      ],
      "source": [
        "X_train = X_train.reshape([-1, 28*28])\n",
        "X_test = X_test.reshape([-1, 28*28])\n",
        "print('Shape of X:', X_train.shape)\n",
        "print('Shape of y:', y_train.shape)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BOs5qIx90e_e"
      },
      "source": [
        "- (b) Pour une meilleure performance du modèle, réduire les pixels des données ```X_train``` et ```X_test``` afin qu'ils soient compris entre 0 et 1.\n",
        "> - Diviser **l'ensemble de l'échantillon** par 255."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "NLGuXoPlF6_Q"
      },
      "outputs": [],
      "source": [
        "X_train = X_train / 255\n",
        "X_test = X_test / 255"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pDQ0TQH705aV"
      },
      "source": [
        "- (c) Transformer les labels de ```y_train``` et ```y_test``` en vecteurs catégoriels binaires (one hot) grâce à la fonction to_categorical du sous-module ```np_utils``` de ```keras```.\n",
        "\n",
        "- (d) Extraire dans des variables appelées respectivement ```num_pixels``` et ```num_classes``` le nombre de colonnes (pixels) de ```X_train``` et le nombre de colonnes (classes) de ```y_test```."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "CYVq19QqF7qB"
      },
      "outputs": [],
      "source": [
        "y_train = np_utils.to_categorical(y_train, dtype = 'int') # Veiller à n'exécuter cette instruction qu'une seule fois\n",
        "y_test = np_utils.to_categorical(y_test, dtype = 'int')   # Veiller à n'exécuter cette instruction qu'une seule fois\n",
        "\n",
        "# La méthode *np_utils.to_categorical* transforme les chiffres en vecteurs\n",
        "# remplis de zéros, sauf à la position du chiffre en question qui vaut 1\n",
        "# Par exemple, 3 devient (0,0,0,1,0,0,0,0,0,0)\n",
        "\n",
        "# Ainsi y_train et y_test deviennent des matrices de mêmes nombres de lignes mais avec autant\n",
        "# de colonnes que de classes (chiffres uniques) qu'elles possèdent.\n",
        "\n",
        "num_pixels = X_train.shape[1]\n",
        "num_classes = y_test.shape[1]"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BlTiqhRqif2n"
      },
      "source": [
        "# <a name=\"batch_size\"></a> II - Influence de la taille des batchs\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OmTA0PJj1lPJ"
      },
      "source": [
        "## <a name=\"fixed_epochs\"></a> A - Nombre fixe d'epoch"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "R55E4aSya7v_"
      },
      "source": [
        "- (a) Définir une liste ```batch_sizes``` contenant les valeurs ```[1, 20, 50, 200, 500, 2000, 4000]```."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "EEQypDu_dadH"
      },
      "outputs": [],
      "source": [
        "batch_sizes = [1, 20, 50, 200, 500, 2000, 4000]"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LvcHdhpIbLOE"
      },
      "source": [
        "Nous allons utiliser le modèle suivant :\n",
        "\n",
        "\n",
        "```\n",
        "  model = Sequential()\n",
        "  model.add(Dense(units = 20, input_dim = num_pixels, kernel_initializer ='normal', activation ='tanh'))             \n",
        "  model.add(Dense(units = num_classes, kernel_initializer ='normal', activation ='softmax'))\n",
        "  model.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])\n",
        "```\n",
        "\n",
        "\n",
        "\n",
        "- (b) Pour chaque taille de batch défini dans ```batch_sizes```, définir et entraîner le modèle sur les données ```X_train``` et ```y_train``` grâce à la méthode ```fit``` :\n",
        ">  - L'entraînement devra se faire sur 5 epochs (paramètre ```epochs```).\n",
        ">\n",
        ">  - La performance du modèle devra être évaluée sur un échantillon de validation contenant 20% des données (paramètre ```validation_split```).\n",
        ">\n",
        ">  - La sortie des entraînements devra être stockée dans une liste nommée ```training_histories_fixed_epoch```.\n",
        "\n",
        "- (c) Stocker dans une liste ```times_fixed_epoch``` la durée de chaque entraînement."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Pc8BGQnQGiU1",
        "outputId": "631c27ec-3a81-4ec7-8097-ec7b0c4879a6"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/5\n",
            "48000/48000 [==============================] - 73s 1ms/step - loss: 0.4925 - accuracy: 0.8680 - val_loss: 0.2450 - val_accuracy: 0.9297\n",
            "Epoch 2/5\n",
            "48000/48000 [==============================] - 60s 1ms/step - loss: 0.2283 - accuracy: 0.9341 - val_loss: 0.2291 - val_accuracy: 0.9340\n",
            "Epoch 3/5\n",
            "48000/48000 [==============================] - 62s 1ms/step - loss: 0.2013 - accuracy: 0.9394 - val_loss: 0.2059 - val_accuracy: 0.9413\n",
            "Epoch 4/5\n",
            "48000/48000 [==============================] - 60s 1ms/step - loss: 0.1788 - accuracy: 0.9470 - val_loss: 0.2039 - val_accuracy: 0.9426\n",
            "Epoch 5/5\n",
            "48000/48000 [==============================] - 60s 1ms/step - loss: 0.1686 - accuracy: 0.9498 - val_loss: 0.1918 - val_accuracy: 0.9453\n",
            "Epoch 1/5\n",
            "2400/2400 [==============================] - 5s 2ms/step - loss: 0.8509 - accuracy: 0.8002 - val_loss: 0.2527 - val_accuracy: 0.9287\n",
            "Epoch 2/5\n",
            "2400/2400 [==============================] - 4s 2ms/step - loss: 0.2456 - accuracy: 0.9294 - val_loss: 0.2097 - val_accuracy: 0.9410\n",
            "Epoch 3/5\n",
            "2400/2400 [==============================] - 4s 2ms/step - loss: 0.1926 - accuracy: 0.9436 - val_loss: 0.1862 - val_accuracy: 0.9477\n",
            "Epoch 4/5\n",
            "2400/2400 [==============================] - 4s 2ms/step - loss: 0.1637 - accuracy: 0.9529 - val_loss: 0.1788 - val_accuracy: 0.9489\n",
            "Epoch 5/5\n",
            "2400/2400 [==============================] - 4s 2ms/step - loss: 0.1517 - accuracy: 0.9558 - val_loss: 0.1722 - val_accuracy: 0.9509\n",
            "Epoch 1/5\n",
            "960/960 [==============================] - 3s 2ms/step - loss: 1.0987 - accuracy: 0.7791 - val_loss: 0.3076 - val_accuracy: 0.9173\n",
            "Epoch 2/5\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.3057 - accuracy: 0.9150 - val_loss: 0.2449 - val_accuracy: 0.9322\n",
            "Epoch 3/5\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.2325 - accuracy: 0.9349 - val_loss: 0.2120 - val_accuracy: 0.9392\n",
            "Epoch 4/5\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.2064 - accuracy: 0.9416 - val_loss: 0.1956 - val_accuracy: 0.9452\n",
            "Epoch 5/5\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.1844 - accuracy: 0.9474 - val_loss: 0.1853 - val_accuracy: 0.9487\n",
            "Epoch 1/5\n",
            "240/240 [==============================] - 1s 4ms/step - loss: 1.6593 - accuracy: 0.6511 - val_loss: 0.5910 - val_accuracy: 0.8810\n",
            "Epoch 2/5\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.5424 - accuracy: 0.8778 - val_loss: 0.3730 - val_accuracy: 0.9057\n",
            "Epoch 3/5\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.3682 - accuracy: 0.9062 - val_loss: 0.3066 - val_accuracy: 0.9177\n",
            "Epoch 4/5\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.3066 - accuracy: 0.9190 - val_loss: 0.2728 - val_accuracy: 0.9251\n",
            "Epoch 5/5\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.2705 - accuracy: 0.9262 - val_loss: 0.2520 - val_accuracy: 0.9298\n",
            "Epoch 1/5\n",
            "96/96 [==============================] - 1s 7ms/step - loss: 1.9594 - accuracy: 0.5268 - val_loss: 1.1011 - val_accuracy: 0.8225\n",
            "Epoch 2/5\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.9725 - accuracy: 0.8313 - val_loss: 0.6325 - val_accuracy: 0.8863\n",
            "Epoch 3/5\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.6062 - accuracy: 0.8784 - val_loss: 0.4679 - val_accuracy: 0.9016\n",
            "Epoch 4/5\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.4641 - accuracy: 0.8949 - val_loss: 0.3873 - val_accuracy: 0.9108\n",
            "Epoch 5/5\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.3890 - accuracy: 0.9087 - val_loss: 0.3411 - val_accuracy: 0.9158\n",
            "Epoch 1/5\n",
            "24/24 [==============================] - 1s 20ms/step - loss: 2.2238 - accuracy: 0.3114 - val_loss: 1.9678 - val_accuracy: 0.6554\n",
            "Epoch 2/5\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 1.8943 - accuracy: 0.6598 - val_loss: 1.6411 - val_accuracy: 0.7418\n",
            "Epoch 3/5\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 1.5826 - accuracy: 0.7385 - val_loss: 1.3455 - val_accuracy: 0.8163\n",
            "Epoch 4/5\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 1.3017 - accuracy: 0.8088 - val_loss: 1.1018 - val_accuracy: 0.8536\n",
            "Epoch 5/5\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 1.0784 - accuracy: 0.8430 - val_loss: 0.9163 - val_accuracy: 0.8706\n",
            "Epoch 1/5\n",
            "12/12 [==============================] - 1s 42ms/step - loss: 2.2557 - accuracy: 0.2042 - val_loss: 2.1132 - val_accuracy: 0.5093\n",
            "Epoch 2/5\n",
            "12/12 [==============================] - 0s 21ms/step - loss: 2.0739 - accuracy: 0.5389 - val_loss: 1.9362 - val_accuracy: 0.6453\n",
            "Epoch 3/5\n",
            "12/12 [==============================] - 0s 22ms/step - loss: 1.9031 - accuracy: 0.6392 - val_loss: 1.7673 - val_accuracy: 0.6753\n",
            "Epoch 4/5\n",
            "12/12 [==============================] - 0s 21ms/step - loss: 1.7402 - accuracy: 0.6741 - val_loss: 1.6046 - val_accuracy: 0.7219\n",
            "Epoch 5/5\n",
            "12/12 [==============================] - 0s 22ms/step - loss: 1.5795 - accuracy: 0.7145 - val_loss: 1.4507 - val_accuracy: 0.7637\n"
          ]
        }
      ],
      "source": [
        "training_histories_fixed_epoch = []\n",
        "times_fixed_epoch = []\n",
        "\n",
        "for batch_size in batch_sizes:\n",
        "  model = Sequential()\n",
        "  model.add(Dense(units = 20, input_dim = num_pixels, kernel_initializer ='normal', activation ='tanh'))\n",
        "  model.add(Dense(units = num_classes, kernel_initializer ='normal', activation ='softmax'))\n",
        "  model.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])\n",
        "  begin = time.time()\n",
        "  training_histories_fixed_epoch.append(model.fit(X_train, y_train, epochs = 5, batch_size = batch_size, validation_split = 0.2))\n",
        "  end = time.time()\n",
        "  times_fixed_epoch.append(end-begin)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XeBgSXysjE82"
      },
      "source": [
        "- (d) Tracer sur un même graphe l'évolution de la précision d'entraînement en fonction de l'epoch pour chaque taille de batch.\n",
        "\n",
        "- (e) Sur un second graphe, tracer le temps d'entraînement du modèle pour chaque taille de batch."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 373
        },
        "id": "94lCK-SdLtPF",
        "outputId": "daa64539-8a02-4349-f0d2-a63c3e2bcd05"
      },
      "outputs": [
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABW4AAAHwCAYAAAA2F/Y+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd3xkd33v/9f3TB/1vtJWr9f22msvLmvvjkMPvZluMBhMsQmkkBByk8tNAkm4QHKTm4Y7EEp+XFoSQxwMgUBIAK17wQ1c8PYmabUq0+d8f3+cM6Mzo5FWuytpJO37+XjMY86cNt85Gu2eeeszn2OstYiIiIiIiIiIiIjI0uE0egAiIiIiIiIiIiIiUk3BrYiIiIiIiIiIiMgSo+BWREREREREREREZIlRcCsiIiIiIiIiIiKyxCi4FREREREREREREVliFNyKiIiIiIiIiIiILDEKbkXktGCMucMY8875XldEREREVhZjzPONMXsXaN+OMeabxpjrFmj/HzHGfGYh9j1fjOcfjDFHjTF3LeLzrjPGTBhjQovwXP9pjHnvSWy3wRhjjTHhhRhXzXNVHY/gmI0x1xhjfjxPz/MxY8w/zse+FnKfMzzPov08RGai4FZEliz/RKJ8c40xmcDjt53Ivqy1L7fWfmG+1xURERGR6ebzPK7RjDHPGGM2zNPuPg78wFp7y6nuqF7AbK39hLX2hAPDRfZs4MXAGmvtZQv1JP7P7UXlx9ba3dbaZmttaaGes5FONGxd6cdjJvMZSossBv3VQESWLGttc3naGPMM8F5r7fdr1zPGhK21xcUc23Kk4yQiIiKLZa7ncacba+1HGj2GJWA98Iy1drLRAxERWepUcSsiy065usAY8/vGmIPAPxhjOowxtxtjjvhfu7rdGLMmsM20r/4YY/7SX/eXxpiXn+S6Zxhj/ssYM26M+b4x5vqZvrYzhzF2+l8b2+8vvy2w7ApjzAPGmDFjzFPGmJf586sqCYJfGwp8tec9xpjdwA/8+V83xhw0xhzzx74lsH3CGPNXxphd/vIf+/P+zRjzmzWv5yFjzOtO9OcnIiIipy/jtQr4A/98ZtgY8zVjTKe/rHzu8i5jzB7/fOjXjDGX+ucdo8aYTwf2dY0x5ifGmE/75y2PG2N+tWb50/552i9nqvT1z3U+7z/fo8Cls4w/5p8X7jbGHDLG3GSMSfjLvm2M+avAul8xxnzOn24zxnzWGHPAGLPPGPNxE/jKvjHmWmPMY/5YHzXGXOzPt8aYTYH1Pu9v2wTcAQyYqUrmAVPzFXJjzGuMMY/4x+4/jTHnBpY9Y4z5sH9sjxljvmqMic/y2t/tj/GoMea7xpj1gWXW/1k94T/X9cYYU2cf7wE+A6T8Mf9J4PU/aYwZMcZ8yxgzMNd91zt2xpgvAeuAf/Wf53+Ymq+9+8frW/5zPmmMuTawz4/5780v+vt9xBizbZZj82L//XfMf4+amuUzHrvZ+L8L5df2tDHmfTOsdy5wU+C4jvrzX2mMud94nyH2GGM+Fthmzm0AjDGbjTHf84/Vz40xb55l3TOMMT/yx/w9oLtm+Q5jzE/9n+WDxpjnB5bN6XfWF/ffs+PGmPuMMc8K7Kf8b0z5PfG64xynup+BAs/1NuP9zg8ZY/7X8Y6XyHxScCsiy9UqoBPvL/bX4f179g/+43VABvj0jFvDduDneCcSfwF8tt7J5RzW/TJwF9AFfAy4epbnPN4YvwQkgS1AL/DXAMaYy4AvAr8HtAPPBZ6Z5XlqPQ84F3ip//gO4Cz/Oe4D/r/Aun8JXAJcjnd8/wfgAl8A3l5eyT8xWg382wmMQ0REROQ3gdfinZ8MAEeB62vW2Y53rnIl8DfA/wJehHeO9GZjzPNq1n0K7zzto8A/G++P4U3A3wEvt9a24J3bPDDDmD4KnOnfXgpUXevAWrvBWvuM//BTwNnAhcAmvPOhP/aXvRu42hjzQj9wugz4oL/s80DR3+Yi4CVAuVDgTXjnke8AWoHXAMMzjLU8pkng5cB+/+vuzdba/cF1jDFnA/8P+G2gB/g2XpAZDaz2ZuBlwBnAVuCaes9njLkC+Ajwen9f/+3vO+hVeKH3Vn+/L61ZjrX2s8CvAYP+mD9qjHkh8El/m35gF/CVuex7pmNnrb0a2A282n+ev6jzsr4C7MV7H74R+IQ/lrLX+Ou0A99ihs8Wxphu4J+BP8R7Hz4F/Epg+VyO3UwO+6+9FXgX8NfGD/WDrLWPUX1c2/1Fk3jHph14JfB+Y8xr5/jc5fE3Ad/D+9zTC7wFuMEYc94Mm3wZuBfvWPwZgd8nY0z588PH8T5rfBj4J2NMzwn+zgJcAXzd38+XgduMMRF/2VPAc4A24E+AfzTG9M9ynGb6DFT2bOAc4FeBPzaBP4CILDhrrW666abbkr/hBZUv8qefD+SB+CzrXwgcDTz+T7yv6IF3QvpkYFkSsMCqE1kXL3wtAsnA8n8E/nGOr6kyRryTVBfoqLPezcBfH++4+I8/Vn5+YIM/1o2zjKHdX6cNL1jOAM+qs14c74PVWf7jvwRuaPT7QjfddNNNN910W/q3mvO4x4BfDSzrBwp4bfzK5y6rA8uHgSsDj/8J+G1/+hpgP2ACy+/C+0N6EzAKvAFIHGd8TwMvCzy+DthbZz2DF4SdGZiXAn4ZePwGYA8wBDzbn9cH5ILjAN4K/NCf/i7wwRnGZoFNgcefBz7uTz+/dpw154J/BHwtsMwB9gHPD/xc3h5Y/hfATTOM4w7gPTX7SgPrA+N8dmD514A/mGFf1wA/Djz+LPAXgcfN/ntiw/H2fZxjV3nf+Y/L768wsBYoAS2B5Z8EPh84jt8PLDsPyMzwPO8Adta8T/Yy9Xli1mNXs6/KGGd4rttmeb1Vx3WGdf4G/3NF7XMx/TPQj/3pK4H/rtnPzcBH6+y//PmoKTDvy4H35O8DX6rZ5rt44e6J/M5+rOaYO8AB4DkzrP8AcMUM77/ZPgOVj9GawLy7gLfMNj7ddJvPmypuRWS5OmKtzZYfGGOSxpib/a+3jAH/BbSbma8ae7A8Ya1N+5PNJ7juADASmAfeiXpdxxnjWn9fR+tsuhbvr8YnqzImY0zIGPMp/6tDY0xV7nb7t3i95/KP9VeBtxtjHLwPG186hTGJiIjI6Wk98C/+16RH8YLcEl64WXYoMJ2p8zh4zrbPWmsDj3cBA9arSL0Sr7rugPHaPm2eYUwDVJ/D7ZphvR68P+LfGxj/d/z5Zf8KhICfW2vLF0BaD0T8cZS3uxmvehFO/VxvJgMEXou11sV7nasD6xwMTKeZ+Xx4PfC3gfGP4AWUJ7Ov441zAi+wn8u+T/bYlc/jxwPzdh3nOeMztBWoev/478fg+2kux64uY8zLjTE7/RYFo8ArqGk9cJzttxtjfmi8Vm3H8H4f5rx9YPzby+P3x/E2vEKWWgN4hSnB/sXB36f1wJtq9vVsoP8Ef2eh+pi7TFVPY4x5h/HazJWf4/xZXveMn4ECTva9LXLKFNyKyHJlax7/Lt7XV7Zba1vx2glATX+peXYA6DTGJAPz1s6y/mxj3OPvq73OdnvwvrpXzyTeB4iyeidQwWN1Fd7Xil6EV2W7ITCGISA7y3N9Ae8k7VeBtLV2cIb1RERERGayB++r0O2BW9xau+8k97e6pt3VOrwqXKy137XWvhivqvdx4NYZ9nGA6nO4dTOsN4QXHG8JjL3NBi7EBvxvvDC63xjzVn/eHryK2+7Adq3W2i2B5TOdf6WZ+Vyv9ny41n68oAwA/zitxau6PVF7gPfV/NwS1tqfnsS+jjfOJrw2ZHMZ52zHbrbjsx/v3LslMG/dHJ+zVtX7J3Ccg2M84WNnjInhVZj/JdBnva/1f5uZP9/Ue71fxmvzsNZa24bX3/VEPx/tAX5UM/5ma+3766x7AOjwf4Zlwd+nPXgVt8F9NVlrPwUn9DsL1cfcAdYA+43XP/hW4DeALv+4PRx43bXH6XifgUQaSsGtiKwULXgn0qPGu8DFRxf6Ca21u4B7gI8ZY6LGmBTw6pMZo7X2AN7XqG4w3kXMIsaYcrD7WeBdxphfNd4FPVYH/vr8APAWf/1teP25ZtOC98FhGO9DwCcCY3CBzwH/13gXawgZY1L+SSN+UOsCf4WqbUVEROTk3AT8bz9cwe9tecUp7K8X+C3/XOhNeH39v22M6TPexV2b8M59JqjuWRn0NeB/+udga/D68E7jnyvditdntNcf/2pjTLnf6nPx+pC+A++r339vjFntn+f9O/BXxphW/3zuzECv3s8AHzbGXGI8m8zUxaseAK7yz8tehtcbuOwQ0GWMaZvldb3SP4eM4BUR5ICTCVtvwjtGW/zX2uYf7/nw//DOdS/0zzs/Adxpp/oKz2a2Y3cI2FhvI2vtHrzj8EljTNwYsxV4D17bsxP1b8AWY8zr/Yrc36I6YD/ZYxcFYsARoGi8CyS/ZJb1DwFrTHUP4xa8yuKs8a6bcdWcX9WU24GzjTFX+79nEeNdMHBan9fA56M/8T8fPZvqz0f/CLzaGPNS/z0dN96Fp9ec4O8swCWBY/7b/jY78VouWLzjhjHmXXgVt2VVx+l4n4FEGk3BrYisFH8DJPD+YroT72tri+FteL3NhvGa7H8V76ShnuON8Wq8fl6P412I4LcBrLV34V+MADgG/IipqoQ/wvvr8FG8xvtfPs54v4j3daV9wKP+OII+DPwMuBvva1x/TvX/FV8ELuDkTmpFRERE/havAvDfjTHjeOci209hf3fiXchsCK/a9Y3W2mG885cP4VVWjuAFnvUqBME7h9oF/BIvYJ3tD9S/DzwJ7DRe26nvA+cYY1rxzpN+w1q7z1r733h/fP8HvwLzHXhB3KN4523fwKsqxFr7dX/sXwbG8fqYdvrP90G84Kv89fTbygOx1j6OF3o+7X8lfCA4UGvtz/EuLvv3/vF5Nd7FuvKzvL66rLX/gnde+BX/dT+Md3G0U2at/T7eOe0/4VVsnol3Aay5bDvbsfsk8If+sflwnc3fivfts/3Av+D1bP3+SYx/CHgT3oXrhvHejz8JLD+pY+e3cfgtvAD+KF7o+q1ZNvkB8Ahw0Bgz5M/7APCn/u/aH/v7OiH+OF6C9zPZj9c24M/xQuV6rsL7nR7BK1T5YmBfe/C+/fcRvGB1D94FmB1O7HcW4Jt4rRWO4n2Oer21tmCtfRSv0GQQL6S9gMDPg/rH6XifgUQaxlS3AxIRkVNhjPkq8Li1dsErfhvBGPMO4Dpr7bMbPRYRERE5vRljrsG7mJLOS0REZEXSXxBERE6B/zWhM/2vvL0M7y/Itx1vu+XIeL18PwDc0uixiIiIiIiIiKx0Cm5FRE7NKuA/8Xow/R3wfmvt/Q0d0QLwe7cdwfu60fHaMYiIiIiIiIjIKVKrBBEREREREREREZElRhW3IiIiIiIiIiIiIkuMglsRERERERERERGRJSbc6AGcqO7ubrthw4ZGD0NERETktHLvvfcOWWt7Gj0OWVg61xYRERFZXLOdZy+74HbDhg3cc889jR6GiIiIyGnFGLOr0WOQhadzbREREZHFNdt5tloliIiIiIiIiIiIiCwxCm5FRERERERERERElhgFtyIiIiIiIiIiIiJLjIJbERERERERERERkSVGwa2IiIiIiIiIiIjIEqPgVkRERERERERERGSJUXArIiIiIiIiIiIissQouBURERERERERERFZYhTcioiIiIiIiIiIiCwxCm5FRERERERERERElhgFtyIiIiIiIiIiIiJLjIJbERERERERERERkSVGwa2IiIiIiIiIiIjIEqPgVkRERERERERERGSJUXArIiIiIiIiIiIissQouBURERERERERERFZYsKNHoCIiIiIzJFbglIB3II3nWhv9IhERERERGSBKLgVERGR04e14BahlPcD0MB0ORCtmp7reiezTR5KxerpUt5/PMO0dadeS9ta+J2HG3csRURERERkQSm4FRERkRMTrPqsF0TWDR+PE17OdZtS3g9Eg9P5OuvNEK66hYU/Pk4EQv5t2nQUQuHAdAQi8TmsVzMdikK8beFfi4iIiIiINIyCWxERkUYph5SlHBTz/nSe+lWaswWWM1V2znW9WSpA64WrwarPhWCc6SGlE6mZDoSc4TjEWvxlYX9Zveno9DB11vUC01XbzbKeEwZjFvb4iKwQG/7g3xo9hIZ75lOvbPQQREREZAlTcCsiIitb8KvxxdxUMFosB6Y5P7wMhqe5qeXBdUsFf/2aoLVqv8H91Xm+4PRCBqBObdg4WwVoBMKxGULKuVSNzhCozrreTNtEwAkt3HFZwlzXUrKWkmtxy/cu0+aVpx1jWNuZbPSwRURERERkgSi4FRGR+WHtVJVnJcCsCTsr0/XCzuNN1wk+pwWtM+wDO3+v0zgQikE46oeRwenoVAAabYZk1/T5oai/fsyfF5m+v2nVoHUqRY9XNbpAVZ+2HB7amlCxMi8QPtYLHYuBUNISCChzlGw2sC7VAWZNaFny9+26Ncv9ea6l7vySS515wbFTZ97Uc7jB561azgzHILDcWoqui3WLlChRskVciri2hGtLYEoY44LxpsHFlKf9+cafjynRkUhy9+98aEF+ziIiIiIi0ngKbkVElhtrZ6jsDFaN1laKnkTYOVt4OlOV6nwyoZqwc4bgM9I6Q5AaC1SSBqdr9lcVntZO1zxfuWq08qOwFEqWQsklX3TJB+4r8wLzCyXrPy5RKFpy5fl5l3x6aptiTRg5FVqCa11KbpaSzUwLLYNBaHVoyYwBZyWAnSV0dO1UKGnnMQOfO9cLLpljuElg2inhGIvjlHCMi+O4GMfFMSWM42LK8/z9GicQkJqp+eXnJjw1bcshKt60peRNT7sVcSvTXpV1yL+dilikB1BwKyIiIiKyUim4FRE5VcUc5CYgP+7f+7dc8H4c8pMn8PX8GapUy7f55ITnEHxGIZI8sarRYNhZL0itV6UanA58Xb5Ymgo9c6XSVABa9MLOXLE6KC2UqkPUyry0S75m28o6lfUK5Iu56fuuDV/97eaTMRANOYQdg+MYQo4hZPxp4z12HKbPq9yDU7m3OI5LOFwTVOLfB6bBxTjFqYDS8e7LAWk5kCxPWz8YLYeS04NKdyqstKWpqlJKuHZqumSLuLZIyZanvfupW4miW6yEnfOp3h5DJkTYCU/dzNR0xInUzI9Vr1tex4SnzQ9uV7WfOs9Vd/kMy2Kh2LwfFxERERERWToU3IrI6adUrA5Za0PXuuHreP0wNjcx96vUO2HvIkrHCz5jzXOrGp2xGnUOVaPBaccBvIrK6gBz9orRfNFbv1ATjuby1duXQ89cZfty8JolX0zXrVItB6blaXeeqzyjIYdo2CESMkTD5WmHaMghVp4OOzTFwpXlUX95JGyIhkL+PFPZj3GKOKEimAKOU8SYApYC1vHuXQpYk6dkC1jylChQtDmKbp6izVFw8+TdHPlSnnwpT9EWKbrVt4JbmDY/H1zH+uu4xdkPgEv95PIkzBhU1oSescqyeN1tKqHkDPs7XigaMTMEonMITCNOhJATwjHO/BwUERERERGReaDgVkSWPtetCVKPF7r61a0zrVPMzu15TcgLUaMt/n0zRJugudebLs+rXWfa4xbvPhzFWj/s9APQclCZqw06iy65OqGoF3D64Wl6lsrSSuBaIl+cpFAarxO+VleoFuc5HQ0700PRSgAaCE1bImFiddaL1ISo0eB0OXANharC12ioHKKWsKYAFHBNHmsLlEwe13phab6UJ1fKkS1lyBVzZEtZ73HRu6/cAssmgssKOXKZqWW5U2gTETIhYqEY8XCcaChKPBSvW10ZD8frBpvTK0EXNxA1C9RLV0RERERE5HSn4FZE5p+1UEjPUq06l8eTU/MKk3N8YjMtULXRZtzWNZQiTRQjTZTCTRRCycotF2oi5yTIOkmyTpIMCTImQdokyLiRSg/ScpBaCUSLLvnx6orUXHBZya8mLR6sCkjn+6v1jqEqDPUqQqvDz3LlaHto+vKpilGnurK0EoJWh6jBdSMh480Lhfwq1MC+Qw6OY7DWUnSLMwaj3uOMt7w4NS/v5ivrZotZxkp+SJr3AtPK9v52lQDV34c9yYuRGQzxcJxYKFYJU2OhGPFQnFg4Rnuk3Zuus6y8Tb1l5W1ioVhl3fKyiBOZ1/eEiIiIiIiIrAwKbkXEC1qL2ZrK1MkTDl1tYL6ZY3BWDCUohpPkQ03kQ0nyTpJcqJWsWUU2kSCTTJDGC1InSTBp40zYOOPEmXBjjLlxjpVijLoxxosRciXIT7rkRr2QtHRCVaRp/1ZfOdAsB5hVIac/3RKfqh6tXhaqTMfK1aYhh2g4UDFat7K0fiBbvg85J1btWHALlXAzGHzmStmqEDRfypMtZckUcxwtr5OtrjCtDUyD4Wx5+1wph2tPPqyuG4r681qjrfSEeirBaLlaNRiUVkLYOaxTrnRVBamIiIiIiIgsBQpuRZarYh6bGyefGaOYHqOYGaOYHcf1b6VsOUj1gliTn8ApTOIUJggVJgkVJwkXJgmXJokW0ziU5vS0OaJk/BA1TZwJm2CCOBNuF2N2tResEmfSJpj0l08SZ8IPXYPz0sRxqd9TsnyRJi/oDNUPQyN+NWnYoSPsEKtXNRp4HIuEjr/OLKHsfAd6JbdUt2q0HHxOlOdnszUVqrUVq9XTleC1zrKSndvPuZ6oE60fioZiNEeb6Qx11l0WnJ5zNWo4TtSJKkQVERERERGR05aCW5FTYK1laCLPyGS+0k+0+uvygR6ihTxudhLy49i8H6TmJypBarjo3SLFNJHSJJFShpibJuamibtp4m6GhM2QJEPSZoiaIgaI+beZ5G3Iq1QlzpgfmHrhaheTdrUXqBInTYKckyDnJL3K11CSQriJYihJMdxEKdKEG2kiFI76YWZoegVqIOhsrlk2tU5oWjBab/uwY5ZsaFdwC0zkJ5jITzBWGGM8P854ftx7nA88Lkw9nshPMFGYIFPMVMLU415AahYRJzLjV/ST4SSdsc6qr+RHQ9GZA9M5VKNGQ1FduElERERERERkESm4FZlFOZjdezTN3qMZ/+ZN7x8ZJzH6BFvt45xhDtJEhmaTpYkMSZOjhwxNZGkyGZrJkjD5OT1nCYeMSZI1cXJO0gtSo0kmQ92MhpoohINBajNupAk32oyNeBfBMrFmTKwZJ96ME28hGk1UVZV2hBz6ymFqIEQ90a/cL2e5Uo7x/HhVqFp+PFGYqASvY/mxyrLKrTBOppiZdf8GQ3OkmZZoCy3RFpqjzfQ399McaSYejk8LTIPVpnOpWI2FYoSc0CIdLRERERERERFpBAW3clqz1jI8ma8KZPceTbNnxLvfN5ohW/D6c8bIc4F5mufGn+S14V+wpfQYybB30axiOEkx0jwVpEY6ILYWos2YaDOleDOTsRZCcf+WaMWJtVRdRIuo9zgUjtNsDM2NPDBLmLWWdDFdHab6geq0eYHK12BQW3ALsz5H2IQrgWs5fO1OdFemq26R6fOaIk2qThURERERERGRU6LgVlY0ay0jfjC7JxDMBqtny8FsWXsywpqOBFu7DNf27eW84qOsm3iA1pGHcdw8WKB9M6x7E6y7HNbtINy+jvAS/Vr/UlNyS1VVrbXtBKoqYWurXwtee4Lj9WmNhWJVwWprtJXVzasrYWxrtLUSuFYeR1sqVbKJcGLJtmkQERERERERkdODgltZ1oLBbG3VbHleplAd8rUlvGD2zJ4mnnd2D2s6EqzpSLIhcpQ14w+Q2H8X7N4JzzwKWHDCMHAR7Pg1WJeCdTsg2dmYF7wEFEqFqurW2nYCte0GaqthJwuTx32OpkhTJURtjbbSm+xlY/vGutWt9apeo6HoIhwJEREREREREZGFo+BWljRrLUfThbqBbHk6na8OZlvjYdZ0JDmju4nnBoLZNR0JVnckaI1HwHVh6Oew66deSHvXTji229tBtAXWXgZbXueFtKsvgWiyAa9+/llryZayM7YTqLqwVp0Lb43nx8mWsrM+h2OcqtC1OdrM2ua1J9RmIOzonyYROX1Ya6FQwM3nsbkc1r/3HuexBf9xZVkem89johHaXv3qRg9fREREREQWiNIRaShrLaPpwvQes7MEsy3xMGs7kmzoauLZm8rBrBfOru5I0JaITH+iYh723w/3DHpB7Z6dkDnqLWvu8yppU78O61PQuwVCS/NXw7Uu6UK6bjuBqnYDM/V7LYxTdIuzPkfYCVdaB5SD1b5k39S8QEuBcjBbnm6JtpAMJ9VmQESWBWstFIvYfN4LSQPBqZvLVQJSmz/OvHzB3y4Qtlb2VW9eflpIezLCq1YpuD3NGGPiwH8BMbzz+G9Yaz9qjDkD+ArQBdwLXG2tzRtjYsAXgUuAYeBKa+0zDRm8iIiIiJywpZlOyYphreVYxgtm94zUr5qdrA1mY2HWdCZZ39XEr2zqrlTLlsPZusFsrewY7LkLdg96t333QtGvFO06Cza/CtZ7/WnpOAMWKWgsusWptgL1wtUZAtdyNexEfgKLnfU5EuFEVf/WjngH61rXeSFrpLl+5WsgpI2FYgpeRWTB2VKpurJ0pmrTqoA0XxWcVoem5Xm1AWlgnWBoWg5MXff4gz2eSAQnEsHEYphoFBOL4cSimEi0Ms/pSOLEYlXzTCyKE41iorHp82Ixb3404m0XDWwXjXrz4vFTH7ssNznghdbaCWNMBPixMeYO4EPAX1trv2KMuQl4D3Cjf3/UWrvJGPMW4M+BKxs1eBERERE5MQpu5ZQEg9l6bQz2Hs0wkauu8GyOhVnTkWBtZ5LUmV2V6RMKZmuNHZgKaXcPwqFHwLpgQtD/LNj2Hq+adu0OaO6Zp1fv9Xt9aOghnj729PEvrJWfIF1MH3efLZGWShVrS7SF/ub+mXu71rQbaI42E3FO4viJyGnDui62UBNkBr6Of2LVprNXlrr5mrA1EJxSnL36f04cxwtJg6FmOSj1A85QS6s/L4ITDawTCwSglTDVD0krAekc5znOqb8WkTmw1lpgwn8Y8W8WeCFwlT//C8DH8ILbK/xpgG8AnzbGGH8/IiIiIrLEKbiVWVlrGcsU2TrmfGYAACAASURBVDNDj9l9RzOMzxDMrulIsmNjV1WP2bUdSVoT4VOr6LQWhp6A3X5/2t2DcPQZb1kkCWsuhef9vldNu+ZSiDad/HNNe2rLU6NPsfPATgYPDHL3wbvJFDOV5SETmhaqbmjdUAlVgy0F6rUbaAo3EXJC8zZeEVm6rLXYTAY3ncadnKy+z2brVJvWVpZWz3PzuToB6fRqU1sonPrgjakKQJ1IdXDqRKM4ySShjo5KFakXptapLA3Mc/z15lRtGo1iwjqNkdOPMSaE1w5hE3A98BQwaq0tn5DtBVb706uBPQDW2qIx5hheO4Whmn1eB1wHsG7duoV+CSIiIiIyR/rEI37FbHUgu2dk5mC2KRqqVMhOBbNT4WxbIjK/X7UvFeDAQ9VBbXrYW5bs9ippL7vO61O76gIIzW/F6ZH0EXYe2Ond9u/kcOYwAOtb1/OaM19Dqj/Flu4ttEZbSYQTajMgskLZUgk3k/HC1clJ3Mly0BqcTk8trw1ka+el0yf1Nf1yeFlVWRqrriQNtbRMryytqTatbHeC1aZONAKRef53XkTmzFpbAi40xrQD/wJsnod93gLcArBt2zZV44qIiIgsEQpuTwNj2QJ7RzIzVs2OZ6uD2WQ0xFo/hN1+RudUtWznAgWztXITsPcuL6Td9VPYew+Uq1o7N8LZL/OqadddDl1nznt/2nQhzb2H7mXwwCCD+wd5cvRJANpj7Wzv306qP0VqIMVA88C8Pq+IzC+bz1eC0tLkJDadpnTccDU9FcTWrGczmeM/qc8kkzhNSZxkEqepiVCyiVB3F9GmdZhkklBTE05TU2V55d6fNvF4dWVp+T4S0dfyRQQAa+2oMeaHQApoN8aE/arbNcA+f7V9wFpgrzEmDLThXaRMRERERJYBBbcrQDmYrd9jNs1YnWC23Lbgsg0dgYt/efftyUWupJo47Pem9YPagz8DWwLjeBW0l7zTq6ZdtwNaVs3705fcEo8MP+K1P9g/yANHHqDoFok6US7uu5hXbXwVqYEUmzs34xgFJiILwVqLzWbnXrE6h2rXObcEcJyq0LR8H+nvnyFcTeIkg/c108mEwlURWRDGmB6g4Ie2CeDFeBcc+yHwRuArwDuBb/qbfMt/POgv/4H624qIiIgsHwpul4HxbKFuIFuedyxTHU4kIiHWdnpB7LYNHVWh7JqOJB2LHcwGWQsjT3tB7S7/QmIjT3nLwglYsw2e87teSLv2Moi1LMAQLHvG91SC2jsP3sl4fhyAczvP5epzr2bHwA4u7r2YeFhX7Bapx7quH5SeRIuAaSHs5Am1DTCRyPQq1aYmwr29NfMCAWttMBuYNrGYvvYvIstFP/AFv8+tA3zNWnu7MeZR4CvGmI8D9wOf9df/LPAlY8yTwAjwlkYMWkREREROjoLbJWAiV/SC2EDV7J7jBLPlvrIXr6sOZtd2NjiYrVUqwqGfTYW0u3fCpNcjlkSnV0l7yTXeff+zIBxdkGGMZke58+CdDO4fZOeBneyb8L5BuKppFS9a9yJSAym292+nM965IM8v0mi2UKgboAZbCFRaCcyhhYBNp+f83CaRmFa5GursILJ2Td0gtdJeoF61azKJiS7MvxMiIkudtfYh4KI6858GLqszPwu8aRGGJiIiIiILQMHtIpjIFdl3NMOekXR1O4NRb3o0XR3MxiNOJYi9aF273282WQlrO5uiSyeYrZWf9HrS7t7pXUxsz91QmPSWta+HM1/oVdOuvxy6zoIF+jpxrpTjgcMPMLh/kMEDgzw2/BgWS3OkmUtXXco7t7yTVH+K9a3rl+6xlNOWtRaby81cnXoSVa42n5/bkxtTtzI1smrVrD1Zp1fA+tOJBCYUWtgDJnIaKZVcirkShZyLW3Jp7U40ekgiIiIiIrJAFNzOg8lcccY2BnuPpjk6SzB74dr2aT1mu5ZyMFtrcsgPaf2K2gMPglsEDPSdDxdeBetTsHYHtK1esGG41uWJo09Ugtr7Dt1HtpQlbMJs7dnK+y98P6n+FOd3n0/Y0dteFpa11qtmHRqiODJCcWiI0vAwxaNHcSdm6Mta2zagVJrbk0UihJJJTJN/wSu/NUC4p/v44WqdeSYeXz7//ogsUdZaioVywOrf8t59sTydLVHMuxRyRQo5t7JO1Ta5EsV89T7c4lR70pbOOO/4xOUNfKUiIiIiIrKQlGDNwWSuyL7RmS7+lWFksrqSLRZ2KkHs1jVtVdWyazqSdDcvo2A2yFo4+sxUNe3unTD0C29ZKAarL4Ff+aDX9mDNpZBoX9DhHJw8WAlq7zxwJyPZEQDObDuTN5z9BlL9Kbat2kZTpGlBxyGnB1sqUTp2bCqEHR6hNDxEcXiE4vAQpaFhL6QdHqI0PILN5erux8Tj09sGtLcTWb16DhfBKi/3pkNNTWobIHIK3JJLIX/8gLVqXiVgnSV0zZfgBC7/5IQNkWiISMy7hf3pZGvUexwLBZY7RGJhwlGHeHNk4Q6OiIiIiIg03IIGt8aYlwF/C4SAz1hrP1WzfD3wOaAH74IJb7fW7l3IMc1VJl/iylsG6waz0UAwe/7qtsr02uUezNZyS3Dokalq2t07YfyAtyze5gW0F17l3Q9cBOHYgg5nIj/B3QfvZvDAIIP7B3lm7BkAuuJdpAZSpPpT7OjfQV9T34KOQ1YON5+nNDJCcWi4OoQdHqE4HJw3TGlkpP7Fs8Jhwp2dhLq7CHd2Edu4sTId7u4i1NVNuKuTUFcX4Y4OTERBi8iJsNZSKrpTFaiBoHRuAWt1BetU6FqiVJzbBfHKvADVmRawJloidQLW6nVqQ9dILEwk5hCOhQiFFqZtkIiIiIiILG8LFtz6V7u9HngxsBe42xjzLWvto4HV/hL4orX2C8aYFwKfBK5eqDGdiHjEoa81XhXMlqtmu5tiOM4KCGZrFTKw714vpN01CHvugvy4t6x1DWx4ttefdt3l0LN5wfrTVobjFnh46GGvqnb/ID8b+hklWyIRTnBx38W88ew3khpIcVb7WSsjKJdT5rUoSFMaGfbCWP8+GMYGp92xsbr7MYkE4a4uwl1dRFavJrF1a3UY69+Hu7pwWlsxC/y7ILIcuK6t+lp/caYwNefOHrAGtivPsydSveoYIvHqwNSrTo3S0uWHrlE/ZJ0xYJ0+LxxxMCvx/34REREREVmyFrLi9jLgSf8qtxhjvgJcAQSD2/OAD/nTPwRuW8DxnBBjDLe+Y1ujh7Gw0iOw507Y5bc92H8/uH4/3t7zYOubvJB23Q5oX7vgw7HW8suxXzK4f5Cd+3dy96G7mSxM4hiHLV1bePf57yY1kOJZPc8iGtLXw08X1nUpHTvmtSeoCmNrpoe9e5vN1t2P09ZWCWNj55xDU1cXoa5Owl3dfhjbSbi72wtjk8lFfpUii8Nai1u0leC0XsA6ra9qVcDq1t2umCtRLJxg9WrEqROwhmhuikyFp9EQ4ZhzQgFrKKw/pIiIiIiIyMqwkMHtamBP4PFeYHvNOg8Cr8drp/A6oMUY02WtHV7AcZ2+Rnd7AW05qD3ymDfficDqiyH1AS+oXXsZJDsXZUjDmWF2HtjphbUHdnIofQiAtS1recUZryA1kOKyVZfRFmtblPHI4rCFAsWRkUrYWglea4NZ/4JeFIvTdxIKEers8ILXri6iG9b7051ee4JgGNvRoV6wsqxY1wtXgxevqhew1oasVVWrleUuhWyx0svVdedevmocU2kNEKxQjSfDRDpiU/PqVLBOha5hv0VAIGCNhlS9KiIiIiIichyNvjjZh4FPG2OuAf4L2AdMu5S6MeY64DqAdevWLeb4li/X9YLZcki7eyeM+e2DY61eOHvBG73+tKsvhkhiUYaVKWa479B9lYuK/eKod3Gz1mgr2/u3V3rVrmlZsyjjkfnjptNVla/BMLY4MuxdvMufVzp2rO4+TCxGuKuLUHc3kVWriJ+/JRDGdk1Nd3cTamtTiwJZskoll3ymSC5drNxXTWcK5NNFchnvVpn274u5af8VzioUcSq9VSt9WOMhmttj1aHrjAHr1HQkPhW6hsKOWtGIiIiIiIg0yEIGt/uA4Pfr1/jzKqy1+/EqbjHGNANvsNaO1u7IWnsLcAvAtm3bTqDT3WmkmIN9901dRGzPTsj64VhLvxfQrvsgrE95bRCc0KIMq+SWeHzk8coFxe4/fD8Ft0DEiXBR70V88OIPkupPsblzM6FFGpPMjbUW99ixmUPYkRFKQ0Pe8pERbDpddz9OS4sfxnYR27SJ8I7tfgjbVR3GdnXjNCUVEsmSsNDBqzEQTYaJJcLEkhGiiTDtfUliiTDRRJho3AtTozOFrlUBa2hl9l0XERERERE5zS1kcHs3cJYx5gy8wPYtwFXBFYwx3cCItdYF/ifwuQUcz8qSGfUuHrZ70Lvtuw9KOW9Z9zlw3mu9sHZ9CtrXeynBItk7vrcS1N518C6O5bwA+eyOs7lq81WkBlJc3HcxifDiVPnKFFssBloUjFAaHpoWxhaHhyrTdVsUOA6hjg6vX2x3F4m1ayvBbOXiXYEqWUctCqQBGhq8VuaH/ceRyuOofx+JhfRHChEREREREZnVggW31tqiMeY3gO8CIeBz1tpHjDF/Ctxjrf0W8Hzgk8YYi9cq4dcXajzL3rF9UyHt7p1w6BHAghOG/gvhsmth/eWwdjs0dS/u0HLHuOvgXZU+tXvGvdbGvclenr/m+aQGUmzv3053YnHHdbpws1mvN+zwEMVhP3gdHvGrZAPzhoYpjU4raAfARCKE/AtzRXp6iW8+txLMhjpr7tvbMSFVR8vCUvAqIiIiIiIip7sF7XFrrf028O2aeX8cmP4G8I2FHMOy5Low9AvYXe5PO+hdWAwg2gxrLoUXfATW7YDV2yCaXNTh5Ut5HjzyoNendv8gj448imtdmiJNXNp3KW87922kBlKc0XqGgo2TYK3FHR+fMYytCmaHhnBnalHQ3Eyoq5NwVzexMzYS2raNcFd3ZV7w4l1Oc7N+VjKvFLyKiIiIiIiInJpGX5xMAIp5OPDAVDXt7kHIHPWWNfV67Q52fMALavsugNDi/tistTwx+kTlgmL3HbqPTDFDyIS4oPsC3rf1faQGUpzffT4RJ7KoY1subKlE6ehRL3gdGqI0MuIFsyN+39iqKtlhbKEwfSfGEGpvr7QiSJx/vteeIHjxru5uwp1+i4J4fPFfqKwYCl5FREREREREGkvBbSNkx2DvXV5Iu2sQ9t0Dxay3rGsTbH6lfzGxFHRuXNT+tGWHJg+x88BOBg8MsnP/ToazwwBsaN3Aaze9llR/iktXXUpztHnRx7bUFEdGyDz4YJ0w1r+g1/AwpaNHwda5rl4kQrizc+riXWedNa1HbCWM7ejAhPUrK3Oj4FVERERERERkeVMKtBjGD3pVtLv8HrWHHgbrgglB/1bY9m4/qN0Bzb0NGeJkYZJ7Dt5TCWqfOvYUAJ3xTrb3byfVnyI1kGJV06qGjG8pKh45wvBnPsvRr34Vm81W5jvJpBe4dnURWb+OxEUX+WGsNy/cNRXMOq2tCq+kLgWvIiIiIiIiIqc3BbfzzVoYfhJ2lfvT/hSOPuMtiyRhzTZ47u95Qe2abRBracgwi26Rh4cergS1Dx15iKItEgvFuKTvEq+qdiDFWR1n4RinIWNcqgqHDjP82c8w+tWvYYtF2l71Ktrf/CbCfX2Eu7pwEolGD1GWmEKuxOihNGNDmUoAq+BVRERERERERGaj4PZUlQpw4CG/P63fozY95C1LdnkB7aXXevf9WyHUmB6w1lp2je2qBLV3H7yb8cI4BsO5Xefyzi3vJDWQ4sLeC4mFYg0Z41JXOHSI4Vs/w+jXvoYtlWi74gq633cd0fXrGz00WQJKJZfxoSyjh9OMHvJvh9OMHsowOZqbtr6CVxERERERERGZjYLbE5WbgL13T1XT7r0HCmlvWccGOOslXsuD9Zd7/WobGKyMZEe488CdXq/a/YMcmDwAwOrm1bxkw0tIDaTYvmo77fH2ho1xOSgcOMDwrbcy+vVvYK2l7bVX0P2+9xFdu7bRQ5NFZq0lPZafCmYPpRk9nPGqaY9kcN2pPsaxpBfErt3cQVtfkvbeJG09CeLNEQWvIiIiIiIiInJcCm6PZ+KwH9L6FbUHHgJbAuNA3/lw0dVeULsuBa39DR1qtpjlvsP3sfPATnbu38ljI48B0BJtYfuq7bz3gveS6k+xpmWNAqM5KOzbx9CttzL6T/8MQPvrXkfXddcRXbO6wSOThZbPFL1q2cNpRg9OhbOjh9MUslMtDEIRh/beBF0DTWy8qIf23iQdq7yQNt7cmOp6EREREREREVkZFNzOJHsMbnkBjHgX6SIch9Xb4Nm/A+tTsOZSiLc1dIiudXl85HEG9w+y88BO7jt0H3k3T9gJc2HPhfzmRb9Jqj/FeV3nEXJCDR3rcpLfu4/hm29m9LbbAGh/w+vpvvZaIqsV2K4kpaLL2FDGC2UPpqdaHBxOkz6Wn1rRQEtnnI6+JKvO7PfC2b4kbX0JWjriGEd/BBERERERERGR+afgdibxNq+K9pJ3+v1pL4RwtNGjYv/E/kpQe+eBOzmaOwrAWR1nceXmK0n1p7ik7xKSkWSDR7r85PfsYejmmzl22zcxxtDxpjfSde21RPobW0ktJ89ay+Rofnrf2YNpxoaz2EBrg3hzhI6+JOu2dNHem6Cjr4m2vgRtPQnCEf3hQ0REREREREQWl4Lb2bz2+kaPgLH8GHcfuNu7qNiBnewa2wVAT6KH56x5Djv6d5AaSNGd6G7wSJev/K5dDN18C8e++U1MKETHW95C13vfQ2TVqkYPTeYolynW9J2d6j9bzE21NghHHNr6knSvbWHTtl7a+5LerTdJvEmtDURERERERERk6VBwu8QUSgUePPKgd0GxA4M8PPQwrnVJhBNcuupS3nLOW0gNpNjYtlF9ak9R/plnGLrxJo7dfjsmHKbjbVfR9Z73EunrbfTQpI5SweXYUGZ6OHsoTWa8UFnPGGjpTtDem2TgrHbae5O0+31nm9tjam0gIiIiIiIiIsuCgtsGs9by1OhTlaD27oN3kylmcIzD+d3nc+0F15IaSLG1eyuRkCoC50Pu6acZuukmxm7/N0w0SufVV9P57ncR6VVg22jWtUyM5mrC2QyjhyYZH85ipzobkGiN0t6bYMPWbi+c9atn27oThCJO416EiIiIiIiIiMg8UHDbAEfSR9h5YKd327+Tw5nDAGxo3cBrznwNqYEUl666lNZoa4NHurLknnqKoRtuZOzb38bE43Recw1d734X4W61mVhs2cnCtKrZ0UMZjh1OUyy4lfXCsRDtvQl6N7Ry9mWrAq0NEsSS+kOGiIiIiIiIiKxcCm4XQbqQ5p5D93hVtfsHeXL0SQA6Yh1s799OaiDFjv4dDDQPNHikK1PuiScYuvFGxu74DiaRoOs976bzXe8i3NXV6KGtaMVCiWOHM9PC2dHDabITgdYGjqG1O057X5I1mzuq+s42tUfVEkRERERERERETksKbhdAyS3xyPAjlaD2gSMPUHSLRJ0oF/ddzKvPfDWp/hTndJ6DY/SV7oWS/fkvGLrxRsa/+12cRIKu976Xzne/i3BHR6OHtmJY1zI+kg20NJiqoh0fyUKgtUGyLUp7b5KNF/ZUVc62dicIhfV7ICIiIiIiIiISpOB2Hlhr2TO+pxLU3nnwTsbz4wCc23kuV593Nan+FBf1XkQ8HG/waFe+7OOPM3T9DYx/73s4TU10ve86Ot/5TgW2pyAzkZ8WzI4eSnPscIZScaq1QSQWor0vyaqNbWxO9dPe510krL03STShf25EREREREREROZKScpJGs2OcufBOxncP8jOAzvZN7EPgP6mfl68/sWk+lNc1n8ZnfHOBo/09JF99FGO3HADE9//D5zmZro/8H463/EOQu3tjR7aslDI+60Nqi4M5t3nJouV9RzH0NqToL0vybrzOqeqZ/uSJFvV2kBEREREREREZD4ouJ2jXCnHA4cfYHD/IIMHBnls+DEsluZIM5etuoxrtlxDaiDFupZ1Cq4WWebhRxi6/nomfvhDnJYWun/91+l8x9WE2toaPbQlx3Ut48PZmr6zXjg7MZKrWrepPUZ7X4JNF/dW9Z1t7Y7jhNTaQERERERERERkISm4nYFrXZ44+kQlqL3v0H1kS1nCJszWnq184MIPkBpIsaVrC2FHh7ERMg89xND1NzDxox/htLXR/Vu/SefVVxNqaWn00BrKWktmvFAnnM1w7EgatzjVeDYa91obDGxqrwpn23oTRON6X4uIiIiIiIiINIqSmRlMFia58vYrKdkSZ7adyRvPfiOpgRSX9F1CU6Sp0cM7rWUeeIAj19/A5H//N6G2Nnp++4N0vP3thJqbGz20RVXIlarDWf8CYccOp8mlA60NQoY2v7XBhgu6KuFse1+SREtEFeIiIiIiIiIiIkuQgtsZtERb+LsX/h2bOzfTm+xt9HAESN93P0PXX8/kT35CqL2dng99iI6rriLUvHKDdLfkMjacraqaLU9Pjla3NmjuiNHel+SsS/sqwWx7X4KWTrU2EBERERERERFZbhTczuK5a57b6CEIkL73Xi+w/ekgoc5Oej/8u3S89a04TSsjsLXWkh7Lc+xwmqMHq8PZsSMZXHeqtUEsGaa9L8maczoCFwVL0NabJBINNfBViIiIiIiIiIjIfFJwK0vW5F13MXT9DaTvvJNQVxe9v/d7dLz1LTjJZKOHdlLy2SLHDmc4emiS0UOZqhYHhWypsl4o7NDWm6Czv4mNF/bQ3pfwKmhXJYk3qbWBiIiIiIiIiMjpQMGtLCnWWtJ33sXQ9deTvvtuQj3d9P7B79Nx5ZU4iUSjh3dcpZLL+FCWo8G+swe9+/Sx/NSKBlo64rT3Jdi8o38qnO1L0twZx3EUzoqIiIiIiIiInM4U3MqSYK0lPTjIkRtuIHPPvYR7euj7yP+k/c1vxonHGz28WT1x9yF+ftdBr7XBUBYbaG0Qb4rQ3pdg3XmdVRcFa+tJEFZrAxERERERERERmYGCW2koay2TP/kpQ9dfT+b++wn39dH3h39I+5veiBOLNXp4syoVXX7y9Sf42Y/20dqToGdtM5su7p3qPdubJN4cafQwRURkBcln0owNHWFs6DDFfJ6zt/9Ko4ckIiIiIiILRMGtNIS1lskf/5ihT19P5sEHCa9aRd8f/xHtb3jDkg9sASZHc3znloc5+PQxLnzxOlKv3YgTcho9LBERWcastWTGjjF25DBjQ4crAe3YEe9+/MhhspMTlfWb2jsU3IqIiIiIrGAKbmVRWWuZ+NGPGLrhRrIPPUR4oJ9VH/soba9/PU402ujhzcmBJ0f5zi0Pk88Wecl7t3DWtr5GD0lERJYBt1RiYmR4KpitCmiPMD50hGI+V7VNNJGgtbuX1p5eBs7aTGtPL63dPf59b4NeiYiIiIiILAYFt7IorLVM/PA/GbrhBrIPP0xk9WpW/emf0P7a12KWSWBrreXhH+3jx197gpauOK/54IV0rW5u9LBERGSJKOSylerYsSOHGR8+EghpjzAxMoy1btU2ybZ2Wrt76Fm7no0XX1oJaVu7e2jt7iXW1IQxumCliIiIiMjpSMGtLChrLRP/8R8cueEGco8+RmTNGvo//me0XXEFJrJ8+r8W8yV+9OWf8/jOg2y4oIsXves8YsnlM34RETk11lqyE+NVQWxtQJsZH6vaxgmFaO7sprWnh7XnnU9rTy8tlWC2l5bubiLRpd8eSEREREREGkPBrSwI67qMf//7DN1wI7nHHyeybh39n/gEba9+1bIKbAHGhjLccfPPGNozwaWvOoNLX7EB46j6SURkJXHdEhMjI5VespX+skNeKDs+dIRCLlu1TTgWq1TI9m3cVJlu8atlmzs7cZxQg16RiIiIiIgsdwpuZV5Z12X837/H0A03kPvFL4iuX0//pz5J26tehQkvv7fbnkdH+PfPPoLrWl75ga1s2Nrd6CGJiMhJKORzjA8NTVXJlnvMDh/x2xgM4ZZKVdskWlpp7emla/VaNjzrYj+Y7akEtPHmFrUxEBERERGRBbP8kjRZkmypxPh3v8vQjTeSe+JJomecwcBf/Dmtr3jFsgxsrbXc/++72XnbU3T0N/Hy911Ae1+y0cMSEZE6rLXkJicDF/w6Eqic9R6nj41WbWOMQ3NnF609Paw+59xA+wI/mO3uIRKPN+gViYiIiIiIKLiVU2RLJcbu+A5DN95I/qmniG7cyMD/+T+0vuLlmNDy/HpoPlvkB194jKfuP8KmS3p5wdWbicb1qyIi0ijWdZkcPVodzPqh7Lgf0uYzmaptwpEoLf5Fvs5cf0bgol/efXNnF84y/X9KREREREROD0qj5KTYYpGxO+5g6MabyD/9NLGzNrH6//4VLS996bINbAGOHpzkjpt+xuihNJe/YRMXvmitvgYrIrLAioVC5QJf5SC2cvGvocOMDw3hlopV28SbW2jp7qGtr5+152+dCma7emjt6SXR2qZ/v0VEREREZFlTcCsnxBaLHLv9doZvvIn8rl3EzjqL1X/z17S85CUYx2n08E7J0w8c4T8+/yhO2OE1H7yQNZs7Gz0kEZEVIZeerKqSLVfNjvttDCZHj4K1UxsYQ3NHJ63dvfRvOoezdzy7ur9sdw/RhNrXiIiIiIjIyqbgVubEFgoc+9fbGbr5Jgq7dhPbvJnVf/e3tLzoRcs+sHVdy923/5J7vv0MvetbeNn7LqClU30NRUTmwlpL+tjoVCgbbGPgB7S59GTVNqFwuNJL9owLL6lUy7b41bItXV2EwpEGvSIREREREZGlQcGtzMoWChz75jcZuvkWCnv2EDvvXNZ8+u9pfuELl31gC5CdLPC9zz3C7kdG2Hx5P89769mEI8u31YOIyHwrFYtMjAxN6y0brJgtFQpV/QVjYwAAIABJREFU28SSTbR299DS08vqc7dM6y+bbG1bEf+HiIiIiIiILCQFt1KXzecZve02hm++hcK+fcS3bKHvhutpfsELVkzPwKG9E9xx00NMHM3xvKvOYctzBlbMaxMRmat8NlN9oa+agHbi6Eh1GwOgqb2D1u5eejacyZnbdkyFst1exWws2dSgVyMiIiIiIrJyKLiVKm4+z7F//heGbrmZ4v4DxC+4gL4/+kOan/e8FRVq/uKug/zwS48TS4Z53e9ezKqNbY0ekojIvLPWkhkfq+kt6134q3wRsOzEeNU2TihMS3c3rd29rL/gIq+lQbm3rN/OIBxRGwMREREREZGFpuBWAC+wHf3GNxi+9TMUDxwg/qyt9H/sYzQ95zkrKrAtlVwG/+kpHvzBHvo3tfHSa8+nqS3W6GGJiJwUt1RiYmTYC2SH61386wjFfK5qm0g8UamM7T/rHL9atocW/+JfTe0dOI5axoiIiIiIiDSagtvTnJvLMfr1bzB8660UDx0icdFF9P/Zn9H0K5evqMAWID2W57u3Psz+J0bZ+oI1XP7GTYRC6rEoIktXIZf1Athy+4JAKDs2dJiJkWGs61Ztk2xrp7W7h56169l40baq3rKt3b3EmppW3L/vIiIiIiIiK5GC29OUm80y+rWveRW2R46QuOQSBj75CZKp1Ir8QH/wl8f4zs0Pk5ss8KJ3ncc521c1ekgiIlXGho6w62f3s/tnD3L0wD7GjhwmMz5WtY5xHFq6vDYGa88932tdUAlle2jp7iES1bcIREREREREVgIFt6cZN5Ph6Fe/yvBn/3/27ju+yvru//jre0bGySBABjNh7wQCCAooWHet21Zra7W1zu721/vurnenv96dd3+1bq12aKu9q23FWUGDW8EESBgiYZMJIfuMz++PcwiBqkQkXCfJ+/l48EiuK9cV3mkTOeed7/l87yRaW0do7lxG/PdPCM2f3y8LW4A1z23n2QfWkzEolQv/Yw55o7O8jiQiQmdbK1vXVlBdvorN5Stp3LENiG/8lTdmHAVjJxwoZBOrZTOHDNEYAxERERERkQFCxe0AEWttpfH+B6i/6y6idXWE5s8n92c/I2PePK+j9ZpoOMaz969j7YqdFE4bwmlXTSctQxvqiIg3YtEou97cQHXFSqrLV7FzQxWxaJRASiqjps1g5qlnUlQ8i6Gji/rtL9JE5Mg550YD9wIFgAG3mdmvnHM3AlcDtYlLv2Fmjybu+TpwFRAFPm9mjx/z4CIiIiJyxFTc9nOxlhYa//Qn6u+6m2hDA6ETjifvl78gNHeu19F61b6Gdh67bTU1m5uYc1YR884Zh8+nIkREjq09u3dRXb6S6vKVbFnzBh0tLeAcBWPHM/dDF1BUMpsRk6cSCOqXSiJyWBHgK2b2unMuC3jNOfdk4mO/MLOfdr/YOTcNuBSYDowAnnLOTTKz6DFNLSIiIiJHTMVtPxVtbqHxj3+k4e67iTY2krFgAbmf/Qyh2bO9jtbrtq9r5PE7VhMJxzjr2mLGleZ5HUlEBoj2lma2ri6numIlm8tXsnf3LgCyhuYxcd5CikpmUThjJqHsQR4nFZG+xsx2AjsT7+9zzlUCI9/llvOA+82sA3jLObcRmAe80OthRUREROSoUHHbz0Sbm2n8/e9puPseonv3knHiieTecD2h0lKvo/U6M+ONp7fy/F/fJCc/nbOuK2bwsAyvY4lIPxaNRNi5cV3XqtpdGzdgFiOYls7o6cXM+eB5FJWUMnj4SI0/EJGjxjk3BigFXgIWAp91zn0CeJX4qtxG4qXui91u28a7F70iIiIikmRU3PYT0X37aLjvPhp+dy+xvXvJWHwSeTfcQPrMmV5HOybCHVGe+X0VG17ZzbhZeZxyxVRS0vXtLSJHl5nRuHM71eXxFbXb1lbQ2daGcz6GTZjI/As/QlFJKcMnTMYf0H+DROToc85lAg8BXzSzJufcb4HvE597+33gZ8Cn3uPnvAa4BqCwsPDoBhYRERGRI6ZnlX1ctKmJht/dS8O99xLbt4/Mk08m94YbSC+e4XW0Y2ZvbStLb6mgfkcLx58/jtlnaGMfETl6Wpv2smX1G1SXr6K6fCX76uP7/wwqGMbURUsoKi5l9IwS0jIyPU4qIv2dcy5IvLT9g5n9FcDMdnf7+O3APxKH24HR3W4flTj3b8zsNuA2gLlz59rRTy4iIiIiR0LFbR8V3bOHhnvvpeHe+4g1N5N5yink3nA96dOnex3tmNpcUcdTd68FB+d8biaF04Z6HUlE+rhIOMyOdWvj4w8qVrH7rTfBjNSMDAqnz2T+BfFVtTkFw7yOKiIDiIv/VvpOoNLMft7t/PDE/FuAC4DVifcfAf7onPs58c3JJgIvH8PIIiIiIvI+9Wpx65w7E/gV4AfuMLObDvl4IfA7ICdxzdfM7NHezNTXRRobafjd72i87/fEWlrIOu1Ucm+4gbSpU72OdkxZzHh16WZe/sdb5I7K5Kxri8nOTfc6loj0QWZG/dZqNieK2m2Vq4l0dODz+xk+cQoLPnwZY0pmUzBuAj6/3+u4IjJwLQQuByqcc6sS574BfNQ5N4v4qITNwLUAZrbGOfdnYC0QAT5jZtFjnlpEREREjlivFbfOOT/wG+A04pshvOKce8TM1na77FvAn83st865acCjwJjeytSXRRobabjrbhr/8Adira1knXEGuTdcT9rkyV5HO+Y62iI8dfdaNpfXMWl+AUs+NoVgisoUEem5lj2NVFes6lpV29LYAMCQEaMoPvl0ikpmMXpaMSnpIY+TiojEmVkZ8HazoN5x0YOZ/RD4Ya+FEhEREZFe1ZsrbucBG81sE4Bz7n7gPOK/9d/PgOzE+4OAHb2Yp0+KNDTQcNddNPzxT1hbG1lnnkHu9deTNmmS19E8Ub+jmaW3VLCvrp0TL5lI8ZJRmmcrIocV7uxge+UaNpevZEv5Smq3bAYgLSubohkzKZpZSlFxKdm5ed4GFREREREREUnozeJ2JLC12/E2YP4h19wIPOGc+xyQAZzai3n6lEhdHfV33kXj/fdj7e1kf/CD5F5/HakTJngdzTMbX6vh6XsrCab6Oe9LpYyYmON1JBFJUhaLUVP9VnxFbflKtq9bSzQcxh8IMHLKNBZ99ArGlJSSP2YczufzOq6IiIiIiIjIv/F6c7KPAveY2c+ccycA9znnZphZrPtFzrlrgGsACgsLPYh57ERqa6m/404aH3gA6+wk++yz44XtuHFeR/NMLBrjxYc3sfKJLQwbl82Z1xSTkZPqdSwRSTL76uu6Rh9UV6yirWkvALmji5h1+tkUlZQyaup0gqlpHicVERERERERObzeLG63A6O7HY9KnOvuKuBMADN7wTmXBuQCNd0vMrPbgNsA5s6da70V2Evh3TXU33kHex74MxYOM+icDzH0uutIHTvW62ieamvu5Ik71rCtqpHpJ43kxI9MxB/Q6jgRgc72NrauqaC6YiXV5ato2B5/kUdoUA5jZs5mTEkphcWzyBw8xOOkIiIiIiIiIu9dbxa3rwATnXNjiRe2lwKXHXLNFuAU4B7n3FQgDajtxUxJJ7x7N/W338GeP/8Zi0YZdO655F57DSljxngdzXM11U0svbWCtqYwJ18+hWkLR3gdSUQ8FItF2b1pI9VvxFfV7lhfRSwaIZCSyqip0yk++TSKZs4md3SRZl+LiIiIiIhIn9drxa2ZRZxznwUeB/zAXWa2xjn3PeBVM3sE+Apwu3PuS8Q3KrvSzPrlitpDhXfupP7229nzlwcxMwaddy65115LSj8fBdFTVS/sZNkf1pGeFeTCr84mvyj78DeJSL+zt2YX1eWrqC5fyZbVb9De0gxA/tjxzPnQ+RQVz2Lk5GkEUlI8TioiIiIiIiJydPXqjFszexR49JBz3+n2/lpgYW9mSDbh7dupu/129jz0VzAj54ILGHrtNaSMGuV1tKQQjcQo+8sGVi/fzsjJgznj09NJz1IhIzJQdLS2sGVNeWJV7Ur27NoJQObQXMYfd3zX+INQ9iCPk4qIiIiIiIj0Lq83JxswOrdtp/7WW9nzt78BkHPhheReczXBkSM9TpY8WvZ08NhtFeza1ETpaYUcf/44fH7NsxXpz6KRCLs2rqe6YiWby1eya+N6LBYjmJrG6OnFlJ55DkUlpQwZMUrjD0RERERERGRAUXHbyzq3bqXu1lvZ+7eHcc6Rc/FF5F59NcERmtfa3Y6Ne3j8ttV0dkQ5/dPTmTi3wOtIItILzIzGnTu6NhTbuuYNOtvacM5HwfgJzD//wxQVlzJ80mT8gaDXcUVEREREREQ8o+K2l3RWV1N3623sffhhnN/P4EsuYejVnyY4bJjX0ZKKmbF6+XbK/ryBrKFpnPuFWQwdmel1LBE5itr2NbFl9RtUl8c3FWuqrQFgUH4BUxYspmhmKYXTZ5KWqZ99ERERERERkf1U3B5lnZs3U/fbW9j7j3/gAgEGX3YZQz99FcECrSA9VKQzyrI/rmPdi7sYUzyUUz85jdSQVtiJ9HWRcJid6yvZXB5fVbv7rY1gRmoog9HTSzju3IsZU1JKzrDhXkcVERERERERSVoqbo+Sjk2bqLvlFpr+8U9cSgpDPv5xhlz1KYL5+V5HS0pNdW0svbWCum3NzDtnLHPPGoPzaX6lSF9kZtRv20J1+SqqK1aydW0FkY4OnM/H8IlTWHDxZRSVzGLY+En4/H6v44qIiIiIiIj0CSpu36eON9+k7ubf0vToo7jUVIZccQVDP/VJAnl5XkdLWlvXNvD4nauxGJx9QwljinO9jiQi71Hr3j1UV6yKjz8oX0lzYwMAg4ePZMaSUykqmc3oacWkhkIeJxURERERERHpm1TcHqGODRuo++1vaVr6GC4tjSGf+iRDP/UpAkOHeh0taZkZrz9ezUsPb2Lw8AzOuq6YnHyVOiJ9Qbizg+1Va7uK2trqtwBIy8yisHgWRcWzGFNSSnaeXmUgIiIiIiIicjSouH2P2tetp+7mm9n3+OO4UIihn76KIZ/8JIEhQ7yOltQ62yP863eVvLmylglz8zn541NISdO3n0iysliM2i2buzYU2165hki4E58/wMjJU1l06ScoKiklf+w4fD6NPxARERERERE52tSc9VB7VRV1v7mZfU8+iS8jg6HXXsuQK68gMHiw19GSXuOuFpbeUsGemjYWXjyBmaeMxjnNsxVJNvsa6uJzastXsmX1G7Tu3QPA0FGFlJx2FmNKShk1dQbBtDSPk4qIiIiIiIj0fypuD6N97Vpqb76Z5qeexpeZydDrr2PoFVfgz8nxOlqfsGlVLU/ds5ZA0Me5X5jFqMkqukWSRbi9na2VFVS/EV9VW79tCwChQTkUFc+iqKSUwuKZZA3RHGoRERERERGRY03F7TuINDay8xvfpPmZZ/BlZZH7mc8w5BOX4x80yOtofUIsZrz89028trSa/KIszry2mKwhWqUn4qVYLErNpje7NhXbvq6SWDRCIJjCyKnTmb7kVIqKZ5FXOAbn83kdV0RERERERGRAU3H7DvxZWUQa6sn93GcZcvnl+LOzvY7UZ7S3hHnyrjVsWdPA1IXDOenSSQSCmoEp4oWm2ho2l79Odfkqtqx+g/bmfQDkjRnH7A+ey5iS2YyYMpVgSqrHSUVERERERESkOxW378AFAoy5/37NYn2P6rbtY+ktFTQ3drDkY5OZtmiE/jcUOYY6WlvZuqac6oqVVJevpHHnDgAyhwxl/Jz5FJXMoqh4FqFBGvciIiIiIiIiksxU3L4LFY7vzfqXd/HMfVWkhgJc8JXZDBunsRIivS0WjbJz43qqy+NzanduqMJiMYKpaYyaNoNZp59NUUkpQ0ZqU0ARERERERGRvkTFrbxv0WiM5x/aSPm/tjFiYg5nXD2DUHaK17FE+iUzY8/unYkNxVayZXU5nW2t4BzDxk1g3nkXU1RSyohJU/AHgl7HFREREREREZEjpOJW3pfWpk4ev301OzbsYeYHRnPCRePx+7WpkcjR1Na8j62r32Bz+Uqqy1fRVLsbgOy8fCYvOJGi4lIKZ5SQnqVZ3CIiIiIiIiL9hYpbOWK7Nu3lsVsr6GiNcOonpzF5/jCvI4n0C9FImB3rq6guX0V1xUp2vbkBzEhJDzF6egnHnXMhRSWzyBmmGdIiIiIiIiIi/ZWKWzkia57bzrMPrCczJ5UL/2MOeaOzvI4k0meZGQ3bt1Fd/jrVFavYuqaCcEc7zudj+ITJnHDRpRSVzGb4hEn4/H6v44qIiIiIiIjIMaDiVt6TSDjKc/evZ+2KnRROG8JpV00nLUNzNEXeq9amvVRXrIpvKla+kuaGegAGDx/BtMWnUFQyi8LpJaSGMjxOKiIiIiIiIiJeUHErPbavoZ3Hbq2gpnofc84qYt454/D59DJtkZ6IRiJsq1ydKGpXUbP5TQDSMjIpnDGTopmlFBWXMii/wOOkIiLSm5xzJwAfB04EhgNtwGrgn8DvzWyvh/FEREREJImouJUe2baukSfuWE0kHOOs64oZNyvP60giSc/M2LG+isqyZax/4Tna9jXh8wcYMXkKCy+5nDElpeSPG4/Pp/EHIiIDgXNuKbADeBj4IVADpAGTgJOBh51zPzezR7xLKSIiIiLJQsWtvCsz442nt/L8X98kJz+ds64rZvAwvXRb5N3Uba2masVyKsuW01S7m0AwhXFz5zNl4UkUFc8iJS3d64giIuKNy82s7pBzzcDriT8/c87lHvtYIiIiIpKMVNzKOwp3RHnmvko2vFrDuNI8TrliKilp+pYReTtNdTVUrXiWqrJl1G7ZjHM+ikpmsfAjH2PCcceTkh7yOqKIiHhsf2nrnMsA2sws5pybBEwBlppZ+G2KXREREREZoNTCydvaU9PK0lsqaNzZwvHnj2P2GUU4p3m2It21Ne9j/QtlVK1YzrbK1QAMnziZk6+8lsknLCIjZ7DHCUVEJEk9C5zonBsMPAG8AlwCfMzTVCIiIiKSVFTcyr/ZXFHHk3etxfngQ5+bSeG0oV5HEkka4Y523nztZSrLlrF51evEohGGjBjFwo98nCkLF5MzbLjXEUVEJPk5M2t1zl0F3GxmP3HOrfI6lIiIiIgkFxW30sVixqtLN/PyP94id1QmZ11bTHauZnGKxKJRqitWUVW2jA2vvEi4vY3MwUMoPescpi5aQv6YcVqRLiIi74Vzzp1AfIXtVYlz2qlSRERERA6i4lYA6GiL8NTda9lcXsfk+cNY8rHJBFL0/EEGLjNj54YqKsuWs+6F52hr2ktqRgZTFpzI1EVLGDl1Oj6ffkZEROSIfBH4OvC/ZrbGOTcOeMbjTCIiIiKSZFTcCvU7mll6SwX76to58ZJJFC8ZqdWDMmDVb9tKZdkyqlYsY2/NbgLBFMbNmceURYsZO2sugWDQ64giItLHmdlyYLlzLpQ43gR83ttUIiIiIpJsVNwOcBtfq+HpeytJSfVz3pdLGTEhx+tIIsfcvvo6qp5/lsqyZdRu3oRzPgqLZ3LCxZcx4bgTSA2FvI4oIiL9SGJMwp1AJlDonJsJXGtmN3ibTERERESSiYrbASoWjfHi3zax8sktDBs3iDOvmUFGTqrXsUSOmfbmZta/VEZV2XK2Vq4GM4ZNmMTJV1zN5AUnkZEz2OuIIiLSf/0SOAN4BMDM3nDOneRtJBERERFJNipuB6C2fZ08fscatq9rZMbikSz68ET8AZ/XsUR6Xbizg02vvUJl2TLeWvkqsWiEwcNHsuDiy5iyaDGDh43wOqKIiAwQZrb1kNFUUa+yiIiIiEhyUnE7wNRUN7H01gramsJ84BNTmbpguNeRRHpVLBply+o3qCxbxsZXXqCzrY2MwUMoPfNDTF20hPyx4zXTWUREjrWtzrkFgDnngsAXgEqPM4mIiIhIklFxO4BUPr+T5X9cR3p2kAu/Opv8omyvI4n0CjNj18b1VK5Yxrrnn6N17x5SQxlMOv5Epi5azKhpM/D5/F7HFBGRges64FfASGA78ATwGU8TiYiIiEjSUXE7AEQjMcr+vIHVz25n1JTBnH7VdNKzUryOJXLUNezYRmXZMqrKlrNn9078wSDjZh/H1EVLGDtrLoEUfd+LiIj3zKwO+JjXOUREREQkuam47eda9nTw2G0V7NrUROnphRx/3jh8fs2zlf5jX0Md655/jsqyZdS89SbO+Rg9o4T5F3yEifMXkBrK8DqiiIgIAM65XwP2Th83s88fwzgiIiIikuRU3PZjOzbs4bHbVxPuiHLG1TOYMCff60giR0V7SzMbXnqeyrJlbF1bAWYUjJvIkk9czeQFJ5I5eIjXEUVERN7Oq4m3C4FpwAOJ4w8Daz1JJCIiIiJJS8VtP2RmVCzbxoq/bCQrN43zvjiLoSMyvY4l8r5EOjvZ9PrLVJYt562VrxCNRBg8fAQnXHQpUxYuYciIkV5HFBEReVdm9jsA59z1wCIziySObwGe8zKbiIiIiCQfFbf9TLgzyvI/rGPdS7sYU5LLqVdOJTUU9DqWyBGJxaJsXV1BZdkyNrz8PJ1trWTkDGbm6WczddESCsZNwDnndUwREZH3ajCQDTQkjjMT50REREREuqi47Uea6tpYemsFdduamXfOWOaeNQbnU6klfYuZsfvNDVSuWM6655+lZU8jKekhJs5fwNSFSxg9oxifz+91TBERkffjJmClc+4ZwAEnATd6mkhEREREko6K235iy9p6nrhzDRicfUMJY4pzvY4k8p407NhO1YplVK1YTuPOHfgDAcaWHsfUE5cwtnQuwZRUryOKiIgcFWZ2t3NuKTA/ceo/zWyXl5lEREREJPmouO3jzIzXH6/mxYc3MXREBmdeW0xOfsjrWCI90tzYwLrnn6WybDm7N20A5yicXsxx517MxPkLSMvQbGYRETMjti9MuKaFyO5WwjWthHe3goP8a2d6HU+OnB+oJf54fJJzbpKZPetxJhERERFJIipu+7DO9ghP31PJplW1TJybz8mXTyWYqpeQS3LraG1hw0vPU1m2jK1rKjCLUTBuAosvv4opC04ic8hQryOKiHjCzIg1dRJOlLORREEbrmnF2iJd17m0AMGCEMHhGR6mlffDOfd/gUuANUAscdoAFbciIiIi0kXFbR/VuKuFpbdUsKemjYUXT2DmKaO1SZMkrUhnJ2+tfJXKFcvY9PorRMNhcgqGM//CS5iy8CSGjhztdUQRkWPGzIju7Tho9ez+ktY6ol3X+UIBAgUhQiW5BPNDBAoyCBaE8GUG9W9+33c+MNnMOrwOIiIiIiLJS8VtH7RpVS1P3bOWQNDHuV+YxajJ2oRYkk8sFmXrmgqqVixnw0vP09HaQmhQDiWnnsnURUsYNn6SigcR6dcsZkT3dMRXz+4vaRPvW2e3gjYzSDA/RKg0n2BBiEB+iGBBCH9miofppZdtAoKAilsREREReUcqbvuQWMx4+ZFNvPZYNflFWZx5bTFZQ9K8jiXSxcyoeetNKsueoer552hpbCAlPZ2J8xYwZeFiCmfMxOfXOA8R6V8sZkQb2w+MOOg26sDCsa7rfFlBggUZZMwtiJez+SECBSH8GUEP04tHWoFVzrmn6VbemtnnvYskIiIiIslGxW0f0d4S5sk717BlbQPTFg7nxEsnEQiqAJPk0LhrB1Vly6lcsZzGHdvw+QOMmz2XKQuXMG7OcQRTUr2OKCLyvlnMiDS0E9ndcnBBW9t2UEHrz04hUBAiY94wAgXxgjaYH8IXUkErXR5J/BEREREReUcqbvuAum37WHpLBc17OljysclMP3Gk15FEaNnTyLrnn6VyxXJ2bVwPzjF66gzmfugCJs1fSFpmptcRRUSOiEWNSH3bQZuDRXa3Eq5rhYh1XecflBovaMflHDTiwJemh1fy7szsd865FGBS4tQ6Mwt7mUlEREREko+eWSS5dS/tYtnvq0jNCHLBl2czbNwgryPJANbR2sqGl5+nasVytlS8gVmM/DHjOenjn2LKgpPIGprrdUQRkR6zSIxIfdvBM2h3txKpa4Not4J2cCrB/BCpk3II5sc3CAvkpauglSPmnFsC/A7YDDhgtHPuCjN71stcIiIiIpJc9IwjSUWjMZ5/aCPl/9rGiIk5nHH1DELZ2qREjr1IOMxbq16l6rllbHr9FSLhTgYVDGP+BR9mysIlDB012uuIIiLvyiIxInVtB1bP1rQS3t1CpK4dYomC1oF/cBrBghDpU4Z0rZ4N5IfwpWg0kRx1PwNON7N1AM65ScCfgDmephIRERGRpKLiNgm17O3g8dtXs3PjXmZ+YDQnXDQev9/ndSwZQCwWY+va1VStWMb6l1bQ0dJCevYgZnzgdKYuWsLwiZNxznkdU0TkIBaOEa5tPXjEQU0rkfo22D+C1kFgaDqB/BDp03K7ytlAXroKWjmWgvtLWwAzW++c0xBkERERETmIitsks2vTXh67tYKO1ginfWoak+YN8zqSDBBmRs3mTVSWLWPd88/S3FBPMC2diccdz9RFSygsnoXPr1JDRLwX64wSqT14xEFkdwuRhnbYP+HAFy9og/kh0otzCeYnZtDmhXBB/TJUPPeqc+4O4PeJ448Br77bDc650cC9QAHx7/TbzOxXzrkhwAPAGOKjFz5iZo0u/hvWXwEfBFqBK83s9V74WkRERESkl/RqceucO5P4A0Y/cIeZ3XTIx38BnJw4DAH5ZpbTm5mSlZmx5rkdPPfAejIHp3LRf84hd1SW17FkANizayeVK5ZRVbachh3b8PkDjC2dw+LLr2L8nHkEU9O8jigiA1SsI0qk9pANwmpaiTZ2L2gdgdx0giMySZ+VT7AgFC9pc9NxARW0krSuBz4DfD5x/Bxw82HuiQBfMbPXnXNZwGvOuSeBK4Gnzewm59zXgK8B/wmcBUxM/JkP/DbxVkRERET6iF4rbp1zfuA3wGnANuAV59xblYj9AAAgAElEQVQjZrZ2/zVm9qVu138OKO2tPMksEo7y7P3rqVyxk8LpQzjtU9NJy9Cr5aT3tOxpZN0LZVStWMbODfFXao6aNoM5Z5/PxPkLSM/K9jihiAwksY5IfFOwmm4F7e5Wons6Dlzkjxe0KaMyCc7OJ9C9oNU4Iel7AsCvzOzn0PW4OfXdbjCzncDOxPv7nHOVwEjgPGBJ4rLfAcuIF7fnAfeamQEvOudynHPDE59HRERERPqA3lxxOw/YaGabAJxz9xN/ALn2Ha7/KPDdXsyTlPY1tPPYrRXUVO9j7gfHcNyHxuLzaXaoHH2dba1sePkFqlYsp7piFRaLkVc0lhMvu5IpCxeTnZvndUQR6edibZFum4MdKGmje7sVtAFHMC9ESlE2weMSG4QVhAgMScf59e+j9BtPA6cCzYnjdOAJYEFPbnbOjSG+4OEloKBbGbuL+CgFiJe6W7vdti1xTsWtiIiISB/Rm8Xt2z1YfNuXZznnioCxwL/e4ePXANcAFBYWHt2UHtq2rpHHb19NNBLjrOuKGTdLxZkcXdFImLdWvU5l2TI2vfoSkXAn2XkFzDvvYqYsXEzu6CKvI4pIPxRrDRNOlLP7V9GGd7cSa+o8cFHARzA/ndSx2QQKMuKrZwtCBAanqaCVgSDNzPaXtphZs3Mu1JMbnXOZwEPAF82sqftmoWZmzjl7x5vf+XP2y8faIiIiIn1dsmxOdinwoJlF3+6DZnYbcBvA3Llz3/OD0WRjZqx6aisv/HUjOQUhzrqumMHDMryOJf2ExWJsq1pDZdkyNry4gvaWZtKzspnxgdOYsnAJIyZNofuTPBGRIxVtCXfNnQ3vbukqaWP7wl3XuKCPQEGItAk58c3B8uOraP2D03B6hYkMXC3Oudn7Nwtzzs0B2g53k3MuSLy0/YOZ/TVxevf+EQjOueFATeL8dmB0t9tHJc79m/72WFtERESkv+jN4rbHDxaJF7ef6cUsSSPcEeVf91Wy8dUaxpfm8YErppKSliz9ufRVZkZt9VtUli2j6vlnaa6vI5iaxoTjjmfqoiUUFs/CH9D3mYi8d2ZGrCV8YPVst7exlm4FbYo/XtBOGhIfb5Aoaf05qSpoRf7dF4G/OOd2AA4YBlzybje4+G9d7wQq98/GTXgEuAK4KfH24W7nP5sYVzYf2Kv5tiIiIiJ9S282Oa8AE51zY4kXtpcClx16kXNuCjAYeKEXsySFPTWtLL2lgsadLZxwwXhKTy/Uykd5X/bW7KKybDlVK5ZTv20LPr+fMTNnc9LHPsmEOfMJpqV5HVFE+ggzI7YvTLimpdsq2nhJG2uNdF3nUv0EC0KkTY0XtPERBxn4B6Xo3zSRHjKzVxKPgScnTq0zs/C73QMsBC4HKpxzqxLnvkG8sP2zc+4qoBr4SOJjjwIfBDYCrcAnj+KXICIiIiLHQK8Vt2YWcc59Fngc8AN3mdka59z3gFfN7JHEpZcC9yd2vO23NlfU8eRda3E+OOdzsxg9bYjXkaSPat27h3UvllFZtoyd66sAGDllOqd++gYmHb+I9KxsjxOKSDIzM2JNnQfPoE0UtdbWraBNCxAsCJE+Ize+ejZR0vqyVdCKvF+JebZfBorM7Grn3ETn3GQz+8c73WNmZcRX576dU97memOAvKJNREREpL/q1ddOm9mjxH/b3/3cdw45vrE3M3jNYsYrj27mlX+8Re7oTM66tpjs3HSvY0kf09nexsZXXqSybBnV5SuxWIzcwjGceNmVTFlwEtl5+V5HFJEkY2ZE93b82+rZ8O5WrOPASHlfKECgIESoJLdrg7BgQQa+zKAKWpHeczfwGnBC4ng78BfgHYtbERERERl4NPSyF3W0hnnq7rVsrqhn8vHDWHLZZAIpfq9jSR8RjYTZ/MbrVJYt581XXyLS2UFWbh7HnXMhUxYtIa9wjNcRRSQJWMyI7ukgXNN6oKRNvG+d3QrazCDB/BCh0vwDM2gLQvgyVNCKeGC8mV3inPsogJm1Ov0gioiIiMghVNz2kvrtzSy9tYJ9de2cdOkkZiweqSfGclgWi7F93Voqy5ax/sUVtDfvIy0rm+mLT2HKosWMnDQV5/N5HVNEPGAxI9rY3jXWoPtGYRaOdV3nywoSLMggY25B1wZhgYIQ/oygh+lF5BCdzrl0wACcc+OBDm8jiYiIiEiyUXHbCza8upt/3VdFSqqf879cyvAJOV5HkiRXW/0WlSvim4ztq6slkJrKhLnHM3XREopKSvEH9KMqMlBYzIg0tCdWzx7YKCxS23ZQQevPTiFQECJj3rD4eINESesLqaAV6QO+CzwGjHbO/YH4xmNXeppIRERERJKO2qCjKBaN8cLfNrHqyS0MGzeIM6+ZQUZOqtexJEntrdlNVaKsrdtajfP5GDNzNid+9ArGz51PSppmIYv0ZxaJEalvOzB7tqaVSE0b4bpWiBzYr9M/KDVe0I7LOXjEQZr+CRfpq8zsSefc68DxxDcc+4KZ1XkcS0RERESSjJ71HSVt+zp5/I41bF/XSPHikSz88ET8Ab2kXQ7W2rSX9S+UUbliOTvWrQVgxORpnPKp65l0wiJC2YM8TigiR1usM0qktu3g8Qa1rUTq22D/AloH/sFpBPNDpE7Kia+eLcggkJeuglakH3LOLQRWmdk/nXMfB77hnPuVmVV7nU1EREREkoeeDR4FNdVNLL2lgrbmMKdcMZUpJwz3OpIkkXB7OxtffZHKsmVUl68kFo0ydFQhiy79BFMWLmZQfoHXEUXkKIi1RQ6Us93eRhu7ja30OQJD4wVt+ozc+AravFC8oNXmlSIDyW+Bmc65mcCXgTuBe4HFnqYSERERkaSi4vZ9qnx+B8v/uJ707CAXfXUOeYVZXkeSJBCNRKguX0ll2TI2vvoikY4OsobmMedDFzB14WLyisZ6HVFEjoCZEWsJd62ajb+NjzuI7es8cGHAEcwLkVKYTXBuYrxBfjqBoek4vRpDRCBiZuacOw/4jZnd6Zy7yutQIiIiIpJcVNweoWgkxnN/3sCaZ7czaspgTv/0dNIzU7yOJR6yWIwd66uoLFvGuhfLaN/XRFpmFtNOPJmpi5YwcvI0nE+FjUhfYGZE93b+2+rZSE0rsdZI13UuxU+gIETaxJyu1bPBghD+wWk4n/PwKxCRJLfPOfd14OPASc45H6CdBUVERETkICpuj0BzYweP3VbB7reamH1GIfPPHYfPr0JuoGqqq+GNJ5dStWI5TbU1BFJSGT93PlMXLWHMzFL8AT0PE0lWFjOije2Edx9S0Na2YR3Rrut8oQCBxHiD+OrZEIGCEP7sFJxTQSsi79klwGXAVWa2yzlXCPy3x5lEREREJMmouH2PdmzYw2O3rybcEeWMq2cwYU6+15HEQzWbN/HQj75D274mxpSUsvCSy5lw3PGkpKV7HU1EurFIjEj9/g3Cum0UVtsKEeu6zpedQjA/RMacAgL56fGCNj+ELyOoglZEjhoz2wX8vNvxFuIzbkVEREREuqi47SEzo/yZbTz/4Eay89I5/4ulDBmR4XUs8dCO9VX89abvEkxL54qf/oahI0d7HUlkwIt1RonUtnWtnN1f0Ebq2yF2oKD1D04lmB8idUJOVzkbzA/hS9c/iyIiIiIiIpIc9Ay1B8KdUZb/YR3rXtrFmJJcTv3kNFL15H5A27L6Df72k++TkTOYD3/7h2TnaeW1yLEUa48cMns2voo22tgO+/tZHwSGpneNONhf0Aby0vGl+D3NLyIiIiIiInI4ah8Po6mujaW3VlC3rZn5545lzpljtOHMAPfmay/x91/cxOBhI7jom98nc/AQryOJ9FvR5v0bhB28ijbW1HngooAjmBsiZVQmwdn58dWzBSECQ9NxAc0fFxERERERkb5Jxe272LKmnifuXAPA2TeUMKY41+NE4rWqFctZ+pufkz9mHBd+/b9Iz8r2OpJIn2dmRJsSBe3uViK1B97GWiJd17kUH4H8EGkTcrpGGwTzQ/iHpOkXaiLSpzjnFgI3AkXEH487wMxsnJe5RERERCS5qLh9By17Onj0txXkFKRz1nXFDMoLeR1JPFb+9OM8efv/Y9SU6Zz/H98hNaTvCZH3wmJGtLG924iDA5uEWUe06zqXHiCYHyJ9ei6BvMTq2fx0/INStUGYiPQXdwJfAl4Dooe5VkREREQGKBW37yAjJ5UPXl/M8Ak5BFM1C3Gge+2ff2PZvXcwZtYczv3y1wmmpnkdSSRpWTRGpL49vmq22wZh4do2iMS6rvNlBQnmhwjNzj94g7DMoApaEenv9prZUq9DiIiIiEhyU3H7LgqnD/U6gnjMzHjhwT/xwoN/ZNL8hXzw8/8HfyDodSyRpGDhKOHaA7NnI7tbCde2Eqlrh5h1XefPSSVYECJ1fE5i9WyIYF46vpB+lkRkwHrGOfffwF+Bjv0nzex17yKJiIiISLJRcSvyDsyM5ffdyWv//BvTF5/K6dd+Dp9fq69l4Im1RxKrZtu6jTloJdrYDvv7WQeBoekE8kOkT8slUBAvZwP5IXwp+rkRETnE/MTbud3OGfABD7KIiIiISJJScSvyNmKxKE/dcTMVTz9O6ZnncPIVV+N82p1e+rdoS/jfV8/ubiXa1HngIr8jmJdOyqhMgrPzu8YbBHLTcQH9jIiI9ISZnex1BhERERFJfoctbp1z5wD/NLPY4a4V6Q+ikQhLf/Nz1j3/LPMvuISFl3xc8zal3zAzYk2dhLvPnk2spo21hLuucyk+Annx8Qbx1bMhAgUhAoPTcH79PIiIvB/OuQLgR8AIMzvLOTcNOMHM7vQ4moiIiIgkkZ6suL0E+KVz7iHgLjOr6uVMIp6JdHby91/exKbXXubEy65k3nkXex1J5IhYzIg2tsdn0O4+uKS1jgMbmLu0AMGCEOnThhLIT+/aJMw/KBXnU0ErItJL7gHuBr6ZOF4PPACouBURERGRLoctbs3s4865bOCjwD3OOSP+QPNPZravtwOKHCud7W08/N/fZ8uaCk656gZmnf5BryOJHJZFY0Tq27tK2f0FbaS2DQsfeKGELzNIMD9EqDT/wAZh+SF8mUGtKBcROfZyzezPzrmvA5hZxDkXPdxNIiIiIjKw9GjGrZk1OeceBNKBLwIXAF91zv2Pmf26NwOKHAvtzc389abvsuvNDZz1mS8z7USNnpPkYuFYfObsQSMO2ojUt0HUuq7z56QSyA+RMS4nsXo2vorWFwp6mF5ERA7R4pwbSmKLR+fc8cBebyOJiIiISLLpyYzbc4FPAhOAe4F5ZlbjnAsBawEVt9Knte7dw4M//DYN27dyzpe+xsR5C7yOJANYrCNCpKbt32bQRhvaE0/vAQeBoekE8tJJnzaEQF4ovoo2L4Qv1e9pfhER6ZEvA48A451zK4A8QPOZREREROQgPVlxexHwCzN7tvtJM2t1zl3VO7FEjo2muloe/MG32Fdfx/n/8R3GzJztdSQZIKItYSK1rYR3H7yKNrq388BFfkcgN52UEZkEZuUTzE8UtEPTcUGfd+FFROR9MbPXnXOLgcmAA9aZWfgwt4mIiIjIANOT4vZGYOf+A+dcOlBgZpvN7OneCibS2xp37eAv3/8mHS0tXPTN7zFqynSvI0k/ZZEY7RsaaV/XGC9qa1uJNR94fu6CPgL5IVLH5Ry0QVhgSDrOr/mzIiL9hXNukZmVQXyuLbDmkI9nA4VmttqLfCIiIiKSXHpS3P4F6P7a8Wji3HG9kkjkGKjbspkHf/htYtEoH/nOjygYN8HrSNLPWCRG+/pG2irqaFtbj3VEcSl+gsNCpE0ZEi9nC0IE80L4c1JxPhW0IiIDwEXOuZ8AjwGvAbVAGvGRZCcDRcBXvIsnIiIiIsmkJ8VtwMy6XrtrZp3OuZRezCTSq3ZtXM9DP/4ugWCQS268iaGjCr2OJP2EheMraw8qa9MCpM/IJb04l7QJObiARhyIiAxUZvYl59wQ4qPIPgwMB9qASuDW/atxRURERESgZ8VtrXPuXDN7BMA5dx5Q17uxRHrHtrWr+d+f/BfpWdlc/K0fklMwzOtI0sdZeP/K2lraKhviZW16oqwtySVtvMpaERE5wMwagNsTf0RERERE3lFPitvrgD845/4f8c0TtgKf6NVUIr3grZWv8sjPfkR2fgEXf+v7ZA3J9TqS9FHxsraB1oo62ruXtcW5hIpzSVVZKyIiIiIiIiLv02GLWzN7EzjeOZeZOG7u9VQiR9m6F8p49Nc/JXd0ERd983uEsgd5HUn6GAtHaV/XeKCs7YziCyXK2pI8UscPwvlV1oqIiIiIiIjI0dGTFbc4584GpgNpzsU30DGz7/ViLpGjZvWyp3jilv9h+KQpXPi175IayvA6kvQR71TWhmbmkV6cq7JWRERERERERHrNYYtb59wtQIj4Trd3ABcDL/dyLpGj4vWlf+eZe26lqKSU877yTYJpaV5HkiTXVdaW19Je1YB1xuJl7axEWTtOZa2IiLx/zrkFwBi6PR43s3s9CyQiIiIiSacnK24XmFmJc67czP7LOfczYGlvBxN5P8yMl//2F8ruv5cJxx3P2V/4TwLBoNexJEnFOuNlbVtFt7I2I0BoVn6irM3B+Z3XMUVEpJ9wzt0HjAdWAdHEaQNU3IqIiIhIl54Ut+2Jt63OuRFAPTC89yKJvD9mxnN/vIdXHnmIqSeezJnXfxGf3+91LEky8bK2gbb9YxDCKmtFROSYmQtMMzPzOoiIiIiIJK+eFLd/d87lAP8NvE58NcDtvZpK5AhZLMbTd/2WN55cyszTPsgpn7oO59PL2iUu1hmlvSpR1lbtL2uDhGYnytqxKmtFROSYWA0MA3Z6HUREREREkte7FrfOOR/wtJntAR5yzv0DSDOzvccknch7EItGeey3v6TyuWc47tyLOPGyK9m/mZ4MXG9b1mbuL2vzSB07SGWtiIgca7nAWufcy0DH/pNmdq53kUREREQk2bxrcWtmMefcb4DSxHEH3R5ciiSLSDjMP3/1Eza+8gKLLv0E887/sErbASzWsb+sraV9XeOBsnZOQWJl7SCcT98fIiLimRu9DiAiIiIiya8noxKeds5dBPxVc7gkGYXb23n4Zz+kunwlJ195LbPPOsfrSOKBeFlbT1t5He3rVdaKiEjyMrPlzrkC4LjEqZfNrMbLTCIiIiKSfHpS3F4LfBmIOOfaAQeYmWX3ajKRHmhvaeZ//+/32Lm+ijOu/yIzlpzqdSQ5hmIdEdorG2itqKN9XSNEYviygoTmFhAqziVljMpaERFJPs65jxDfP2IZ8cfWv3bOfdXMHvQ0mIiIiIgklcMWt2aWdSyCiLxXrU17eeiH36FuazUf+uJ/MOn4RV5HkmOgq6xNrKyNl7UpZBxXQKg4j5Qx2SprRUQk2X0TOG7/KlvnXB7wFKDiVkRERES6HLa4dc6d9HbnzezZox9HpGf2NdTx4A++TVPNbs7/6rcYWzrX60jSiw4uaxsgYgfK2pI8UopU1oqISJ/iO2Q0Qj3g8yqMiIiIiCSnnoxK+Gq399OAecBrwAd6JZHIYezZvYsHf/BN2vY1ceE3/ovR04q9jiS9INbebQzC/rI2O4XMecNJL85VWSsiIn3ZY865x4E/JY4vAR71MI+IiIiIJKGejEo4aKcn59xo4Je9lkjkXdRv28KDP/gWkXCYD3/rhwybMMnrSHIUxdojtFU20FZeS/uGxoPL2pJcUgpV1oqISN9nZl9NbP67MHHqNjP7Xy8ziYiIiEjy6cmK20NtA6Ye7SAih7N700Ye+tF38Pn9XPLdH5NbOMbrSHIUxNojtK2tp60iMbM2avizU8icn1hZq7JWRET6ITN7CHjI6xwiIiIikrx6MuP214AlDn3ALOD13gwlcqjtVWv56003kpqRwYe//UMGDxvhdSR5H2Jt3craDYmydlAKmccPJ70kj5TRWSprRUSk33HOlZnZIufcPg48vgZwgJlZtkfRRERERCQJ9WTF7avd3o8AfzKzFb2UR+TfbC5fycM//QFZQ3K5+Fs/IDs3z+tIcgTevqxNJfOEEfGVtSprRUSknzOzRYm3WV5nEREREZHk15Pi9kGg3cyiAM45v3MuZGatvRtNBDa8/Dz//NVPGDJyNBd943tk5Az2OpK8B7HWMG1rG2irqKV9456Dy9qSXFJGqawVEZGBxzl3n5ldfrhzIiIiIjKw9aS4fRo4FWhOHKcDTwALeiuUCMDa557hsZt/wbDxE7nwa/9FWmam15GkB+JlbWJl7f6yNieVzAXdVtY6lbUiIjKgTe9+4JwLAHM8yiIiIiIiSaonxW2ame0vbTGzZudcqBczibDqiUd5+s6bKZxRwnlf/TYpaeleR5J3EWsN07amntaKOjo27oHYgbI2VJJHcFSmyloRERnwnHNfB74BpDvnmvafBjqB2zwLJiIiIiJJqSfFbYtzbraZvQ7gnJsDtPXkkzvnzgR+BfiBO8zspre55iPAjcQ3aHjDzC7rYXbpp15++EGe++M9jJszj3O++DUCKSleR5K38Y5l7aIRhIpV1oqIiBzKzH4M/Ng592Mz+7rXeUREREQkufWkuP0i8Bfn3A7iKwKGAZcc7ibnnB/4DXAasA14xTn3iJmt7XbNRODrwEIza3TO5R/B1yD9hJmx4oHf89L/PsDkBSdx1me+jD/Qk29ROVaiLWHa1x5S1g5OJXPRSELFuSprRUREemapc+6kQ0+a2bNehBERERGR5HTYVszMXnHOTQEmJ06tM7NwDz73PGCjmW0CcM7dD5wHrO12zdXAb8ysMfF31byX8NJ/WCzGM7+7nZWP/Z3iD5zOqVd/Bp/P73UsIVHWrqmntaKWjjf3QAz8Q9LIPDFR1o5UWSsiIvIefbXb+2nEHze/BnzAmzgiIiIikowOW9w65z4D/MHMVieOBzvnPmpmNx/m1pHA1m7H24D5h1wzKfE5VxAfp3CjmT3W0/DSP8SiUZ649desWf4Uc84+n8WXX6Ui0GPRljBta+poq6g7qKzNOnEU6SprRURE3hczO6f7sXNuNPBLj+KIiIiISJLqyevQrzaz3+w/SIw0uBo4XHHb079/IrAEGAU865wrNrM93S9yzl0DXANQWFh4FP5aSRbRSJhH/+enrH9pBSdcfBknXPxRFYIeiTZ30ramPl7WbkqUtUPTyDppFOnFeQRHZOj/GxERkd6xDZjqdQgRERERSS49KW79zjlnZgZds2t7slvUdmB0t+NRiXPdbQNeSoxeeMs5t554kftK94vM7DYSO+3OnTvXevB3Sx8Q7mjn7z//MW+teo3Fl1/F3A9d4HWkAeftytrA0DSyThodX1mrslZEROSoc879mvjGvAA+YBbwuneJRERERCQZ9aS4fQx4wDl3a+L4WmBpD+57BZjonBtLvLC9FLjskGv+BnwUuNs5l0t8dMKmngSXvq2jtZW//eR7bKtaw2nXfJaSU870OtKAEW3upG11PW0VtXRs2gsGgdx0shYnytrhKmtFRER62avd3o8AfzKzFYe7yTl3F/AhoMbMZiTO3Uh834jaxGXfMLNHEx/7OnAVEAU+b2aPH7WvQERERER6XU+K2/8kPqbgusRxOTDscDeZWcQ591ngceLza+8yszXOue8Br5rZI4mPne6cW0v8AeVXzaz+CL4O6UPa9jXx0I++S231Js7+3P9hysLFXkfq96L7OuMza8vr6HirW1m7RGWtiIiIBx4E2s0sCvFXtDnnQmbWepj77gH+H3DvIed/YWY/7X7COTeN+MKJ6cAI4Cnn3KT9f6eIiIiIJL/DFrdmFnPOvQSMBz4C5AIP9eSTJ37b/+gh577T7X0Dvpz4IwNAc2MDD/7gW+zZvZNzv/JNxs+Z53Wkfiu6r5O21YkNxlTWioiIJJOngVOB5sRxOvD/2bv3uC7Lw//jr4uDCAqC4IGUUouQU7CF2dSaDUk2Z67UOaelWVqarq1Zc1szc7asb63VsrmW/TxMN9RWluugppurzCTCiSiZJioQyUnOyOH6/QExTTyg4v0B3s/Hwwefz31f932/P1CP7t5eXPcGYNCZDrLWbjXG9DnHa4wC/m6traJ+SbLPgOuAbecTWEREREQuvdMWt8aYq6lfxmA8kAckAVhrb7o00aStKT76JWsW/JqywkJumzOPy6NinI7U5jRZ1nbzxvemkPoHjPX0UVkrIiLivI7W2q9KW6y1pcYYnws430xjzB3UL8Hwc2ttIdAL+PCEMUcatomIiIhIK3GmGbd7gf8A37fWfgZgjPnZJUklbU5B9hHWLHiY6soKxjy8gMuu7u90pDbjq7K2/L95HD94clnrc003PHqorBUREXExZcaYb1prUwCMMdcCFed5rj8Bv6X+YWe/BZ4GpjTnBMaYadQvjcbll19+njFERERE5GI7U3F7G/XrYm0xxrwN/B1Q+yPN9uXBA6x97DcYY/jh3Mfp3qef05FavdrihrJ211GOHyyuL2u7e+P7ncvxiQ5SWSsiIuLafgqsMcZkU39/3RMYdz4nstbmfvXaGPMXYH3D2ywg5IShvRu2NXWOF4EXAeLi4uz55BARERGRi++0xa219jXgNWNMJ+rXyPop0N0Y8yfgVWvthkuUUVqx7E/38I+F8/Ds6M3YhxfQ9bLeTkdqtWqLq6jYlUf5rjyOZ35V1vrUl7XXBOHZo5PTEUVEROQcWGt3GGP6A2ENmzKstdXncy5jTLC1Nqfh7a1AWsPr14FVxpjfU/9wslDgowuILSIiIiKX2Lk8nKwMWEX9jV8AMBb4BfUPUBA5rUNpO3ntyd/SKSCAsQ8/hl+37k5HanVqj1VR3rBmbWNZ28MHv/jL6x8wprJWRESk1WlYz/YB4Apr7VRjTKgxJsxau/4sx/0NGAoEGWOOAI8AQ40xsdQvlXAQuAfAWrvbGLMaSAdqgPustbUt9ZlERERE5OI7a3F7omTXlDwAACAASURBVIYHHTT+KpXI6ez/eDtvPLOQgJ6XMfrXv6VzQFenI7UatceqKN91QlmLyloREZHWzhjzfeBfDQ8l+3/Ax8C3GnZnAWv43zIHTbLWjm9i85IzjH8MeOy8AouIiIiI45pV3Iqci73v/5u3Fv2e7n36cdsvH8Xb18/pSC6v5lj9MginlLXDLsf7mm54dr+QB02LiIiICzgALAYmAldaa8cZY8YDWGvLjRanFxEREZGvUXErF9V/332bjX9ZRO/+kfzgobl4+ahwPJ2aoq/K2qMcP1QCgGdPH/wSrqifWauyVkREpM2w1qYbY37Z8Pa4Mcab+uUNMMZcCVQ5Fk5EREREXJKKW7lokte/yr9XLKFv7LWMfOCXeHp1dDqSy2m6rO1UX9ZeE4RnN5W1IiIibZW19nDDy3nA20CIMWYlMBi406lcIiIiIuKaVNzKBbPWsm3tKrat/RtXDxzM934yG3cPT6djuYyaosr/LYPwVVkb3Am/mxtm1qqsFRERaVestRuMMR8D1wMGuN9am+dwLBERERFxMSpu5YJYa/n3iiV8/M/XiPz2MG6+ZxZu7u5Ox3JcTeEJZe3hE8ra4VfgHd0NzyBvhxOKiIiIU4wxK4CZ1tp/Nry/whiTZK2NdziaiIiIiLgQFbdy3urqatn0l0Xs2ryBb3x3JDfdMRXj5uZ0LMfUFFRSkZZH+a48qr8qay/rhN/wPvUza1XWioiISL33gO3GmAeAXsCDwM+djSQiIiIirkbFrZyX2poa3nr+aTK2/YfrbxvHoB9OpL0+DLlibwHFmzKpPlIKgGevzvgl9sEnKggPlbUiIiLyNdbaPxtjdgNbgDzgG9baLxyOJSIiIiIuRsWtNFvN8eO88czjHEjZwQ0/nsx1o8Y4HckxVZnF5P81HQ//jvVlbXQQHoEqa0VEROT0jDG3A78B7gCuAd40xtxprd3pbDIRERERcSUqbqVZjldWsO7/fsuh3bsYdvcMYhK+53Qkx9QUVpK/Ih33Ll50mx6Deyc9kE1ERETOyWhgiLX2S+BvxphXgWVArLOxRERERMSVqLiVc1ZZWso/Fj7CF/v38d37HiDihpucjuSYuqoa8pbuxtbU0W3aNSptRURE5JxZa38AYIzxsdaWW2s/MsZc53QuEREREXEt7fdJUtIsZUWFrH50Dl9+vp+RD/yyXZe2ts5S8LcMao6WE/jjcDy7+zgdSURERFoRY8y3jDHpwN6G9zHAH5xNJSIiIiKuRsWtnFVx3pckzZtDYW4OP/jFI4QO+JbTkRx17K3PqdxbgP/IK+l4dYDTcURERKT1+QMwHMgHaFjb9kZHE4mIiIiIy9FSCXJGhV9ks+a3v6aqrIwxv/otvfpHOB3JUWU7vqD0P1l0+lYwnb91mdNxREREpJWy1h42xpy4qdapLCIiIiLimlTcymnlHTrI2sd+Q11tLT+c+zt69LvK6UiOqtxfROGrn+EV6o//9690Oo6IiIi0XoeNMYMAa4zxBO4H9jicSURERERcjJZKkCZ98dmnJD36S4wxjJv3RLsvbWvyKihYuQePwI4E/jgc427OfpCIiIhI0+4F7gN6AVlAbMN7EREREZFGmnErpzicvotXn5iPj58fY3/zGF2693Q6kqPqyqvJW7YbgKBJkbh5618bEREROX/W2jxggtM5RERERMS1qYGSk3z+STKvP/07/Lr3YMzDv8W3a5DTkRxla+vIX7WXmoJKut0VhUeQt9ORREREpJUyxjwCWKDUWvt7p/OIiIiIiGtTcSuNMra9x5t/fIqgy69g9K/m4+PXxelIjit64wBVnxURMDoUr37+TscRERGR1u1gw9cKJ0OIiIiISOug4lYASNuykQ1//iPBV/fntjmP4OXTyelIjiv9IJuyD3PofGMvOg1o38tFiIiIyIWz1i5zOoOIiIiItB4qboWUt15ny9IXueKabzDq57/Gs2NHpyM5rvLTQore2E/H8K50SezrdBwRERFpA4wxb1C/VEKTrLW3XMI4IiIiIuLiVNy2Y9Zatr+6mveTVnDVgG8x4v6H8PD0dDqW46pzy8hfuQfPHp3o+qMwjJtxOpKIiIi0DU81fL0N6An8teH9eCDXkUQiIiIi4rJU3LZT1lr+s2opO15/hYgbbmL49J/i5u7udCzH1ZZVk7csHePpRuCkCNy89K+IiIiIXBzW2n8DGGOettbGnbDrDWNMskOxRERERMRFqZVqh2xdHe++/Cd2bnyLmITvET/lXoybm9OxHGdr6sj/azq1xVV0m3YNHgFaMkJERERaRCdjTD9r7QEAY0xfQA8YEBEREZGTqLhtZ+pqa3n7T39gz3+2MGDUGG4YPwljtBSAtZbCVz/j+OfFdP1RGF6X+zkdSURERNqunwH/MsYcAAxwBTDN2UgiIiIi4mpU3LYjNdXV/PPZJ/hsx4cM+dEdDLz1h05HchmlW7Mo/zgX3++E4BPb3ek4IiIi0oZZa982xoQC/Rs27bXWVjmZSURERERcj4rbdqK6spJ1Tz9G5n8/4Tt33sM3Ekc6HcllVKTnc+ztz/GODsJv2BVOxxEREZF2oKGo3el0DhERERFxXSpu24HKslJefWI+OZ/uZfj0nxI1dJjTkVzG8exSCv6+F89enQkYezXGTctGiIiIiIiIiIiI81TctnHlxcd45bG55B3O5Ps/fYirrx/idCSXUVtynPxl6bh19CDojgjcOrg7HUlERERERERERAQAN6cDSMspKcgj6ZFfUJB1mB88+LBK2xPY6lryl6dTV15N4KRI3P28nI4kIiIi7YQxZrAxplPD64nGmN8bY7Rek4iIiIicRMVtG1WU+wVJj/yC0sJ8Rv9qPn2/Eed0JJdhraVg7T6OHy6h67gwOvTq7HQkERERaV/+BJQbY2KAnwP7geXORhIRERERV6Pitg3KP3KIvz/yEFXl5Yx9+DF6R0Q5HcmllGw+TMXOo/gNvwLvqCCn44iIiEj7U2OttcAo4Hlr7SLA1+FMIiIiIuJitMZtG5N74DNe+d1c3NzdGffI4wRd3sfpSC6l/L9HKd6Yic83uuM7NMTpOCIiItI+lRhjfgncDtxgjHEDPB3OJCIiIiIuRjNu25Aje3ezev6v8OzYkXGPPqHS9muOHy6hYPWndLjCj4DRoRhjnI4kIiIi7dM4oAqYYq39AugN/J+zkURERETE1ai4bSMO7kzhlcfm0sk/gHHzniCg52VOR3IpNceqyFuejruvJ4G3h2M89I++iIiIOKOhrF0FBBhjRgLHrbVa41ZERERETqL2qg3Y99EHvPbkfAKCL+NHjz6BX1A3pyO5lLrjteQv242tqiVoUiTunTs4HUlERETaMWPM3cBHwG3AGOBDY8wUZ1OJiIiIiKvRGretXPrWzbz9pz/Q88pQbpvzKB07d3Y6kkuxdZaCpAyqc8oInBSJZ89OTkcSEREReRD4hrU2H8AYEwh8ALzsaCoRERERcSkqblux1A1v8u6SF7g86hpGPfgbOnT0djqSyynekEnl7ny6jOiHd/+uTscRERERAcgHSk54X9KwTURERESkkYrbVuqjdWv5z6ql9Lv2Okb+dA4eHfTr/19X9nEuJf86TKfretJ5iNb8FREREZfxGbDdGLMOsMAo4L/GmAcArLW/dzKciIiIiLgGFbetjLWW95NWsP3V1fQf/G0SZ/wMdw/9GL+u6uAxCv+xD69+XfAfdSXGGKcjiYiIiHxlf8Ofr6xr+OrrQBYRERERcVFq/FoRW1fHlmV/4ZO33yA6fjjD7p6Bm5u707FcTk1BJfkr0vHw9yJwYjjGXc/gExEREddhrX3U6QwiIiIi4vpU3LYSdbW1bPjzH9n9701cO+IHfPv2uzSLtAl1lTXkLduNrYXAyZG4+Xg6HUlERETkJMaYOODXwBWccD9urb3GsVAiIiIi4nJU3LYCtTXVvPncU3y6/X0GjZ3A9aN/pNK2CbbOUvC3vdQcLSfozig8u/k4HUlERESkKSuBB4FdQJ3DWURERETERam4dXHVVZW8/vvHOZj6MUPvuJtrR/zA6Ugu69g/D1CZUYj/D66iY2iA03FERERETueotfZ1p0OIiIiIiGtTcevCqsrLefWJR8nKSCdh2iyuiR/udCSXVbo9h9L3s+k86DI6Xx/sdBwRERGRM3nEGPMS8C5Q9dVGa+0/nIskIiIiIq5Gxa2Lqigp5pXfPcLRzAOM+MmD9B90o9ORXFblZ0UUrduP19UBdBnRz+k4IiIiImdzJ9Af8OR/SyVYQMWtiIiIiDRq0eLWGJMIPAu4Ay9Zaxd+bf9k4P+ArIZNz1trX2rJTK1BaWEBaxc8TFFuDrf8/Ndcee11TkdyWdVHy8lfuQePIG8Cf9wf4661f0VERMTlDbDWhjkdQkRERERcW4sVt8YYd2ARkAAcAXYYY1631qZ/bWiStXZmS+VobY59mcvaBQ9TVlTIbXMe5fIoPVz4dOrKq8lflo5xg6BJEbh11ARyERERaRU+MMZENHFfLCIiIiLSqCWbruuAz6y1BwCMMX8HRgG6QT2NguwjrFnwMNWVFYx5eAGXXd3f6Uguy9bWkb9yDzWFlXS7OxqPQG+nI4mIiIicq+uBVGPM59SvcWsAa63V39iLiIiISKOWLG57AYdPeH8EGNjEuNHGmBuBT4GfWWsPNzGmzfvy4AHWPvYbjDGMe2Qh3a7o63Qkl2Wtpej1/VTtP0bAmKvx6tvF6UgiIiIizZHodAARERERcX1uDl//DaBPw+yCjcCypgYZY6YZY5KNMclHjx69pAEvhexP97B6/i9x9/Rk3DyVtmdT+kE2Zdu/oPO3e9MprofTcURERESaxVqbCYQA32l4XY7z9+UiIiIi4mJa8gYxi/ob0q/05n8PIQPAWptvra1qePsScG1TJ7LWvmitjbPWxnXr1q1Fwjolc1cqaxf8Bm9fP8Y/+iRdL+vtdCSXVpFRwLH1B+gYEUiX4X2cjiMiIiLSbMaYR4BfAL9s2OQJ/NW5RCIiIiLiilqyuN0BhBpj+hpjOgA/Al4/cYAxJviEt7cAe1owj8vZ//F2Xn3iUbp078GPHn0Sv27dnY7k0qpzyyhYtRfPnp3oOi4M42acjiQiIiJyPm6l/t63DMBamw34OppIRERERFxOi61xa62tMcbMBN4B3IGXrbW7jTHzgWRr7evAT4wxtwA1QAEwuaXyuJo97/+bt55/mh59r+S2X83Hu7Pu1c+ktvQ4ecvSMR3cCJwUiZuXu9ORRERERM7XcWutNcZYAGNMJ6cDiYiIiIjracmHk2GtfRN482vb5p7w+pf871fE2o3/vvs2G/+yiN7hkdz60Fw6ePs4Hcml2Zo68lfsobb4ON3vuQYPfy+nI4mIiIhciNXGmD8D/saYqcAU6pcNExERERFp1KLFrZwqef2r/HvFEvrGXsvIB36Jp1dHpyO5NGsthf/Yx/HMYrqOD6NDiGYmi4iISOtmrX3KGJMAFANhwFxr7UaHY4mIiIiIi1Fxe4lYa9m2dhXb1v6Nq68fwvdm/Rx3D0+nY7m8kn8foTzlS3zjL8cnRmsAi4iISOtnjHnCWvsLYGMT20REREREgJZ9OJk0sNby7xUvsW3t34gcOowR9z+o0vYcVOzOo/idg3hfE4TfsMudjiMiIiJysSQ0se27lzyFiIiIiLg0FbctrK6ulo0v/pGP/7mOb3x3JMPv+Qlubnqw1tkczyql4O8ZePb2pevYqzHGOB1JRERE5IIYY6YbY3YBYcaY/57w53Pgv+dw/MvGmC+NMWknbOtqjNlojNnX8DWgYbsxxjxnjPms4RrfbLlPJiIiIiItQcVtC6qtqeHN555i1+YNXH/bOG6aNA3jpm/52dQWHyd/+W7cfDwIuj0C46miW0RERNqEVcBI4PWGr1/9udZaO/Ecjl8KJH5t2xzgXWttKPBuw3uon8Eb2vBnGvCnCw0vIiIiIpeW1rhtIdXHq1j/zEIOpOzghh9P5rpRY5yO1CrY6lrylu+mrryGbvfG4O7XwelIIiIiIheFtfYYcAwYf57HbzXG9Pna5lHA0IbXy4B/Ab9o2L7cWmuBD40x/saYYGttzvlcW0REREQuPRW3LeB4RTmv/d8CDqfvYtjdM4hJ+J7TkVoFay0Faz6lOquUwInhdOjV2elIIiIiIq6uxwll7BdAj4bXvYDDJ4w70rDtlOLWGDON+lm5XH65nisgIiIi4ir0e/sXWUVpCWsX/IYje9L43n0PqLRthuJNh6j4bx5+w/vgHRnkdBwRERGRVqVhdq09j+NetNbGWWvjunXr1gLJREREROR8aMbtRVRWVMgrj/2GguwjjHzgl4QO+JbTkVqN8p1fUvLuIXy+2R3fb/d2Oo6IiIhIa5H71RIIxphg4MuG7VlAyAnjejdsExEREZFWQjNuL5LivC9JmjeHwtwcfvCLR1TaNkPVoWIK1nxKhz5+BNwWijHG6UgiIiIircXrwKSG15OAdSdsv8PUux44pvVtRURERFoXzbi9CApzsliz4GGqysoY86vf0qt/hNORWo2aoiryl6fj7udF4MRwjIf+LkFERESkKcaYv1H/ILIgY8wR4BFgIbDaGHMXkAn8sGH4m8D3gM+AcuDOSx5YRERERC6IitsLdPTQQdYueBhbV8cPH3mcHn2vdDpSq1FXVUv+st3Y6jq6TY3GvXMHpyOJiIiIuCxr7fjT7IpvYqwF7mvZRCIiIiLSklTcXoCczzL4x+8ewaNDB8bMfZzA3iFnP0gAsHWWgqQMqr8oI2hyJJ49OjkdSURERERERERExGWouD1Ph9N38eoT8/Hx82Psbx6jS/eeTkdqVY69c5DK9Hy6jOxHx7CuTscRERERERERERFxKSpuz8OBT3bwxtOP49e9B2Me/i2+XYOcjtSqlCXnUvrvI3Qa2JPOgy5zOo6IiIiIiIiIiIjLUXHbTBnb3uPNP/4fQZf3YfSv5uPj18XpSK1K1efHKHx1H15X+eN/y5UYY5yOJCIiIiIiIiIi4nJU3DbDri0b2Pjn57ksrD+3/uIRvHy0Lmtz1ORXkL8iHY+AjgT+uD/G3c3pSCIiIiIiIiIiIi5Jxe05SnnrdbYsfZErrvkGo37+azw7dnQ6UqtSV1lD3rLdWAuBkyNx8/F0OpKIiIiIiIiIiIjLUnF7FtZatr+6mveTVnDVgG8x4v6H8PBU6dgcttaSv2ovNXmVBN0VhWeQt9ORRETkPFRXV3PkyBEqKyudjiItqGPHjvTu3RtP3e+IiIiIiDhKxe0ZWGvZuvL/kfzGP4i44SaGT/8pbu7uTsdqdY798wBVnxbif9tVdLzS3+k4IiJyno4cOYKvry99+vTRGuVtlLWW/Px8jhw5Qt++fZ2OIyIiIiLSrmmR0dOwdXW8u+QFkt/4BzE3jyBxxs9U2p6H0g9zKP0gm86DL6PzdcFOxxERkQtQWVlJYGCgSts2zBhDYGCgZlWLiIiIiLgAzbg9jfLiY+z/+CMGjBrDDeMn6X9Sz0PlvkKKXv+MjmEBdBnRz+k4IiJyEei/h22ffsYiIiIiIq5Bxe1pdPIP4PYnnsPHr4vTUVql6qPl5K/ci0c3H7qO749x0/8EioiIiIiIiIiInCstlXAGKm3PT21ZNflLd2PcDUGTInHrqL8fEBERERERERERaQ4Vt3JR2Zo6ClbuoaaoisDbw/Ho2tHpSCIi0kYcPHiQqKioZh2zdOlSsrOzzzpm5syZFxKNuXPnsmnTpgs6x5lMmTKF7t27N/vzi4iIiIhI66XiVi4aay1Fr++n6sAxAkaH4tVHM5ZFRMRZ51LcXgzz589n2LBhLXb+yZMn8/bbb7fY+UVERERExPXod9jloil9L5uyj77Ad2gInb7Zw+k4IiLSgh59Yzfp2cUX9ZwRl/nxyMjIM46pqalhwoQJpKSkEBkZyfLly/Hx8WH+/Pm88cYbVFRUMGjQIP785z/zyiuvkJyczIQJE/D29mbbtm2kpaVx//33U1ZWhpeXF++++y4A2dnZJCYmsn//fm699VaefPLJJq9fW1vLXXfdRXJyMsYYpkyZws9+9jMmT57M97//ffr06cPdd9/dODYtLQ1rLfv37+e+++7j6NGj+Pj48Je//IX+/fuf8/fmxhtv5ODBg+c8XkREREREWj/NuJWLomJvAcfePIB3ZCB+N1/hdBwREWmjMjIymDFjBnv27MHPz48XXngBgJkzZ7Jjxw7S0tKoqKhg/fr1jBkzhri4OFauXElqairu7u6MGzeOZ599lp07d7Jp0ya8vb0BSE1NJSkpiV27dpGUlMThw4ebvH5qaipZWVmkpaWxa9cu7rzzzpP2x8XFkZqaSmpqKomJicyePRuAadOm8cc//pGPP/6Yp556ihkzZpxy7uTk5MbSV0RERERERDNu5YJVf1FGwaq9eF7WmYBxYRg343QkERFpYWebGdtSQkJCGDx4MAATJ07kueeeY/bs2WzZsoUnn3yS8vJyCgoKiIyMZOTIkScdm5GRQXBwMAMGDADAz8+vcV98fDxdutQv8RMREUFmZiYhISGnXL9fv34cOHCAWbNmMWLECG6++eYmcyYlJZGSksKGDRsoLS3lgw8+YOzYsY37q6qqTjkmLi6Ol156qZnfERERERERaatU3MoFqS09Tt7S3RgvdwLviMCtg7vTkUREpA0zxpzyvrKykhkzZpCcnExISAjz5s2jsrKyWef18vJqfO3u7k5NTU2T4wICAti5cyfvvPMOixcvZvXq1bz88ssnjUlLS2PevHls3boVd3d36urq8Pf3JzU1tVmZRERERESkfdNSCXLebHUd+Sv2UFtaTdAdEXh08Tr7QSIiIhfg0KFDbNu2DYBVq1YxZMiQxpI2KCiI0tJS1q5d2zje19eXkpISAMLCwsjJyWHHjh0AlJSUnLagPZ28vDzq6uoYPXo0CxYsICUl5aT9RUVFjB8/nuXLl9OtWzegfmZv3759WbNmDVD/MM+dO3eex6cXEREREZH2RMWtnBdrLYX/2MfxzGK6/vBqOoT4Oh1JRETagbCwMBYtWkR4eDiFhYVMnz4df39/pk6dSlRUFMOHD29cCgFg8uTJ3HvvvcTGxlJbW0tSUhKzZs0iJiaGhISEZs/MzcrKYujQocTGxjJx4kQef/zxk/avW7eOzMxMpk6dSmxsLLGxsQCsXLmSJUuWEBMTQ2RkJOvWrTvl3Gda43b8+PF861vfIiMjg969e7NkyZJm5RYRERERkdbHWGudztAscXFxNjk52ekY7V7xlkMUv5OJX8IV+MVf7nQcERG5BPbs2UN4eLjTMeQSaOpnbYz52Fob51AkuUQu5b12nzn/vCTXcWUHF45wOoKIiIg47Ez32ZpxK81WkZZH8TuZeMd2w/c7pz64RURERERERERERC6MHk4mzXI8q5SCpAw6hPjSdfTVpzwkRkREpK0YOHAgVVVVJ21bsWIF0dHRDiUSEREREZH2RMWtnLPa4irylu3GzceTwDsiMJ6asC0iIm3X9u3bnY4gIiIiIiLtmJo3OSd1x2vJW5aOrawhcHIk7r4dnI4kIiIiIiIiIiLSZqm4lbOydZbCNZ9SnV1K1x/1p0NwJ6cjiYiIiIiIiIiItGkqbuWsijdlUrErjy7f7Yt3RKDTcURERERERERERNo8FbdyRuWpX1Ky+TA+1/ag8w29nI4jIiIiIiIiIiLSLqi4ldOqyiymYO2ndOjrR8CtV2GMcTqSiIi0YwcPHiQqKqpZxyxdupTs7Oyzjpk5c+aFRGPu3Lls2rTpgs5xOocPH+amm24iIiKCyMhInn322cZ9BQUFJCQkEBoaSkJCAoWFhS2SQURERERELj0Vt9KkmsJK8lek4+7nReDECIyH/lEREZHW51yK24th/vz5DBs2rEXO7eHhwdNPP016ejoffvghixYtIj09HYCFCxcSHx/Pvn37iI+PZ+HChS2SQURERERELj0PpwOI66mrqiF/2W5sTR3dpl2DeydPpyOJiIireWsOfLHr4p6zZzR898zFY01NDRMmTCAlJYXIyEiWL1+Oj48P8+fP54033qCiooJBgwbx5z//mVdeeYXk5GQmTJiAt7c327ZtIy0tjfvvv5+ysjK8vLx49913AcjOziYxMZH9+/dz66238uSTTzZ5/draWu666y6Sk5MxxjBlyhR+9rOfMXnyZL7//e/Tp08f7r777saxaWlpWGvZv38/9913H0ePHsXHx4e//OUv9O/f/5y+LcHBwQQHBwPg6+tLeHg4WVlZREREsG7dOv71r38BMGnSJIYOHcoTTzxxTucVERERERHXpmmUchJbZyn4ewbVX5YT+ONwPLv7OB1JRESkUUZGBjNmzGDPnj34+fnxwgsvADBz5kx27NhBWloaFRUVrF+/njFjxhAXF8fKlStJTU3F3d2dcePG8eyzz7Jz5042bdqEt7c3AKmpqSQlJbFr1y6SkpI4fPhwk9dPTU0lKyuLtLQ0du3axZ133nnS/ri4OFJTU0lNTSUxMZHZs2cDMG3aNP74xz/y8ccf89RTTzFjxoxTzp2cnNxY+p7OwYMH+eSTTxg4cCAAubm5jaVuz549yc3NbcZ3U0REREREXJlm3MpJjr39OZV7CvC/5Uo6Xh3gdBwREXFVZ5kZ21JCQkIYPHgwABMnTuS5555j9uzZbNmyhSeffJLy8nIKCgqIjIxk5MiRJx2bkZFBcHAwAwYMAMDPz69xX3x8PF26dAEgIiKCzMxMQkJCTrl+v379OHDgALNmzWLEiBHcfPPNTeZMSkoiJSWFDRs2UFpaygcffMDYsWMb91dVVZ1yTFxcHC+99NJpP3tpaSmjR4/mD3/4w0nZv2KM0Xr0IiIiIiJtiIpbaVS24wtKt2bR6fpgOg+6zOk4IiIip/h6MWmMobKykhkzZpCcnExISAjz5s2jsrKyWef18vJqPKaCwgAAIABJREFUfO3u7k5NTU2T4wICAti5cyfvvPMOixcvZvXq1bz88ssnjUlLS2PevHls3boVd3d36urq8Pf3JzU1tVmZTlRdXc3o0aOZMGECt912W+P2Hj16kJOTQ3BwMDk5OXTv3v28ryEiIiIiIq5FSyUIAFUHiih89TO8Qv3xH3ml03FERESadOjQIbZt2wbAqlWrGDJkSGNJGxQURGlpKWvXrm0c7+vrS0lJCQBhYWHk5OSwY8cOAEpKSk5b0J5OXl4edXV1jB49mgULFpCSknLS/qKiIsaPH8/y5cvp1q0bUD+zt2/fvqxZswYAay07d+4852taa7nrrrsIDw/ngQceOGnfLbfcwrJlywBYtmwZo0aNatbnERERERER16XiVqjJryD/r3vwCOxI4I/DMe76NUsREXFNYWFhLFq0iPDwcAoLC5k+fTr+/v5MnTqVqKgohg8f3rgUAsDkyZO59957iY2Npba2lqSkJGbNmkVMTAwJCQnNnpmblZXF0KFDiY2NZeLEiTz++OMn7V+3bh2ZmZlMnTqV2NhYYmNjAVi5ciVLliwhJiaGyMhI1q1bd8q5T7fG7fvvv8+KFSvYvHlz4znffPNNAObMmcPGjRsJDQ1l06ZNzJkzp1mfR0REREREXJex1jqdoVni4uJscnKy0zHajLqKGr58IZW6smq6z4jFI8jb6UgiIuKi9uzZQ3h4uNMx5BJo6mdtjPnYWhvnUCS5RC7lvXafOf+8JNdxZQcXjnA6goiIiDjsTPfZmnHbjtlaS/6qPdTkV9J1QrhKWxERERERERERERehh5O1Y0Xr91O1r4iA0aF0vNLf6TgiIiIuZeDAgVRVVZ20bcWKFURHRzuUSERERERE2pMWLW6NMYnAs4A78JK1duFpxo0G1gIDrLVaB+ESKN2WTdm2HDrf0ItOA3o6HUdERMTlbN++3ekIIiIiIiLSjrXYUgnGGHdgEfBdIAIYb4yJaGKcL3A/oP87ukQqPy2k6I39dOzflS7f7et0HBEREREREREREfmallzj9jrgM2vtAWvtceDvwKgmxv0WeAJo3mOd5bxUf1lO/qo9eHb3oev4MIybcTqSiIiIiIiIiIiIfE1LFre9gMMnvD/SsK2RMeabQIi1Vo+UvQRqy6rJW7Yb4+FG4KRI3Ly0xLGIiIiIiIiIiIgrasni9oyMMW7A74Gfn8PYacaYZGNM8tGjR1s+XBtka+rI/2s6tceqCLw9Ao+Ajk5HEhERaZaDBw8SFRXVrGOWLl1Kdnb2WcfMnDnzQqIxd+5cNm3adEHnOJM+ffoQHR1NbGwscXFxjdsLCgpISEggNDSUhIQECgsLWyyDiIiIiIhcWi1Z3GYBISe8792w7Su+QBTwL2PMQeB64HVjTBxfY6190VobZ62N69atWwtGbpustRS+9hnHPy+m6+ir8brCz+lIIiIil8S5FLcXw/z58xk2bFiLXmPLli2kpqaSnPy/57guXLiQ+Ph49u3bR3x8PAsXNvkcWBERERERaYVasrjdAYQaY/oaYzoAPwJe/2qntfaYtTbIWtvHWtsH+BC4xVqb3PTp5HyV/ieL8uRcfL8Tgs83ujsdR0RE5LzV1NQwYcIEwsPDGTNmDOXl5UB9cTpgwACioqKYNm0a1lrWrl1LcnIyEyZMIDY2loqKCnbs2MGgQYOIiYnhuuuuo6SkBIDs7GwSExMJDQ3loYceOu31a2trmTx5MlFRUURHR/PMM88AMHny5MbrxcbGEhsbS3R0NMbUryW/f/9+EhMTufbaa7nhhhvYu3fvRfl+rFu3jkmTJgEwadIkXnvttYtyXhERERERcV6LLXJqra0xxswE3gHcgZettbuNMfOBZGvt62c+g1wMFen5HHvrc7yjg/AbdoXTcUREpI144qMn2FtwccrHr/Tv2p9fXPeLM47JyMhgyZIlDB48mClTpvDCCy8we/ZsZs6cydy5cwG4/fbbWb9+PWPGjOH555/nqaeeIi4ujuPHjzNu3DiSkpIYMGAAxcXFeHt7A5Camsonn3yCl5cXYWFhzJo1i5CQkFOun5qaSlZWFmlpaQAUFRWdtD8uLo7U1FQAHnzwQRITEwGYNm0aixcvJjQ0lO3btzNjxgw2b9580rHJycksXryYl1566ZTrGmO4+eabMcZwzz33MG3aNAByc3MJDg4GoGfPnuTm5p75mywiIiIiIq1Giz6dylr7JvDm17bNPc3YoS2ZpT06nlNGwd8z8LysMwFjr8a4GacjiYiIXJCQkBAGDx4MwMSJE3nuueeYPXs2W7Zs4cknn6S8vJyCggIiIyMZOXLkScdmZGQQHBzMgAEDAPDz+9/SQfHx8XTp0gWAiIgIMjMzmyxu+/Xrx4EDB5g1axYjRozg5ptvbjJnUlISKSkpbNiwgdLSUj744APGjh3buL+qquqUY+Li4posbQHee+89evXqxZdffklCQgL9+/fnxhtvPGmMMaZxhq+IiIiIiLR+LVrcinNqS46Tv2w3pqM7QXdE4NbB3elIIiLShpxtZmxL+XoxaYyhsrKSGTNmkJycTEhICPPmzaOysrJZ5/Xy8mp87e7uTk1NTZPjAgIC2LlzJ++88w6LFy9m9erVvPzyyyeNSUtLY968eWzduhV3d3fq6urw9/dvnIl7Pnr16gVA9+7dufXWW/noo4+48cYb6dGjBzk5OQQHB5OTk0P37loSSURERESkrWjJNW7FIba6jvwV6dSVVRN0RwTuXbzOfpCIiEgrcOjQIbZt2wbAqlWrGDJkSGNJGxQURGlpKWvXrm0c7+vr27iObVhYGDk5OezYsQOAkpKS0xa0p5OXl0ddXR2jR49mwYIFpKSknLS/qKiI8ePHs3z5cr56oKqfnx99+/ZlzZo1QP1DQ3fu3HnO1ywrK2v8DGVlZWzYsIGoqCgAbrnlFpYtWwbAsmXLGDVqVLM+j4iIiIiIuC4Vt22MtZaCVz7l+KESuo4Lo0NvX6cjiYiIXDRhYWEsWrSI8PBwCgsLmT59Ov7+/kydOpWoqCiGDx/euBQC1D807N577yU2Npba2lqSkpKYNWsWMTExJCQkNHtmblZWFkOHDiU2NpaJEyfy+OOPn7R/3bp1ZGZmMnXq1MaHlAGsXLmSJUuWEBMTQ2RkJOvWrTvl3MnJydx9992nbM/NzWXIkCGND1QbMWJE49q5c+bMYePGjYSGhrJp0ybmzJnTrM8jIiIiIiKuy1hrnc7QLHFxcTY5OdnpGC6r+N1DFG/MxG/4FfjddLnTcUREpA3Zs2cP4eHhTseQS6Cpn7Ux5mNrbZxDkeQSuZT32n3m/POSXMeVHVw4wukIIiIi4rAz3Wdrxm0bUv7foxRvzMTnG93xHXrqA1VERERERERERESkddDDydqI40dKKFzzKR2u8CPgtlA9VVpEROQCDRw4kKqqqpO2rVixgujoaIcSiYiIiIhIe6Litg2oPVZF3rJ03Dp5Enh7OMZTE6lFREQu1Pbt252OICIiIiIi7Zgavlau7ngtecvTsVW1BE2OxL1zB6cjiYiIiIiIiIiIyAXSjNtWzNZZCpMyqM4uJXBSJJ49OzkdSUREREQcYIw5CJQAtUCNtTbOGNMVSAL6AAeBH1prC53KKCIiIiLNoxm3rVjxxkwqdufT5Xt98e7f1ek4IiIiIuKsm6y1sSc8lXgO8K61NhR4t+G9iIiIiLQSKm5bqbKUXEq2HKbTgJ50HtLL6TgiIiIi4npGAcsaXi8DfuBgFhERERFpJhW3rVBVZjGFr+zDq18X/EddiTHG6UgiIiIt7uDBg0RFRTXrmKVLl5KdnX3WMTNnzryQaMydO5dNmzZd0DlO5/Dhw9x0001EREQQGRnJs88+27ivoKCAhIQEQkNDSUhIoLCw/rfgrbX85Cc/4aqrruKaa64hJSWlRbKJS7HABmPMx8aYaQ3belhrcxpefwH0cCaaiIiIiJwPFbetTE1BJfnL0/Hw9yJwYjjGQz9CERGR0zmX4vZimD9/PsOGDWuRc3t4ePD000+Tnp7Ohx9+yKJFi0hPTwdg4cKFxMfHs2/fPuLj41m4cCEAb731Fvv27WPfvn28+OKLTJ8+vUWyiUsZYq39JvBd4D5jzI0n7rTWWurL3VMYY6YZY5KNMclHjx69BFFFRERE5Fyo9WtF6ipryFu2G1tbR+DkSNx8PJ2OJCIicknV1NQwYcIEwsPDGTNmDOXl5UB9cTpgwACioqKYNm0a1lrWrl1LcnIyEyZMIDY2loqKCnbs2MGgQYOIiYnhuuuuo6SkBIDs7GwSExMJDQ3loYceOu31a2trmTx5MlFRUURHR/PMM88AMHny5MbrxcbGEhsbS3R0dONvxezfv5/ExESuvfZabrjhBvbu3XvOnzk4OJhvfvObAPj6+hIeHk5WVhYA69atY9KkSQBMmjSJ1157rXH7HXfcgTGG66+/nqKiInJycpq+gLQJ1tqshq9fAq8C1wG5xphggIavX57m2BettXHW2rhu3bpdqsgiIiIichYeTgeQc2PrLAV/z6DmaDlBd0bh2c3H6UgiItKOffG731G159zLx3PhFd6fnr/61RnHZGRksGTJEgYPHsyUKVN44YUXmD17NjNnzmTu3LkA3H777axfv54xY8bw/PPP89RTTxEXF8fx48cZN24cSUlJDBgwgOLiYry9vQFITU3lk08+wcvLi7CwMGbNmkVISMgp109NTSUrK4u0tDQAioqKTtofFxdHamoqAA8++CCJiYkATJs2jcWLFxMaGsr27duZMWMGmzdvPunY5ORkFi9ezEsvvXTaz3/w4EE++eQTBg4cCEBubi7BwcEA9OzZk9zcXACysrJOyt+7d2+ysrIax0rbYozpBLhZa0saXt8MzAdeByYBCxu+rnMupYiIiIg0l2bcthLH3vycyr0F+N9yJR1DA5yOIyIi4oiQkBAGDx4MwMSJE3nvvfcA2LJlCwMHDiQ6OprNmzeze/fuU47NyMggODiYAQMGAODn54eHR/3fYcfHx9OlSxc6duxIREQEmZmZTV6/X79+HDhwgFmzZvH222/j5+fX5LikpCRSUlJYuHAhpaWlfPDBB4wdO5bY2FjuueeeJme/xsXFnbG0LS0tZfTo0fzhD39o8rrGGK173371AN4zxuwEPgL+aa19m/rCNsEYsw8Y1vBeRERERFoJzbhtBUo/yqH0vSw6D7qMztdf5nQcERGRs86MbSlfLyaNMVRWVjJjxgySk5MJCQlh3rx5VFZWNuu8Xl5eja/d3d2pqalpclxAQAA7d+7knXfeYfHixaxevZqXX375pDFpaWnMmzePrVu34u7uTl1dHf7+/o0zcc9HdXU1o0ePZsKECdx2222N23v06EFOTg7BwcHk5OTQvXt3AHr16sXhw4cbxx05coRevXqd9/XFtVlrDwAxTWzPB+IvfSIRERERuRg049bFVe4voui1/XhdHUCXEf2cjiMiIuKoQ4cOsW3bNgBWrVrFkCFDGkvaoKAgSktLWbt2beN4X1/fxnVsw8LCyMnJYceOHQCUlJSctqA9nby8POrq6hg9ejQLFiwgJSXlpP1FRUWMHz+e5cuX89VaoX5+fvTt25c1a9YAYK1l586d53xNay133XUX4eHhPPDAAyftu+WWW1i2bBkAy5YtY9SoUY3bly9fjrWWDz/8kC5dumiZBBERERGRVkbFrQurzqsg/6978AjqSOCP+2Pc9euPIiLSvoWFhbFo0SLCw8MpLCxk+vTp+Pv7M3XqVKKiohg+fHjjUghQ/9Cwe++9l9jYWGpra0lKSmLWrFnExMSQkJDQ7Jm5WVlZDB06lNjYWCZOnMjjjz9+0v5169aRmZnJ1KlTGx9SBrBy5UqWLFlCTEwMkZGRrFt36lKjycnJ3H333adsf//991mxYgWbN29uPOebb74JwJw5c9i4cSOhoaFs2rSJOXPmAPC9732Pfv36cdVVVzF16lReeOGFZn1OERERERFxnrHWOp2hWeLi4mxycrLTMVpcXXk1X76wk7ryarrfF4tHoLfTkUREpJ3bs2cP4eHhTseQS6Cpn7Ux5mNrbZxDkeQSuZT32n3m/POSXMeVHVw4wukIIiIi4rAz3Wdrxq0LsrV15K/aS01hJYG3R6i0FRERERERERERaWf0cDIXY62l6I0DVH1WRMCYq/Hq28XpSCIiIu3SwIEDqaqqOmnbihUriI6OdiiRiIiIiIi0JypuXUzZB9mUfZhD52/3plNcD6fjiIiItFvbt293OoKIiIiIiLRjWirBhVRmFFC0/gAdIwLpMryP03FERERERERERETEISpuXUR1bhn5q/bi2bMTXceFYdyM05FERERERERERETEISpuXUBtWTV5y9Ixnm4ETorAzcvd6UgiIiIiIiIiIiLiIBW3DrM1deSvSKe2uIrAOyLw8O/odCQRERERERERERFxmIpbB1lrKXz1M44fLKbr2KvxutzP6UgiIiIu6+DBg0RFRTXrmKVLl5KdnX3WMTNnzryQaMydO5dNmzZd0DnOpE+fPkRHRxMbG0tcXFzj9oKCAhISEggNDSUhIYHCwkKg/h7jJz/5CVdddRXXXHMNKSkpLZZNRERERERahopbB5VuPUL5x7n4xl+OT0x3p+OIiIi0OedS3F4M8+fPZ9iwYS16jS1btpCamkpycnLjtoULFxIfH8++ffuIj49n4cKFALz11lvs27ePffv28eKLLzJ9+vQWzSYiIiIiIhefiluHVOzO49jbB/G+Jgi/+MudjiMiItIq1NTUMGHCBMLDwxkzZgzl5eVAfXE6YMAAoqKimDZtGtZa1q5dS3JyMhMmTCA2NpaKigp27NjBoEGDiImJ4brrrqOkpASA7OxsEhMTCQ0N5aGHHjrt9Wtra5k8eTJRUVFER0fzzDPPADB58uTG68XGxhIbG0t0dDTG1D9sdP/+/SQmJnLttddyww03sHfv3ovy/Vi3bh2TJk0CYNKkSbz22muN2++44w6MMVx//fUUFRWRk5NzUa4pIiIiIiKXhofTAdqj49mlFCRl4NmrM13HXo1xM05HEhERaZb/rP6UvMOlF/WcQSGdueGHV59xTEZGBkuWLGHw4MFMmTKFF154gdmzZzNz5kzmzp0LwO2338769esZM2YMzz//PE899RRxcXEcP36ccePGkZSUxIABAyguLsbb2xuA1NRUPvnkE7y8vAgLC2PWrFmEhISccv3U1FSysrJIS0sDoKio6KT9cXFxpKamAvDggw+SmJgIwLRp01i8eDGhoaFs376dGTNmsHnz5pOOTU5OZvHixbz00kunXNcYw80334wxhnvuuYdp06YBkJubS3BwMAA9e/YkNzcXgKysrJPy9+7dm6ysrMaxIiIiIiLi+jTj9hKrLTlO/rJ03Lw9CLojEuPp7nQkERGRViMkJITBgwcDMHHiRN577z2gfhmBgQMHEh0dzebNm9m9e/cpx2ZkZBAcHMyAAQMA8PPzw8Oj/u+w4+Pj6dKlCx07diQiIoLMzMwmr9+vXz8OHDjArFmzePvtt/Hza3p9+qSkJFJSUli4cCGlpaV88MEHjB07ltjYWO65554mZ7/GxcU1WdoCvPfee6SkpPDWW2+xaNEitm7desoYY0zjDF8REREREWn9NOP2ErLVteQtT6euvJpu98bg7tfB6UgiIiLn5WwzY1vK14tJYwyVlZXMmDGD5ORkQkJCmDdvHpWVlc06r5eXV+Nrd3d3ampqmhwXEBDAzp07eeedd1i8eDGrV6/m5ZdfPmlMWloa8+bNY+vWrbi7u1NXV4e/v3/jTNzz0atXLwC6d+/OrbfeykcffcSNN95Ijx49yMnJITg4mJycHLp37944/vDhw43HHzlypPEcIiIiIiLSOmjG7SViraVg7T6qD5fQdVwYHXp1djqSiIhIq3Po0CG2bdsGwKpVqxgyZEhjSRsUFERpaSlr165tHO/r69u4jm1YWBg5OTns2LEDgJKSktMWtKeTl5dHXV0do0ePZsGCBaSkpJy0v6ioiPHjx7N8+XK6devG/2/v/qOjqs99j78fA0Ki/BRUChxBQyMJFFqjiLq6UESpKBRBAeHIL2EBglYv98o5PbXIsQW51hY9eP2FGhBKENsGkSpFpT1LEQkYJIAcwIIQqEIABQyRJM/9YzZpyC8mJJlJmM9rrazM7P3dez/z8GX2d57s+W4IXdnbsWNHXn/9dSA0Jti4cWPYxzx+/Hjxazh+/DgrV66kS5cuAPTv35+0tDQA0tLSGDBgQPHy+fPn4+589NFHNGvWTNMkiIiIiIjUM7riNkKOvvsFeRsP0LRvB+K7tIp2OCIiIvVSUlISc+fOZcyYMSQnJzNx4kQSEhIYN24cXbp04dJLLy2eCgFCNw2bMGEC8fHxrFmzhvT0dKZMmUJeXh7x8fGsWrWqSsfPyclh9OjRFBUVATBz5szT1mdkZLB7927GjRtXvCwrK4uFCxcyceJEHn/8cU6ePMnQoUPp1q3badtWNMftl19+ycCBA4HQzdnuueee4rlzp02bxt133828efO47LLLWLJkCQC33XYbK1asIDExkYSEBF555ZUqvU4REREREYk+c/dox1AlqampnpmZGe0wquTbjQc49PvPSPjRxbS46/uaf05EROqlrVu30rlz52iHIRFQ3r+1ma1399QohSQREsmxdodpb0XkOHXZrln9oh2CiIiIRFll42xNlVDLvttzlEOv/w/nd2hKizs7qWgrIiIiIiIiIiIiZ6SpEmpRwZF8Ds7fTFyThlw0ojPWQHVyERGR+qJHjx7k5+eftmzBggV07do1ShGJiIiIiEgsUeG2lhTlF5Kbthn/rojW93Ul7sLzox2SiIiIVMHatWujHYKIiIiIiMQwXQJaC7zIOZS+jZP/OE7Le66k4SUXRDskERERERERERERqUdUuK0F36zcxYktuTTrdznxSS2jHY6IiIiIiIiIiIjUMyrc1rDj67/k6Oq9XNDjUi68/nvRDkdERERERERERETqIRVua1D+rq85/IftNLqiGc37X4GZRTskERERERERERERqYdUuK0hBbl55C7YQoMWjbloeGcsTqkVERGpSbt27aJLly5V2ubVV19l3759Z2wzefLk6oTGo48+yqpVq6q1j4rs2bOHG2+8keTkZFJSUpgzZ07xukOHDtGnTx86depEnz59OHz4MADuzgMPPEBiYiI/+MEP2LBhQ/E2aWlpdOrUiU6dOpGWllYrMYuIiIiISPWpulgDik4UcDBtC14EF41M5ryEhtEOSURERAivcFsTZsyYwc0331wr+27QoAG/+c1v2LJlCx999BFz585ly5YtAMyaNYvevXuzfft2evfuzaxZswD485//zPbt29m+fTsvvPACEydOBEKF3scee4y1a9fy8ccf89hjjxUXe0VEREREpG5pEO0A6jsvdA79/jMKDubRakwXGrZOiHZIIiIite79V1/gq92f1+g+L77scm4cNb7SNgUFBQwfPpwNGzaQkpLC/PnzSUhIYMaMGbz55pvk5eVx3XXX8fzzz/PGG2+QmZnJ8OHDiY+PZ82aNWRnZ/Pggw9y/PhxGjVqxLvvvgvAvn376Nu3Lzt37mTgwIHMnj273OMXFhYyduxYMjMzMTPGjBnDQw89xKhRo7j99tvp0KED9913X3Hb7Oxs3J2dO3dy//33c+DAARISEnjxxRe58sorw8pLmzZtaNOmDQBNmjShc+fO5OTkkJycTEZGBqtXrwZg5MiR9OrViyeeeIKMjAzuvfdezIxrr72WI0eOsH//flavXk2fPn1o2TJ089Q+ffrw9ttvM2zYsLBiERERERGRyNEVt9X09YrPObHtMM0HXEHjxObRDkdEROSctm3bNiZNmsTWrVtp2rQpzz77LACTJ09m3bp1ZGdnk5eXx/Llyxk8eDCpqaksXLiQrKws4uLiGDJkCHPmzGHjxo2sWrWK+Ph4ALKyskhPT2fTpk2kp6ezZ8+eco+flZVFTk4O2dnZbNq0idGjR5+2PjU1laysLLKysujbty9Tp04FYPz48TzzzDOsX7+eJ598kkmTJpXZd2ZmZnHRtyK7du3ik08+oUePHgB8+eWXxUXdSy+9lC+//BKAnJwc2rdvX7xdu3btyMnJqXC5iIiIiIjUPbrithqOrd3PsQ/2ceH13+PCHm2iHY6IiEjEnOnK2NrSvn17rr/+egBGjBjB008/zdSpU3n//feZPXs23377LYcOHSIlJYU77rjjtG23bdtGmzZtuPrqqwFo2rRp8brevXvTrFkzAJKTk9m9e/dpBc5TLr/8cj7//HOmTJlCv379uOWWW8qNMz09nQ0bNrBy5UqOHTvGhx9+yF133VW8Pj8/v8w2qampvPTSSxW+9mPHjjFo0CB+97vfnRb7KWamG6OKiIiIiJxDdMXtWTqx4zBHMnbQOKkFzfpdHu1wREREYkLpwqSZceLECSZNmsTSpUvZtGkT48aN48SJE1Xab6NGjYofx8XFUVBQUG67Fi1asHHjRnr16sVzzz1X7hWy2dnZTJ8+ncWLFxMXF0dRURHNmzcvvhI3KyuLrVu3Vim+kydPMmjQIIYPH86dd95ZvPySSy5h//79AOzfv5+LL74YgLZt25521fDevXtp27ZthctFRERERKTuUeH2LJw88C25r31Gg9YJtBx2JXaerm4RERGJhC+++II1a9YAsGjRIm644YbiIm2rVq04duwYS5cuLW7fpEkTjh49CkBSUhL79+9n3bp1ABw9erTCAm1FDh48SFFREYMGDeLxxx9nw4YNp60/cuQIw4YNY/78+bRu3RoIXdnbsWNHXn/9dQDcnY0bN4Z9THdn7NixdO7cmYcffvi0df379yctLQ2AtLQ0BgwYULx8/vz5uDsfffQRzZo1o02bNtx6662sXLmSw4cPc/jwYVauXMmtt97ROVzIAAAYKUlEQVRapRyIiIiIiEhkaKqEKir69iS5aVuwOGg1MoXzGiuFIiIikZKUlMTcuXMZM2YMycnJTJw4kYSEBMaNG0eXLl249NJLi6dCABg1ahQTJkwovjlZeno6U6ZMIS8vj/j4eFatWlWl4+fk5DB69GiKiooAmDlz5mnrMzIy2L17N+PGjStelpWVxcKFC5k4cSKPP/44J0+eZOjQoXTr1u20bTMzM3nuuefKTJfwwQcfsGDBArp27Ur37t0B+PWvf81tt93GtGnTuPvuu5k3bx6XXXYZS5YsAeC2225jxYoVJCYmkpCQwCuvvAJAy5Yt+cUvflGco0cffbT4RmUiUj91mPZWtEOIul2z+kU7BBERkVph7l57OzfrC8wB4oCX3H1WqfUTgPuBQuAYMN7dt1S2z9TUVM/MzKyliCvnhUUcfDmb/F3f0HpcVxp1aBaVOERERKJh69atdO7cOdphSASU929tZuvdPTVKIUmERHKsrYJjzRQclUcVbkVEpH6rbJxda1MlmFkcMBf4CZAMDDOz5FLNFrl7V3fvDswGnqqteKrL3TmSsZP8nV/T4s5OKtqKiIiIiIiIiIhIranN7/lfA+xw988BzGwxMAAovqLW3b8p0f4CoPYu/62mYx/s4/jH/6BJr3ZccNUl0Q5HREREalmPHj3Iz88/bdmpKQtERERERERqW20WbtsCe0o83wv0KN3IzO4HHgbOB26qxXjOWt62Q3z91uc0TrmIprd0iHY4IiIiEgFr166NdggiIiIiIhLDam2qhHC5+1x3vwJ4BPiP8tqY2XgzyzSzzAMHDkQ0vpP/OM6hRZ/RsM0FtByShJ1nET2+iIiIiIiIiIiIxJ7aLNzmAO1LPG8XLKvIYuCn5a1w9xfcPdXdU1u3bl2DIVau8Nh3HEzbjJ0fx0UjUzjv/LiIHVtERERERERERERiV20WbtcBncyso5mdDwwFlpVsYGadSjztB2yvxXiqxAuKyF2wlcKjJ2l1bzINmjWKdkgiIiIiIiIiIiISI2ptjlt3LzCzycA7QBzwsrtvNrMZQKa7LwMmm9nNwEngMDCytuKpqqL8QnCn5d3f5/z2TaIdjoiIiIiIiIiIiMSQWp3j1t1XuPv33f0Kd/9VsOzRoGiLuz/o7inu3t3db3T3zbUZT1XEXdCQ1hO6kfCDyE3NICIiIhXbtWsXXbp0qdI2r776Kvv27Ttjm8mTJ1cnNB599FFWrVpVrX2cSWFhIT/84Q+5/fbbi5f9/e9/p0ePHiQmJjJkyBC+++47APLz8xkyZAiJiYn06NGDXbt2FW8zc+ZMEhMTSUpK4p133qnVmEVERERE5OxF/eZkdZluRCYiIlK/hVO4rQkzZszg5ptvrtVjzJkzh86dO5+27JFHHuGhhx5ix44dtGjRgnnz5gEwb948WrRowY4dO3jooYd45JFHANiyZQuLFy9m8+bNvP3220yaNInCwsJajVtERERERM5OrU2VICIiIueuI2/u5Lt9x2t0n+d/7wKa33FFpW0KCgoYPnw4GzZsICUlhfnz55OQkMCMGTN48803ycvL47rrruP555/njTfeIDMzk+HDhxMfH8+aNWvIzs7mwQcf5Pjx4zRq1Ih3330XgH379tG3b1927tzJwIEDmT17drnHLywsZOzYsWRmZmJmjBkzhoceeohRo0Zx++2306FDB+67777ittnZ2bg7O3fu5P777+fAgQMkJCTw4osvcuWVV4adm7179/LWW2/x85//nKeeegoAd+e9995j0aJFAIwcOZLp06czceJEMjIymD59OgCDBw9m8uTJuDsZGRkMHTqURo0a0bFjRxITE/n444/p2bNn2LGIiIiIiEhk6IpbERERqTe2bdvGpEmT2Lp1K02bNuXZZ58FYPLkyaxbt47s7Gzy8vJYvnw5gwcPJjU1lYULF5KVlUVcXBxDhgxhzpw5bNy4kVWrVhEfHw9AVlYW6enpbNq0ifT0dPbs2VPu8bOyssjJySE7O5tNmzYxevTo09anpqaSlZVFVlYWffv2ZerUqQCMHz+eZ555hvXr1/Pkk08yadKkMvvOzMwsLvqW9rOf/YzZs2dz3nn/HLrl5ubSvHlzGjQI/R2+Xbt25OTkAJCTk0P79u0BaNCgAc2aNSM3N/e05aW3ERERERGRukVX3IqIiEiVnenK2NrSvn17rr/+egBGjBjB008/zdSpU3n//feZPXs23377LYcOHSIlJYU77rjjtG23bdtGmzZtuPrqqwFo2rRp8brevXvTrFkzAJKTk9m9e/dpBc5TLr/8cj7//HOmTJlCv379uOWWW8qNMz09nQ0bNrBy5UqOHTvGhx9+yF133VW8Pj8/v8w2qampvPTSS2WWL1++nIsvvpirrrqK1atXnyFDIiIiIiJyrlDhVkREROoNMyvz/MSJE0yaNInMzEzat2/P9OnTOXHiRJX226hRo+LHcXFxFBQUlNuuRYsWbNy4kXfeeYfnnnuOJUuW8PLLL5/WJjs7m+nTp/O3v/2NuLg4ioqKaN68OVlZWVWK6ZQPPviAZcuWsWLFCk6cOME333zDiBEjWLBgAUeOHKGgoIAGDRqwd+9e2rZtC0Dbtm3Zs2cP7dq1o6CggK+//pqLLrqoePkpJbcREREREZG6RYVbERERqTe++OIL1qxZQ8+ePVm0aBE33HBDcZG2VatWHDt2jKVLlzJ48GAAmjRpwtGjRwFISkpi//79rFu3jquvvpqjR48WT5UQroMHD3L++eczaNAgkpKSGDFixGnrjxw5wrBhw5g/fz6tW7cGQlf2duzYkddff5277roLd+fTTz+lW7duYR1z5syZzJw5E4DVq1fz5JNP8tprrwFw4403snTpUoYOHUpaWhoDBgwAoH///qSlpdGzZ0+WLl3KTTfdhJnRv39/7rnnHh5++GH27dvH9u3bueaaa6qUAxGRc02HaW9FO4So2zWrX7RDEBGRcmiOWxEREak3kpKSmDt3Lp07d+bw4cNMnDiR5s2bM27cOLp06cKtt95aPBUCwKhRo5gwYQLdu3ensLCQ9PR0pkyZQrdu3ejTp0+Vr8zNycmhV69edO/enREjRhQXVE/JyMhg9+7djBs3ju7du9O9e3cAFi5cyLx58+jWrRspKSlkZGSU2Xdlc9xW5IknnuCpp54iMTGR3Nxcxo4dC8DYsWPJzc0lMTGRp556ilmzZgGQkpLC3XffTXJyMn379mXu3LnExcVV6ZgiIiIiIhIZ5u7RjqFKUlNTPTMzM9phiIiIxJytW7fSuXPnaIchEVDev7WZrXf31CiFJBESybG2rnKsmasclcfq51E51BW3IiLRVNk4W1fcioiIiIiIiIiIiNQxmuNWREREpBw9evQgPz//tGULFiyga9euUYpIRERERERiiQq3IiIiIuVYu3ZttEMQERGpFzTdhKabEJHaoakSREREJGz1bW58qTr9G4uIiIiI1A0q3IqIiEhYGjduTG5urgp75zB3Jzc3l8aNG0c7FBERERGRmKepEkRERCQs7dq1Y+/evRw4cCDaoUgtaty4Me3atYt2GCIiIiIiMU+FWxEREQlLw4YN6dixY7TDEJEqMLO+wBwgDnjJ3WdFOSQREamA5gqu/lzBymHNzLesPNadeas1VYKIiIiIyDnIzOKAucBPgGRgmJklRzcqEREREQmXCrciIiIiIuema4Ad7v65u38HLAYGRDkmEREREQmTCrciIiIiIuemtsCeEs/3BstEREREpB6w+nZnaDM7AOyO4CFbAQcjeLz6QDkpSzkpSzkpSzkpn/JSlnJSlnJSVqRzcpm7t47g8aSazGww0Nfd7wue/yvQw90nl2o3HhgfPE0CtkU00OjSe0v1KYc1Q3msPuWw+pTDmqE8Vl+s5bDCcXa9uzlZpD8wmFmmu6dG8ph1nXJSlnJSlnJSlnJSPuWlLOWkLOWkLOVEwpADtC/xvF2w7DTu/gLwQqSCqkv0/6j6lMOaoTxWn3JYfcphzVAeq085/CdNlSAiIiIicm5aB3Qys45mdj4wFFgW5ZhEREREJEz17opbERERERE5M3cvMLPJwDtAHPCyu2+OclgiIiIiEiYVbs8sJr82dgbKSVnKSVnKSVnKSfmUl7KUk7KUk7KUEzkjd18BrIh2HHWY/h9Vn3JYM5TH6lMOq085rBnKY/Uph4F6d3MyERERERERERERkXOd5rgVERERERERERERqWNUuAXM7GUz+8rMsitYb2b2tJntMLNPzexHkY4x0sLISS8z+9rMsoKfRyMdY6SZWXsze9/MtpjZZjN7sJw2MdVXwsxJTPUVM2tsZh+b2cYgJ4+V06aRmaUH/WStmXWIfKSRE2ZORpnZgRL95L5oxBppZhZnZp+Y2fJy1sVUPznlDDmJ1X6yy8w2Ba85s5z1MXXuEakJZxrrSvkqGvuZWUsz+4uZbQ9+t4h2rHVZee/rymHlqtr3dG6sWFX6n/IYUpP9z8xGBu23m9nIaL2maCk91rfQDVTXBrlKt9DNVCv9LGRm/xYs32Zmt0bnlUSOCrchrwJ9K1n/E6BT8DMe+H8RiCnaXqXynAD8t7t3D35mRCCmaCsA/pe7JwPXAvebWXKpNrHWV8LJCcRWX8kHbnL3bkB3oK+ZXVuqzVjgsLsnAr8FnohwjJEWTk4A0kv0k5ciG2LUPAhsrWBdrPWTUyrLCcRmPwG4MXjNqeWsi7Vzj0hNeJUzj3WlrIrGftOAd929E/Bu8FwqV/p9XTmsXFX7ns6NlQu3/ymPITXS/8ysJfBLoAdwDfDLGPwjTemx/hPAb4PPPIcJfQaCCj4LBXkfCqQQOo8/a2ZxEYo9KlS4Bdz9b8ChSpoMAOZ7yEdAczNrE5nooiOMnMQcd9/v7huCx0cJvdm0LdUspvpKmDmJKcG//bHgacPgp/Rk4gOAtODxUqC3mVmEQoy4MHMSc8ysHdAPqKj4GFP9BMLKiZQvps49IjVBY92zU8nYr+Q5Kw34aXQirNeUw0qcRd/TubFqlMdK1GD/uxX4i7sfcvfDwF+IoT8ilh7rB59tbiL0WQfK5rC8z0IDgMXunu/ufwd2ECqCn7NUuA1PW2BPied7ifHiVKCnhb76/GczS4l2MJEUXKb/Q2BtqVUx21cqyQnEWF8Jvv6RBXxF6MRcYT9x9wLga+CiyEYZWWHkBGBQ8FWipWbWPsIhRsPvgP8DFFWwPub6CWfOCcReP4HQHzpWmtl6MxtfzvqYPfeISPSUGvtd4u77g1X/AC6JUlj1RXnv68phmMLsezo3Vqwq/U95LKWa/S/W81l6rH8RcCT4rAOn56Oiz0Ixl0MVbuVsbQAuC776/AzwpyjHEzFmdiHwBvAzd/8m2vHUBWfIScz1FXcvdPfuQDvgGjPrEu2Yoi2MnLwJdHD3HxD6y3Na6X2cS8zsduArd18f7VjqijBzElP9pIQb3P1HhL52d7+Z/TjaAYlIbKts7Ofujr5ZcyaVvq8rhxVT36sR6n9nSf3v7Onzz9lT4TY8OUDJq3raBctilrt/c+qrz+6+AmhoZq2iHFatM7OGhN6oF7r7H8ppEnN95Uw5idW+AuDuR4D3Kfv1l+J+YmYNgGZAbmSji46KcuLuue6eHzx9Cbgq0rFF2PVAfzPbBSwGbjKz10q1ibV+csacxGA/AcDdc4LfXwF/pOzXwWLu3CMi0VPB2O/LU1+fDn5/Fa346oMK3teVwzOoYt/TubECVex/ymOghvpfLOezzFgfmENoGokGQZuS+ajos1DM5VCF2/AsA+4N7gx4LfB1icvhY5KZXXpqrkUzu4ZQXzqXCwqn5l+ZB2x196cqaBZTfSWcnMRaXzGz1mbWPHgcD/QBPivVbBlw6g6ig4H3gr/QnpPCyUmpubL6U/nNqeo9d/83d2/n7h0ITa7/nruPKNUspvpJODmJtX4CYGYXmFmTU4+BW4DsUs1i6twjItFTydiv5DlrJJAR6djqi0re15XDSpxF39O5sRxn0f+UR2q0/70D3GJmLSx0U7JbgmXnvArG+sMJXdQzOGhWOoflfRZaBgw1s0Zm1pHQDeA+jtDLiIoGZ25y7jOz3wO9gFZmtpfQXf4aArj7c8AK4DZCkx5/C4yOTqSRE0ZOBgMTzawAyAOGnssFhcD1wL8Cmyw0VyfAvwP/AjHbV8LJSaz1lTZAmoXubHkesMTdl5vZDCDT3ZcROukvMLMdhG6MMjR64UZEODl5wMz6E7pj6yFgVNSijaIY7yflUj/hEuCPwd+/GgCL3P1tM5sAMXvuEam28sa67j4vulHVCxWN/WYBS8xsLLAbuDtK8dUHFb2vr0M5rExV+57OjeWrav9THkNqpP+5+yEz+09gXdBuhrvH+o0yHwEWm9njwCeEPgNBBZ+F3H2zmS0BthD6THC/uxdGPuzIsXO7fiIiIiIiIiIiIiJS/2iqBBEREREREREREZE6RoVbERERERERERERkTpGhVsRERERERERERGROkaFWxEREREREREREZE6RoVbERERERERERERkTpGhVsRkTCZWaGZZZX4mVaD++5gZtk1tT8RERERkZpmZs3NbFKYbT8MfhePc82sl5ktr8bxXzWzwWe7fbCPUWb2X1Xc5t/DaFPt2ERESlPhVkQkfHnu3r3Ez6xoByQiIiIiEkHNgbAKt+5+XS3HEklnLNyKiNQGFW5FRKrJzHaZ2Wwz22RmH5tZYrC8g5m9Z2afmtm7ZvYvwfJLzOyPZrYx+Dk1qI0zsxfNbLOZrTSz+KD9A2a2JdjP4ii9TBERERGRWcAVwbfP/q+ZXRiMczcEY+EBpxqa2bHKdmRmF5jZy8H4+ZOS25ZoY2b2X2a2zcxWAReXWHeVmf3VzNab2Ttm1iZYHs7Yub2ZrTaz7Wb2yxL7/FOwv81mNj5YNguID17zwmDZvcH+N5rZghL7/bGZfWhmn+vqWxGpCQ2iHYCISD0Sb2ZZJZ7PdPf04PHX7t7VzO4FfgfcDjwDpLl7mpmNAZ4Gfhr8/qu7DzSzOOBCoAXQCRjm7uPMbAkwCHgNmAZ0dPd8M2seiRcqIiIiIlKOaUAXd+8OYGYNgIHu/o2ZtQI+MrNl7u5h7OvnwHvuPiYY435sZqvc/XiJNgOBJCAZuATYArxsZg0JjbUHuPsBMxsC/AoYQ3hj52uALsC3wDoze8vdM4Ex7n4ouIBinZm94e7TzGxyidecAvwHcJ27HzSzliX22wa4AbgSWAYsDSMPIiIVUuFWRCR8eacGbOX4fYnfvw0e9wTuDB4vAGYHj28C7gVw90LgazNrAfzd3U8VhtcDHYLHnwILzexPwJ9q4HWIiIiIiNQEA35tZj8GioC2hAqs/whj21uA/mY2NXjeGPgXYGuJNj8Gfh+MmfeZ2XvB8iRChde/mBlAHLA/WBfO2Pkv7p4LYGZ/IFRszQQeMLOBQZv2hC6syC217U3A6+5+EMDdD5VY9yd3LwK2mNklYeRARKRSKtyKiNQMr+BxVeSXeFwIxAeP+xEatN4B/NzMurp7wVkeQ0RERESkpgwHWgNXuftJM9tFqAAbDgMGufu2sziuAZvdvWc568IZO5cer7uZ9QJuBnq6+7dmtprwX8spJcfzVsVtRUTK0By3IiI1Y0iJ32uCxx8CQ4PHw4H/Dh6/C0wEMLM4M2tW0U7N7Dygvbu/DzwCNCM0tYKIiIiISKQdBZqUeN4M+Coo2t4IXFaFfb0DTLHgklkz+2E5bf4GDAnGzG2AG4Pl24DWZtYz2LahmaVUYezcx8xaBlMi/BT4IGh7OCjaXglcW6L9yWB6BoD3gLvM7KLg2CWnShARqVG64lZEJHyl57h9292nBY9bmNmnhP7KPixYNgV4xcz+N3AAGB0sfxB4wczGErqydiL//GpXaXHAa0Fx14Cn3f1Ijb0iEREREZEwuXuumX1gZtnAn4EngDfNbBOhqQY+q8Lu/pPQvSE+DQqufyd0n4iS/khoaoItwBcEF0i4+3fBzb+eDsbJDYJ9/Q/hjZ0/Bt4A2gGvuXtm8BommNlWQoXhj0q0fyGIc4O7DzezXwF/NbNC4BNgVBVet4hI2Cy8OcNFRKQiwVfCUk/NcyUiIiIiIiIiUl2aKkFERERERERERESkjtEVtyIiIiIiIiIiIiJ1jK64FREREREREREREaljVLgVERERERERERERqWNUuBURERERERERERGpY1S4FREREREREREREaljVLgVERERERERERERqWNUuBURERERERERERGpY/4/bLnzvOsJjOsAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 1728x576 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.figure(figsize=(24,8))\n",
        "\n",
        "plt.subplot(121)\n",
        "plt.title('Training accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "for i in range(len(batch_sizes)):\n",
        "  plt.plot(np.arange(1,6,1), training_histories_fixed_epoch[i].history['accuracy'], label='batch_size : {}'.format(batch_sizes[i]))\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(122)\n",
        "plt.title(\"Temps d'exécution en fonction de la taille des batch\")\n",
        "plt.xlabel(\"taille des batch\")\n",
        "plt.ylabel(\"temps d'exécution (en secondes)\")\n",
        "plt.bar(np.arange(1,8,1), times_fixed_epoch)\n",
        "plt.xticks(ticks=np.arange(1,8,1), labels=batch_sizes);"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "w69hPP04idZO"
      },
      "source": [
        "## <a name=\"fixed_iterations\"></a> B - Nombre fixe d'itérations"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Zh5zwPLoj_bh"
      },
      "source": [
        "- (f) Définir une liste ```batch_sizes_2``` contenant les valeurs ```[20, 50, 200, 500, 2000, 4000]```.\n",
        "\n",
        "- (g) Définir une liste ```epochs``` contenant les valeurs de ```batch_size_2``` divisées par 5.\n",
        ">  - Cela permettra d'entrainer le modèle sur un nombre fixe d'itération plutôt qu'un nombre fixe d'epochs.\n",
        ">\n",
        ">  - Attention à ce que les valeurs de ```epochs``` soient bien des valeurs entières."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "IHnRjZIiddeY"
      },
      "outputs": [],
      "source": [
        "batch_sizes_2 = [20, 50, 200, 500, 2000, 4000]\n",
        "epochs = [np.int(x/5) for x in batch_sizes_2]"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_7QHYcGlj5Fa"
      },
      "source": [
        "- (h) Pour chaque taille de batch défini dans ```batch_sizes```, définir et entraîner le modèle sur les données ```X_train``` et ```y_train``` grâce à la méthode ```fit``` :\n",
        ">  - L'entraînement devra se faire avec les éléments de la liste ```epoch``` (paramètre ```epochs```).\n",
        ">\n",
        ">  - La perfomance du modèle devra être évaluée sur un échantillon de validation contenant 20% des données (paramètre ```validation_split```).\n",
        ">\n",
        ">  - La sortie des entraînements devra être stockée dans une liste nommée ```training_histories_fixed_iterations```.\n",
        "\n",
        "- (i) Stocker dans une liste ```times_fixed_iterations``` la durée de chaque entraînement."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AqRHKHFJMiWM",
        "outputId": "4cbdb394-9599-47a1-c1a3-67741ba7bac8"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/4\n",
            "2400/2400 [==============================] - 4s 2ms/step - loss: 0.8414 - accuracy: 0.8149 - val_loss: 0.2539 - val_accuracy: 0.9248\n",
            "Epoch 2/4\n",
            "2400/2400 [==============================] - 4s 2ms/step - loss: 0.2469 - accuracy: 0.9282 - val_loss: 0.2182 - val_accuracy: 0.9371\n",
            "Epoch 3/4\n",
            "2400/2400 [==============================] - 4s 2ms/step - loss: 0.2045 - accuracy: 0.9424 - val_loss: 0.2016 - val_accuracy: 0.9425\n",
            "Epoch 4/4\n",
            "2400/2400 [==============================] - 4s 2ms/step - loss: 0.1802 - accuracy: 0.9479 - val_loss: 0.1925 - val_accuracy: 0.9457\n",
            "Epoch 1/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 1.1064 - accuracy: 0.7620 - val_loss: 0.3106 - val_accuracy: 0.9160\n",
            "Epoch 2/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.3054 - accuracy: 0.9164 - val_loss: 0.2459 - val_accuracy: 0.9321\n",
            "Epoch 3/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.2442 - accuracy: 0.9314 - val_loss: 0.2213 - val_accuracy: 0.9374\n",
            "Epoch 4/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.2057 - accuracy: 0.9421 - val_loss: 0.2059 - val_accuracy: 0.9414\n",
            "Epoch 5/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.1879 - accuracy: 0.9471 - val_loss: 0.2014 - val_accuracy: 0.9418\n",
            "Epoch 6/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.1799 - accuracy: 0.9491 - val_loss: 0.1935 - val_accuracy: 0.9442\n",
            "Epoch 7/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.1662 - accuracy: 0.9531 - val_loss: 0.1907 - val_accuracy: 0.9450\n",
            "Epoch 8/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.1514 - accuracy: 0.9565 - val_loss: 0.1794 - val_accuracy: 0.9482\n",
            "Epoch 9/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.1438 - accuracy: 0.9606 - val_loss: 0.1796 - val_accuracy: 0.9488\n",
            "Epoch 10/10\n",
            "960/960 [==============================] - 2s 2ms/step - loss: 0.1356 - accuracy: 0.9621 - val_loss: 0.1732 - val_accuracy: 0.9515\n",
            "Epoch 1/40\n",
            "240/240 [==============================] - 1s 4ms/step - loss: 1.6554 - accuracy: 0.6306 - val_loss: 0.5838 - val_accuracy: 0.8857\n",
            "Epoch 2/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.5290 - accuracy: 0.8857 - val_loss: 0.3681 - val_accuracy: 0.9090\n",
            "Epoch 3/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.3653 - accuracy: 0.9079 - val_loss: 0.3021 - val_accuracy: 0.9202\n",
            "Epoch 4/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.3013 - accuracy: 0.9179 - val_loss: 0.2701 - val_accuracy: 0.9241\n",
            "Epoch 5/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.2683 - accuracy: 0.9259 - val_loss: 0.2487 - val_accuracy: 0.9303\n",
            "Epoch 6/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.2443 - accuracy: 0.9322 - val_loss: 0.2354 - val_accuracy: 0.9328\n",
            "Epoch 7/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.2235 - accuracy: 0.9366 - val_loss: 0.2215 - val_accuracy: 0.9370\n",
            "Epoch 8/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.2084 - accuracy: 0.9411 - val_loss: 0.2143 - val_accuracy: 0.9391\n",
            "Epoch 9/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.2009 - accuracy: 0.9426 - val_loss: 0.2057 - val_accuracy: 0.9411\n",
            "Epoch 10/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1849 - accuracy: 0.9471 - val_loss: 0.1981 - val_accuracy: 0.9434\n",
            "Epoch 11/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1760 - accuracy: 0.9487 - val_loss: 0.1939 - val_accuracy: 0.9451\n",
            "Epoch 12/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1714 - accuracy: 0.9519 - val_loss: 0.1904 - val_accuracy: 0.9457\n",
            "Epoch 13/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1600 - accuracy: 0.9540 - val_loss: 0.1872 - val_accuracy: 0.9444\n",
            "Epoch 14/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1549 - accuracy: 0.9552 - val_loss: 0.1835 - val_accuracy: 0.9465\n",
            "Epoch 15/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1480 - accuracy: 0.9575 - val_loss: 0.1806 - val_accuracy: 0.9477\n",
            "Epoch 16/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1457 - accuracy: 0.9582 - val_loss: 0.1772 - val_accuracy: 0.9492\n",
            "Epoch 17/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1376 - accuracy: 0.9609 - val_loss: 0.1753 - val_accuracy: 0.9480\n",
            "Epoch 18/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1322 - accuracy: 0.9623 - val_loss: 0.1751 - val_accuracy: 0.9483\n",
            "Epoch 19/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1315 - accuracy: 0.9627 - val_loss: 0.1722 - val_accuracy: 0.9485\n",
            "Epoch 20/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1240 - accuracy: 0.9652 - val_loss: 0.1706 - val_accuracy: 0.9506\n",
            "Epoch 21/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1232 - accuracy: 0.9650 - val_loss: 0.1736 - val_accuracy: 0.9484\n",
            "Epoch 22/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1189 - accuracy: 0.9663 - val_loss: 0.1696 - val_accuracy: 0.9497\n",
            "Epoch 23/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1138 - accuracy: 0.9675 - val_loss: 0.1683 - val_accuracy: 0.9501\n",
            "Epoch 24/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1121 - accuracy: 0.9678 - val_loss: 0.1683 - val_accuracy: 0.9505\n",
            "Epoch 25/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1094 - accuracy: 0.9689 - val_loss: 0.1684 - val_accuracy: 0.9503\n",
            "Epoch 26/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1099 - accuracy: 0.9686 - val_loss: 0.1687 - val_accuracy: 0.9512\n",
            "Epoch 27/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1073 - accuracy: 0.9690 - val_loss: 0.1671 - val_accuracy: 0.9517\n",
            "Epoch 28/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1032 - accuracy: 0.9711 - val_loss: 0.1664 - val_accuracy: 0.9513\n",
            "Epoch 29/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.1011 - accuracy: 0.9713 - val_loss: 0.1663 - val_accuracy: 0.9511\n",
            "Epoch 30/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0973 - accuracy: 0.9724 - val_loss: 0.1666 - val_accuracy: 0.9519\n",
            "Epoch 31/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0989 - accuracy: 0.9705 - val_loss: 0.1667 - val_accuracy: 0.9510\n",
            "Epoch 32/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0954 - accuracy: 0.9735 - val_loss: 0.1669 - val_accuracy: 0.9529\n",
            "Epoch 33/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0902 - accuracy: 0.9741 - val_loss: 0.1683 - val_accuracy: 0.9517\n",
            "Epoch 34/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0923 - accuracy: 0.9731 - val_loss: 0.1656 - val_accuracy: 0.9524\n",
            "Epoch 35/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0862 - accuracy: 0.9756 - val_loss: 0.1675 - val_accuracy: 0.9534\n",
            "Epoch 36/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0889 - accuracy: 0.9749 - val_loss: 0.1681 - val_accuracy: 0.9518\n",
            "Epoch 37/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0857 - accuracy: 0.9758 - val_loss: 0.1684 - val_accuracy: 0.9517\n",
            "Epoch 38/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0839 - accuracy: 0.9768 - val_loss: 0.1677 - val_accuracy: 0.9525\n",
            "Epoch 39/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0829 - accuracy: 0.9764 - val_loss: 0.1686 - val_accuracy: 0.9538\n",
            "Epoch 40/40\n",
            "240/240 [==============================] - 1s 3ms/step - loss: 0.0822 - accuracy: 0.9765 - val_loss: 0.1675 - val_accuracy: 0.9522\n",
            "Epoch 1/100\n",
            "96/96 [==============================] - 1s 7ms/step - loss: 1.9820 - accuracy: 0.5459 - val_loss: 1.1161 - val_accuracy: 0.8384\n",
            "Epoch 2/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.9783 - accuracy: 0.8415 - val_loss: 0.6382 - val_accuracy: 0.8830\n",
            "Epoch 3/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.6128 - accuracy: 0.8751 - val_loss: 0.4692 - val_accuracy: 0.8980\n",
            "Epoch 4/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.4675 - accuracy: 0.8946 - val_loss: 0.3904 - val_accuracy: 0.9083\n",
            "Epoch 5/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.3971 - accuracy: 0.9031 - val_loss: 0.3437 - val_accuracy: 0.9155\n",
            "Epoch 6/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.3491 - accuracy: 0.9100 - val_loss: 0.3129 - val_accuracy: 0.9197\n",
            "Epoch 7/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.3142 - accuracy: 0.9186 - val_loss: 0.2922 - val_accuracy: 0.9228\n",
            "Epoch 8/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2927 - accuracy: 0.9228 - val_loss: 0.2761 - val_accuracy: 0.9261\n",
            "Epoch 9/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2663 - accuracy: 0.9304 - val_loss: 0.2635 - val_accuracy: 0.9291\n",
            "Epoch 10/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2649 - accuracy: 0.9289 - val_loss: 0.2533 - val_accuracy: 0.9315\n",
            "Epoch 11/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2479 - accuracy: 0.9332 - val_loss: 0.2448 - val_accuracy: 0.9339\n",
            "Epoch 12/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2406 - accuracy: 0.9352 - val_loss: 0.2371 - val_accuracy: 0.9359\n",
            "Epoch 13/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2308 - accuracy: 0.9359 - val_loss: 0.2312 - val_accuracy: 0.9371\n",
            "Epoch 14/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2191 - accuracy: 0.9397 - val_loss: 0.2254 - val_accuracy: 0.9375\n",
            "Epoch 15/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2116 - accuracy: 0.9418 - val_loss: 0.2217 - val_accuracy: 0.9388\n",
            "Epoch 16/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2026 - accuracy: 0.9450 - val_loss: 0.2173 - val_accuracy: 0.9401\n",
            "Epoch 17/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.2018 - accuracy: 0.9440 - val_loss: 0.2116 - val_accuracy: 0.9414\n",
            "Epoch 18/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1974 - accuracy: 0.9435 - val_loss: 0.2088 - val_accuracy: 0.9419\n",
            "Epoch 19/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1893 - accuracy: 0.9473 - val_loss: 0.2059 - val_accuracy: 0.9418\n",
            "Epoch 20/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1811 - accuracy: 0.9495 - val_loss: 0.2022 - val_accuracy: 0.9433\n",
            "Epoch 21/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1784 - accuracy: 0.9501 - val_loss: 0.2009 - val_accuracy: 0.9452\n",
            "Epoch 22/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1702 - accuracy: 0.9529 - val_loss: 0.1969 - val_accuracy: 0.9445\n",
            "Epoch 23/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1687 - accuracy: 0.9521 - val_loss: 0.1941 - val_accuracy: 0.9467\n",
            "Epoch 24/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1662 - accuracy: 0.9540 - val_loss: 0.1929 - val_accuracy: 0.9457\n",
            "Epoch 25/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1580 - accuracy: 0.9563 - val_loss: 0.1894 - val_accuracy: 0.9468\n",
            "Epoch 26/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1580 - accuracy: 0.9556 - val_loss: 0.1889 - val_accuracy: 0.9470\n",
            "Epoch 27/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1524 - accuracy: 0.9565 - val_loss: 0.1856 - val_accuracy: 0.9482\n",
            "Epoch 28/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1567 - accuracy: 0.9567 - val_loss: 0.1837 - val_accuracy: 0.9482\n",
            "Epoch 29/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1502 - accuracy: 0.9583 - val_loss: 0.1825 - val_accuracy: 0.9482\n",
            "Epoch 30/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1445 - accuracy: 0.9587 - val_loss: 0.1810 - val_accuracy: 0.9492\n",
            "Epoch 31/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1438 - accuracy: 0.9595 - val_loss: 0.1804 - val_accuracy: 0.9489\n",
            "Epoch 32/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1413 - accuracy: 0.9600 - val_loss: 0.1777 - val_accuracy: 0.9500\n",
            "Epoch 33/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1374 - accuracy: 0.9612 - val_loss: 0.1776 - val_accuracy: 0.9492\n",
            "Epoch 34/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1356 - accuracy: 0.9613 - val_loss: 0.1759 - val_accuracy: 0.9505\n",
            "Epoch 35/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1296 - accuracy: 0.9632 - val_loss: 0.1755 - val_accuracy: 0.9499\n",
            "Epoch 36/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1295 - accuracy: 0.9638 - val_loss: 0.1752 - val_accuracy: 0.9500\n",
            "Epoch 37/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1265 - accuracy: 0.9655 - val_loss: 0.1737 - val_accuracy: 0.9511\n",
            "Epoch 38/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1275 - accuracy: 0.9625 - val_loss: 0.1727 - val_accuracy: 0.9505\n",
            "Epoch 39/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1236 - accuracy: 0.9660 - val_loss: 0.1731 - val_accuracy: 0.9508\n",
            "Epoch 40/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1204 - accuracy: 0.9655 - val_loss: 0.1713 - val_accuracy: 0.9515\n",
            "Epoch 41/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1157 - accuracy: 0.9684 - val_loss: 0.1716 - val_accuracy: 0.9507\n",
            "Epoch 42/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1208 - accuracy: 0.9669 - val_loss: 0.1703 - val_accuracy: 0.9514\n",
            "Epoch 43/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1133 - accuracy: 0.9682 - val_loss: 0.1711 - val_accuracy: 0.9509\n",
            "Epoch 44/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1117 - accuracy: 0.9686 - val_loss: 0.1707 - val_accuracy: 0.9513\n",
            "Epoch 45/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1142 - accuracy: 0.9686 - val_loss: 0.1701 - val_accuracy: 0.9518\n",
            "Epoch 46/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1086 - accuracy: 0.9689 - val_loss: 0.1693 - val_accuracy: 0.9513\n",
            "Epoch 47/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1098 - accuracy: 0.9690 - val_loss: 0.1689 - val_accuracy: 0.9522\n",
            "Epoch 48/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1061 - accuracy: 0.9704 - val_loss: 0.1692 - val_accuracy: 0.9523\n",
            "Epoch 49/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1064 - accuracy: 0.9698 - val_loss: 0.1683 - val_accuracy: 0.9523\n",
            "Epoch 50/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1036 - accuracy: 0.9710 - val_loss: 0.1695 - val_accuracy: 0.9514\n",
            "Epoch 51/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1056 - accuracy: 0.9701 - val_loss: 0.1693 - val_accuracy: 0.9511\n",
            "Epoch 52/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1027 - accuracy: 0.9707 - val_loss: 0.1692 - val_accuracy: 0.9523\n",
            "Epoch 53/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0992 - accuracy: 0.9727 - val_loss: 0.1668 - val_accuracy: 0.9520\n",
            "Epoch 54/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0985 - accuracy: 0.9729 - val_loss: 0.1692 - val_accuracy: 0.9520\n",
            "Epoch 55/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0981 - accuracy: 0.9730 - val_loss: 0.1673 - val_accuracy: 0.9527\n",
            "Epoch 56/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.1010 - accuracy: 0.9726 - val_loss: 0.1685 - val_accuracy: 0.9525\n",
            "Epoch 57/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0987 - accuracy: 0.9724 - val_loss: 0.1686 - val_accuracy: 0.9523\n",
            "Epoch 58/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0951 - accuracy: 0.9739 - val_loss: 0.1692 - val_accuracy: 0.9528\n",
            "Epoch 59/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0930 - accuracy: 0.9751 - val_loss: 0.1683 - val_accuracy: 0.9530\n",
            "Epoch 60/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0935 - accuracy: 0.9751 - val_loss: 0.1705 - val_accuracy: 0.9515\n",
            "Epoch 61/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0933 - accuracy: 0.9739 - val_loss: 0.1682 - val_accuracy: 0.9528\n",
            "Epoch 62/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0915 - accuracy: 0.9753 - val_loss: 0.1693 - val_accuracy: 0.9523\n",
            "Epoch 63/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0868 - accuracy: 0.9766 - val_loss: 0.1689 - val_accuracy: 0.9530\n",
            "Epoch 64/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0882 - accuracy: 0.9754 - val_loss: 0.1699 - val_accuracy: 0.9521\n",
            "Epoch 65/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0903 - accuracy: 0.9754 - val_loss: 0.1695 - val_accuracy: 0.9518\n",
            "Epoch 66/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0868 - accuracy: 0.9765 - val_loss: 0.1704 - val_accuracy: 0.9519\n",
            "Epoch 67/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0879 - accuracy: 0.9760 - val_loss: 0.1703 - val_accuracy: 0.9516\n",
            "Epoch 68/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0851 - accuracy: 0.9764 - val_loss: 0.1696 - val_accuracy: 0.9518\n",
            "Epoch 69/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0852 - accuracy: 0.9771 - val_loss: 0.1695 - val_accuracy: 0.9528\n",
            "Epoch 70/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0832 - accuracy: 0.9777 - val_loss: 0.1702 - val_accuracy: 0.9523\n",
            "Epoch 71/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0800 - accuracy: 0.9792 - val_loss: 0.1725 - val_accuracy: 0.9523\n",
            "Epoch 72/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0789 - accuracy: 0.9786 - val_loss: 0.1722 - val_accuracy: 0.9513\n",
            "Epoch 73/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0786 - accuracy: 0.9793 - val_loss: 0.1722 - val_accuracy: 0.9528\n",
            "Epoch 74/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0808 - accuracy: 0.9775 - val_loss: 0.1713 - val_accuracy: 0.9532\n",
            "Epoch 75/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0795 - accuracy: 0.9782 - val_loss: 0.1719 - val_accuracy: 0.9530\n",
            "Epoch 76/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0777 - accuracy: 0.9801 - val_loss: 0.1744 - val_accuracy: 0.9506\n",
            "Epoch 77/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0786 - accuracy: 0.9782 - val_loss: 0.1736 - val_accuracy: 0.9508\n",
            "Epoch 78/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0740 - accuracy: 0.9800 - val_loss: 0.1754 - val_accuracy: 0.9498\n",
            "Epoch 79/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0744 - accuracy: 0.9804 - val_loss: 0.1729 - val_accuracy: 0.9517\n",
            "Epoch 80/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0745 - accuracy: 0.9814 - val_loss: 0.1736 - val_accuracy: 0.9513\n",
            "Epoch 81/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0746 - accuracy: 0.9800 - val_loss: 0.1742 - val_accuracy: 0.9513\n",
            "Epoch 82/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0703 - accuracy: 0.9816 - val_loss: 0.1736 - val_accuracy: 0.9522\n",
            "Epoch 83/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0687 - accuracy: 0.9818 - val_loss: 0.1770 - val_accuracy: 0.9507\n",
            "Epoch 84/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0718 - accuracy: 0.9808 - val_loss: 0.1747 - val_accuracy: 0.9517\n",
            "Epoch 85/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0709 - accuracy: 0.9812 - val_loss: 0.1751 - val_accuracy: 0.9518\n",
            "Epoch 86/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0712 - accuracy: 0.9806 - val_loss: 0.1751 - val_accuracy: 0.9515\n",
            "Epoch 87/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0711 - accuracy: 0.9813 - val_loss: 0.1767 - val_accuracy: 0.9510\n",
            "Epoch 88/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0662 - accuracy: 0.9829 - val_loss: 0.1773 - val_accuracy: 0.9516\n",
            "Epoch 89/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0686 - accuracy: 0.9813 - val_loss: 0.1763 - val_accuracy: 0.9517\n",
            "Epoch 90/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0683 - accuracy: 0.9828 - val_loss: 0.1765 - val_accuracy: 0.9516\n",
            "Epoch 91/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0642 - accuracy: 0.9838 - val_loss: 0.1763 - val_accuracy: 0.9517\n",
            "Epoch 92/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0655 - accuracy: 0.9836 - val_loss: 0.1785 - val_accuracy: 0.9503\n",
            "Epoch 93/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0655 - accuracy: 0.9828 - val_loss: 0.1777 - val_accuracy: 0.9507\n",
            "Epoch 94/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0662 - accuracy: 0.9827 - val_loss: 0.1785 - val_accuracy: 0.9509\n",
            "Epoch 95/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0610 - accuracy: 0.9844 - val_loss: 0.1803 - val_accuracy: 0.9503\n",
            "Epoch 96/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0627 - accuracy: 0.9843 - val_loss: 0.1797 - val_accuracy: 0.9507\n",
            "Epoch 97/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0625 - accuracy: 0.9840 - val_loss: 0.1801 - val_accuracy: 0.9509\n",
            "Epoch 98/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0658 - accuracy: 0.9837 - val_loss: 0.1813 - val_accuracy: 0.9503\n",
            "Epoch 99/100\n",
            "96/96 [==============================] - 0s 5ms/step - loss: 0.0630 - accuracy: 0.9846 - val_loss: 0.1812 - val_accuracy: 0.9504\n",
            "Epoch 100/100\n",
            "96/96 [==============================] - 0s 4ms/step - loss: 0.0608 - accuracy: 0.9839 - val_loss: 0.1806 - val_accuracy: 0.9507\n",
            "Epoch 1/400\n",
            "24/24 [==============================] - 1s 21ms/step - loss: 2.2259 - accuracy: 0.2987 - val_loss: 1.9700 - val_accuracy: 0.6313\n",
            "Epoch 2/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 1.8941 - accuracy: 0.6452 - val_loss: 1.6400 - val_accuracy: 0.7149\n",
            "Epoch 3/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 1.5750 - accuracy: 0.7161 - val_loss: 1.3396 - val_accuracy: 0.7747\n",
            "Epoch 4/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 1.2966 - accuracy: 0.7688 - val_loss: 1.1007 - val_accuracy: 0.8203\n",
            "Epoch 5/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 1.0734 - accuracy: 0.8127 - val_loss: 0.9240 - val_accuracy: 0.8474\n",
            "Epoch 6/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.9093 - accuracy: 0.8406 - val_loss: 0.7949 - val_accuracy: 0.8654\n",
            "Epoch 7/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.7944 - accuracy: 0.8545 - val_loss: 0.6985 - val_accuracy: 0.8765\n",
            "Epoch 8/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.6968 - accuracy: 0.8705 - val_loss: 0.6235 - val_accuracy: 0.8857\n",
            "Epoch 9/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.6280 - accuracy: 0.8801 - val_loss: 0.5649 - val_accuracy: 0.8934\n",
            "Epoch 10/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.5743 - accuracy: 0.8862 - val_loss: 0.5181 - val_accuracy: 0.8977\n",
            "Epoch 11/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.5285 - accuracy: 0.8907 - val_loss: 0.4804 - val_accuracy: 0.9003\n",
            "Epoch 12/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.4906 - accuracy: 0.8974 - val_loss: 0.4494 - val_accuracy: 0.9048\n",
            "Epoch 13/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.4532 - accuracy: 0.9030 - val_loss: 0.4242 - val_accuracy: 0.9073\n",
            "Epoch 14/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.4319 - accuracy: 0.9045 - val_loss: 0.4028 - val_accuracy: 0.9110\n",
            "Epoch 15/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.4112 - accuracy: 0.9074 - val_loss: 0.3848 - val_accuracy: 0.9120\n",
            "Epoch 16/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.3895 - accuracy: 0.9116 - val_loss: 0.3698 - val_accuracy: 0.9133\n",
            "Epoch 17/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.3752 - accuracy: 0.9142 - val_loss: 0.3566 - val_accuracy: 0.9145\n",
            "Epoch 18/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.3664 - accuracy: 0.9138 - val_loss: 0.3451 - val_accuracy: 0.9153\n",
            "Epoch 19/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.3519 - accuracy: 0.9166 - val_loss: 0.3350 - val_accuracy: 0.9171\n",
            "Epoch 20/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.3423 - accuracy: 0.9161 - val_loss: 0.3255 - val_accuracy: 0.9189\n",
            "Epoch 21/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.3294 - accuracy: 0.9196 - val_loss: 0.3174 - val_accuracy: 0.9204\n",
            "Epoch 22/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.3151 - accuracy: 0.9227 - val_loss: 0.3097 - val_accuracy: 0.9207\n",
            "Epoch 23/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.3138 - accuracy: 0.9216 - val_loss: 0.3036 - val_accuracy: 0.9222\n",
            "Epoch 24/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.3021 - accuracy: 0.9253 - val_loss: 0.2978 - val_accuracy: 0.9230\n",
            "Epoch 25/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2983 - accuracy: 0.9253 - val_loss: 0.2918 - val_accuracy: 0.9234\n",
            "Epoch 26/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2909 - accuracy: 0.9270 - val_loss: 0.2870 - val_accuracy: 0.9247\n",
            "Epoch 27/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2800 - accuracy: 0.9293 - val_loss: 0.2822 - val_accuracy: 0.9254\n",
            "Epoch 28/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.2801 - accuracy: 0.9287 - val_loss: 0.2784 - val_accuracy: 0.9258\n",
            "Epoch 29/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2743 - accuracy: 0.9299 - val_loss: 0.2738 - val_accuracy: 0.9261\n",
            "Epoch 30/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2712 - accuracy: 0.9303 - val_loss: 0.2701 - val_accuracy: 0.9273\n",
            "Epoch 31/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.2629 - accuracy: 0.9309 - val_loss: 0.2666 - val_accuracy: 0.9270\n",
            "Epoch 32/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2588 - accuracy: 0.9327 - val_loss: 0.2629 - val_accuracy: 0.9277\n",
            "Epoch 33/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.2581 - accuracy: 0.9329 - val_loss: 0.2602 - val_accuracy: 0.9285\n",
            "Epoch 34/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2484 - accuracy: 0.9352 - val_loss: 0.2572 - val_accuracy: 0.9287\n",
            "Epoch 35/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.2493 - accuracy: 0.9339 - val_loss: 0.2545 - val_accuracy: 0.9295\n",
            "Epoch 36/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2402 - accuracy: 0.9370 - val_loss: 0.2516 - val_accuracy: 0.9310\n",
            "Epoch 37/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2432 - accuracy: 0.9356 - val_loss: 0.2489 - val_accuracy: 0.9310\n",
            "Epoch 38/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2403 - accuracy: 0.9353 - val_loss: 0.2462 - val_accuracy: 0.9324\n",
            "Epoch 39/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2369 - accuracy: 0.9367 - val_loss: 0.2440 - val_accuracy: 0.9332\n",
            "Epoch 40/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2336 - accuracy: 0.9385 - val_loss: 0.2423 - val_accuracy: 0.9337\n",
            "Epoch 41/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2285 - accuracy: 0.9393 - val_loss: 0.2399 - val_accuracy: 0.9344\n",
            "Epoch 42/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2231 - accuracy: 0.9417 - val_loss: 0.2381 - val_accuracy: 0.9346\n",
            "Epoch 43/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2215 - accuracy: 0.9399 - val_loss: 0.2358 - val_accuracy: 0.9357\n",
            "Epoch 44/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2244 - accuracy: 0.9402 - val_loss: 0.2342 - val_accuracy: 0.9356\n",
            "Epoch 45/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.2194 - accuracy: 0.9406 - val_loss: 0.2320 - val_accuracy: 0.9359\n",
            "Epoch 46/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2144 - accuracy: 0.9426 - val_loss: 0.2303 - val_accuracy: 0.9365\n",
            "Epoch 47/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2150 - accuracy: 0.9417 - val_loss: 0.2288 - val_accuracy: 0.9363\n",
            "Epoch 48/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2166 - accuracy: 0.9413 - val_loss: 0.2276 - val_accuracy: 0.9374\n",
            "Epoch 49/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2109 - accuracy: 0.9438 - val_loss: 0.2257 - val_accuracy: 0.9370\n",
            "Epoch 50/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2042 - accuracy: 0.9455 - val_loss: 0.2242 - val_accuracy: 0.9371\n",
            "Epoch 51/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2056 - accuracy: 0.9460 - val_loss: 0.2226 - val_accuracy: 0.9382\n",
            "Epoch 52/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.1973 - accuracy: 0.9471 - val_loss: 0.2213 - val_accuracy: 0.9384\n",
            "Epoch 53/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.2023 - accuracy: 0.9453 - val_loss: 0.2202 - val_accuracy: 0.9387\n",
            "Epoch 54/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1961 - accuracy: 0.9473 - val_loss: 0.2198 - val_accuracy: 0.9382\n",
            "Epoch 55/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1959 - accuracy: 0.9486 - val_loss: 0.2178 - val_accuracy: 0.9391\n",
            "Epoch 56/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1897 - accuracy: 0.9486 - val_loss: 0.2168 - val_accuracy: 0.9392\n",
            "Epoch 57/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1904 - accuracy: 0.9484 - val_loss: 0.2157 - val_accuracy: 0.9392\n",
            "Epoch 58/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1952 - accuracy: 0.9475 - val_loss: 0.2146 - val_accuracy: 0.9402\n",
            "Epoch 59/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1872 - accuracy: 0.9493 - val_loss: 0.2133 - val_accuracy: 0.9392\n",
            "Epoch 60/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1850 - accuracy: 0.9509 - val_loss: 0.2126 - val_accuracy: 0.9404\n",
            "Epoch 61/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1835 - accuracy: 0.9505 - val_loss: 0.2111 - val_accuracy: 0.9401\n",
            "Epoch 62/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1804 - accuracy: 0.9519 - val_loss: 0.2106 - val_accuracy: 0.9402\n",
            "Epoch 63/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1813 - accuracy: 0.9513 - val_loss: 0.2100 - val_accuracy: 0.9406\n",
            "Epoch 64/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1789 - accuracy: 0.9509 - val_loss: 0.2088 - val_accuracy: 0.9404\n",
            "Epoch 65/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1805 - accuracy: 0.9502 - val_loss: 0.2077 - val_accuracy: 0.9411\n",
            "Epoch 66/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1739 - accuracy: 0.9535 - val_loss: 0.2071 - val_accuracy: 0.9407\n",
            "Epoch 67/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1763 - accuracy: 0.9527 - val_loss: 0.2056 - val_accuracy: 0.9413\n",
            "Epoch 68/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1695 - accuracy: 0.9550 - val_loss: 0.2054 - val_accuracy: 0.9413\n",
            "Epoch 69/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1733 - accuracy: 0.9528 - val_loss: 0.2044 - val_accuracy: 0.9415\n",
            "Epoch 70/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1744 - accuracy: 0.9517 - val_loss: 0.2033 - val_accuracy: 0.9414\n",
            "Epoch 71/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1753 - accuracy: 0.9521 - val_loss: 0.2028 - val_accuracy: 0.9416\n",
            "Epoch 72/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1688 - accuracy: 0.9550 - val_loss: 0.2028 - val_accuracy: 0.9416\n",
            "Epoch 73/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1673 - accuracy: 0.9551 - val_loss: 0.2022 - val_accuracy: 0.9420\n",
            "Epoch 74/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1678 - accuracy: 0.9542 - val_loss: 0.2009 - val_accuracy: 0.9426\n",
            "Epoch 75/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1659 - accuracy: 0.9542 - val_loss: 0.2001 - val_accuracy: 0.9423\n",
            "Epoch 76/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1625 - accuracy: 0.9562 - val_loss: 0.1999 - val_accuracy: 0.9425\n",
            "Epoch 77/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1663 - accuracy: 0.9547 - val_loss: 0.1989 - val_accuracy: 0.9433\n",
            "Epoch 78/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1605 - accuracy: 0.9566 - val_loss: 0.1993 - val_accuracy: 0.9427\n",
            "Epoch 79/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1595 - accuracy: 0.9564 - val_loss: 0.1978 - val_accuracy: 0.9431\n",
            "Epoch 80/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1585 - accuracy: 0.9567 - val_loss: 0.1975 - val_accuracy: 0.9446\n",
            "Epoch 81/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1586 - accuracy: 0.9562 - val_loss: 0.1966 - val_accuracy: 0.9442\n",
            "Epoch 82/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1592 - accuracy: 0.9552 - val_loss: 0.1963 - val_accuracy: 0.9438\n",
            "Epoch 83/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1557 - accuracy: 0.9566 - val_loss: 0.1953 - val_accuracy: 0.9444\n",
            "Epoch 84/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1544 - accuracy: 0.9585 - val_loss: 0.1952 - val_accuracy: 0.9441\n",
            "Epoch 85/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1523 - accuracy: 0.9588 - val_loss: 0.1944 - val_accuracy: 0.9446\n",
            "Epoch 86/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1535 - accuracy: 0.9568 - val_loss: 0.1941 - val_accuracy: 0.9446\n",
            "Epoch 87/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1521 - accuracy: 0.9585 - val_loss: 0.1936 - val_accuracy: 0.9450\n",
            "Epoch 88/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1494 - accuracy: 0.9588 - val_loss: 0.1940 - val_accuracy: 0.9448\n",
            "Epoch 89/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1442 - accuracy: 0.9607 - val_loss: 0.1926 - val_accuracy: 0.9445\n",
            "Epoch 90/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1502 - accuracy: 0.9589 - val_loss: 0.1923 - val_accuracy: 0.9451\n",
            "Epoch 91/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1453 - accuracy: 0.9594 - val_loss: 0.1914 - val_accuracy: 0.9447\n",
            "Epoch 92/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1464 - accuracy: 0.9605 - val_loss: 0.1913 - val_accuracy: 0.9450\n",
            "Epoch 93/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1464 - accuracy: 0.9597 - val_loss: 0.1913 - val_accuracy: 0.9448\n",
            "Epoch 94/400\n",
            "24/24 [==============================] - 0s 11ms/step - loss: 0.1466 - accuracy: 0.9603 - val_loss: 0.1906 - val_accuracy: 0.9451\n",
            "Epoch 95/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1432 - accuracy: 0.9617 - val_loss: 0.1906 - val_accuracy: 0.9448\n",
            "Epoch 96/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1428 - accuracy: 0.9606 - val_loss: 0.1903 - val_accuracy: 0.9460\n",
            "Epoch 97/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1418 - accuracy: 0.9612 - val_loss: 0.1889 - val_accuracy: 0.9455\n",
            "Epoch 98/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1397 - accuracy: 0.9615 - val_loss: 0.1887 - val_accuracy: 0.9458\n",
            "Epoch 99/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1354 - accuracy: 0.9635 - val_loss: 0.1880 - val_accuracy: 0.9460\n",
            "Epoch 100/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1392 - accuracy: 0.9614 - val_loss: 0.1881 - val_accuracy: 0.9460\n",
            "Epoch 101/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1352 - accuracy: 0.9641 - val_loss: 0.1879 - val_accuracy: 0.9460\n",
            "Epoch 102/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1347 - accuracy: 0.9638 - val_loss: 0.1869 - val_accuracy: 0.9459\n",
            "Epoch 103/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1375 - accuracy: 0.9621 - val_loss: 0.1868 - val_accuracy: 0.9462\n",
            "Epoch 104/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1320 - accuracy: 0.9651 - val_loss: 0.1867 - val_accuracy: 0.9462\n",
            "Epoch 105/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1363 - accuracy: 0.9636 - val_loss: 0.1870 - val_accuracy: 0.9461\n",
            "Epoch 106/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1325 - accuracy: 0.9633 - val_loss: 0.1866 - val_accuracy: 0.9457\n",
            "Epoch 107/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1320 - accuracy: 0.9648 - val_loss: 0.1862 - val_accuracy: 0.9460\n",
            "Epoch 108/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1338 - accuracy: 0.9640 - val_loss: 0.1854 - val_accuracy: 0.9459\n",
            "Epoch 109/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1288 - accuracy: 0.9647 - val_loss: 0.1852 - val_accuracy: 0.9459\n",
            "Epoch 110/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1313 - accuracy: 0.9647 - val_loss: 0.1855 - val_accuracy: 0.9462\n",
            "Epoch 111/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1294 - accuracy: 0.9658 - val_loss: 0.1845 - val_accuracy: 0.9467\n",
            "Epoch 112/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1303 - accuracy: 0.9652 - val_loss: 0.1850 - val_accuracy: 0.9467\n",
            "Epoch 113/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1275 - accuracy: 0.9661 - val_loss: 0.1843 - val_accuracy: 0.9466\n",
            "Epoch 114/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1242 - accuracy: 0.9664 - val_loss: 0.1840 - val_accuracy: 0.9465\n",
            "Epoch 115/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1266 - accuracy: 0.9663 - val_loss: 0.1839 - val_accuracy: 0.9462\n",
            "Epoch 116/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1268 - accuracy: 0.9660 - val_loss: 0.1833 - val_accuracy: 0.9466\n",
            "Epoch 117/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1275 - accuracy: 0.9656 - val_loss: 0.1833 - val_accuracy: 0.9467\n",
            "Epoch 118/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1235 - accuracy: 0.9669 - val_loss: 0.1830 - val_accuracy: 0.9468\n",
            "Epoch 119/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1220 - accuracy: 0.9678 - val_loss: 0.1829 - val_accuracy: 0.9463\n",
            "Epoch 120/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1244 - accuracy: 0.9672 - val_loss: 0.1826 - val_accuracy: 0.9463\n",
            "Epoch 121/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1231 - accuracy: 0.9674 - val_loss: 0.1825 - val_accuracy: 0.9468\n",
            "Epoch 122/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1197 - accuracy: 0.9675 - val_loss: 0.1821 - val_accuracy: 0.9471\n",
            "Epoch 123/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1201 - accuracy: 0.9667 - val_loss: 0.1824 - val_accuracy: 0.9465\n",
            "Epoch 124/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1234 - accuracy: 0.9665 - val_loss: 0.1817 - val_accuracy: 0.9469\n",
            "Epoch 125/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1210 - accuracy: 0.9685 - val_loss: 0.1820 - val_accuracy: 0.9464\n",
            "Epoch 126/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1198 - accuracy: 0.9676 - val_loss: 0.1820 - val_accuracy: 0.9472\n",
            "Epoch 127/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1179 - accuracy: 0.9678 - val_loss: 0.1810 - val_accuracy: 0.9469\n",
            "Epoch 128/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1166 - accuracy: 0.9683 - val_loss: 0.1816 - val_accuracy: 0.9473\n",
            "Epoch 129/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1170 - accuracy: 0.9682 - val_loss: 0.1809 - val_accuracy: 0.9473\n",
            "Epoch 130/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1166 - accuracy: 0.9694 - val_loss: 0.1812 - val_accuracy: 0.9472\n",
            "Epoch 131/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1141 - accuracy: 0.9694 - val_loss: 0.1817 - val_accuracy: 0.9468\n",
            "Epoch 132/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1154 - accuracy: 0.9682 - val_loss: 0.1807 - val_accuracy: 0.9473\n",
            "Epoch 133/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1168 - accuracy: 0.9684 - val_loss: 0.1808 - val_accuracy: 0.9485\n",
            "Epoch 134/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1137 - accuracy: 0.9694 - val_loss: 0.1807 - val_accuracy: 0.9483\n",
            "Epoch 135/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1134 - accuracy: 0.9699 - val_loss: 0.1804 - val_accuracy: 0.9485\n",
            "Epoch 136/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1153 - accuracy: 0.9690 - val_loss: 0.1805 - val_accuracy: 0.9481\n",
            "Epoch 137/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1139 - accuracy: 0.9701 - val_loss: 0.1801 - val_accuracy: 0.9482\n",
            "Epoch 138/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1112 - accuracy: 0.9699 - val_loss: 0.1801 - val_accuracy: 0.9481\n",
            "Epoch 139/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1110 - accuracy: 0.9701 - val_loss: 0.1804 - val_accuracy: 0.9483\n",
            "Epoch 140/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1132 - accuracy: 0.9704 - val_loss: 0.1803 - val_accuracy: 0.9476\n",
            "Epoch 141/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1110 - accuracy: 0.9704 - val_loss: 0.1799 - val_accuracy: 0.9482\n",
            "Epoch 142/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1089 - accuracy: 0.9717 - val_loss: 0.1797 - val_accuracy: 0.9477\n",
            "Epoch 143/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1095 - accuracy: 0.9706 - val_loss: 0.1803 - val_accuracy: 0.9479\n",
            "Epoch 144/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1105 - accuracy: 0.9698 - val_loss: 0.1803 - val_accuracy: 0.9477\n",
            "Epoch 145/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1088 - accuracy: 0.9708 - val_loss: 0.1797 - val_accuracy: 0.9475\n",
            "Epoch 146/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1066 - accuracy: 0.9722 - val_loss: 0.1798 - val_accuracy: 0.9481\n",
            "Epoch 147/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1041 - accuracy: 0.9726 - val_loss: 0.1794 - val_accuracy: 0.9483\n",
            "Epoch 148/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1090 - accuracy: 0.9712 - val_loss: 0.1798 - val_accuracy: 0.9486\n",
            "Epoch 149/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1088 - accuracy: 0.9722 - val_loss: 0.1799 - val_accuracy: 0.9481\n",
            "Epoch 150/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1062 - accuracy: 0.9716 - val_loss: 0.1792 - val_accuracy: 0.9477\n",
            "Epoch 151/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1043 - accuracy: 0.9720 - val_loss: 0.1797 - val_accuracy: 0.9482\n",
            "Epoch 152/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1083 - accuracy: 0.9715 - val_loss: 0.1794 - val_accuracy: 0.9485\n",
            "Epoch 153/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1022 - accuracy: 0.9731 - val_loss: 0.1795 - val_accuracy: 0.9482\n",
            "Epoch 154/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1058 - accuracy: 0.9722 - val_loss: 0.1795 - val_accuracy: 0.9481\n",
            "Epoch 155/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1057 - accuracy: 0.9718 - val_loss: 0.1794 - val_accuracy: 0.9481\n",
            "Epoch 156/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1012 - accuracy: 0.9739 - val_loss: 0.1794 - val_accuracy: 0.9490\n",
            "Epoch 157/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1012 - accuracy: 0.9739 - val_loss: 0.1800 - val_accuracy: 0.9474\n",
            "Epoch 158/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1018 - accuracy: 0.9733 - val_loss: 0.1792 - val_accuracy: 0.9490\n",
            "Epoch 159/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0995 - accuracy: 0.9739 - val_loss: 0.1794 - val_accuracy: 0.9487\n",
            "Epoch 160/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.1007 - accuracy: 0.9732 - val_loss: 0.1787 - val_accuracy: 0.9476\n",
            "Epoch 161/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.1025 - accuracy: 0.9728 - val_loss: 0.1800 - val_accuracy: 0.9491\n",
            "Epoch 162/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0998 - accuracy: 0.9730 - val_loss: 0.1795 - val_accuracy: 0.9491\n",
            "Epoch 163/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0998 - accuracy: 0.9747 - val_loss: 0.1790 - val_accuracy: 0.9488\n",
            "Epoch 164/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0997 - accuracy: 0.9740 - val_loss: 0.1801 - val_accuracy: 0.9479\n",
            "Epoch 165/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0983 - accuracy: 0.9740 - val_loss: 0.1792 - val_accuracy: 0.9488\n",
            "Epoch 166/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0943 - accuracy: 0.9760 - val_loss: 0.1794 - val_accuracy: 0.9488\n",
            "Epoch 167/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0980 - accuracy: 0.9743 - val_loss: 0.1792 - val_accuracy: 0.9491\n",
            "Epoch 168/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0961 - accuracy: 0.9745 - val_loss: 0.1798 - val_accuracy: 0.9489\n",
            "Epoch 169/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0950 - accuracy: 0.9750 - val_loss: 0.1797 - val_accuracy: 0.9480\n",
            "Epoch 170/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0956 - accuracy: 0.9756 - val_loss: 0.1795 - val_accuracy: 0.9491\n",
            "Epoch 171/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0986 - accuracy: 0.9742 - val_loss: 0.1801 - val_accuracy: 0.9485\n",
            "Epoch 172/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0927 - accuracy: 0.9765 - val_loss: 0.1794 - val_accuracy: 0.9483\n",
            "Epoch 173/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0954 - accuracy: 0.9754 - val_loss: 0.1799 - val_accuracy: 0.9490\n",
            "Epoch 174/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0925 - accuracy: 0.9757 - val_loss: 0.1802 - val_accuracy: 0.9486\n",
            "Epoch 175/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0938 - accuracy: 0.9759 - val_loss: 0.1801 - val_accuracy: 0.9485\n",
            "Epoch 176/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0939 - accuracy: 0.9754 - val_loss: 0.1801 - val_accuracy: 0.9492\n",
            "Epoch 177/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0906 - accuracy: 0.9765 - val_loss: 0.1799 - val_accuracy: 0.9486\n",
            "Epoch 178/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0939 - accuracy: 0.9749 - val_loss: 0.1809 - val_accuracy: 0.9489\n",
            "Epoch 179/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0903 - accuracy: 0.9768 - val_loss: 0.1801 - val_accuracy: 0.9476\n",
            "Epoch 180/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0894 - accuracy: 0.9768 - val_loss: 0.1806 - val_accuracy: 0.9494\n",
            "Epoch 181/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0904 - accuracy: 0.9765 - val_loss: 0.1799 - val_accuracy: 0.9492\n",
            "Epoch 182/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0927 - accuracy: 0.9761 - val_loss: 0.1803 - val_accuracy: 0.9488\n",
            "Epoch 183/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0912 - accuracy: 0.9767 - val_loss: 0.1799 - val_accuracy: 0.9493\n",
            "Epoch 184/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0896 - accuracy: 0.9768 - val_loss: 0.1807 - val_accuracy: 0.9488\n",
            "Epoch 185/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0880 - accuracy: 0.9775 - val_loss: 0.1806 - val_accuracy: 0.9489\n",
            "Epoch 186/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0865 - accuracy: 0.9780 - val_loss: 0.1804 - val_accuracy: 0.9492\n",
            "Epoch 187/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0864 - accuracy: 0.9778 - val_loss: 0.1808 - val_accuracy: 0.9484\n",
            "Epoch 188/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0897 - accuracy: 0.9772 - val_loss: 0.1806 - val_accuracy: 0.9484\n",
            "Epoch 189/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0872 - accuracy: 0.9775 - val_loss: 0.1802 - val_accuracy: 0.9492\n",
            "Epoch 190/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0850 - accuracy: 0.9783 - val_loss: 0.1806 - val_accuracy: 0.9493\n",
            "Epoch 191/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0877 - accuracy: 0.9769 - val_loss: 0.1806 - val_accuracy: 0.9489\n",
            "Epoch 192/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0861 - accuracy: 0.9775 - val_loss: 0.1817 - val_accuracy: 0.9491\n",
            "Epoch 193/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0871 - accuracy: 0.9776 - val_loss: 0.1812 - val_accuracy: 0.9496\n",
            "Epoch 194/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0860 - accuracy: 0.9782 - val_loss: 0.1808 - val_accuracy: 0.9487\n",
            "Epoch 195/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0849 - accuracy: 0.9784 - val_loss: 0.1810 - val_accuracy: 0.9489\n",
            "Epoch 196/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0855 - accuracy: 0.9782 - val_loss: 0.1813 - val_accuracy: 0.9494\n",
            "Epoch 197/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0825 - accuracy: 0.9785 - val_loss: 0.1811 - val_accuracy: 0.9491\n",
            "Epoch 198/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0823 - accuracy: 0.9785 - val_loss: 0.1819 - val_accuracy: 0.9493\n",
            "Epoch 199/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0854 - accuracy: 0.9789 - val_loss: 0.1813 - val_accuracy: 0.9485\n",
            "Epoch 200/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0820 - accuracy: 0.9790 - val_loss: 0.1814 - val_accuracy: 0.9486\n",
            "Epoch 201/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0832 - accuracy: 0.9790 - val_loss: 0.1813 - val_accuracy: 0.9489\n",
            "Epoch 202/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0848 - accuracy: 0.9781 - val_loss: 0.1819 - val_accuracy: 0.9487\n",
            "Epoch 203/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0826 - accuracy: 0.9785 - val_loss: 0.1823 - val_accuracy: 0.9491\n",
            "Epoch 204/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0807 - accuracy: 0.9790 - val_loss: 0.1820 - val_accuracy: 0.9486\n",
            "Epoch 205/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0808 - accuracy: 0.9786 - val_loss: 0.1819 - val_accuracy: 0.9490\n",
            "Epoch 206/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0826 - accuracy: 0.9792 - val_loss: 0.1823 - val_accuracy: 0.9488\n",
            "Epoch 207/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0842 - accuracy: 0.9785 - val_loss: 0.1821 - val_accuracy: 0.9486\n",
            "Epoch 208/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0800 - accuracy: 0.9799 - val_loss: 0.1825 - val_accuracy: 0.9488\n",
            "Epoch 209/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0790 - accuracy: 0.9796 - val_loss: 0.1821 - val_accuracy: 0.9486\n",
            "Epoch 210/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0795 - accuracy: 0.9800 - val_loss: 0.1827 - val_accuracy: 0.9487\n",
            "Epoch 211/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0792 - accuracy: 0.9799 - val_loss: 0.1824 - val_accuracy: 0.9491\n",
            "Epoch 212/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0784 - accuracy: 0.9808 - val_loss: 0.1827 - val_accuracy: 0.9497\n",
            "Epoch 213/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0772 - accuracy: 0.9803 - val_loss: 0.1835 - val_accuracy: 0.9486\n",
            "Epoch 214/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0772 - accuracy: 0.9798 - val_loss: 0.1832 - val_accuracy: 0.9483\n",
            "Epoch 215/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0777 - accuracy: 0.9796 - val_loss: 0.1841 - val_accuracy: 0.9487\n",
            "Epoch 216/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0774 - accuracy: 0.9808 - val_loss: 0.1839 - val_accuracy: 0.9494\n",
            "Epoch 217/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0739 - accuracy: 0.9817 - val_loss: 0.1836 - val_accuracy: 0.9489\n",
            "Epoch 218/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0774 - accuracy: 0.9808 - val_loss: 0.1841 - val_accuracy: 0.9495\n",
            "Epoch 219/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0759 - accuracy: 0.9811 - val_loss: 0.1837 - val_accuracy: 0.9489\n",
            "Epoch 220/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0770 - accuracy: 0.9807 - val_loss: 0.1844 - val_accuracy: 0.9500\n",
            "Epoch 221/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0745 - accuracy: 0.9812 - val_loss: 0.1841 - val_accuracy: 0.9490\n",
            "Epoch 222/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0745 - accuracy: 0.9816 - val_loss: 0.1845 - val_accuracy: 0.9492\n",
            "Epoch 223/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0733 - accuracy: 0.9809 - val_loss: 0.1839 - val_accuracy: 0.9489\n",
            "Epoch 224/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0776 - accuracy: 0.9813 - val_loss: 0.1846 - val_accuracy: 0.9494\n",
            "Epoch 225/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0744 - accuracy: 0.9815 - val_loss: 0.1842 - val_accuracy: 0.9491\n",
            "Epoch 226/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0760 - accuracy: 0.9808 - val_loss: 0.1848 - val_accuracy: 0.9492\n",
            "Epoch 227/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0697 - accuracy: 0.9829 - val_loss: 0.1857 - val_accuracy: 0.9489\n",
            "Epoch 228/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0705 - accuracy: 0.9823 - val_loss: 0.1848 - val_accuracy: 0.9492\n",
            "Epoch 229/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0759 - accuracy: 0.9803 - val_loss: 0.1845 - val_accuracy: 0.9504\n",
            "Epoch 230/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0741 - accuracy: 0.9811 - val_loss: 0.1856 - val_accuracy: 0.9502\n",
            "Epoch 231/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0722 - accuracy: 0.9816 - val_loss: 0.1853 - val_accuracy: 0.9497\n",
            "Epoch 232/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0706 - accuracy: 0.9827 - val_loss: 0.1857 - val_accuracy: 0.9494\n",
            "Epoch 233/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0719 - accuracy: 0.9810 - val_loss: 0.1845 - val_accuracy: 0.9500\n",
            "Epoch 234/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0712 - accuracy: 0.9820 - val_loss: 0.1860 - val_accuracy: 0.9492\n",
            "Epoch 235/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0726 - accuracy: 0.9819 - val_loss: 0.1859 - val_accuracy: 0.9490\n",
            "Epoch 236/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0727 - accuracy: 0.9812 - val_loss: 0.1861 - val_accuracy: 0.9498\n",
            "Epoch 237/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0715 - accuracy: 0.9818 - val_loss: 0.1858 - val_accuracy: 0.9507\n",
            "Epoch 238/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0683 - accuracy: 0.9829 - val_loss: 0.1869 - val_accuracy: 0.9498\n",
            "Epoch 239/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0723 - accuracy: 0.9811 - val_loss: 0.1869 - val_accuracy: 0.9488\n",
            "Epoch 240/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0698 - accuracy: 0.9829 - val_loss: 0.1862 - val_accuracy: 0.9491\n",
            "Epoch 241/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0689 - accuracy: 0.9835 - val_loss: 0.1871 - val_accuracy: 0.9489\n",
            "Epoch 242/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0691 - accuracy: 0.9828 - val_loss: 0.1871 - val_accuracy: 0.9497\n",
            "Epoch 243/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0690 - accuracy: 0.9823 - val_loss: 0.1879 - val_accuracy: 0.9492\n",
            "Epoch 244/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0679 - accuracy: 0.9829 - val_loss: 0.1880 - val_accuracy: 0.9494\n",
            "Epoch 245/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0714 - accuracy: 0.9822 - val_loss: 0.1867 - val_accuracy: 0.9496\n",
            "Epoch 246/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0678 - accuracy: 0.9830 - val_loss: 0.1871 - val_accuracy: 0.9488\n",
            "Epoch 247/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0690 - accuracy: 0.9828 - val_loss: 0.1879 - val_accuracy: 0.9490\n",
            "Epoch 248/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0661 - accuracy: 0.9836 - val_loss: 0.1876 - val_accuracy: 0.9503\n",
            "Epoch 249/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0667 - accuracy: 0.9834 - val_loss: 0.1890 - val_accuracy: 0.9483\n",
            "Epoch 250/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0667 - accuracy: 0.9834 - val_loss: 0.1877 - val_accuracy: 0.9496\n",
            "Epoch 251/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0666 - accuracy: 0.9834 - val_loss: 0.1878 - val_accuracy: 0.9500\n",
            "Epoch 252/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0657 - accuracy: 0.9832 - val_loss: 0.1879 - val_accuracy: 0.9498\n",
            "Epoch 253/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0667 - accuracy: 0.9833 - val_loss: 0.1887 - val_accuracy: 0.9491\n",
            "Epoch 254/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0652 - accuracy: 0.9846 - val_loss: 0.1893 - val_accuracy: 0.9506\n",
            "Epoch 255/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0648 - accuracy: 0.9851 - val_loss: 0.1886 - val_accuracy: 0.9491\n",
            "Epoch 256/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0666 - accuracy: 0.9838 - val_loss: 0.1900 - val_accuracy: 0.9494\n",
            "Epoch 257/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0635 - accuracy: 0.9843 - val_loss: 0.1891 - val_accuracy: 0.9496\n",
            "Epoch 258/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0640 - accuracy: 0.9841 - val_loss: 0.1908 - val_accuracy: 0.9487\n",
            "Epoch 259/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0651 - accuracy: 0.9841 - val_loss: 0.1906 - val_accuracy: 0.9481\n",
            "Epoch 260/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0662 - accuracy: 0.9833 - val_loss: 0.1900 - val_accuracy: 0.9503\n",
            "Epoch 261/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0652 - accuracy: 0.9840 - val_loss: 0.1908 - val_accuracy: 0.9490\n",
            "Epoch 262/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0617 - accuracy: 0.9848 - val_loss: 0.1904 - val_accuracy: 0.9509\n",
            "Epoch 263/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0642 - accuracy: 0.9841 - val_loss: 0.1912 - val_accuracy: 0.9497\n",
            "Epoch 264/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0651 - accuracy: 0.9839 - val_loss: 0.1909 - val_accuracy: 0.9498\n",
            "Epoch 265/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0648 - accuracy: 0.9836 - val_loss: 0.1914 - val_accuracy: 0.9505\n",
            "Epoch 266/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0605 - accuracy: 0.9854 - val_loss: 0.1913 - val_accuracy: 0.9498\n",
            "Epoch 267/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0623 - accuracy: 0.9848 - val_loss: 0.1911 - val_accuracy: 0.9495\n",
            "Epoch 268/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0613 - accuracy: 0.9848 - val_loss: 0.1917 - val_accuracy: 0.9498\n",
            "Epoch 269/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0598 - accuracy: 0.9865 - val_loss: 0.1923 - val_accuracy: 0.9493\n",
            "Epoch 270/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0625 - accuracy: 0.9844 - val_loss: 0.1930 - val_accuracy: 0.9488\n",
            "Epoch 271/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0627 - accuracy: 0.9845 - val_loss: 0.1920 - val_accuracy: 0.9492\n",
            "Epoch 272/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0614 - accuracy: 0.9854 - val_loss: 0.1925 - val_accuracy: 0.9492\n",
            "Epoch 273/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0607 - accuracy: 0.9854 - val_loss: 0.1926 - val_accuracy: 0.9500\n",
            "Epoch 274/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0636 - accuracy: 0.9846 - val_loss: 0.1929 - val_accuracy: 0.9496\n",
            "Epoch 275/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0608 - accuracy: 0.9854 - val_loss: 0.1933 - val_accuracy: 0.9495\n",
            "Epoch 276/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0580 - accuracy: 0.9864 - val_loss: 0.1935 - val_accuracy: 0.9485\n",
            "Epoch 277/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0587 - accuracy: 0.9860 - val_loss: 0.1935 - val_accuracy: 0.9498\n",
            "Epoch 278/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0617 - accuracy: 0.9852 - val_loss: 0.1956 - val_accuracy: 0.9491\n",
            "Epoch 279/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0591 - accuracy: 0.9851 - val_loss: 0.1940 - val_accuracy: 0.9493\n",
            "Epoch 280/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0602 - accuracy: 0.9858 - val_loss: 0.1944 - val_accuracy: 0.9490\n",
            "Epoch 281/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0573 - accuracy: 0.9854 - val_loss: 0.1942 - val_accuracy: 0.9503\n",
            "Epoch 282/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0569 - accuracy: 0.9861 - val_loss: 0.1948 - val_accuracy: 0.9499\n",
            "Epoch 283/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0586 - accuracy: 0.9858 - val_loss: 0.1949 - val_accuracy: 0.9498\n",
            "Epoch 284/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0588 - accuracy: 0.9857 - val_loss: 0.1949 - val_accuracy: 0.9493\n",
            "Epoch 285/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0562 - accuracy: 0.9865 - val_loss: 0.1947 - val_accuracy: 0.9499\n",
            "Epoch 286/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0552 - accuracy: 0.9869 - val_loss: 0.1954 - val_accuracy: 0.9492\n",
            "Epoch 287/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0563 - accuracy: 0.9871 - val_loss: 0.1952 - val_accuracy: 0.9492\n",
            "Epoch 288/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0581 - accuracy: 0.9867 - val_loss: 0.1960 - val_accuracy: 0.9494\n",
            "Epoch 289/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0587 - accuracy: 0.9859 - val_loss: 0.1959 - val_accuracy: 0.9492\n",
            "Epoch 290/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0588 - accuracy: 0.9860 - val_loss: 0.1964 - val_accuracy: 0.9496\n",
            "Epoch 291/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0548 - accuracy: 0.9873 - val_loss: 0.1963 - val_accuracy: 0.9495\n",
            "Epoch 292/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0550 - accuracy: 0.9876 - val_loss: 0.1967 - val_accuracy: 0.9491\n",
            "Epoch 293/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0562 - accuracy: 0.9867 - val_loss: 0.1972 - val_accuracy: 0.9493\n",
            "Epoch 294/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0581 - accuracy: 0.9859 - val_loss: 0.1967 - val_accuracy: 0.9499\n",
            "Epoch 295/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0562 - accuracy: 0.9863 - val_loss: 0.1961 - val_accuracy: 0.9489\n",
            "Epoch 296/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0563 - accuracy: 0.9864 - val_loss: 0.1981 - val_accuracy: 0.9490\n",
            "Epoch 297/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0558 - accuracy: 0.9865 - val_loss: 0.1973 - val_accuracy: 0.9498\n",
            "Epoch 298/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0569 - accuracy: 0.9872 - val_loss: 0.1979 - val_accuracy: 0.9496\n",
            "Epoch 299/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0557 - accuracy: 0.9876 - val_loss: 0.1978 - val_accuracy: 0.9490\n",
            "Epoch 300/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0526 - accuracy: 0.9876 - val_loss: 0.1982 - val_accuracy: 0.9491\n",
            "Epoch 301/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0553 - accuracy: 0.9875 - val_loss: 0.1992 - val_accuracy: 0.9488\n",
            "Epoch 302/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0527 - accuracy: 0.9876 - val_loss: 0.1990 - val_accuracy: 0.9486\n",
            "Epoch 303/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0529 - accuracy: 0.9879 - val_loss: 0.1990 - val_accuracy: 0.9488\n",
            "Epoch 304/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0551 - accuracy: 0.9872 - val_loss: 0.1992 - val_accuracy: 0.9496\n",
            "Epoch 305/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0533 - accuracy: 0.9876 - val_loss: 0.1995 - val_accuracy: 0.9485\n",
            "Epoch 306/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0550 - accuracy: 0.9873 - val_loss: 0.1999 - val_accuracy: 0.9485\n",
            "Epoch 307/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0527 - accuracy: 0.9878 - val_loss: 0.2007 - val_accuracy: 0.9490\n",
            "Epoch 308/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0521 - accuracy: 0.9879 - val_loss: 0.2003 - val_accuracy: 0.9489\n",
            "Epoch 309/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0525 - accuracy: 0.9878 - val_loss: 0.2002 - val_accuracy: 0.9490\n",
            "Epoch 310/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0540 - accuracy: 0.9876 - val_loss: 0.2015 - val_accuracy: 0.9486\n",
            "Epoch 311/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0531 - accuracy: 0.9879 - val_loss: 0.2018 - val_accuracy: 0.9487\n",
            "Epoch 312/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0520 - accuracy: 0.9880 - val_loss: 0.2018 - val_accuracy: 0.9484\n",
            "Epoch 313/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0508 - accuracy: 0.9877 - val_loss: 0.2025 - val_accuracy: 0.9489\n",
            "Epoch 314/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0507 - accuracy: 0.9884 - val_loss: 0.2014 - val_accuracy: 0.9489\n",
            "Epoch 315/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0517 - accuracy: 0.9878 - val_loss: 0.2032 - val_accuracy: 0.9482\n",
            "Epoch 316/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0507 - accuracy: 0.9885 - val_loss: 0.2018 - val_accuracy: 0.9494\n",
            "Epoch 317/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0498 - accuracy: 0.9886 - val_loss: 0.2024 - val_accuracy: 0.9491\n",
            "Epoch 318/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0504 - accuracy: 0.9887 - val_loss: 0.2035 - val_accuracy: 0.9487\n",
            "Epoch 319/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0514 - accuracy: 0.9881 - val_loss: 0.2033 - val_accuracy: 0.9483\n",
            "Epoch 320/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0520 - accuracy: 0.9881 - val_loss: 0.2029 - val_accuracy: 0.9484\n",
            "Epoch 321/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0524 - accuracy: 0.9883 - val_loss: 0.2040 - val_accuracy: 0.9480\n",
            "Epoch 322/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0490 - accuracy: 0.9881 - val_loss: 0.2033 - val_accuracy: 0.9484\n",
            "Epoch 323/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0501 - accuracy: 0.9889 - val_loss: 0.2038 - val_accuracy: 0.9491\n",
            "Epoch 324/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0504 - accuracy: 0.9887 - val_loss: 0.2042 - val_accuracy: 0.9483\n",
            "Epoch 325/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0513 - accuracy: 0.9882 - val_loss: 0.2047 - val_accuracy: 0.9481\n",
            "Epoch 326/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0498 - accuracy: 0.9887 - val_loss: 0.2058 - val_accuracy: 0.9480\n",
            "Epoch 327/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0492 - accuracy: 0.9890 - val_loss: 0.2050 - val_accuracy: 0.9486\n",
            "Epoch 328/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0492 - accuracy: 0.9889 - val_loss: 0.2055 - val_accuracy: 0.9479\n",
            "Epoch 329/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0523 - accuracy: 0.9872 - val_loss: 0.2059 - val_accuracy: 0.9482\n",
            "Epoch 330/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0474 - accuracy: 0.9894 - val_loss: 0.2064 - val_accuracy: 0.9477\n",
            "Epoch 331/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0494 - accuracy: 0.9889 - val_loss: 0.2066 - val_accuracy: 0.9488\n",
            "Epoch 332/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0504 - accuracy: 0.9883 - val_loss: 0.2061 - val_accuracy: 0.9483\n",
            "Epoch 333/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0480 - accuracy: 0.9895 - val_loss: 0.2069 - val_accuracy: 0.9482\n",
            "Epoch 334/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0480 - accuracy: 0.9894 - val_loss: 0.2067 - val_accuracy: 0.9481\n",
            "Epoch 335/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0474 - accuracy: 0.9895 - val_loss: 0.2073 - val_accuracy: 0.9492\n",
            "Epoch 336/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0480 - accuracy: 0.9899 - val_loss: 0.2075 - val_accuracy: 0.9481\n",
            "Epoch 337/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0457 - accuracy: 0.9900 - val_loss: 0.2077 - val_accuracy: 0.9477\n",
            "Epoch 338/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0473 - accuracy: 0.9899 - val_loss: 0.2086 - val_accuracy: 0.9479\n",
            "Epoch 339/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0471 - accuracy: 0.9895 - val_loss: 0.2079 - val_accuracy: 0.9482\n",
            "Epoch 340/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0490 - accuracy: 0.9891 - val_loss: 0.2093 - val_accuracy: 0.9478\n",
            "Epoch 341/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0467 - accuracy: 0.9902 - val_loss: 0.2086 - val_accuracy: 0.9485\n",
            "Epoch 342/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0484 - accuracy: 0.9895 - val_loss: 0.2090 - val_accuracy: 0.9478\n",
            "Epoch 343/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0463 - accuracy: 0.9898 - val_loss: 0.2093 - val_accuracy: 0.9477\n",
            "Epoch 344/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0464 - accuracy: 0.9901 - val_loss: 0.2103 - val_accuracy: 0.9482\n",
            "Epoch 345/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0459 - accuracy: 0.9903 - val_loss: 0.2103 - val_accuracy: 0.9478\n",
            "Epoch 346/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0452 - accuracy: 0.9900 - val_loss: 0.2107 - val_accuracy: 0.9481\n",
            "Epoch 347/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0458 - accuracy: 0.9904 - val_loss: 0.2105 - val_accuracy: 0.9475\n",
            "Epoch 348/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0460 - accuracy: 0.9902 - val_loss: 0.2109 - val_accuracy: 0.9480\n",
            "Epoch 349/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0439 - accuracy: 0.9899 - val_loss: 0.2116 - val_accuracy: 0.9474\n",
            "Epoch 350/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0450 - accuracy: 0.9905 - val_loss: 0.2110 - val_accuracy: 0.9483\n",
            "Epoch 351/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0427 - accuracy: 0.9910 - val_loss: 0.2115 - val_accuracy: 0.9474\n",
            "Epoch 352/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0428 - accuracy: 0.9910 - val_loss: 0.2111 - val_accuracy: 0.9478\n",
            "Epoch 353/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0434 - accuracy: 0.9910 - val_loss: 0.2117 - val_accuracy: 0.9477\n",
            "Epoch 354/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0440 - accuracy: 0.9908 - val_loss: 0.2130 - val_accuracy: 0.9479\n",
            "Epoch 355/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0429 - accuracy: 0.9910 - val_loss: 0.2123 - val_accuracy: 0.9478\n",
            "Epoch 356/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0454 - accuracy: 0.9906 - val_loss: 0.2130 - val_accuracy: 0.9477\n",
            "Epoch 357/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0431 - accuracy: 0.9912 - val_loss: 0.2135 - val_accuracy: 0.9487\n",
            "Epoch 358/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0423 - accuracy: 0.9914 - val_loss: 0.2143 - val_accuracy: 0.9475\n",
            "Epoch 359/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0435 - accuracy: 0.9910 - val_loss: 0.2141 - val_accuracy: 0.9471\n",
            "Epoch 360/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0441 - accuracy: 0.9904 - val_loss: 0.2147 - val_accuracy: 0.9470\n",
            "Epoch 361/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0420 - accuracy: 0.9910 - val_loss: 0.2159 - val_accuracy: 0.9472\n",
            "Epoch 362/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0421 - accuracy: 0.9912 - val_loss: 0.2151 - val_accuracy: 0.9476\n",
            "Epoch 363/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0442 - accuracy: 0.9904 - val_loss: 0.2151 - val_accuracy: 0.9466\n",
            "Epoch 364/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0409 - accuracy: 0.9920 - val_loss: 0.2166 - val_accuracy: 0.9469\n",
            "Epoch 365/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0423 - accuracy: 0.9908 - val_loss: 0.2155 - val_accuracy: 0.9469\n",
            "Epoch 366/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0425 - accuracy: 0.9915 - val_loss: 0.2163 - val_accuracy: 0.9470\n",
            "Epoch 367/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0404 - accuracy: 0.9918 - val_loss: 0.2167 - val_accuracy: 0.9462\n",
            "Epoch 368/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0406 - accuracy: 0.9914 - val_loss: 0.2166 - val_accuracy: 0.9467\n",
            "Epoch 369/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0406 - accuracy: 0.9919 - val_loss: 0.2169 - val_accuracy: 0.9469\n",
            "Epoch 370/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0393 - accuracy: 0.9919 - val_loss: 0.2179 - val_accuracy: 0.9469\n",
            "Epoch 371/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0403 - accuracy: 0.9915 - val_loss: 0.2187 - val_accuracy: 0.9472\n",
            "Epoch 372/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0415 - accuracy: 0.9918 - val_loss: 0.2179 - val_accuracy: 0.9478\n",
            "Epoch 373/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0405 - accuracy: 0.9919 - val_loss: 0.2185 - val_accuracy: 0.9473\n",
            "Epoch 374/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0404 - accuracy: 0.9917 - val_loss: 0.2189 - val_accuracy: 0.9473\n",
            "Epoch 375/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0430 - accuracy: 0.9911 - val_loss: 0.2185 - val_accuracy: 0.9471\n",
            "Epoch 376/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0420 - accuracy: 0.9913 - val_loss: 0.2193 - val_accuracy: 0.9464\n",
            "Epoch 377/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0424 - accuracy: 0.9915 - val_loss: 0.2210 - val_accuracy: 0.9473\n",
            "Epoch 378/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0401 - accuracy: 0.9921 - val_loss: 0.2195 - val_accuracy: 0.9470\n",
            "Epoch 379/400\n",
            "24/24 [==============================] - 0s 12ms/step - loss: 0.0426 - accuracy: 0.9913 - val_loss: 0.2205 - val_accuracy: 0.9465\n",
            "Epoch 380/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0406 - accuracy: 0.9921 - val_loss: 0.2208 - val_accuracy: 0.9470\n",
            "Epoch 381/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0399 - accuracy: 0.9915 - val_loss: 0.2209 - val_accuracy: 0.9467\n",
            "Epoch 382/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0422 - accuracy: 0.9918 - val_loss: 0.2204 - val_accuracy: 0.9472\n",
            "Epoch 383/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0386 - accuracy: 0.9923 - val_loss: 0.2205 - val_accuracy: 0.9468\n",
            "Epoch 384/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0404 - accuracy: 0.9916 - val_loss: 0.2224 - val_accuracy: 0.9473\n",
            "Epoch 385/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0385 - accuracy: 0.9923 - val_loss: 0.2233 - val_accuracy: 0.9469\n",
            "Epoch 386/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0375 - accuracy: 0.9922 - val_loss: 0.2226 - val_accuracy: 0.9471\n",
            "Epoch 387/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0383 - accuracy: 0.9923 - val_loss: 0.2233 - val_accuracy: 0.9461\n",
            "Epoch 388/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0394 - accuracy: 0.9920 - val_loss: 0.2229 - val_accuracy: 0.9469\n",
            "Epoch 389/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0405 - accuracy: 0.9920 - val_loss: 0.2235 - val_accuracy: 0.9468\n",
            "Epoch 390/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0385 - accuracy: 0.9920 - val_loss: 0.2231 - val_accuracy: 0.9468\n",
            "Epoch 391/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0379 - accuracy: 0.9928 - val_loss: 0.2240 - val_accuracy: 0.9470\n",
            "Epoch 392/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0381 - accuracy: 0.9920 - val_loss: 0.2237 - val_accuracy: 0.9471\n",
            "Epoch 393/400\n",
            "24/24 [==============================] - 0s 14ms/step - loss: 0.0396 - accuracy: 0.9918 - val_loss: 0.2252 - val_accuracy: 0.9468\n",
            "Epoch 394/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0381 - accuracy: 0.9926 - val_loss: 0.2250 - val_accuracy: 0.9466\n",
            "Epoch 395/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0381 - accuracy: 0.9926 - val_loss: 0.2264 - val_accuracy: 0.9467\n",
            "Epoch 396/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0390 - accuracy: 0.9921 - val_loss: 0.2269 - val_accuracy: 0.9458\n",
            "Epoch 397/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0394 - accuracy: 0.9921 - val_loss: 0.2265 - val_accuracy: 0.9456\n",
            "Epoch 398/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0360 - accuracy: 0.9930 - val_loss: 0.2277 - val_accuracy: 0.9463\n",
            "Epoch 399/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0383 - accuracy: 0.9925 - val_loss: 0.2273 - val_accuracy: 0.9463\n",
            "Epoch 400/400\n",
            "24/24 [==============================] - 0s 13ms/step - loss: 0.0372 - accuracy: 0.9926 - val_loss: 0.2269 - val_accuracy: 0.9467\n",
            "Epoch 1/800\n",
            "12/12 [==============================] - 1s 42ms/step - loss: 2.2684 - accuracy: 0.2195 - val_loss: 2.1352 - val_accuracy: 0.5752\n",
            "Epoch 2/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 2.0976 - accuracy: 0.6163 - val_loss: 1.9578 - val_accuracy: 0.6985\n",
            "Epoch 3/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 1.9234 - accuracy: 0.6828 - val_loss: 1.7811 - val_accuracy: 0.7232\n",
            "Epoch 4/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 1.7527 - accuracy: 0.7103 - val_loss: 1.6127 - val_accuracy: 0.7547\n",
            "Epoch 5/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 1.5890 - accuracy: 0.7434 - val_loss: 1.4585 - val_accuracy: 0.7890\n",
            "Epoch 6/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 1.4421 - accuracy: 0.7733 - val_loss: 1.3207 - val_accuracy: 0.8141\n",
            "Epoch 7/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 1.3098 - accuracy: 0.7972 - val_loss: 1.1992 - val_accuracy: 0.8267\n",
            "Epoch 8/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 1.1945 - accuracy: 0.8153 - val_loss: 1.0931 - val_accuracy: 0.8387\n",
            "Epoch 9/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 1.0950 - accuracy: 0.8264 - val_loss: 1.0020 - val_accuracy: 0.8482\n",
            "Epoch 10/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 1.0106 - accuracy: 0.8331 - val_loss: 0.9236 - val_accuracy: 0.8557\n",
            "Epoch 11/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.9328 - accuracy: 0.8418 - val_loss: 0.8565 - val_accuracy: 0.8602\n",
            "Epoch 12/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.8661 - accuracy: 0.8527 - val_loss: 0.7983 - val_accuracy: 0.8664\n",
            "Epoch 13/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.8124 - accuracy: 0.8578 - val_loss: 0.7476 - val_accuracy: 0.8705\n",
            "Epoch 14/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.7597 - accuracy: 0.8646 - val_loss: 0.7031 - val_accuracy: 0.8753\n",
            "Epoch 15/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.7194 - accuracy: 0.8683 - val_loss: 0.6641 - val_accuracy: 0.8803\n",
            "Epoch 16/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.6792 - accuracy: 0.8730 - val_loss: 0.6295 - val_accuracy: 0.8848\n",
            "Epoch 17/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.6469 - accuracy: 0.8752 - val_loss: 0.5982 - val_accuracy: 0.8870\n",
            "Epoch 18/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.6129 - accuracy: 0.8814 - val_loss: 0.5706 - val_accuracy: 0.8898\n",
            "Epoch 19/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.5897 - accuracy: 0.8831 - val_loss: 0.5455 - val_accuracy: 0.8932\n",
            "Epoch 20/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.5597 - accuracy: 0.8873 - val_loss: 0.5234 - val_accuracy: 0.8950\n",
            "Epoch 21/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.5362 - accuracy: 0.8896 - val_loss: 0.5031 - val_accuracy: 0.8982\n",
            "Epoch 22/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.5136 - accuracy: 0.8937 - val_loss: 0.4849 - val_accuracy: 0.8993\n",
            "Epoch 23/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.4973 - accuracy: 0.8954 - val_loss: 0.4685 - val_accuracy: 0.9018\n",
            "Epoch 24/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.4848 - accuracy: 0.8964 - val_loss: 0.4535 - val_accuracy: 0.9034\n",
            "Epoch 25/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.4685 - accuracy: 0.8975 - val_loss: 0.4400 - val_accuracy: 0.9038\n",
            "Epoch 26/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.4530 - accuracy: 0.8998 - val_loss: 0.4277 - val_accuracy: 0.9051\n",
            "Epoch 27/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.4415 - accuracy: 0.9022 - val_loss: 0.4163 - val_accuracy: 0.9068\n",
            "Epoch 28/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.4275 - accuracy: 0.9035 - val_loss: 0.4058 - val_accuracy: 0.9078\n",
            "Epoch 29/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.4163 - accuracy: 0.9050 - val_loss: 0.3964 - val_accuracy: 0.9088\n",
            "Epoch 30/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.4049 - accuracy: 0.9061 - val_loss: 0.3875 - val_accuracy: 0.9094\n",
            "Epoch 31/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.4006 - accuracy: 0.9064 - val_loss: 0.3792 - val_accuracy: 0.9111\n",
            "Epoch 32/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3899 - accuracy: 0.9081 - val_loss: 0.3716 - val_accuracy: 0.9122\n",
            "Epoch 33/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.3850 - accuracy: 0.9088 - val_loss: 0.3644 - val_accuracy: 0.9123\n",
            "Epoch 34/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.3723 - accuracy: 0.9111 - val_loss: 0.3578 - val_accuracy: 0.9137\n",
            "Epoch 35/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3635 - accuracy: 0.9131 - val_loss: 0.3516 - val_accuracy: 0.9145\n",
            "Epoch 36/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3614 - accuracy: 0.9123 - val_loss: 0.3459 - val_accuracy: 0.9146\n",
            "Epoch 37/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3507 - accuracy: 0.9147 - val_loss: 0.3404 - val_accuracy: 0.9150\n",
            "Epoch 38/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3462 - accuracy: 0.9146 - val_loss: 0.3353 - val_accuracy: 0.9156\n",
            "Epoch 39/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.3397 - accuracy: 0.9144 - val_loss: 0.3304 - val_accuracy: 0.9167\n",
            "Epoch 40/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.3368 - accuracy: 0.9164 - val_loss: 0.3258 - val_accuracy: 0.9162\n",
            "Epoch 41/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.3299 - accuracy: 0.9175 - val_loss: 0.3214 - val_accuracy: 0.9166\n",
            "Epoch 42/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3235 - accuracy: 0.9197 - val_loss: 0.3173 - val_accuracy: 0.9177\n",
            "Epoch 43/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3239 - accuracy: 0.9188 - val_loss: 0.3133 - val_accuracy: 0.9178\n",
            "Epoch 44/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3178 - accuracy: 0.9198 - val_loss: 0.3097 - val_accuracy: 0.9192\n",
            "Epoch 45/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.3122 - accuracy: 0.9207 - val_loss: 0.3063 - val_accuracy: 0.9192\n",
            "Epoch 46/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.3082 - accuracy: 0.9220 - val_loss: 0.3026 - val_accuracy: 0.9201\n",
            "Epoch 47/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.3074 - accuracy: 0.9207 - val_loss: 0.2995 - val_accuracy: 0.9210\n",
            "Epoch 48/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.3017 - accuracy: 0.9225 - val_loss: 0.2962 - val_accuracy: 0.9217\n",
            "Epoch 49/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.3001 - accuracy: 0.9220 - val_loss: 0.2935 - val_accuracy: 0.9225\n",
            "Epoch 50/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.2959 - accuracy: 0.9231 - val_loss: 0.2904 - val_accuracy: 0.9229\n",
            "Epoch 51/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2900 - accuracy: 0.9241 - val_loss: 0.2877 - val_accuracy: 0.9234\n",
            "Epoch 52/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2852 - accuracy: 0.9258 - val_loss: 0.2849 - val_accuracy: 0.9246\n",
            "Epoch 53/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2891 - accuracy: 0.9249 - val_loss: 0.2826 - val_accuracy: 0.9243\n",
            "Epoch 54/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.2834 - accuracy: 0.9252 - val_loss: 0.2801 - val_accuracy: 0.9255\n",
            "Epoch 55/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2810 - accuracy: 0.9254 - val_loss: 0.2775 - val_accuracy: 0.9259\n",
            "Epoch 56/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2801 - accuracy: 0.9266 - val_loss: 0.2755 - val_accuracy: 0.9272\n",
            "Epoch 57/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.2727 - accuracy: 0.9278 - val_loss: 0.2730 - val_accuracy: 0.9269\n",
            "Epoch 58/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.2690 - accuracy: 0.9298 - val_loss: 0.2712 - val_accuracy: 0.9280\n",
            "Epoch 59/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2678 - accuracy: 0.9286 - val_loss: 0.2688 - val_accuracy: 0.9283\n",
            "Epoch 60/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.2624 - accuracy: 0.9307 - val_loss: 0.2667 - val_accuracy: 0.9289\n",
            "Epoch 61/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.2654 - accuracy: 0.9297 - val_loss: 0.2648 - val_accuracy: 0.9293\n",
            "Epoch 62/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.2604 - accuracy: 0.9327 - val_loss: 0.2630 - val_accuracy: 0.9296\n",
            "Epoch 63/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.2600 - accuracy: 0.9304 - val_loss: 0.2613 - val_accuracy: 0.9304\n",
            "Epoch 64/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2575 - accuracy: 0.9301 - val_loss: 0.2592 - val_accuracy: 0.9306\n",
            "Epoch 65/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2539 - accuracy: 0.9329 - val_loss: 0.2577 - val_accuracy: 0.9310\n",
            "Epoch 66/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2514 - accuracy: 0.9336 - val_loss: 0.2557 - val_accuracy: 0.9307\n",
            "Epoch 67/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2450 - accuracy: 0.9344 - val_loss: 0.2543 - val_accuracy: 0.9309\n",
            "Epoch 68/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2420 - accuracy: 0.9359 - val_loss: 0.2526 - val_accuracy: 0.9319\n",
            "Epoch 69/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2459 - accuracy: 0.9339 - val_loss: 0.2514 - val_accuracy: 0.9320\n",
            "Epoch 70/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.2415 - accuracy: 0.9347 - val_loss: 0.2497 - val_accuracy: 0.9325\n",
            "Epoch 71/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2428 - accuracy: 0.9346 - val_loss: 0.2481 - val_accuracy: 0.9330\n",
            "Epoch 72/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2381 - accuracy: 0.9359 - val_loss: 0.2469 - val_accuracy: 0.9331\n",
            "Epoch 73/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2373 - accuracy: 0.9356 - val_loss: 0.2453 - val_accuracy: 0.9336\n",
            "Epoch 74/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.2352 - accuracy: 0.9367 - val_loss: 0.2442 - val_accuracy: 0.9333\n",
            "Epoch 75/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.2328 - accuracy: 0.9376 - val_loss: 0.2425 - val_accuracy: 0.9341\n",
            "Epoch 76/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2295 - accuracy: 0.9396 - val_loss: 0.2413 - val_accuracy: 0.9337\n",
            "Epoch 77/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.2289 - accuracy: 0.9376 - val_loss: 0.2401 - val_accuracy: 0.9344\n",
            "Epoch 78/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2293 - accuracy: 0.9388 - val_loss: 0.2390 - val_accuracy: 0.9345\n",
            "Epoch 79/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2227 - accuracy: 0.9396 - val_loss: 0.2377 - val_accuracy: 0.9350\n",
            "Epoch 80/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2288 - accuracy: 0.9382 - val_loss: 0.2365 - val_accuracy: 0.9348\n",
            "Epoch 81/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2216 - accuracy: 0.9403 - val_loss: 0.2355 - val_accuracy: 0.9356\n",
            "Epoch 82/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2209 - accuracy: 0.9399 - val_loss: 0.2345 - val_accuracy: 0.9353\n",
            "Epoch 83/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2236 - accuracy: 0.9388 - val_loss: 0.2330 - val_accuracy: 0.9362\n",
            "Epoch 84/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2148 - accuracy: 0.9425 - val_loss: 0.2322 - val_accuracy: 0.9358\n",
            "Epoch 85/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.2135 - accuracy: 0.9427 - val_loss: 0.2312 - val_accuracy: 0.9359\n",
            "Epoch 86/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2160 - accuracy: 0.9422 - val_loss: 0.2300 - val_accuracy: 0.9368\n",
            "Epoch 87/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2162 - accuracy: 0.9417 - val_loss: 0.2290 - val_accuracy: 0.9368\n",
            "Epoch 88/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.2150 - accuracy: 0.9420 - val_loss: 0.2284 - val_accuracy: 0.9368\n",
            "Epoch 89/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2101 - accuracy: 0.9436 - val_loss: 0.2271 - val_accuracy: 0.9377\n",
            "Epoch 90/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2109 - accuracy: 0.9430 - val_loss: 0.2263 - val_accuracy: 0.9377\n",
            "Epoch 91/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.2087 - accuracy: 0.9437 - val_loss: 0.2253 - val_accuracy: 0.9382\n",
            "Epoch 92/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2086 - accuracy: 0.9438 - val_loss: 0.2245 - val_accuracy: 0.9383\n",
            "Epoch 93/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2058 - accuracy: 0.9450 - val_loss: 0.2234 - val_accuracy: 0.9384\n",
            "Epoch 94/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2044 - accuracy: 0.9441 - val_loss: 0.2229 - val_accuracy: 0.9384\n",
            "Epoch 95/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2018 - accuracy: 0.9458 - val_loss: 0.2218 - val_accuracy: 0.9382\n",
            "Epoch 96/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.2033 - accuracy: 0.9459 - val_loss: 0.2210 - val_accuracy: 0.9389\n",
            "Epoch 97/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.2010 - accuracy: 0.9459 - val_loss: 0.2201 - val_accuracy: 0.9391\n",
            "Epoch 98/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.2026 - accuracy: 0.9453 - val_loss: 0.2198 - val_accuracy: 0.9391\n",
            "Epoch 99/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1979 - accuracy: 0.9475 - val_loss: 0.2188 - val_accuracy: 0.9398\n",
            "Epoch 100/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.1987 - accuracy: 0.9463 - val_loss: 0.2178 - val_accuracy: 0.9400\n",
            "Epoch 101/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1975 - accuracy: 0.9465 - val_loss: 0.2172 - val_accuracy: 0.9404\n",
            "Epoch 102/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1922 - accuracy: 0.9489 - val_loss: 0.2163 - val_accuracy: 0.9398\n",
            "Epoch 103/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1938 - accuracy: 0.9488 - val_loss: 0.2157 - val_accuracy: 0.9408\n",
            "Epoch 104/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1919 - accuracy: 0.9485 - val_loss: 0.2151 - val_accuracy: 0.9398\n",
            "Epoch 105/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1921 - accuracy: 0.9489 - val_loss: 0.2143 - val_accuracy: 0.9413\n",
            "Epoch 106/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.1914 - accuracy: 0.9487 - val_loss: 0.2137 - val_accuracy: 0.9406\n",
            "Epoch 107/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1904 - accuracy: 0.9489 - val_loss: 0.2129 - val_accuracy: 0.9409\n",
            "Epoch 108/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1845 - accuracy: 0.9505 - val_loss: 0.2125 - val_accuracy: 0.9412\n",
            "Epoch 109/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1884 - accuracy: 0.9496 - val_loss: 0.2116 - val_accuracy: 0.9413\n",
            "Epoch 110/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1870 - accuracy: 0.9494 - val_loss: 0.2113 - val_accuracy: 0.9416\n",
            "Epoch 111/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1847 - accuracy: 0.9510 - val_loss: 0.2105 - val_accuracy: 0.9417\n",
            "Epoch 112/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1858 - accuracy: 0.9502 - val_loss: 0.2098 - val_accuracy: 0.9417\n",
            "Epoch 113/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1803 - accuracy: 0.9520 - val_loss: 0.2092 - val_accuracy: 0.9425\n",
            "Epoch 114/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1819 - accuracy: 0.9505 - val_loss: 0.2088 - val_accuracy: 0.9423\n",
            "Epoch 115/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1821 - accuracy: 0.9520 - val_loss: 0.2081 - val_accuracy: 0.9419\n",
            "Epoch 116/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1798 - accuracy: 0.9520 - val_loss: 0.2074 - val_accuracy: 0.9427\n",
            "Epoch 117/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1795 - accuracy: 0.9523 - val_loss: 0.2069 - val_accuracy: 0.9427\n",
            "Epoch 118/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1769 - accuracy: 0.9519 - val_loss: 0.2067 - val_accuracy: 0.9425\n",
            "Epoch 119/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1792 - accuracy: 0.9523 - val_loss: 0.2059 - val_accuracy: 0.9429\n",
            "Epoch 120/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1765 - accuracy: 0.9535 - val_loss: 0.2056 - val_accuracy: 0.9426\n",
            "Epoch 121/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1763 - accuracy: 0.9523 - val_loss: 0.2048 - val_accuracy: 0.9433\n",
            "Epoch 122/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.1774 - accuracy: 0.9528 - val_loss: 0.2045 - val_accuracy: 0.9437\n",
            "Epoch 123/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1753 - accuracy: 0.9537 - val_loss: 0.2039 - val_accuracy: 0.9434\n",
            "Epoch 124/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1688 - accuracy: 0.9557 - val_loss: 0.2033 - val_accuracy: 0.9434\n",
            "Epoch 125/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1735 - accuracy: 0.9527 - val_loss: 0.2031 - val_accuracy: 0.9434\n",
            "Epoch 126/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1687 - accuracy: 0.9559 - val_loss: 0.2021 - val_accuracy: 0.9438\n",
            "Epoch 127/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1744 - accuracy: 0.9534 - val_loss: 0.2020 - val_accuracy: 0.9436\n",
            "Epoch 128/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1674 - accuracy: 0.9559 - val_loss: 0.2013 - val_accuracy: 0.9436\n",
            "Epoch 129/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1685 - accuracy: 0.9551 - val_loss: 0.2010 - val_accuracy: 0.9434\n",
            "Epoch 130/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1662 - accuracy: 0.9554 - val_loss: 0.2003 - val_accuracy: 0.9438\n",
            "Epoch 131/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1652 - accuracy: 0.9566 - val_loss: 0.2002 - val_accuracy: 0.9438\n",
            "Epoch 132/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1661 - accuracy: 0.9554 - val_loss: 0.1997 - val_accuracy: 0.9439\n",
            "Epoch 133/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1667 - accuracy: 0.9559 - val_loss: 0.1992 - val_accuracy: 0.9440\n",
            "Epoch 134/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1666 - accuracy: 0.9563 - val_loss: 0.1988 - val_accuracy: 0.9437\n",
            "Epoch 135/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1637 - accuracy: 0.9569 - val_loss: 0.1983 - val_accuracy: 0.9438\n",
            "Epoch 136/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1597 - accuracy: 0.9582 - val_loss: 0.1980 - val_accuracy: 0.9441\n",
            "Epoch 137/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1611 - accuracy: 0.9564 - val_loss: 0.1975 - val_accuracy: 0.9442\n",
            "Epoch 138/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1595 - accuracy: 0.9579 - val_loss: 0.1970 - val_accuracy: 0.9442\n",
            "Epoch 139/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1637 - accuracy: 0.9560 - val_loss: 0.1968 - val_accuracy: 0.9442\n",
            "Epoch 140/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1636 - accuracy: 0.9574 - val_loss: 0.1962 - val_accuracy: 0.9448\n",
            "Epoch 141/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.1581 - accuracy: 0.9582 - val_loss: 0.1959 - val_accuracy: 0.9445\n",
            "Epoch 142/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1585 - accuracy: 0.9587 - val_loss: 0.1954 - val_accuracy: 0.9448\n",
            "Epoch 143/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1582 - accuracy: 0.9579 - val_loss: 0.1951 - val_accuracy: 0.9449\n",
            "Epoch 144/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1563 - accuracy: 0.9587 - val_loss: 0.1946 - val_accuracy: 0.9452\n",
            "Epoch 145/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1552 - accuracy: 0.9592 - val_loss: 0.1942 - val_accuracy: 0.9449\n",
            "Epoch 146/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1552 - accuracy: 0.9592 - val_loss: 0.1939 - val_accuracy: 0.9448\n",
            "Epoch 147/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1549 - accuracy: 0.9590 - val_loss: 0.1936 - val_accuracy: 0.9456\n",
            "Epoch 148/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1572 - accuracy: 0.9584 - val_loss: 0.1933 - val_accuracy: 0.9452\n",
            "Epoch 149/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1583 - accuracy: 0.9574 - val_loss: 0.1926 - val_accuracy: 0.9452\n",
            "Epoch 150/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1533 - accuracy: 0.9597 - val_loss: 0.1925 - val_accuracy: 0.9457\n",
            "Epoch 151/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1541 - accuracy: 0.9588 - val_loss: 0.1921 - val_accuracy: 0.9453\n",
            "Epoch 152/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1538 - accuracy: 0.9597 - val_loss: 0.1920 - val_accuracy: 0.9458\n",
            "Epoch 153/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1512 - accuracy: 0.9598 - val_loss: 0.1915 - val_accuracy: 0.9457\n",
            "Epoch 154/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1496 - accuracy: 0.9605 - val_loss: 0.1911 - val_accuracy: 0.9452\n",
            "Epoch 155/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1473 - accuracy: 0.9605 - val_loss: 0.1906 - val_accuracy: 0.9458\n",
            "Epoch 156/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1516 - accuracy: 0.9591 - val_loss: 0.1904 - val_accuracy: 0.9453\n",
            "Epoch 157/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1481 - accuracy: 0.9610 - val_loss: 0.1902 - val_accuracy: 0.9454\n",
            "Epoch 158/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1493 - accuracy: 0.9609 - val_loss: 0.1899 - val_accuracy: 0.9454\n",
            "Epoch 159/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1468 - accuracy: 0.9617 - val_loss: 0.1895 - val_accuracy: 0.9451\n",
            "Epoch 160/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1482 - accuracy: 0.9603 - val_loss: 0.1890 - val_accuracy: 0.9460\n",
            "Epoch 161/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1467 - accuracy: 0.9614 - val_loss: 0.1890 - val_accuracy: 0.9457\n",
            "Epoch 162/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1432 - accuracy: 0.9632 - val_loss: 0.1886 - val_accuracy: 0.9458\n",
            "Epoch 163/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1445 - accuracy: 0.9618 - val_loss: 0.1883 - val_accuracy: 0.9459\n",
            "Epoch 164/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1461 - accuracy: 0.9617 - val_loss: 0.1879 - val_accuracy: 0.9457\n",
            "Epoch 165/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1408 - accuracy: 0.9626 - val_loss: 0.1874 - val_accuracy: 0.9460\n",
            "Epoch 166/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1427 - accuracy: 0.9624 - val_loss: 0.1874 - val_accuracy: 0.9458\n",
            "Epoch 167/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1426 - accuracy: 0.9617 - val_loss: 0.1872 - val_accuracy: 0.9457\n",
            "Epoch 168/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1425 - accuracy: 0.9626 - val_loss: 0.1868 - val_accuracy: 0.9461\n",
            "Epoch 169/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1427 - accuracy: 0.9619 - val_loss: 0.1866 - val_accuracy: 0.9459\n",
            "Epoch 170/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1452 - accuracy: 0.9615 - val_loss: 0.1862 - val_accuracy: 0.9467\n",
            "Epoch 171/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1399 - accuracy: 0.9630 - val_loss: 0.1865 - val_accuracy: 0.9463\n",
            "Epoch 172/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1395 - accuracy: 0.9622 - val_loss: 0.1856 - val_accuracy: 0.9458\n",
            "Epoch 173/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1375 - accuracy: 0.9635 - val_loss: 0.1855 - val_accuracy: 0.9469\n",
            "Epoch 174/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1377 - accuracy: 0.9638 - val_loss: 0.1852 - val_accuracy: 0.9461\n",
            "Epoch 175/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1382 - accuracy: 0.9640 - val_loss: 0.1847 - val_accuracy: 0.9467\n",
            "Epoch 176/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1363 - accuracy: 0.9638 - val_loss: 0.1849 - val_accuracy: 0.9463\n",
            "Epoch 177/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1361 - accuracy: 0.9645 - val_loss: 0.1842 - val_accuracy: 0.9475\n",
            "Epoch 178/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1323 - accuracy: 0.9653 - val_loss: 0.1844 - val_accuracy: 0.9470\n",
            "Epoch 179/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1338 - accuracy: 0.9647 - val_loss: 0.1840 - val_accuracy: 0.9465\n",
            "Epoch 180/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1383 - accuracy: 0.9627 - val_loss: 0.1837 - val_accuracy: 0.9473\n",
            "Epoch 181/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1340 - accuracy: 0.9653 - val_loss: 0.1835 - val_accuracy: 0.9474\n",
            "Epoch 182/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1352 - accuracy: 0.9639 - val_loss: 0.1836 - val_accuracy: 0.9465\n",
            "Epoch 183/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1352 - accuracy: 0.9643 - val_loss: 0.1831 - val_accuracy: 0.9481\n",
            "Epoch 184/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1307 - accuracy: 0.9653 - val_loss: 0.1830 - val_accuracy: 0.9472\n",
            "Epoch 185/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.1316 - accuracy: 0.9654 - val_loss: 0.1827 - val_accuracy: 0.9476\n",
            "Epoch 186/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1312 - accuracy: 0.9649 - val_loss: 0.1824 - val_accuracy: 0.9482\n",
            "Epoch 187/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1314 - accuracy: 0.9647 - val_loss: 0.1824 - val_accuracy: 0.9480\n",
            "Epoch 188/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1305 - accuracy: 0.9644 - val_loss: 0.1824 - val_accuracy: 0.9483\n",
            "Epoch 189/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1294 - accuracy: 0.9662 - val_loss: 0.1818 - val_accuracy: 0.9471\n",
            "Epoch 190/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1282 - accuracy: 0.9660 - val_loss: 0.1817 - val_accuracy: 0.9487\n",
            "Epoch 191/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1257 - accuracy: 0.9668 - val_loss: 0.1817 - val_accuracy: 0.9473\n",
            "Epoch 192/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1304 - accuracy: 0.9650 - val_loss: 0.1814 - val_accuracy: 0.9477\n",
            "Epoch 193/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1290 - accuracy: 0.9656 - val_loss: 0.1813 - val_accuracy: 0.9474\n",
            "Epoch 194/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1293 - accuracy: 0.9654 - val_loss: 0.1811 - val_accuracy: 0.9486\n",
            "Epoch 195/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1286 - accuracy: 0.9662 - val_loss: 0.1810 - val_accuracy: 0.9482\n",
            "Epoch 196/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1244 - accuracy: 0.9669 - val_loss: 0.1808 - val_accuracy: 0.9480\n",
            "Epoch 197/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1251 - accuracy: 0.9670 - val_loss: 0.1807 - val_accuracy: 0.9489\n",
            "Epoch 198/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1254 - accuracy: 0.9669 - val_loss: 0.1805 - val_accuracy: 0.9482\n",
            "Epoch 199/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1286 - accuracy: 0.9662 - val_loss: 0.1804 - val_accuracy: 0.9489\n",
            "Epoch 200/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1262 - accuracy: 0.9672 - val_loss: 0.1801 - val_accuracy: 0.9482\n",
            "Epoch 201/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1250 - accuracy: 0.9675 - val_loss: 0.1801 - val_accuracy: 0.9480\n",
            "Epoch 202/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1227 - accuracy: 0.9675 - val_loss: 0.1799 - val_accuracy: 0.9490\n",
            "Epoch 203/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1235 - accuracy: 0.9679 - val_loss: 0.1796 - val_accuracy: 0.9487\n",
            "Epoch 204/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1228 - accuracy: 0.9678 - val_loss: 0.1798 - val_accuracy: 0.9492\n",
            "Epoch 205/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1242 - accuracy: 0.9677 - val_loss: 0.1795 - val_accuracy: 0.9482\n",
            "Epoch 206/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1233 - accuracy: 0.9677 - val_loss: 0.1793 - val_accuracy: 0.9483\n",
            "Epoch 207/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1213 - accuracy: 0.9683 - val_loss: 0.1791 - val_accuracy: 0.9490\n",
            "Epoch 208/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1224 - accuracy: 0.9680 - val_loss: 0.1792 - val_accuracy: 0.9483\n",
            "Epoch 209/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1230 - accuracy: 0.9674 - val_loss: 0.1788 - val_accuracy: 0.9485\n",
            "Epoch 210/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1213 - accuracy: 0.9685 - val_loss: 0.1788 - val_accuracy: 0.9488\n",
            "Epoch 211/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1192 - accuracy: 0.9687 - val_loss: 0.1787 - val_accuracy: 0.9492\n",
            "Epoch 212/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1189 - accuracy: 0.9689 - val_loss: 0.1785 - val_accuracy: 0.9492\n",
            "Epoch 213/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1193 - accuracy: 0.9686 - val_loss: 0.1787 - val_accuracy: 0.9486\n",
            "Epoch 214/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1178 - accuracy: 0.9691 - val_loss: 0.1783 - val_accuracy: 0.9489\n",
            "Epoch 215/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1213 - accuracy: 0.9680 - val_loss: 0.1781 - val_accuracy: 0.9488\n",
            "Epoch 216/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1189 - accuracy: 0.9687 - val_loss: 0.1782 - val_accuracy: 0.9489\n",
            "Epoch 217/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1179 - accuracy: 0.9685 - val_loss: 0.1780 - val_accuracy: 0.9493\n",
            "Epoch 218/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1190 - accuracy: 0.9688 - val_loss: 0.1780 - val_accuracy: 0.9485\n",
            "Epoch 219/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1194 - accuracy: 0.9696 - val_loss: 0.1779 - val_accuracy: 0.9493\n",
            "Epoch 220/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1202 - accuracy: 0.9688 - val_loss: 0.1776 - val_accuracy: 0.9492\n",
            "Epoch 221/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1159 - accuracy: 0.9699 - val_loss: 0.1782 - val_accuracy: 0.9485\n",
            "Epoch 222/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1153 - accuracy: 0.9701 - val_loss: 0.1778 - val_accuracy: 0.9488\n",
            "Epoch 223/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1171 - accuracy: 0.9689 - val_loss: 0.1777 - val_accuracy: 0.9491\n",
            "Epoch 224/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1181 - accuracy: 0.9694 - val_loss: 0.1776 - val_accuracy: 0.9491\n",
            "Epoch 225/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1143 - accuracy: 0.9703 - val_loss: 0.1774 - val_accuracy: 0.9497\n",
            "Epoch 226/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1160 - accuracy: 0.9698 - val_loss: 0.1776 - val_accuracy: 0.9484\n",
            "Epoch 227/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1154 - accuracy: 0.9698 - val_loss: 0.1772 - val_accuracy: 0.9492\n",
            "Epoch 228/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1143 - accuracy: 0.9696 - val_loss: 0.1775 - val_accuracy: 0.9487\n",
            "Epoch 229/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1139 - accuracy: 0.9696 - val_loss: 0.1772 - val_accuracy: 0.9489\n",
            "Epoch 230/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1100 - accuracy: 0.9710 - val_loss: 0.1771 - val_accuracy: 0.9495\n",
            "Epoch 231/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1139 - accuracy: 0.9703 - val_loss: 0.1769 - val_accuracy: 0.9490\n",
            "Epoch 232/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1154 - accuracy: 0.9696 - val_loss: 0.1773 - val_accuracy: 0.9492\n",
            "Epoch 233/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1086 - accuracy: 0.9712 - val_loss: 0.1768 - val_accuracy: 0.9492\n",
            "Epoch 234/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1081 - accuracy: 0.9709 - val_loss: 0.1772 - val_accuracy: 0.9492\n",
            "Epoch 235/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1100 - accuracy: 0.9715 - val_loss: 0.1773 - val_accuracy: 0.9488\n",
            "Epoch 236/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1111 - accuracy: 0.9706 - val_loss: 0.1767 - val_accuracy: 0.9492\n",
            "Epoch 237/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1125 - accuracy: 0.9704 - val_loss: 0.1766 - val_accuracy: 0.9495\n",
            "Epoch 238/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1100 - accuracy: 0.9705 - val_loss: 0.1769 - val_accuracy: 0.9491\n",
            "Epoch 239/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1106 - accuracy: 0.9705 - val_loss: 0.1766 - val_accuracy: 0.9492\n",
            "Epoch 240/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.1108 - accuracy: 0.9711 - val_loss: 0.1769 - val_accuracy: 0.9489\n",
            "Epoch 241/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1125 - accuracy: 0.9714 - val_loss: 0.1769 - val_accuracy: 0.9491\n",
            "Epoch 242/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1086 - accuracy: 0.9717 - val_loss: 0.1767 - val_accuracy: 0.9486\n",
            "Epoch 243/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1104 - accuracy: 0.9712 - val_loss: 0.1766 - val_accuracy: 0.9493\n",
            "Epoch 244/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1073 - accuracy: 0.9719 - val_loss: 0.1765 - val_accuracy: 0.9488\n",
            "Epoch 245/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1068 - accuracy: 0.9717 - val_loss: 0.1764 - val_accuracy: 0.9487\n",
            "Epoch 246/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1066 - accuracy: 0.9720 - val_loss: 0.1768 - val_accuracy: 0.9487\n",
            "Epoch 247/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1073 - accuracy: 0.9717 - val_loss: 0.1765 - val_accuracy: 0.9488\n",
            "Epoch 248/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1058 - accuracy: 0.9729 - val_loss: 0.1766 - val_accuracy: 0.9491\n",
            "Epoch 249/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1061 - accuracy: 0.9718 - val_loss: 0.1763 - val_accuracy: 0.9489\n",
            "Epoch 250/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1075 - accuracy: 0.9718 - val_loss: 0.1767 - val_accuracy: 0.9489\n",
            "Epoch 251/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1037 - accuracy: 0.9728 - val_loss: 0.1761 - val_accuracy: 0.9493\n",
            "Epoch 252/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1076 - accuracy: 0.9717 - val_loss: 0.1767 - val_accuracy: 0.9491\n",
            "Epoch 253/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1047 - accuracy: 0.9726 - val_loss: 0.1762 - val_accuracy: 0.9496\n",
            "Epoch 254/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1051 - accuracy: 0.9718 - val_loss: 0.1762 - val_accuracy: 0.9493\n",
            "Epoch 255/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1052 - accuracy: 0.9722 - val_loss: 0.1762 - val_accuracy: 0.9488\n",
            "Epoch 256/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1003 - accuracy: 0.9736 - val_loss: 0.1763 - val_accuracy: 0.9492\n",
            "Epoch 257/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1038 - accuracy: 0.9731 - val_loss: 0.1762 - val_accuracy: 0.9493\n",
            "Epoch 258/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1023 - accuracy: 0.9723 - val_loss: 0.1764 - val_accuracy: 0.9487\n",
            "Epoch 259/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1027 - accuracy: 0.9728 - val_loss: 0.1763 - val_accuracy: 0.9492\n",
            "Epoch 260/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1010 - accuracy: 0.9730 - val_loss: 0.1761 - val_accuracy: 0.9488\n",
            "Epoch 261/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.1027 - accuracy: 0.9731 - val_loss: 0.1764 - val_accuracy: 0.9486\n",
            "Epoch 262/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1006 - accuracy: 0.9734 - val_loss: 0.1762 - val_accuracy: 0.9495\n",
            "Epoch 263/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.1014 - accuracy: 0.9734 - val_loss: 0.1763 - val_accuracy: 0.9488\n",
            "Epoch 264/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0989 - accuracy: 0.9741 - val_loss: 0.1762 - val_accuracy: 0.9490\n",
            "Epoch 265/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0986 - accuracy: 0.9745 - val_loss: 0.1763 - val_accuracy: 0.9489\n",
            "Epoch 266/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1009 - accuracy: 0.9736 - val_loss: 0.1763 - val_accuracy: 0.9491\n",
            "Epoch 267/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0993 - accuracy: 0.9742 - val_loss: 0.1761 - val_accuracy: 0.9498\n",
            "Epoch 268/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1004 - accuracy: 0.9734 - val_loss: 0.1762 - val_accuracy: 0.9488\n",
            "Epoch 269/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.1002 - accuracy: 0.9748 - val_loss: 0.1764 - val_accuracy: 0.9486\n",
            "Epoch 270/800\n",
            "12/12 [==============================] - 0s 23ms/step - loss: 0.1004 - accuracy: 0.9730 - val_loss: 0.1766 - val_accuracy: 0.9493\n",
            "Epoch 271/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1012 - accuracy: 0.9739 - val_loss: 0.1761 - val_accuracy: 0.9490\n",
            "Epoch 272/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0976 - accuracy: 0.9741 - val_loss: 0.1763 - val_accuracy: 0.9488\n",
            "Epoch 273/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1003 - accuracy: 0.9740 - val_loss: 0.1760 - val_accuracy: 0.9492\n",
            "Epoch 274/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.1003 - accuracy: 0.9739 - val_loss: 0.1761 - val_accuracy: 0.9489\n",
            "Epoch 275/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0982 - accuracy: 0.9745 - val_loss: 0.1764 - val_accuracy: 0.9490\n",
            "Epoch 276/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0950 - accuracy: 0.9752 - val_loss: 0.1765 - val_accuracy: 0.9498\n",
            "Epoch 277/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0994 - accuracy: 0.9743 - val_loss: 0.1762 - val_accuracy: 0.9493\n",
            "Epoch 278/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0975 - accuracy: 0.9741 - val_loss: 0.1764 - val_accuracy: 0.9489\n",
            "Epoch 279/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0959 - accuracy: 0.9751 - val_loss: 0.1765 - val_accuracy: 0.9490\n",
            "Epoch 280/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0953 - accuracy: 0.9752 - val_loss: 0.1764 - val_accuracy: 0.9488\n",
            "Epoch 281/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0947 - accuracy: 0.9751 - val_loss: 0.1765 - val_accuracy: 0.9492\n",
            "Epoch 282/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0951 - accuracy: 0.9750 - val_loss: 0.1767 - val_accuracy: 0.9491\n",
            "Epoch 283/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0928 - accuracy: 0.9755 - val_loss: 0.1761 - val_accuracy: 0.9493\n",
            "Epoch 284/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0950 - accuracy: 0.9761 - val_loss: 0.1770 - val_accuracy: 0.9486\n",
            "Epoch 285/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0952 - accuracy: 0.9755 - val_loss: 0.1766 - val_accuracy: 0.9488\n",
            "Epoch 286/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0969 - accuracy: 0.9750 - val_loss: 0.1766 - val_accuracy: 0.9489\n",
            "Epoch 287/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0967 - accuracy: 0.9751 - val_loss: 0.1770 - val_accuracy: 0.9498\n",
            "Epoch 288/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0945 - accuracy: 0.9761 - val_loss: 0.1768 - val_accuracy: 0.9482\n",
            "Epoch 289/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0915 - accuracy: 0.9763 - val_loss: 0.1767 - val_accuracy: 0.9488\n",
            "Epoch 290/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0954 - accuracy: 0.9755 - val_loss: 0.1768 - val_accuracy: 0.9493\n",
            "Epoch 291/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0927 - accuracy: 0.9761 - val_loss: 0.1768 - val_accuracy: 0.9490\n",
            "Epoch 292/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0906 - accuracy: 0.9763 - val_loss: 0.1769 - val_accuracy: 0.9489\n",
            "Epoch 293/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0952 - accuracy: 0.9755 - val_loss: 0.1766 - val_accuracy: 0.9502\n",
            "Epoch 294/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0936 - accuracy: 0.9756 - val_loss: 0.1767 - val_accuracy: 0.9494\n",
            "Epoch 295/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0938 - accuracy: 0.9763 - val_loss: 0.1767 - val_accuracy: 0.9493\n",
            "Epoch 296/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0913 - accuracy: 0.9755 - val_loss: 0.1771 - val_accuracy: 0.9483\n",
            "Epoch 297/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0919 - accuracy: 0.9756 - val_loss: 0.1769 - val_accuracy: 0.9496\n",
            "Epoch 298/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0900 - accuracy: 0.9767 - val_loss: 0.1769 - val_accuracy: 0.9491\n",
            "Epoch 299/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0914 - accuracy: 0.9766 - val_loss: 0.1771 - val_accuracy: 0.9490\n",
            "Epoch 300/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0925 - accuracy: 0.9759 - val_loss: 0.1772 - val_accuracy: 0.9490\n",
            "Epoch 301/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0908 - accuracy: 0.9767 - val_loss: 0.1771 - val_accuracy: 0.9492\n",
            "Epoch 302/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0921 - accuracy: 0.9759 - val_loss: 0.1769 - val_accuracy: 0.9491\n",
            "Epoch 303/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0898 - accuracy: 0.9762 - val_loss: 0.1775 - val_accuracy: 0.9490\n",
            "Epoch 304/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0882 - accuracy: 0.9774 - val_loss: 0.1771 - val_accuracy: 0.9486\n",
            "Epoch 305/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0895 - accuracy: 0.9769 - val_loss: 0.1775 - val_accuracy: 0.9485\n",
            "Epoch 306/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0871 - accuracy: 0.9774 - val_loss: 0.1776 - val_accuracy: 0.9493\n",
            "Epoch 307/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0881 - accuracy: 0.9774 - val_loss: 0.1771 - val_accuracy: 0.9492\n",
            "Epoch 308/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0889 - accuracy: 0.9769 - val_loss: 0.1774 - val_accuracy: 0.9486\n",
            "Epoch 309/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0885 - accuracy: 0.9774 - val_loss: 0.1780 - val_accuracy: 0.9488\n",
            "Epoch 310/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0875 - accuracy: 0.9773 - val_loss: 0.1774 - val_accuracy: 0.9496\n",
            "Epoch 311/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0901 - accuracy: 0.9760 - val_loss: 0.1778 - val_accuracy: 0.9491\n",
            "Epoch 312/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0889 - accuracy: 0.9769 - val_loss: 0.1776 - val_accuracy: 0.9493\n",
            "Epoch 313/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0862 - accuracy: 0.9781 - val_loss: 0.1784 - val_accuracy: 0.9488\n",
            "Epoch 314/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0899 - accuracy: 0.9779 - val_loss: 0.1775 - val_accuracy: 0.9493\n",
            "Epoch 315/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0858 - accuracy: 0.9782 - val_loss: 0.1776 - val_accuracy: 0.9493\n",
            "Epoch 316/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0882 - accuracy: 0.9775 - val_loss: 0.1781 - val_accuracy: 0.9488\n",
            "Epoch 317/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0874 - accuracy: 0.9776 - val_loss: 0.1777 - val_accuracy: 0.9497\n",
            "Epoch 318/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0861 - accuracy: 0.9784 - val_loss: 0.1785 - val_accuracy: 0.9481\n",
            "Epoch 319/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0859 - accuracy: 0.9776 - val_loss: 0.1780 - val_accuracy: 0.9492\n",
            "Epoch 320/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0855 - accuracy: 0.9785 - val_loss: 0.1780 - val_accuracy: 0.9493\n",
            "Epoch 321/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0857 - accuracy: 0.9781 - val_loss: 0.1784 - val_accuracy: 0.9489\n",
            "Epoch 322/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0833 - accuracy: 0.9789 - val_loss: 0.1784 - val_accuracy: 0.9496\n",
            "Epoch 323/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0855 - accuracy: 0.9788 - val_loss: 0.1784 - val_accuracy: 0.9491\n",
            "Epoch 324/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0877 - accuracy: 0.9778 - val_loss: 0.1788 - val_accuracy: 0.9490\n",
            "Epoch 325/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0869 - accuracy: 0.9781 - val_loss: 0.1783 - val_accuracy: 0.9491\n",
            "Epoch 326/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0826 - accuracy: 0.9782 - val_loss: 0.1785 - val_accuracy: 0.9495\n",
            "Epoch 327/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0853 - accuracy: 0.9777 - val_loss: 0.1788 - val_accuracy: 0.9484\n",
            "Epoch 328/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0824 - accuracy: 0.9793 - val_loss: 0.1784 - val_accuracy: 0.9489\n",
            "Epoch 329/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0848 - accuracy: 0.9782 - val_loss: 0.1789 - val_accuracy: 0.9492\n",
            "Epoch 330/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0838 - accuracy: 0.9784 - val_loss: 0.1785 - val_accuracy: 0.9496\n",
            "Epoch 331/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0839 - accuracy: 0.9788 - val_loss: 0.1790 - val_accuracy: 0.9495\n",
            "Epoch 332/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0853 - accuracy: 0.9789 - val_loss: 0.1789 - val_accuracy: 0.9495\n",
            "Epoch 333/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0842 - accuracy: 0.9784 - val_loss: 0.1791 - val_accuracy: 0.9493\n",
            "Epoch 334/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0850 - accuracy: 0.9780 - val_loss: 0.1791 - val_accuracy: 0.9495\n",
            "Epoch 335/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0818 - accuracy: 0.9792 - val_loss: 0.1792 - val_accuracy: 0.9496\n",
            "Epoch 336/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0817 - accuracy: 0.9791 - val_loss: 0.1793 - val_accuracy: 0.9491\n",
            "Epoch 337/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0820 - accuracy: 0.9788 - val_loss: 0.1792 - val_accuracy: 0.9497\n",
            "Epoch 338/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0817 - accuracy: 0.9793 - val_loss: 0.1794 - val_accuracy: 0.9493\n",
            "Epoch 339/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0797 - accuracy: 0.9800 - val_loss: 0.1798 - val_accuracy: 0.9499\n",
            "Epoch 340/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0819 - accuracy: 0.9793 - val_loss: 0.1794 - val_accuracy: 0.9495\n",
            "Epoch 341/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0803 - accuracy: 0.9787 - val_loss: 0.1794 - val_accuracy: 0.9493\n",
            "Epoch 342/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0816 - accuracy: 0.9792 - val_loss: 0.1797 - val_accuracy: 0.9498\n",
            "Epoch 343/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0798 - accuracy: 0.9804 - val_loss: 0.1796 - val_accuracy: 0.9498\n",
            "Epoch 344/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0812 - accuracy: 0.9797 - val_loss: 0.1796 - val_accuracy: 0.9492\n",
            "Epoch 345/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0792 - accuracy: 0.9800 - val_loss: 0.1801 - val_accuracy: 0.9495\n",
            "Epoch 346/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0801 - accuracy: 0.9796 - val_loss: 0.1802 - val_accuracy: 0.9496\n",
            "Epoch 347/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0784 - accuracy: 0.9804 - val_loss: 0.1800 - val_accuracy: 0.9496\n",
            "Epoch 348/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0797 - accuracy: 0.9794 - val_loss: 0.1802 - val_accuracy: 0.9495\n",
            "Epoch 349/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0798 - accuracy: 0.9798 - val_loss: 0.1804 - val_accuracy: 0.9499\n",
            "Epoch 350/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0798 - accuracy: 0.9795 - val_loss: 0.1804 - val_accuracy: 0.9491\n",
            "Epoch 351/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0790 - accuracy: 0.9795 - val_loss: 0.1803 - val_accuracy: 0.9490\n",
            "Epoch 352/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0782 - accuracy: 0.9796 - val_loss: 0.1805 - val_accuracy: 0.9498\n",
            "Epoch 353/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0804 - accuracy: 0.9803 - val_loss: 0.1807 - val_accuracy: 0.9491\n",
            "Epoch 354/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0780 - accuracy: 0.9802 - val_loss: 0.1807 - val_accuracy: 0.9490\n",
            "Epoch 355/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0763 - accuracy: 0.9814 - val_loss: 0.1807 - val_accuracy: 0.9492\n",
            "Epoch 356/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0801 - accuracy: 0.9806 - val_loss: 0.1810 - val_accuracy: 0.9491\n",
            "Epoch 357/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0797 - accuracy: 0.9801 - val_loss: 0.1809 - val_accuracy: 0.9496\n",
            "Epoch 358/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0784 - accuracy: 0.9806 - val_loss: 0.1814 - val_accuracy: 0.9488\n",
            "Epoch 359/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0778 - accuracy: 0.9805 - val_loss: 0.1811 - val_accuracy: 0.9491\n",
            "Epoch 360/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0801 - accuracy: 0.9799 - val_loss: 0.1815 - val_accuracy: 0.9494\n",
            "Epoch 361/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0762 - accuracy: 0.9807 - val_loss: 0.1813 - val_accuracy: 0.9491\n",
            "Epoch 362/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0763 - accuracy: 0.9810 - val_loss: 0.1815 - val_accuracy: 0.9493\n",
            "Epoch 363/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0763 - accuracy: 0.9816 - val_loss: 0.1812 - val_accuracy: 0.9493\n",
            "Epoch 364/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0753 - accuracy: 0.9814 - val_loss: 0.1822 - val_accuracy: 0.9488\n",
            "Epoch 365/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0777 - accuracy: 0.9806 - val_loss: 0.1818 - val_accuracy: 0.9482\n",
            "Epoch 366/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0771 - accuracy: 0.9812 - val_loss: 0.1820 - val_accuracy: 0.9493\n",
            "Epoch 367/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0772 - accuracy: 0.9807 - val_loss: 0.1822 - val_accuracy: 0.9494\n",
            "Epoch 368/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0741 - accuracy: 0.9815 - val_loss: 0.1818 - val_accuracy: 0.9492\n",
            "Epoch 369/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0752 - accuracy: 0.9812 - val_loss: 0.1823 - val_accuracy: 0.9493\n",
            "Epoch 370/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0743 - accuracy: 0.9823 - val_loss: 0.1822 - val_accuracy: 0.9488\n",
            "Epoch 371/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0767 - accuracy: 0.9809 - val_loss: 0.1825 - val_accuracy: 0.9489\n",
            "Epoch 372/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0724 - accuracy: 0.9823 - val_loss: 0.1823 - val_accuracy: 0.9485\n",
            "Epoch 373/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0725 - accuracy: 0.9821 - val_loss: 0.1830 - val_accuracy: 0.9493\n",
            "Epoch 374/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0724 - accuracy: 0.9823 - val_loss: 0.1827 - val_accuracy: 0.9488\n",
            "Epoch 375/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0747 - accuracy: 0.9816 - val_loss: 0.1828 - val_accuracy: 0.9487\n",
            "Epoch 376/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0735 - accuracy: 0.9814 - val_loss: 0.1829 - val_accuracy: 0.9482\n",
            "Epoch 377/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0748 - accuracy: 0.9816 - val_loss: 0.1835 - val_accuracy: 0.9487\n",
            "Epoch 378/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0719 - accuracy: 0.9825 - val_loss: 0.1831 - val_accuracy: 0.9485\n",
            "Epoch 379/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0729 - accuracy: 0.9816 - val_loss: 0.1834 - val_accuracy: 0.9485\n",
            "Epoch 380/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0734 - accuracy: 0.9826 - val_loss: 0.1833 - val_accuracy: 0.9488\n",
            "Epoch 381/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0736 - accuracy: 0.9819 - val_loss: 0.1837 - val_accuracy: 0.9492\n",
            "Epoch 382/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0725 - accuracy: 0.9822 - val_loss: 0.1836 - val_accuracy: 0.9488\n",
            "Epoch 383/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0733 - accuracy: 0.9816 - val_loss: 0.1837 - val_accuracy: 0.9481\n",
            "Epoch 384/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0718 - accuracy: 0.9828 - val_loss: 0.1838 - val_accuracy: 0.9487\n",
            "Epoch 385/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0707 - accuracy: 0.9825 - val_loss: 0.1841 - val_accuracy: 0.9486\n",
            "Epoch 386/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0720 - accuracy: 0.9818 - val_loss: 0.1842 - val_accuracy: 0.9485\n",
            "Epoch 387/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0695 - accuracy: 0.9834 - val_loss: 0.1849 - val_accuracy: 0.9486\n",
            "Epoch 388/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0708 - accuracy: 0.9830 - val_loss: 0.1842 - val_accuracy: 0.9488\n",
            "Epoch 389/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0729 - accuracy: 0.9824 - val_loss: 0.1842 - val_accuracy: 0.9486\n",
            "Epoch 390/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0723 - accuracy: 0.9825 - val_loss: 0.1850 - val_accuracy: 0.9492\n",
            "Epoch 391/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0698 - accuracy: 0.9824 - val_loss: 0.1847 - val_accuracy: 0.9480\n",
            "Epoch 392/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0700 - accuracy: 0.9830 - val_loss: 0.1851 - val_accuracy: 0.9487\n",
            "Epoch 393/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0703 - accuracy: 0.9831 - val_loss: 0.1851 - val_accuracy: 0.9485\n",
            "Epoch 394/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0710 - accuracy: 0.9828 - val_loss: 0.1851 - val_accuracy: 0.9483\n",
            "Epoch 395/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0677 - accuracy: 0.9838 - val_loss: 0.1854 - val_accuracy: 0.9483\n",
            "Epoch 396/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0706 - accuracy: 0.9829 - val_loss: 0.1858 - val_accuracy: 0.9488\n",
            "Epoch 397/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0705 - accuracy: 0.9831 - val_loss: 0.1857 - val_accuracy: 0.9484\n",
            "Epoch 398/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0698 - accuracy: 0.9828 - val_loss: 0.1860 - val_accuracy: 0.9488\n",
            "Epoch 399/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0700 - accuracy: 0.9822 - val_loss: 0.1856 - val_accuracy: 0.9484\n",
            "Epoch 400/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0678 - accuracy: 0.9834 - val_loss: 0.1859 - val_accuracy: 0.9486\n",
            "Epoch 401/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0685 - accuracy: 0.9834 - val_loss: 0.1861 - val_accuracy: 0.9485\n",
            "Epoch 402/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0683 - accuracy: 0.9834 - val_loss: 0.1865 - val_accuracy: 0.9483\n",
            "Epoch 403/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0673 - accuracy: 0.9839 - val_loss: 0.1862 - val_accuracy: 0.9487\n",
            "Epoch 404/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0708 - accuracy: 0.9827 - val_loss: 0.1866 - val_accuracy: 0.9488\n",
            "Epoch 405/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0656 - accuracy: 0.9841 - val_loss: 0.1867 - val_accuracy: 0.9486\n",
            "Epoch 406/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0674 - accuracy: 0.9833 - val_loss: 0.1872 - val_accuracy: 0.9488\n",
            "Epoch 407/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0667 - accuracy: 0.9839 - val_loss: 0.1867 - val_accuracy: 0.9484\n",
            "Epoch 408/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0679 - accuracy: 0.9835 - val_loss: 0.1873 - val_accuracy: 0.9481\n",
            "Epoch 409/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0687 - accuracy: 0.9830 - val_loss: 0.1869 - val_accuracy: 0.9487\n",
            "Epoch 410/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0684 - accuracy: 0.9836 - val_loss: 0.1872 - val_accuracy: 0.9484\n",
            "Epoch 411/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0667 - accuracy: 0.9839 - val_loss: 0.1874 - val_accuracy: 0.9483\n",
            "Epoch 412/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0673 - accuracy: 0.9836 - val_loss: 0.1878 - val_accuracy: 0.9486\n",
            "Epoch 413/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0660 - accuracy: 0.9839 - val_loss: 0.1880 - val_accuracy: 0.9481\n",
            "Epoch 414/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0660 - accuracy: 0.9842 - val_loss: 0.1879 - val_accuracy: 0.9484\n",
            "Epoch 415/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0680 - accuracy: 0.9831 - val_loss: 0.1880 - val_accuracy: 0.9481\n",
            "Epoch 416/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0638 - accuracy: 0.9851 - val_loss: 0.1881 - val_accuracy: 0.9483\n",
            "Epoch 417/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0670 - accuracy: 0.9836 - val_loss: 0.1885 - val_accuracy: 0.9484\n",
            "Epoch 418/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0671 - accuracy: 0.9844 - val_loss: 0.1882 - val_accuracy: 0.9480\n",
            "Epoch 419/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0667 - accuracy: 0.9842 - val_loss: 0.1889 - val_accuracy: 0.9492\n",
            "Epoch 420/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0637 - accuracy: 0.9844 - val_loss: 0.1886 - val_accuracy: 0.9477\n",
            "Epoch 421/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0651 - accuracy: 0.9844 - val_loss: 0.1885 - val_accuracy: 0.9482\n",
            "Epoch 422/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0671 - accuracy: 0.9840 - val_loss: 0.1896 - val_accuracy: 0.9483\n",
            "Epoch 423/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0656 - accuracy: 0.9845 - val_loss: 0.1891 - val_accuracy: 0.9480\n",
            "Epoch 424/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0647 - accuracy: 0.9846 - val_loss: 0.1897 - val_accuracy: 0.9482\n",
            "Epoch 425/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0639 - accuracy: 0.9852 - val_loss: 0.1892 - val_accuracy: 0.9483\n",
            "Epoch 426/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0626 - accuracy: 0.9856 - val_loss: 0.1897 - val_accuracy: 0.9482\n",
            "Epoch 427/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0659 - accuracy: 0.9834 - val_loss: 0.1899 - val_accuracy: 0.9479\n",
            "Epoch 428/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0634 - accuracy: 0.9852 - val_loss: 0.1899 - val_accuracy: 0.9473\n",
            "Epoch 429/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0620 - accuracy: 0.9856 - val_loss: 0.1902 - val_accuracy: 0.9481\n",
            "Epoch 430/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0632 - accuracy: 0.9853 - val_loss: 0.1904 - val_accuracy: 0.9479\n",
            "Epoch 431/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0643 - accuracy: 0.9849 - val_loss: 0.1908 - val_accuracy: 0.9479\n",
            "Epoch 432/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0627 - accuracy: 0.9855 - val_loss: 0.1906 - val_accuracy: 0.9473\n",
            "Epoch 433/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0647 - accuracy: 0.9849 - val_loss: 0.1906 - val_accuracy: 0.9481\n",
            "Epoch 434/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0633 - accuracy: 0.9849 - val_loss: 0.1909 - val_accuracy: 0.9478\n",
            "Epoch 435/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0634 - accuracy: 0.9851 - val_loss: 0.1913 - val_accuracy: 0.9482\n",
            "Epoch 436/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0623 - accuracy: 0.9855 - val_loss: 0.1913 - val_accuracy: 0.9473\n",
            "Epoch 437/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0632 - accuracy: 0.9858 - val_loss: 0.1912 - val_accuracy: 0.9478\n",
            "Epoch 438/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0641 - accuracy: 0.9847 - val_loss: 0.1914 - val_accuracy: 0.9475\n",
            "Epoch 439/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0625 - accuracy: 0.9852 - val_loss: 0.1917 - val_accuracy: 0.9477\n",
            "Epoch 440/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0614 - accuracy: 0.9859 - val_loss: 0.1919 - val_accuracy: 0.9479\n",
            "Epoch 441/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0613 - accuracy: 0.9856 - val_loss: 0.1921 - val_accuracy: 0.9485\n",
            "Epoch 442/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0591 - accuracy: 0.9867 - val_loss: 0.1921 - val_accuracy: 0.9480\n",
            "Epoch 443/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0607 - accuracy: 0.9862 - val_loss: 0.1922 - val_accuracy: 0.9480\n",
            "Epoch 444/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0625 - accuracy: 0.9854 - val_loss: 0.1926 - val_accuracy: 0.9480\n",
            "Epoch 445/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0601 - accuracy: 0.9863 - val_loss: 0.1927 - val_accuracy: 0.9481\n",
            "Epoch 446/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0608 - accuracy: 0.9861 - val_loss: 0.1930 - val_accuracy: 0.9482\n",
            "Epoch 447/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0617 - accuracy: 0.9852 - val_loss: 0.1928 - val_accuracy: 0.9473\n",
            "Epoch 448/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0611 - accuracy: 0.9860 - val_loss: 0.1933 - val_accuracy: 0.9481\n",
            "Epoch 449/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0615 - accuracy: 0.9859 - val_loss: 0.1937 - val_accuracy: 0.9482\n",
            "Epoch 450/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0614 - accuracy: 0.9860 - val_loss: 0.1935 - val_accuracy: 0.9476\n",
            "Epoch 451/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0622 - accuracy: 0.9855 - val_loss: 0.1941 - val_accuracy: 0.9480\n",
            "Epoch 452/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0608 - accuracy: 0.9858 - val_loss: 0.1943 - val_accuracy: 0.9479\n",
            "Epoch 453/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0594 - accuracy: 0.9866 - val_loss: 0.1941 - val_accuracy: 0.9480\n",
            "Epoch 454/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0608 - accuracy: 0.9866 - val_loss: 0.1939 - val_accuracy: 0.9479\n",
            "Epoch 455/800\n",
            "12/12 [==============================] - 0s 24ms/step - loss: 0.0601 - accuracy: 0.9863 - val_loss: 0.1943 - val_accuracy: 0.9473\n",
            "Epoch 456/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0575 - accuracy: 0.9868 - val_loss: 0.1943 - val_accuracy: 0.9477\n",
            "Epoch 457/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0591 - accuracy: 0.9864 - val_loss: 0.1951 - val_accuracy: 0.9473\n",
            "Epoch 458/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0595 - accuracy: 0.9862 - val_loss: 0.1951 - val_accuracy: 0.9475\n",
            "Epoch 459/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0608 - accuracy: 0.9855 - val_loss: 0.1954 - val_accuracy: 0.9473\n",
            "Epoch 460/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0589 - accuracy: 0.9866 - val_loss: 0.1953 - val_accuracy: 0.9475\n",
            "Epoch 461/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0585 - accuracy: 0.9868 - val_loss: 0.1954 - val_accuracy: 0.9481\n",
            "Epoch 462/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0612 - accuracy: 0.9858 - val_loss: 0.1959 - val_accuracy: 0.9474\n",
            "Epoch 463/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0569 - accuracy: 0.9869 - val_loss: 0.1957 - val_accuracy: 0.9478\n",
            "Epoch 464/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0584 - accuracy: 0.9864 - val_loss: 0.1957 - val_accuracy: 0.9468\n",
            "Epoch 465/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0587 - accuracy: 0.9865 - val_loss: 0.1965 - val_accuracy: 0.9472\n",
            "Epoch 466/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0587 - accuracy: 0.9865 - val_loss: 0.1963 - val_accuracy: 0.9477\n",
            "Epoch 467/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0581 - accuracy: 0.9869 - val_loss: 0.1967 - val_accuracy: 0.9477\n",
            "Epoch 468/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0564 - accuracy: 0.9872 - val_loss: 0.1970 - val_accuracy: 0.9477\n",
            "Epoch 469/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0590 - accuracy: 0.9863 - val_loss: 0.1969 - val_accuracy: 0.9477\n",
            "Epoch 470/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0558 - accuracy: 0.9877 - val_loss: 0.1971 - val_accuracy: 0.9473\n",
            "Epoch 471/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0559 - accuracy: 0.9876 - val_loss: 0.1973 - val_accuracy: 0.9470\n",
            "Epoch 472/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0595 - accuracy: 0.9864 - val_loss: 0.1975 - val_accuracy: 0.9471\n",
            "Epoch 473/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0576 - accuracy: 0.9868 - val_loss: 0.1978 - val_accuracy: 0.9479\n",
            "Epoch 474/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0605 - accuracy: 0.9864 - val_loss: 0.1978 - val_accuracy: 0.9471\n",
            "Epoch 475/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0588 - accuracy: 0.9868 - val_loss: 0.1984 - val_accuracy: 0.9467\n",
            "Epoch 476/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0567 - accuracy: 0.9874 - val_loss: 0.1982 - val_accuracy: 0.9473\n",
            "Epoch 477/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0575 - accuracy: 0.9874 - val_loss: 0.1985 - val_accuracy: 0.9474\n",
            "Epoch 478/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0551 - accuracy: 0.9877 - val_loss: 0.1992 - val_accuracy: 0.9473\n",
            "Epoch 479/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0563 - accuracy: 0.9876 - val_loss: 0.1992 - val_accuracy: 0.9468\n",
            "Epoch 480/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0557 - accuracy: 0.9875 - val_loss: 0.1995 - val_accuracy: 0.9473\n",
            "Epoch 481/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0572 - accuracy: 0.9872 - val_loss: 0.1997 - val_accuracy: 0.9473\n",
            "Epoch 482/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0550 - accuracy: 0.9881 - val_loss: 0.1997 - val_accuracy: 0.9467\n",
            "Epoch 483/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0577 - accuracy: 0.9874 - val_loss: 0.1993 - val_accuracy: 0.9471\n",
            "Epoch 484/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0553 - accuracy: 0.9882 - val_loss: 0.1994 - val_accuracy: 0.9477\n",
            "Epoch 485/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0561 - accuracy: 0.9876 - val_loss: 0.2004 - val_accuracy: 0.9474\n",
            "Epoch 486/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0557 - accuracy: 0.9876 - val_loss: 0.2003 - val_accuracy: 0.9472\n",
            "Epoch 487/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0559 - accuracy: 0.9878 - val_loss: 0.2004 - val_accuracy: 0.9474\n",
            "Epoch 488/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0540 - accuracy: 0.9882 - val_loss: 0.2010 - val_accuracy: 0.9477\n",
            "Epoch 489/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0552 - accuracy: 0.9873 - val_loss: 0.2011 - val_accuracy: 0.9465\n",
            "Epoch 490/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0557 - accuracy: 0.9880 - val_loss: 0.2012 - val_accuracy: 0.9469\n",
            "Epoch 491/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0545 - accuracy: 0.9877 - val_loss: 0.2014 - val_accuracy: 0.9475\n",
            "Epoch 492/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0540 - accuracy: 0.9874 - val_loss: 0.2013 - val_accuracy: 0.9467\n",
            "Epoch 493/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0540 - accuracy: 0.9881 - val_loss: 0.2018 - val_accuracy: 0.9469\n",
            "Epoch 494/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0544 - accuracy: 0.9881 - val_loss: 0.2017 - val_accuracy: 0.9476\n",
            "Epoch 495/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0534 - accuracy: 0.9882 - val_loss: 0.2025 - val_accuracy: 0.9471\n",
            "Epoch 496/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0552 - accuracy: 0.9876 - val_loss: 0.2022 - val_accuracy: 0.9473\n",
            "Epoch 497/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0536 - accuracy: 0.9881 - val_loss: 0.2031 - val_accuracy: 0.9477\n",
            "Epoch 498/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0535 - accuracy: 0.9884 - val_loss: 0.2025 - val_accuracy: 0.9478\n",
            "Epoch 499/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0547 - accuracy: 0.9878 - val_loss: 0.2027 - val_accuracy: 0.9468\n",
            "Epoch 500/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0540 - accuracy: 0.9885 - val_loss: 0.2035 - val_accuracy: 0.9469\n",
            "Epoch 501/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0528 - accuracy: 0.9880 - val_loss: 0.2033 - val_accuracy: 0.9465\n",
            "Epoch 502/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0522 - accuracy: 0.9882 - val_loss: 0.2037 - val_accuracy: 0.9468\n",
            "Epoch 503/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0546 - accuracy: 0.9876 - val_loss: 0.2035 - val_accuracy: 0.9469\n",
            "Epoch 504/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0536 - accuracy: 0.9882 - val_loss: 0.2044 - val_accuracy: 0.9467\n",
            "Epoch 505/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0553 - accuracy: 0.9874 - val_loss: 0.2039 - val_accuracy: 0.9466\n",
            "Epoch 506/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0530 - accuracy: 0.9880 - val_loss: 0.2044 - val_accuracy: 0.9471\n",
            "Epoch 507/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0511 - accuracy: 0.9888 - val_loss: 0.2048 - val_accuracy: 0.9471\n",
            "Epoch 508/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0517 - accuracy: 0.9887 - val_loss: 0.2047 - val_accuracy: 0.9467\n",
            "Epoch 509/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0512 - accuracy: 0.9881 - val_loss: 0.2052 - val_accuracy: 0.9468\n",
            "Epoch 510/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0516 - accuracy: 0.9885 - val_loss: 0.2049 - val_accuracy: 0.9463\n",
            "Epoch 511/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0537 - accuracy: 0.9881 - val_loss: 0.2052 - val_accuracy: 0.9466\n",
            "Epoch 512/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0506 - accuracy: 0.9886 - val_loss: 0.2058 - val_accuracy: 0.9467\n",
            "Epoch 513/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0515 - accuracy: 0.9885 - val_loss: 0.2059 - val_accuracy: 0.9467\n",
            "Epoch 514/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0511 - accuracy: 0.9888 - val_loss: 0.2061 - val_accuracy: 0.9470\n",
            "Epoch 515/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0528 - accuracy: 0.9880 - val_loss: 0.2060 - val_accuracy: 0.9462\n",
            "Epoch 516/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0528 - accuracy: 0.9888 - val_loss: 0.2069 - val_accuracy: 0.9464\n",
            "Epoch 517/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0522 - accuracy: 0.9888 - val_loss: 0.2073 - val_accuracy: 0.9467\n",
            "Epoch 518/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0544 - accuracy: 0.9884 - val_loss: 0.2072 - val_accuracy: 0.9461\n",
            "Epoch 519/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0526 - accuracy: 0.9886 - val_loss: 0.2071 - val_accuracy: 0.9467\n",
            "Epoch 520/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0525 - accuracy: 0.9884 - val_loss: 0.2076 - val_accuracy: 0.9468\n",
            "Epoch 521/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0509 - accuracy: 0.9891 - val_loss: 0.2075 - val_accuracy: 0.9465\n",
            "Epoch 522/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0515 - accuracy: 0.9891 - val_loss: 0.2075 - val_accuracy: 0.9467\n",
            "Epoch 523/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0504 - accuracy: 0.9886 - val_loss: 0.2084 - val_accuracy: 0.9462\n",
            "Epoch 524/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0503 - accuracy: 0.9889 - val_loss: 0.2081 - val_accuracy: 0.9470\n",
            "Epoch 525/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0511 - accuracy: 0.9891 - val_loss: 0.2087 - val_accuracy: 0.9467\n",
            "Epoch 526/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0497 - accuracy: 0.9888 - val_loss: 0.2089 - val_accuracy: 0.9461\n",
            "Epoch 527/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0493 - accuracy: 0.9898 - val_loss: 0.2082 - val_accuracy: 0.9467\n",
            "Epoch 528/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0482 - accuracy: 0.9896 - val_loss: 0.2098 - val_accuracy: 0.9462\n",
            "Epoch 529/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0496 - accuracy: 0.9897 - val_loss: 0.2095 - val_accuracy: 0.9470\n",
            "Epoch 530/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0501 - accuracy: 0.9892 - val_loss: 0.2094 - val_accuracy: 0.9461\n",
            "Epoch 531/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0492 - accuracy: 0.9896 - val_loss: 0.2101 - val_accuracy: 0.9467\n",
            "Epoch 532/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0484 - accuracy: 0.9897 - val_loss: 0.2106 - val_accuracy: 0.9459\n",
            "Epoch 533/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0495 - accuracy: 0.9895 - val_loss: 0.2103 - val_accuracy: 0.9462\n",
            "Epoch 534/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0485 - accuracy: 0.9900 - val_loss: 0.2108 - val_accuracy: 0.9463\n",
            "Epoch 535/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0504 - accuracy: 0.9899 - val_loss: 0.2109 - val_accuracy: 0.9462\n",
            "Epoch 536/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0493 - accuracy: 0.9898 - val_loss: 0.2109 - val_accuracy: 0.9458\n",
            "Epoch 537/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0482 - accuracy: 0.9900 - val_loss: 0.2110 - val_accuracy: 0.9463\n",
            "Epoch 538/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0521 - accuracy: 0.9888 - val_loss: 0.2114 - val_accuracy: 0.9458\n",
            "Epoch 539/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0501 - accuracy: 0.9898 - val_loss: 0.2118 - val_accuracy: 0.9457\n",
            "Epoch 540/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0488 - accuracy: 0.9901 - val_loss: 0.2116 - val_accuracy: 0.9457\n",
            "Epoch 541/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0488 - accuracy: 0.9900 - val_loss: 0.2122 - val_accuracy: 0.9463\n",
            "Epoch 542/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0486 - accuracy: 0.9892 - val_loss: 0.2122 - val_accuracy: 0.9460\n",
            "Epoch 543/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0488 - accuracy: 0.9899 - val_loss: 0.2128 - val_accuracy: 0.9462\n",
            "Epoch 544/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0488 - accuracy: 0.9898 - val_loss: 0.2129 - val_accuracy: 0.9461\n",
            "Epoch 545/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0459 - accuracy: 0.9902 - val_loss: 0.2130 - val_accuracy: 0.9458\n",
            "Epoch 546/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0484 - accuracy: 0.9897 - val_loss: 0.2130 - val_accuracy: 0.9461\n",
            "Epoch 547/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0482 - accuracy: 0.9898 - val_loss: 0.2139 - val_accuracy: 0.9462\n",
            "Epoch 548/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0477 - accuracy: 0.9902 - val_loss: 0.2138 - val_accuracy: 0.9461\n",
            "Epoch 549/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0476 - accuracy: 0.9900 - val_loss: 0.2143 - val_accuracy: 0.9457\n",
            "Epoch 550/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0466 - accuracy: 0.9907 - val_loss: 0.2143 - val_accuracy: 0.9460\n",
            "Epoch 551/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0477 - accuracy: 0.9902 - val_loss: 0.2146 - val_accuracy: 0.9456\n",
            "Epoch 552/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0457 - accuracy: 0.9898 - val_loss: 0.2146 - val_accuracy: 0.9457\n",
            "Epoch 553/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0463 - accuracy: 0.9906 - val_loss: 0.2150 - val_accuracy: 0.9454\n",
            "Epoch 554/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0490 - accuracy: 0.9900 - val_loss: 0.2158 - val_accuracy: 0.9455\n",
            "Epoch 555/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0467 - accuracy: 0.9901 - val_loss: 0.2154 - val_accuracy: 0.9460\n",
            "Epoch 556/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0475 - accuracy: 0.9901 - val_loss: 0.2156 - val_accuracy: 0.9461\n",
            "Epoch 557/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0475 - accuracy: 0.9904 - val_loss: 0.2159 - val_accuracy: 0.9454\n",
            "Epoch 558/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0461 - accuracy: 0.9903 - val_loss: 0.2159 - val_accuracy: 0.9453\n",
            "Epoch 559/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0457 - accuracy: 0.9906 - val_loss: 0.2169 - val_accuracy: 0.9457\n",
            "Epoch 560/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0458 - accuracy: 0.9906 - val_loss: 0.2167 - val_accuracy: 0.9453\n",
            "Epoch 561/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0428 - accuracy: 0.9913 - val_loss: 0.2173 - val_accuracy: 0.9449\n",
            "Epoch 562/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0465 - accuracy: 0.9905 - val_loss: 0.2169 - val_accuracy: 0.9453\n",
            "Epoch 563/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0456 - accuracy: 0.9909 - val_loss: 0.2174 - val_accuracy: 0.9454\n",
            "Epoch 564/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0449 - accuracy: 0.9908 - val_loss: 0.2181 - val_accuracy: 0.9448\n",
            "Epoch 565/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0445 - accuracy: 0.9912 - val_loss: 0.2178 - val_accuracy: 0.9452\n",
            "Epoch 566/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0444 - accuracy: 0.9910 - val_loss: 0.2187 - val_accuracy: 0.9448\n",
            "Epoch 567/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0441 - accuracy: 0.9906 - val_loss: 0.2183 - val_accuracy: 0.9453\n",
            "Epoch 568/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0476 - accuracy: 0.9901 - val_loss: 0.2189 - val_accuracy: 0.9442\n",
            "Epoch 569/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0452 - accuracy: 0.9905 - val_loss: 0.2188 - val_accuracy: 0.9446\n",
            "Epoch 570/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0436 - accuracy: 0.9914 - val_loss: 0.2199 - val_accuracy: 0.9448\n",
            "Epoch 571/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0453 - accuracy: 0.9907 - val_loss: 0.2193 - val_accuracy: 0.9445\n",
            "Epoch 572/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0430 - accuracy: 0.9918 - val_loss: 0.2195 - val_accuracy: 0.9448\n",
            "Epoch 573/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0443 - accuracy: 0.9907 - val_loss: 0.2202 - val_accuracy: 0.9450\n",
            "Epoch 574/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0456 - accuracy: 0.9905 - val_loss: 0.2202 - val_accuracy: 0.9446\n",
            "Epoch 575/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0443 - accuracy: 0.9911 - val_loss: 0.2205 - val_accuracy: 0.9446\n",
            "Epoch 576/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0471 - accuracy: 0.9906 - val_loss: 0.2208 - val_accuracy: 0.9443\n",
            "Epoch 577/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0443 - accuracy: 0.9904 - val_loss: 0.2208 - val_accuracy: 0.9444\n",
            "Epoch 578/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0422 - accuracy: 0.9914 - val_loss: 0.2220 - val_accuracy: 0.9444\n",
            "Epoch 579/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0432 - accuracy: 0.9910 - val_loss: 0.2212 - val_accuracy: 0.9446\n",
            "Epoch 580/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0413 - accuracy: 0.9918 - val_loss: 0.2216 - val_accuracy: 0.9442\n",
            "Epoch 581/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0423 - accuracy: 0.9921 - val_loss: 0.2221 - val_accuracy: 0.9439\n",
            "Epoch 582/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0449 - accuracy: 0.9909 - val_loss: 0.2227 - val_accuracy: 0.9438\n",
            "Epoch 583/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0439 - accuracy: 0.9910 - val_loss: 0.2225 - val_accuracy: 0.9443\n",
            "Epoch 584/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0438 - accuracy: 0.9911 - val_loss: 0.2228 - val_accuracy: 0.9443\n",
            "Epoch 585/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0433 - accuracy: 0.9913 - val_loss: 0.2228 - val_accuracy: 0.9440\n",
            "Epoch 586/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0417 - accuracy: 0.9917 - val_loss: 0.2230 - val_accuracy: 0.9447\n",
            "Epoch 587/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0426 - accuracy: 0.9917 - val_loss: 0.2232 - val_accuracy: 0.9441\n",
            "Epoch 588/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0427 - accuracy: 0.9916 - val_loss: 0.2243 - val_accuracy: 0.9433\n",
            "Epoch 589/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0437 - accuracy: 0.9912 - val_loss: 0.2241 - val_accuracy: 0.9441\n",
            "Epoch 590/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0441 - accuracy: 0.9911 - val_loss: 0.2242 - val_accuracy: 0.9444\n",
            "Epoch 591/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0430 - accuracy: 0.9913 - val_loss: 0.2244 - val_accuracy: 0.9437\n",
            "Epoch 592/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0435 - accuracy: 0.9913 - val_loss: 0.2255 - val_accuracy: 0.9443\n",
            "Epoch 593/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0442 - accuracy: 0.9907 - val_loss: 0.2246 - val_accuracy: 0.9436\n",
            "Epoch 594/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0426 - accuracy: 0.9909 - val_loss: 0.2253 - val_accuracy: 0.9442\n",
            "Epoch 595/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0416 - accuracy: 0.9919 - val_loss: 0.2257 - val_accuracy: 0.9438\n",
            "Epoch 596/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0426 - accuracy: 0.9915 - val_loss: 0.2255 - val_accuracy: 0.9441\n",
            "Epoch 597/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0411 - accuracy: 0.9920 - val_loss: 0.2260 - val_accuracy: 0.9442\n",
            "Epoch 598/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0426 - accuracy: 0.9915 - val_loss: 0.2263 - val_accuracy: 0.9439\n",
            "Epoch 599/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0420 - accuracy: 0.9915 - val_loss: 0.2265 - val_accuracy: 0.9440\n",
            "Epoch 600/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0431 - accuracy: 0.9914 - val_loss: 0.2270 - val_accuracy: 0.9435\n",
            "Epoch 601/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0416 - accuracy: 0.9918 - val_loss: 0.2272 - val_accuracy: 0.9439\n",
            "Epoch 602/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0411 - accuracy: 0.9915 - val_loss: 0.2275 - val_accuracy: 0.9440\n",
            "Epoch 603/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0414 - accuracy: 0.9915 - val_loss: 0.2277 - val_accuracy: 0.9438\n",
            "Epoch 604/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0413 - accuracy: 0.9915 - val_loss: 0.2280 - val_accuracy: 0.9435\n",
            "Epoch 605/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0437 - accuracy: 0.9915 - val_loss: 0.2285 - val_accuracy: 0.9439\n",
            "Epoch 606/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0425 - accuracy: 0.9915 - val_loss: 0.2285 - val_accuracy: 0.9432\n",
            "Epoch 607/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0405 - accuracy: 0.9921 - val_loss: 0.2294 - val_accuracy: 0.9434\n",
            "Epoch 608/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0405 - accuracy: 0.9923 - val_loss: 0.2289 - val_accuracy: 0.9437\n",
            "Epoch 609/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0421 - accuracy: 0.9915 - val_loss: 0.2286 - val_accuracy: 0.9434\n",
            "Epoch 610/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0411 - accuracy: 0.9916 - val_loss: 0.2291 - val_accuracy: 0.9440\n",
            "Epoch 611/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0425 - accuracy: 0.9920 - val_loss: 0.2301 - val_accuracy: 0.9435\n",
            "Epoch 612/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0400 - accuracy: 0.9923 - val_loss: 0.2301 - val_accuracy: 0.9433\n",
            "Epoch 613/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0428 - accuracy: 0.9917 - val_loss: 0.2302 - val_accuracy: 0.9429\n",
            "Epoch 614/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0401 - accuracy: 0.9920 - val_loss: 0.2302 - val_accuracy: 0.9435\n",
            "Epoch 615/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0415 - accuracy: 0.9913 - val_loss: 0.2309 - val_accuracy: 0.9433\n",
            "Epoch 616/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0403 - accuracy: 0.9922 - val_loss: 0.2310 - val_accuracy: 0.9432\n",
            "Epoch 617/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0415 - accuracy: 0.9919 - val_loss: 0.2314 - val_accuracy: 0.9424\n",
            "Epoch 618/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0383 - accuracy: 0.9925 - val_loss: 0.2310 - val_accuracy: 0.9434\n",
            "Epoch 619/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0414 - accuracy: 0.9921 - val_loss: 0.2323 - val_accuracy: 0.9435\n",
            "Epoch 620/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0391 - accuracy: 0.9927 - val_loss: 0.2322 - val_accuracy: 0.9432\n",
            "Epoch 621/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0401 - accuracy: 0.9924 - val_loss: 0.2320 - val_accuracy: 0.9433\n",
            "Epoch 622/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0397 - accuracy: 0.9922 - val_loss: 0.2325 - val_accuracy: 0.9430\n",
            "Epoch 623/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0413 - accuracy: 0.9920 - val_loss: 0.2329 - val_accuracy: 0.9433\n",
            "Epoch 624/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0404 - accuracy: 0.9923 - val_loss: 0.2329 - val_accuracy: 0.9431\n",
            "Epoch 625/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0421 - accuracy: 0.9919 - val_loss: 0.2335 - val_accuracy: 0.9433\n",
            "Epoch 626/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0383 - accuracy: 0.9926 - val_loss: 0.2332 - val_accuracy: 0.9436\n",
            "Epoch 627/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0391 - accuracy: 0.9925 - val_loss: 0.2352 - val_accuracy: 0.9432\n",
            "Epoch 628/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0390 - accuracy: 0.9927 - val_loss: 0.2340 - val_accuracy: 0.9435\n",
            "Epoch 629/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0389 - accuracy: 0.9924 - val_loss: 0.2342 - val_accuracy: 0.9430\n",
            "Epoch 630/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0395 - accuracy: 0.9925 - val_loss: 0.2343 - val_accuracy: 0.9432\n",
            "Epoch 631/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0392 - accuracy: 0.9927 - val_loss: 0.2357 - val_accuracy: 0.9433\n",
            "Epoch 632/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0387 - accuracy: 0.9929 - val_loss: 0.2354 - val_accuracy: 0.9430\n",
            "Epoch 633/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0385 - accuracy: 0.9926 - val_loss: 0.2353 - val_accuracy: 0.9431\n",
            "Epoch 634/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0383 - accuracy: 0.9928 - val_loss: 0.2356 - val_accuracy: 0.9428\n",
            "Epoch 635/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0374 - accuracy: 0.9926 - val_loss: 0.2356 - val_accuracy: 0.9430\n",
            "Epoch 636/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0402 - accuracy: 0.9926 - val_loss: 0.2362 - val_accuracy: 0.9432\n",
            "Epoch 637/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0387 - accuracy: 0.9929 - val_loss: 0.2364 - val_accuracy: 0.9424\n",
            "Epoch 638/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0396 - accuracy: 0.9926 - val_loss: 0.2371 - val_accuracy: 0.9429\n",
            "Epoch 639/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0394 - accuracy: 0.9923 - val_loss: 0.2370 - val_accuracy: 0.9435\n",
            "Epoch 640/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0364 - accuracy: 0.9931 - val_loss: 0.2369 - val_accuracy: 0.9435\n",
            "Epoch 641/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0383 - accuracy: 0.9927 - val_loss: 0.2379 - val_accuracy: 0.9427\n",
            "Epoch 642/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0386 - accuracy: 0.9930 - val_loss: 0.2382 - val_accuracy: 0.9433\n",
            "Epoch 643/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0385 - accuracy: 0.9930 - val_loss: 0.2381 - val_accuracy: 0.9433\n",
            "Epoch 644/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0376 - accuracy: 0.9932 - val_loss: 0.2381 - val_accuracy: 0.9432\n",
            "Epoch 645/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0369 - accuracy: 0.9931 - val_loss: 0.2385 - val_accuracy: 0.9430\n",
            "Epoch 646/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0403 - accuracy: 0.9925 - val_loss: 0.2391 - val_accuracy: 0.9425\n",
            "Epoch 647/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0391 - accuracy: 0.9925 - val_loss: 0.2391 - val_accuracy: 0.9433\n",
            "Epoch 648/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0374 - accuracy: 0.9933 - val_loss: 0.2391 - val_accuracy: 0.9430\n",
            "Epoch 649/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0372 - accuracy: 0.9930 - val_loss: 0.2393 - val_accuracy: 0.9433\n",
            "Epoch 650/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0374 - accuracy: 0.9929 - val_loss: 0.2398 - val_accuracy: 0.9432\n",
            "Epoch 651/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0372 - accuracy: 0.9928 - val_loss: 0.2401 - val_accuracy: 0.9429\n",
            "Epoch 652/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0364 - accuracy: 0.9930 - val_loss: 0.2408 - val_accuracy: 0.9427\n",
            "Epoch 653/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0375 - accuracy: 0.9934 - val_loss: 0.2403 - val_accuracy: 0.9431\n",
            "Epoch 654/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0383 - accuracy: 0.9931 - val_loss: 0.2411 - val_accuracy: 0.9428\n",
            "Epoch 655/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0381 - accuracy: 0.9927 - val_loss: 0.2408 - val_accuracy: 0.9427\n",
            "Epoch 656/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0370 - accuracy: 0.9930 - val_loss: 0.2411 - val_accuracy: 0.9427\n",
            "Epoch 657/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0363 - accuracy: 0.9934 - val_loss: 0.2422 - val_accuracy: 0.9432\n",
            "Epoch 658/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0375 - accuracy: 0.9930 - val_loss: 0.2425 - val_accuracy: 0.9433\n",
            "Epoch 659/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0361 - accuracy: 0.9934 - val_loss: 0.2417 - val_accuracy: 0.9427\n",
            "Epoch 660/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0375 - accuracy: 0.9928 - val_loss: 0.2425 - val_accuracy: 0.9424\n",
            "Epoch 661/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0344 - accuracy: 0.9938 - val_loss: 0.2428 - val_accuracy: 0.9430\n",
            "Epoch 662/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0358 - accuracy: 0.9934 - val_loss: 0.2429 - val_accuracy: 0.9428\n",
            "Epoch 663/800\n",
            "12/12 [==============================] - 0s 26ms/step - loss: 0.0350 - accuracy: 0.9935 - val_loss: 0.2438 - val_accuracy: 0.9423\n",
            "Epoch 664/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0361 - accuracy: 0.9938 - val_loss: 0.2432 - val_accuracy: 0.9427\n",
            "Epoch 665/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0354 - accuracy: 0.9933 - val_loss: 0.2435 - val_accuracy: 0.9425\n",
            "Epoch 666/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0358 - accuracy: 0.9938 - val_loss: 0.2442 - val_accuracy: 0.9423\n",
            "Epoch 667/800\n",
            "12/12 [==============================] - 0s 25ms/step - loss: 0.0348 - accuracy: 0.9937 - val_loss: 0.2444 - val_accuracy: 0.9422\n",
            "Epoch 668/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0373 - accuracy: 0.9929 - val_loss: 0.2441 - val_accuracy: 0.9420\n",
            "Epoch 669/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0363 - accuracy: 0.9935 - val_loss: 0.2450 - val_accuracy: 0.9426\n",
            "Epoch 670/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0354 - accuracy: 0.9937 - val_loss: 0.2453 - val_accuracy: 0.9423\n",
            "Epoch 671/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0335 - accuracy: 0.9938 - val_loss: 0.2456 - val_accuracy: 0.9427\n",
            "Epoch 672/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0357 - accuracy: 0.9935 - val_loss: 0.2452 - val_accuracy: 0.9425\n",
            "Epoch 673/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0346 - accuracy: 0.9941 - val_loss: 0.2463 - val_accuracy: 0.9425\n",
            "Epoch 674/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0332 - accuracy: 0.9942 - val_loss: 0.2459 - val_accuracy: 0.9424\n",
            "Epoch 675/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0382 - accuracy: 0.9930 - val_loss: 0.2469 - val_accuracy: 0.9423\n",
            "Epoch 676/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0384 - accuracy: 0.9932 - val_loss: 0.2467 - val_accuracy: 0.9417\n",
            "Epoch 677/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0344 - accuracy: 0.9938 - val_loss: 0.2471 - val_accuracy: 0.9426\n",
            "Epoch 678/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0365 - accuracy: 0.9936 - val_loss: 0.2472 - val_accuracy: 0.9425\n",
            "Epoch 679/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0340 - accuracy: 0.9944 - val_loss: 0.2477 - val_accuracy: 0.9426\n",
            "Epoch 680/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0335 - accuracy: 0.9941 - val_loss: 0.2476 - val_accuracy: 0.9419\n",
            "Epoch 681/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0348 - accuracy: 0.9938 - val_loss: 0.2482 - val_accuracy: 0.9419\n",
            "Epoch 682/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0327 - accuracy: 0.9942 - val_loss: 0.2487 - val_accuracy: 0.9417\n",
            "Epoch 683/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0330 - accuracy: 0.9943 - val_loss: 0.2481 - val_accuracy: 0.9421\n",
            "Epoch 684/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0341 - accuracy: 0.9941 - val_loss: 0.2488 - val_accuracy: 0.9425\n",
            "Epoch 685/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0322 - accuracy: 0.9946 - val_loss: 0.2492 - val_accuracy: 0.9419\n",
            "Epoch 686/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0332 - accuracy: 0.9945 - val_loss: 0.2500 - val_accuracy: 0.9417\n",
            "Epoch 687/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0351 - accuracy: 0.9939 - val_loss: 0.2505 - val_accuracy: 0.9415\n",
            "Epoch 688/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0334 - accuracy: 0.9944 - val_loss: 0.2495 - val_accuracy: 0.9420\n",
            "Epoch 689/800\n",
            "12/12 [==============================] - 0s 31ms/step - loss: 0.0355 - accuracy: 0.9940 - val_loss: 0.2509 - val_accuracy: 0.9419\n",
            "Epoch 690/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0336 - accuracy: 0.9940 - val_loss: 0.2504 - val_accuracy: 0.9423\n",
            "Epoch 691/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0337 - accuracy: 0.9945 - val_loss: 0.2508 - val_accuracy: 0.9408\n",
            "Epoch 692/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0339 - accuracy: 0.9940 - val_loss: 0.2508 - val_accuracy: 0.9419\n",
            "Epoch 693/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0327 - accuracy: 0.9944 - val_loss: 0.2513 - val_accuracy: 0.9418\n",
            "Epoch 694/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0346 - accuracy: 0.9939 - val_loss: 0.2518 - val_accuracy: 0.9417\n",
            "Epoch 695/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0310 - accuracy: 0.9946 - val_loss: 0.2512 - val_accuracy: 0.9419\n",
            "Epoch 696/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0334 - accuracy: 0.9942 - val_loss: 0.2523 - val_accuracy: 0.9417\n",
            "Epoch 697/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0343 - accuracy: 0.9939 - val_loss: 0.2523 - val_accuracy: 0.9417\n",
            "Epoch 698/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0321 - accuracy: 0.9946 - val_loss: 0.2527 - val_accuracy: 0.9414\n",
            "Epoch 699/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0346 - accuracy: 0.9942 - val_loss: 0.2529 - val_accuracy: 0.9416\n",
            "Epoch 700/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0364 - accuracy: 0.9938 - val_loss: 0.2529 - val_accuracy: 0.9416\n",
            "Epoch 701/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0325 - accuracy: 0.9945 - val_loss: 0.2529 - val_accuracy: 0.9417\n",
            "Epoch 702/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0343 - accuracy: 0.9945 - val_loss: 0.2538 - val_accuracy: 0.9418\n",
            "Epoch 703/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0315 - accuracy: 0.9945 - val_loss: 0.2542 - val_accuracy: 0.9417\n",
            "Epoch 704/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0353 - accuracy: 0.9938 - val_loss: 0.2546 - val_accuracy: 0.9413\n",
            "Epoch 705/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0336 - accuracy: 0.9940 - val_loss: 0.2542 - val_accuracy: 0.9418\n",
            "Epoch 706/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0333 - accuracy: 0.9947 - val_loss: 0.2545 - val_accuracy: 0.9411\n",
            "Epoch 707/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0327 - accuracy: 0.9946 - val_loss: 0.2548 - val_accuracy: 0.9413\n",
            "Epoch 708/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0311 - accuracy: 0.9949 - val_loss: 0.2552 - val_accuracy: 0.9419\n",
            "Epoch 709/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0318 - accuracy: 0.9946 - val_loss: 0.2561 - val_accuracy: 0.9411\n",
            "Epoch 710/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0327 - accuracy: 0.9945 - val_loss: 0.2559 - val_accuracy: 0.9411\n",
            "Epoch 711/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0316 - accuracy: 0.9947 - val_loss: 0.2559 - val_accuracy: 0.9408\n",
            "Epoch 712/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0315 - accuracy: 0.9950 - val_loss: 0.2563 - val_accuracy: 0.9413\n",
            "Epoch 713/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0311 - accuracy: 0.9948 - val_loss: 0.2566 - val_accuracy: 0.9411\n",
            "Epoch 714/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0319 - accuracy: 0.9949 - val_loss: 0.2569 - val_accuracy: 0.9416\n",
            "Epoch 715/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0329 - accuracy: 0.9944 - val_loss: 0.2570 - val_accuracy: 0.9414\n",
            "Epoch 716/800\n",
            "12/12 [==============================] - 0s 29ms/step - loss: 0.0323 - accuracy: 0.9943 - val_loss: 0.2574 - val_accuracy: 0.9417\n",
            "Epoch 717/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0323 - accuracy: 0.9948 - val_loss: 0.2575 - val_accuracy: 0.9412\n",
            "Epoch 718/800\n",
            "12/12 [==============================] - 0s 28ms/step - loss: 0.0311 - accuracy: 0.9949 - val_loss: 0.2573 - val_accuracy: 0.9412\n",
            "Epoch 719/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0310 - accuracy: 0.9947 - val_loss: 0.2583 - val_accuracy: 0.9410\n",
            "Epoch 720/800\n",
            "12/12 [==============================] - 0s 27ms/step - loss: 0.0320 - accuracy: 0.9948 - val_loss: 0.2582 - val_accuracy: 0.9414\n",
            "Epoch 721/800\n",
            "12/12 [==============================] - 0s 30ms/step - loss: 0.0310 - accuracy: 0.9948 - val_loss: 0.2589 - val_accuracy: 0.9409\n",
            "Epoch 722/800\n",
            " 5/12 [===========>..................] - ETA: 0s - loss: 0.0307 - accuracy: 0.9951"
          ]
        }
      ],
      "source": [
        "training_histories_fixed_iterations = []\n",
        "times_fixed_iterations = []\n",
        "\n",
        "for i in range(len(batch_sizes_2)):\n",
        "  model = Sequential()\n",
        "  model.add(Dense(units = 20, input_dim = num_pixels, kernel_initializer ='normal', activation ='tanh'))\n",
        "  model.add(Dense(units = num_classes, kernel_initializer ='normal', activation ='softmax'))\n",
        "  model.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])\n",
        "  begin = time.time()\n",
        "  training_histories_fixed_iterations.append(model.fit(X_train, y_train, epochs = epochs[i], batch_size = batch_sizes_2[i], validation_split = 0.2))\n",
        "  end = time.time()\n",
        "  times_fixed_iterations.append(end-begin)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Qh-vtuCZn6Av"
      },
      "source": [
        "- (j) Tracer sur un même graphe l'évolution de la précision d'entraînement en fonction des itérations pour chaque taille de batch.\n",
        "\n",
        "- (k) Sur un second graphe, tracer le temps d'entraînement du modèle pour chaque taille de batch."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "background_save": true,
          "base_uri": "https://localhost:8080/",
          "height": 373
        },
        "id": "oIQKUNYSe2Uf",
        "outputId": "66be80d5-79fe-4091-99be-da923eb2e503"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1728x576 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.figure(figsize=(24,8))\n",
        "\n",
        "plt.subplot(121)\n",
        "plt.title('Training accuracy sur 9600 itérations en fonction de la taille des batchs')\n",
        "plt.xlabel('Iterations')\n",
        "plt.ylabel('Accuracy')\n",
        "for i in range(len(batch_sizes_2)):\n",
        "  plt.plot(np.arange(0,9601, 9600/epochs[i]), [0]+training_histories_fixed_iterations[i].history['accuracy'], label='batch_size : {}'.format(batch_sizes_2[i]))\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(122)\n",
        "plt.title(\"Temps d'exécution pour 9600 itérations en fonction de la taille des batch\")\n",
        "plt.xlabel(\"taille des batch\")\n",
        "plt.ylabel(\"temps d'exécution (en secondes)\")\n",
        "plt.bar(np.arange(1,7,1), times_fixed_iterations)\n",
        "plt.xticks(ticks=np.arange(1,7,1), labels=batch_sizes_2);"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Qh1Uy7s6oCUh"
      },
      "source": [
        "On remarque que plus on entraîne notre modèle, plus la précision d'entraînement augmente. On peut alors penser que plus on entraîne notre modèle, meilleure sera la prédiction.\n",
        "\n",
        "Ceci est très loin de la vérité. C'est même un des problèmes les plus classiques en deep learning, qu'on appelle **le surentraînement** (ou **overfitting**)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "59323HlKintj"
      },
      "source": [
        "# <a name=\"overfitting\"></a> III - Illustration du surentraînement"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mOd24PFQp0j1"
      },
      "source": [
        "- (a) Instancier le même modèle que lors de la partie précédente.\n",
        "\n",
        "- (b) Entraîner le modèle sur les données ```X_train``` et ```y_train``` grâce à la méthode ```fit``` :\n",
        ">  - L'entraînement devra se faire sur **100** epochs (paramètre ```epochs```).\n",
        ">\n",
        ">  - Les batchs devront avoir une taille de 20 (paramètre ```batch_size```)\n",
        ">\n",
        ">  - La perfomance du modèle devra être évaluée sur un échantillon de validation contenant 20% des données (paramètre ```validation_split```).\n",
        ">\n",
        ">  - La sortie de l'entraînement devra être stockée dans une variable nommée ```training_history```."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "vE19R7o1Ur97"
      },
      "outputs": [],
      "source": [
        "model = Sequential()\n",
        "model.add(Dense(units = 20, input_dim = num_pixels, kernel_initializer ='normal', activation ='tanh'))\n",
        "model.add(Dense(units = num_classes, kernel_initializer ='normal', activation ='softmax'))\n",
        "\n",
        "model.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "raYxnMdjUyHW",
        "outputId": "87be0ca2-54e1-417c-9784-7459fba08636"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "2400/2400 [==============================] - 4s 1ms/step - loss: 0.8484 - accuracy: 0.8110 - val_loss: 0.2528 - val_accuracy: 0.9282\n",
            "Epoch 2/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.2474 - accuracy: 0.9297 - val_loss: 0.2065 - val_accuracy: 0.9413\n",
            "Epoch 3/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1949 - accuracy: 0.9458 - val_loss: 0.1906 - val_accuracy: 0.9444\n",
            "Epoch 4/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1656 - accuracy: 0.9512 - val_loss: 0.1790 - val_accuracy: 0.9496\n",
            "Epoch 5/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1539 - accuracy: 0.9550 - val_loss: 0.1783 - val_accuracy: 0.9495\n",
            "Epoch 6/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1359 - accuracy: 0.9601 - val_loss: 0.1776 - val_accuracy: 0.9488\n",
            "Epoch 7/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1397 - accuracy: 0.9580 - val_loss: 0.1713 - val_accuracy: 0.9505\n",
            "Epoch 8/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1232 - accuracy: 0.9628 - val_loss: 0.1689 - val_accuracy: 0.9504\n",
            "Epoch 9/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1143 - accuracy: 0.9663 - val_loss: 0.1696 - val_accuracy: 0.9502\n",
            "Epoch 10/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1130 - accuracy: 0.9661 - val_loss: 0.1711 - val_accuracy: 0.9500\n",
            "Epoch 11/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1064 - accuracy: 0.9690 - val_loss: 0.1721 - val_accuracy: 0.9502\n",
            "Epoch 12/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1007 - accuracy: 0.9706 - val_loss: 0.1715 - val_accuracy: 0.9498\n",
            "Epoch 13/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0983 - accuracy: 0.9699 - val_loss: 0.1744 - val_accuracy: 0.9496\n",
            "Epoch 14/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0931 - accuracy: 0.9718 - val_loss: 0.1779 - val_accuracy: 0.9494\n",
            "Epoch 15/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0866 - accuracy: 0.9753 - val_loss: 0.1781 - val_accuracy: 0.9495\n",
            "Epoch 16/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0855 - accuracy: 0.9743 - val_loss: 0.1829 - val_accuracy: 0.9482\n",
            "Epoch 17/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0845 - accuracy: 0.9750 - val_loss: 0.1792 - val_accuracy: 0.9515\n",
            "Epoch 18/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0812 - accuracy: 0.9760 - val_loss: 0.1792 - val_accuracy: 0.9504\n",
            "Epoch 19/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0795 - accuracy: 0.9765 - val_loss: 0.1807 - val_accuracy: 0.9494\n",
            "Epoch 20/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0733 - accuracy: 0.9786 - val_loss: 0.1836 - val_accuracy: 0.9484\n",
            "Epoch 21/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0735 - accuracy: 0.9788 - val_loss: 0.1898 - val_accuracy: 0.9472\n",
            "Epoch 22/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0778 - accuracy: 0.9773 - val_loss: 0.1862 - val_accuracy: 0.9489\n",
            "Epoch 23/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0727 - accuracy: 0.9788 - val_loss: 0.1887 - val_accuracy: 0.9469\n",
            "Epoch 24/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0693 - accuracy: 0.9796 - val_loss: 0.1856 - val_accuracy: 0.9501\n",
            "Epoch 25/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0668 - accuracy: 0.9803 - val_loss: 0.1905 - val_accuracy: 0.9492\n",
            "Epoch 26/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0666 - accuracy: 0.9806 - val_loss: 0.1909 - val_accuracy: 0.9489\n",
            "Epoch 27/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0653 - accuracy: 0.9806 - val_loss: 0.1941 - val_accuracy: 0.9479\n",
            "Epoch 28/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0626 - accuracy: 0.9824 - val_loss: 0.1996 - val_accuracy: 0.9482\n",
            "Epoch 29/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0625 - accuracy: 0.9825 - val_loss: 0.1988 - val_accuracy: 0.9483\n",
            "Epoch 30/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0598 - accuracy: 0.9825 - val_loss: 0.2047 - val_accuracy: 0.9479\n",
            "Epoch 31/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0581 - accuracy: 0.9835 - val_loss: 0.2021 - val_accuracy: 0.9474\n",
            "Epoch 32/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0567 - accuracy: 0.9842 - val_loss: 0.2067 - val_accuracy: 0.9465\n",
            "Epoch 33/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0577 - accuracy: 0.9834 - val_loss: 0.2072 - val_accuracy: 0.9469\n",
            "Epoch 34/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0561 - accuracy: 0.9849 - val_loss: 0.2078 - val_accuracy: 0.9475\n",
            "Epoch 35/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0553 - accuracy: 0.9847 - val_loss: 0.2137 - val_accuracy: 0.9446\n",
            "Epoch 36/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0501 - accuracy: 0.9852 - val_loss: 0.2109 - val_accuracy: 0.9470\n",
            "Epoch 37/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0522 - accuracy: 0.9847 - val_loss: 0.2127 - val_accuracy: 0.9463\n",
            "Epoch 38/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0504 - accuracy: 0.9856 - val_loss: 0.2201 - val_accuracy: 0.9464\n",
            "Epoch 39/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0500 - accuracy: 0.9861 - val_loss: 0.2304 - val_accuracy: 0.9437\n",
            "Epoch 40/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0489 - accuracy: 0.9853 - val_loss: 0.2205 - val_accuracy: 0.9456\n",
            "Epoch 41/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0462 - accuracy: 0.9866 - val_loss: 0.2390 - val_accuracy: 0.9416\n",
            "Epoch 42/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0460 - accuracy: 0.9867 - val_loss: 0.2300 - val_accuracy: 0.9450\n",
            "Epoch 43/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0443 - accuracy: 0.9878 - val_loss: 0.2350 - val_accuracy: 0.9443\n",
            "Epoch 44/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0442 - accuracy: 0.9871 - val_loss: 0.2327 - val_accuracy: 0.9461\n",
            "Epoch 45/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0451 - accuracy: 0.9868 - val_loss: 0.2350 - val_accuracy: 0.9437\n",
            "Epoch 46/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0421 - accuracy: 0.9889 - val_loss: 0.2380 - val_accuracy: 0.9444\n",
            "Epoch 47/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0466 - accuracy: 0.9876 - val_loss: 0.2417 - val_accuracy: 0.9448\n",
            "Epoch 48/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0414 - accuracy: 0.9888 - val_loss: 0.2443 - val_accuracy: 0.9432\n",
            "Epoch 49/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0392 - accuracy: 0.9895 - val_loss: 0.2465 - val_accuracy: 0.9440\n",
            "Epoch 50/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0381 - accuracy: 0.9896 - val_loss: 0.2536 - val_accuracy: 0.9407\n",
            "Epoch 51/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0401 - accuracy: 0.9886 - val_loss: 0.2568 - val_accuracy: 0.9424\n",
            "Epoch 52/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0379 - accuracy: 0.9901 - val_loss: 0.2514 - val_accuracy: 0.9427\n",
            "Epoch 53/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0376 - accuracy: 0.9891 - val_loss: 0.2612 - val_accuracy: 0.9417\n",
            "Epoch 54/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0372 - accuracy: 0.9906 - val_loss: 0.2630 - val_accuracy: 0.9417\n",
            "Epoch 55/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0396 - accuracy: 0.9894 - val_loss: 0.2557 - val_accuracy: 0.9437\n",
            "Epoch 56/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0372 - accuracy: 0.9903 - val_loss: 0.2723 - val_accuracy: 0.9408\n",
            "Epoch 57/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0362 - accuracy: 0.9900 - val_loss: 0.2705 - val_accuracy: 0.9410\n",
            "Epoch 58/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0348 - accuracy: 0.9907 - val_loss: 0.2677 - val_accuracy: 0.9436\n",
            "Epoch 59/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0350 - accuracy: 0.9910 - val_loss: 0.2700 - val_accuracy: 0.9408\n",
            "Epoch 60/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0334 - accuracy: 0.9914 - val_loss: 0.2709 - val_accuracy: 0.9431\n",
            "Epoch 61/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0332 - accuracy: 0.9907 - val_loss: 0.2768 - val_accuracy: 0.9418\n",
            "Epoch 62/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0317 - accuracy: 0.9923 - val_loss: 0.2746 - val_accuracy: 0.9432\n",
            "Epoch 63/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0336 - accuracy: 0.9909 - val_loss: 0.2836 - val_accuracy: 0.9417\n",
            "Epoch 64/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0300 - accuracy: 0.9923 - val_loss: 0.2832 - val_accuracy: 0.9412\n",
            "Epoch 65/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0324 - accuracy: 0.9914 - val_loss: 0.2826 - val_accuracy: 0.9404\n",
            "Epoch 66/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0324 - accuracy: 0.9917 - val_loss: 0.2864 - val_accuracy: 0.9405\n",
            "Epoch 67/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0290 - accuracy: 0.9927 - val_loss: 0.2851 - val_accuracy: 0.9413\n",
            "Epoch 68/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0292 - accuracy: 0.9916 - val_loss: 0.2916 - val_accuracy: 0.9410\n",
            "Epoch 69/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0307 - accuracy: 0.9918 - val_loss: 0.2898 - val_accuracy: 0.9413\n",
            "Epoch 70/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0303 - accuracy: 0.9919 - val_loss: 0.2943 - val_accuracy: 0.9403\n",
            "Epoch 71/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0289 - accuracy: 0.9933 - val_loss: 0.2934 - val_accuracy: 0.9409\n",
            "Epoch 72/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0298 - accuracy: 0.9916 - val_loss: 0.2969 - val_accuracy: 0.9409\n",
            "Epoch 73/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0308 - accuracy: 0.9920 - val_loss: 0.2995 - val_accuracy: 0.9401\n",
            "Epoch 74/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0282 - accuracy: 0.9926 - val_loss: 0.3004 - val_accuracy: 0.9413\n",
            "Epoch 75/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0265 - accuracy: 0.9933 - val_loss: 0.3147 - val_accuracy: 0.9376\n",
            "Epoch 76/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0288 - accuracy: 0.9928 - val_loss: 0.3030 - val_accuracy: 0.9396\n",
            "Epoch 77/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0269 - accuracy: 0.9931 - val_loss: 0.3064 - val_accuracy: 0.9411\n",
            "Epoch 78/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0253 - accuracy: 0.9937 - val_loss: 0.3055 - val_accuracy: 0.9402\n",
            "Epoch 79/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0234 - accuracy: 0.9939 - val_loss: 0.3085 - val_accuracy: 0.9402\n",
            "Epoch 80/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0232 - accuracy: 0.9943 - val_loss: 0.3133 - val_accuracy: 0.9418\n",
            "Epoch 81/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0268 - accuracy: 0.9928 - val_loss: 0.3167 - val_accuracy: 0.9396\n",
            "Epoch 82/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0242 - accuracy: 0.9938 - val_loss: 0.3200 - val_accuracy: 0.9379\n",
            "Epoch 83/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0237 - accuracy: 0.9940 - val_loss: 0.3174 - val_accuracy: 0.9407\n",
            "Epoch 84/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0265 - accuracy: 0.9931 - val_loss: 0.3286 - val_accuracy: 0.9383\n",
            "Epoch 85/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0221 - accuracy: 0.9943 - val_loss: 0.3351 - val_accuracy: 0.9368\n",
            "Epoch 86/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0230 - accuracy: 0.9941 - val_loss: 0.3228 - val_accuracy: 0.9384\n",
            "Epoch 87/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0252 - accuracy: 0.9935 - val_loss: 0.3214 - val_accuracy: 0.9406\n",
            "Epoch 88/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0234 - accuracy: 0.9946 - val_loss: 0.3308 - val_accuracy: 0.9391\n",
            "Epoch 89/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0222 - accuracy: 0.9944 - val_loss: 0.3324 - val_accuracy: 0.9391\n",
            "Epoch 90/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0242 - accuracy: 0.9939 - val_loss: 0.3373 - val_accuracy: 0.9383\n",
            "Epoch 91/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0217 - accuracy: 0.9950 - val_loss: 0.3378 - val_accuracy: 0.9369\n",
            "Epoch 92/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0201 - accuracy: 0.9953 - val_loss: 0.3339 - val_accuracy: 0.9377\n",
            "Epoch 93/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0241 - accuracy: 0.9939 - val_loss: 0.3378 - val_accuracy: 0.9399\n",
            "Epoch 94/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0231 - accuracy: 0.9941 - val_loss: 0.3413 - val_accuracy: 0.9385\n",
            "Epoch 95/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0201 - accuracy: 0.9948 - val_loss: 0.3470 - val_accuracy: 0.9380\n",
            "Epoch 96/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0211 - accuracy: 0.9951 - val_loss: 0.3478 - val_accuracy: 0.9382\n",
            "Epoch 97/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0213 - accuracy: 0.9947 - val_loss: 0.3529 - val_accuracy: 0.9381\n",
            "Epoch 98/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0203 - accuracy: 0.9944 - val_loss: 0.3502 - val_accuracy: 0.9382\n",
            "Epoch 99/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0198 - accuracy: 0.9950 - val_loss: 0.3531 - val_accuracy: 0.9382\n",
            "Epoch 100/100\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0191 - accuracy: 0.9951 - val_loss: 0.3456 - val_accuracy: 0.9399\n"
          ]
        }
      ],
      "source": [
        "training_history = model.fit(X_train, y_train, epochs = 100, batch_size = 20, validation_split = 0.2)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7fUeK0OnqkFn"
      },
      "source": [
        "- (c) Tracer sur un même graphe l'évolution des précisions d'entraînement et de validation. Que remarquez-vous?"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 499
        },
        "id": "DzuYnrmaX8zM",
        "outputId": "ea0cf86d-0a8e-4886-c49f-31439f6a1ecf"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.figure(figsize=(12,8))\n",
        "\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "\n",
        "plt.plot(np.arange(1 , 101, 1), training_history.history['accuracy'], label = 'Training Accuracy', color = 'blue')\n",
        "plt.plot(np.arange(1 , 101, 1), training_history.history['val_accuracy'], label = 'Validation Accuracy', color = 'red')\n",
        "plt.legend();"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yiMhbtA-mG5-"
      },
      "source": [
        "# <a name=\"train_size\"></a> IV - Influence du nombre d'images dans le jeu d'entraînement"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-kvNJ8ySr1AK"
      },
      "source": [
        "- (a) Définir une liste ```train_sizes``` contenant les valeurs ```[10, 200, 1000, 5000, 15000, 60000]```."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "GBYbgfUPpRez"
      },
      "outputs": [],
      "source": [
        "train_sizes = [10, 200, 1000, 5000, 15000, 60000]"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Zpmk11I_uKBl"
      },
      "source": [
        "- (b) Pour chaque élément n de ```train_sizes```, créer un échantillon de taille n de ```X_train``` et de ```y_train```. Stocker ces échantillons dans des listes ```X_train_samples``` et ```y_train_samples```.\n",
        ">  - Attention à ce que les éléments sélectionnés soient les mêmes pour les échantillons de ```X_train``` et ```y_train```."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "TPH_bExXmoVY"
      },
      "outputs": [],
      "source": [
        "X_train_samples = []\n",
        "y_train_samples = []\n",
        "\n",
        "for train_size in train_sizes:\n",
        "  indexes = np.random.randint(X_train.shape[0], size=train_size)\n",
        "  X_train_samples.append(X_train[indexes])\n",
        "  y_train_samples.append(y_train[indexes])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PtUlE1-mxCF_"
      },
      "source": [
        "- (c) Pour chaque jeu d'entraînement, définir et entraîner le modèle que dans les parties précédentes grâce à la méthode ```fit``` :\n",
        ">  - L'entraînement devra se faire sur 10 epochs (paramètre ```epochs```).\n",
        ">  \n",
        ">  - Les batchs devront avoir une taille de 20 (paramètre ```batch_size```)\n",
        ">\n",
        ">  - La perfomance du modèle devra être évaluée sur un échantillon de validation contenant 20% des données (paramètre ```validation_split```).\n",
        ">\n",
        ">  - La sortie des entraînements devra être stockée dans une liste nommée ```training_histories```.\n",
        "  \n",
        "- (d) Stocker dans une liste ```times``` la durée de chaque entraînement."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PMU_JLbBmLtz",
        "outputId": "49de7d49-657c-4a92-997d-426800df36d3"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1/1 [==============================] - 1s 723ms/step - loss: 2.3237 - accuracy: 0.0000e+00 - val_loss: 2.3376 - val_accuracy: 0.0000e+00\n",
            "Epoch 2/10\n",
            "1/1 [==============================] - 0s 44ms/step - loss: 2.2844 - accuracy: 0.1250 - val_loss: 2.3420 - val_accuracy: 0.0000e+00\n",
            "Epoch 3/10\n",
            "1/1 [==============================] - 0s 42ms/step - loss: 2.2458 - accuracy: 0.2500 - val_loss: 2.3457 - val_accuracy: 0.0000e+00\n",
            "Epoch 4/10\n",
            "1/1 [==============================] - 0s 43ms/step - loss: 2.2081 - accuracy: 0.5000 - val_loss: 2.3487 - val_accuracy: 0.0000e+00\n",
            "Epoch 5/10\n",
            "1/1 [==============================] - 0s 40ms/step - loss: 2.1713 - accuracy: 0.5000 - val_loss: 2.3512 - val_accuracy: 0.0000e+00\n",
            "Epoch 6/10\n",
            "1/1 [==============================] - 0s 43ms/step - loss: 2.1357 - accuracy: 0.6250 - val_loss: 2.3531 - val_accuracy: 0.0000e+00\n",
            "Epoch 7/10\n",
            "1/1 [==============================] - 0s 52ms/step - loss: 2.1012 - accuracy: 0.7500 - val_loss: 2.3548 - val_accuracy: 0.0000e+00\n",
            "Epoch 8/10\n",
            "1/1 [==============================] - 0s 47ms/step - loss: 2.0679 - accuracy: 0.7500 - val_loss: 2.3560 - val_accuracy: 0.5000\n",
            "Epoch 9/10\n",
            "1/1 [==============================] - 0s 45ms/step - loss: 2.0357 - accuracy: 0.7500 - val_loss: 2.3570 - val_accuracy: 0.5000\n",
            "Epoch 10/10\n",
            "1/1 [==============================] - 0s 38ms/step - loss: 2.0046 - accuracy: 0.7500 - val_loss: 2.3576 - val_accuracy: 0.5000\n",
            "Epoch 1/10\n",
            "8/8 [==============================] - 1s 26ms/step - loss: 2.2776 - accuracy: 0.1458 - val_loss: 2.2674 - val_accuracy: 0.2500\n",
            "Epoch 2/10\n",
            "8/8 [==============================] - 0s 7ms/step - loss: 2.1871 - accuracy: 0.4169 - val_loss: 2.2125 - val_accuracy: 0.4250\n",
            "Epoch 3/10\n",
            "8/8 [==============================] - 0s 6ms/step - loss: 2.1014 - accuracy: 0.5395 - val_loss: 2.1542 - val_accuracy: 0.5000\n",
            "Epoch 4/10\n",
            "8/8 [==============================] - 0s 7ms/step - loss: 2.0434 - accuracy: 0.7122 - val_loss: 2.0933 - val_accuracy: 0.5750\n",
            "Epoch 5/10\n",
            "8/8 [==============================] - 0s 7ms/step - loss: 1.9350 - accuracy: 0.7566 - val_loss: 2.0318 - val_accuracy: 0.5250\n",
            "Epoch 6/10\n",
            "8/8 [==============================] - 0s 6ms/step - loss: 1.8625 - accuracy: 0.6495 - val_loss: 1.9677 - val_accuracy: 0.5750\n",
            "Epoch 7/10\n",
            "8/8 [==============================] - 0s 7ms/step - loss: 1.7846 - accuracy: 0.6891 - val_loss: 1.9062 - val_accuracy: 0.6250\n",
            "Epoch 8/10\n",
            "8/8 [==============================] - 0s 6ms/step - loss: 1.6839 - accuracy: 0.7697 - val_loss: 1.8415 - val_accuracy: 0.6000\n",
            "Epoch 9/10\n",
            "8/8 [==============================] - 0s 7ms/step - loss: 1.6411 - accuracy: 0.7430 - val_loss: 1.7879 - val_accuracy: 0.6000\n",
            "Epoch 10/10\n",
            "8/8 [==============================] - 0s 7ms/step - loss: 1.5348 - accuracy: 0.7229 - val_loss: 1.7275 - val_accuracy: 0.6500\n",
            "Epoch 1/10\n",
            "40/40 [==============================] - 1s 6ms/step - loss: 2.2345 - accuracy: 0.2787 - val_loss: 2.0685 - val_accuracy: 0.5500\n",
            "Epoch 2/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 1.9255 - accuracy: 0.6863 - val_loss: 1.7514 - val_accuracy: 0.6500\n",
            "Epoch 3/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 1.6000 - accuracy: 0.7141 - val_loss: 1.4774 - val_accuracy: 0.6950\n",
            "Epoch 4/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 1.2910 - accuracy: 0.7843 - val_loss: 1.2607 - val_accuracy: 0.7250\n",
            "Epoch 5/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 1.0465 - accuracy: 0.8222 - val_loss: 1.1127 - val_accuracy: 0.7650\n",
            "Epoch 6/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 0.9183 - accuracy: 0.8403 - val_loss: 0.9916 - val_accuracy: 0.7750\n",
            "Epoch 7/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 0.7896 - accuracy: 0.8699 - val_loss: 0.8884 - val_accuracy: 0.7850\n",
            "Epoch 8/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 0.6900 - accuracy: 0.8727 - val_loss: 0.8141 - val_accuracy: 0.8300\n",
            "Epoch 9/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 0.5708 - accuracy: 0.9054 - val_loss: 0.7664 - val_accuracy: 0.8350\n",
            "Epoch 10/10\n",
            "40/40 [==============================] - 0s 2ms/step - loss: 0.5402 - accuracy: 0.9110 - val_loss: 0.7198 - val_accuracy: 0.8450\n",
            "Epoch 1/10\n",
            "200/200 [==============================] - 1s 2ms/step - loss: 1.9013 - accuracy: 0.5392 - val_loss: 1.0386 - val_accuracy: 0.7660\n",
            "Epoch 2/10\n",
            "200/200 [==============================] - 0s 1ms/step - loss: 0.8643 - accuracy: 0.8239 - val_loss: 0.6501 - val_accuracy: 0.8590\n",
            "Epoch 3/10\n",
            "200/200 [==============================] - 0s 1ms/step - loss: 0.5367 - accuracy: 0.8858 - val_loss: 0.5060 - val_accuracy: 0.8820\n",
            "Epoch 4/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.4174 - accuracy: 0.9098 - val_loss: 0.4426 - val_accuracy: 0.8900\n",
            "Epoch 5/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.3490 - accuracy: 0.9210 - val_loss: 0.3947 - val_accuracy: 0.8960\n",
            "Epoch 6/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.2943 - accuracy: 0.9298 - val_loss: 0.3670 - val_accuracy: 0.9050\n",
            "Epoch 7/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.2551 - accuracy: 0.9436 - val_loss: 0.3521 - val_accuracy: 0.9010\n",
            "Epoch 8/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.2278 - accuracy: 0.9459 - val_loss: 0.3309 - val_accuracy: 0.9040\n",
            "Epoch 9/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.2104 - accuracy: 0.9514 - val_loss: 0.3252 - val_accuracy: 0.9060\n",
            "Epoch 10/10\n",
            "200/200 [==============================] - 0s 2ms/step - loss: 0.1877 - accuracy: 0.9538 - val_loss: 0.3139 - val_accuracy: 0.9070\n",
            "Epoch 1/10\n",
            "600/600 [==============================] - 1s 2ms/step - loss: 1.4659 - accuracy: 0.6697 - val_loss: 0.4589 - val_accuracy: 0.8973\n",
            "Epoch 2/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.3895 - accuracy: 0.9075 - val_loss: 0.3200 - val_accuracy: 0.9127\n",
            "Epoch 3/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.2876 - accuracy: 0.9217 - val_loss: 0.2723 - val_accuracy: 0.9230\n",
            "Epoch 4/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.2279 - accuracy: 0.9375 - val_loss: 0.2454 - val_accuracy: 0.9267\n",
            "Epoch 5/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.1878 - accuracy: 0.9466 - val_loss: 0.2277 - val_accuracy: 0.9323\n",
            "Epoch 6/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.1736 - accuracy: 0.9521 - val_loss: 0.2179 - val_accuracy: 0.9340\n",
            "Epoch 7/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.1498 - accuracy: 0.9593 - val_loss: 0.2094 - val_accuracy: 0.9400\n",
            "Epoch 8/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.1323 - accuracy: 0.9654 - val_loss: 0.2136 - val_accuracy: 0.9347\n",
            "Epoch 9/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.1213 - accuracy: 0.9675 - val_loss: 0.2017 - val_accuracy: 0.9403\n",
            "Epoch 10/10\n",
            "600/600 [==============================] - 1s 1ms/step - loss: 0.1145 - accuracy: 0.9708 - val_loss: 0.1943 - val_accuracy: 0.9433\n",
            "Epoch 1/10\n",
            "2400/2400 [==============================] - 4s 1ms/step - loss: 0.8570 - accuracy: 0.8074 - val_loss: 0.2550 - val_accuracy: 0.9299\n",
            "Epoch 2/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.2388 - accuracy: 0.9354 - val_loss: 0.2037 - val_accuracy: 0.9417\n",
            "Epoch 3/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1883 - accuracy: 0.9465 - val_loss: 0.1844 - val_accuracy: 0.9477\n",
            "Epoch 4/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1593 - accuracy: 0.9542 - val_loss: 0.1628 - val_accuracy: 0.9535\n",
            "Epoch 5/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1374 - accuracy: 0.9590 - val_loss: 0.1519 - val_accuracy: 0.9557\n",
            "Epoch 6/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1270 - accuracy: 0.9639 - val_loss: 0.1448 - val_accuracy: 0.9570\n",
            "Epoch 7/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1056 - accuracy: 0.9701 - val_loss: 0.1404 - val_accuracy: 0.9599\n",
            "Epoch 8/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1008 - accuracy: 0.9713 - val_loss: 0.1361 - val_accuracy: 0.9595\n",
            "Epoch 9/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0915 - accuracy: 0.9745 - val_loss: 0.1323 - val_accuracy: 0.9618\n",
            "Epoch 10/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.0858 - accuracy: 0.9762 - val_loss: 0.1271 - val_accuracy: 0.9625\n"
          ]
        }
      ],
      "source": [
        "training_histories = []\n",
        "times = []\n",
        "\n",
        "for i in range(len(train_sizes)):\n",
        "  model = Sequential()\n",
        "  model.add(Dense(units = 20, input_dim = num_pixels, kernel_initializer ='normal', activation ='tanh'))\n",
        "  model.add(Dense(units = num_classes, kernel_initializer ='normal', activation ='softmax'))\n",
        "  model.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])\n",
        "  begin = time.time()\n",
        "  training_histories.append(model.fit(X_train_samples[i], y_train_samples[i], epochs = 10, batch_size = 20, validation_split = 0.2))\n",
        "  end = time.time()\n",
        "  times.append(end-begin)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LiTc9MMJx0Mi"
      },
      "source": [
        "- (e) Tracer sur un même graphe l'évolution de la précision d'entraînement en fonction des epochs pour chaque jeu d'entraînement.\n",
        "\n",
        "- (f) Sur un deuxième graphe, tracer l'évolution de la précision de validation en fonction des epochs pour chaque jeu d'entraînement.\n",
        "\n",
        "- (g) Sur un dernier graphe, tracer le temps d'entraînement du modèle pour chaque jeu d'entraînement."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 405
        },
        "id": "T4iGmA-8xWyD",
        "outputId": "7f55a210-6a93-4999-9862-0804fe8c6c83"
      },
      "outputs": [
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABXUAAAGECAYAAAB56sP2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd5wcd33/8ddne79TPZ3KSXKRXLAtF1wAC2Nj4wI4kAQILUDAAUKCA06ABAIkhBYIkODQg+mO80sgxLYsGxeIMc1yEbaxjbEsySqn4ut322a+vz++39md3ds73UlXdNLn+XjMY/rsd2fvZnbe+53viDEGpZRSSimllFJKKaWUUnNDZLYLoJRSSimllFJKKaWUUmriNNRVSimllFJKKaWUUkqpOURDXaWUUkoppZRSSimllJpDNNRVSimllFJKKaWUUkqpOURDXaWUUkoppZRSSimllJpDNNRVSimllFJKKaWUUkqpOURDXaUOkohsEJE/nupllVJKqSOZiBgROc4Nf1FEPjCRZQ/idV4jIrcebDmVUupoISIXiMjT07TtiIj8j4hcNU3b/xsR+ep0bHuqiPV1EekRkV/O4Ot2icigiERn4LXuEpE3T/frHIrp/Dsf4/XeICJ3jzHviMoHROQ6EfnIbJejmYhkROQpEfncbJdlumioq44q7qQWdL6IjITGXzOZbRljLjPGfGOql1VKKaUOZyJyi4j8fYvpV4rIbhGJTXRbxpi3GmP+YQrKtMoFwLXXNsZ8xxhzyaFuWymlJmIqrzNmmwtBVk3R5j4C3GGM+fKhbqhVKGeM+agx5rAOE4HnARcDy40xZ0/Xi7jP7YXBuDFmmzEmZ4zxpus11cSJiAmGj+R8wP2f3jVF2zroH/ed9wAfA04RkROnokwzYTIh+YS/dCt1JDDG5IJhEXkKeLMx5kfNy4lIzBhTncmyzUW6n5RS6qj0DeAfReSDxhgTmv464Dt6Xpheeu5V6vA00euMo40x5m9muwyHgZXAU8aYodkuiJo4EYlqIH54G+87kYgIsMsY8yV359YJwG9mtIAzQGvqKkX9V18ReY+I7Aa+LiLzRORGEdnrbpW5UUSWh9ap3WIS3FohIp9yy24RkcsOctnVIvITERkQkR+JyLUi8u0xyn2gMs53t/rsdPN/EJp3pYg8ICL9IvI7EbnUTW/4hVdEPhS8fqgm1J+IyDbgDjf9P13trD5X9pND66dF5NMistXNv9tNu0lE/rzp/WwWkZdN9vNTSik1o34ALADODyaIyDzgxcA3ReRsEfmZiPSKyC4R+byIJFptqLkmgoj8lVtnp4i8qWnZK0Tkfnfe2i4iHwrN/onr97pacedJ022PIvIcEfmVOxf9SkSeE5p3l4j8g4j81J1/bxWRhWOUWc+9SqkJE9v8wHvd//x+EblBROa7ecH/9xvdca1HRN4qIs92/5u9IvL50Lbe4I5Tn3f/24+KyEVN8590x7EtMkYNYXc8uM693iPAs8cpf1Lsdcs2EekW22xO2s27WUQ+HVr2ehH5dzfcJiJfc8f0HSLyEQk1AyAibxGR37iyPiIiZ7jpDTXzgvOEiGSBDcBSqdeAXho+XrrlXyoiD7t9d5eEaue5Y+01bt/2ich/iEhqnPf+JlfGHhHZKCIrQ/OM+6x+617rWhGRFtv4E+CrwHmuzB8Ovf8nROQZEfmhiCyd6LZb7TsR+RbQBfyve52/lqa7WNz++qF7zSdE5C2hbX7I/W1+0233YRE5a5x9c7H7++tzf6PStK3wZzLqbpqmbY37uUxgX73d7asBsefyY0XkHrHn2huk6TuI2CY79rnXfU1o+nUi8gX3dz0EvMDts/8Se87fIiJ/Mc4+WeDK1y+2mY1jx1m2OR/4qYh8xn3eT4r9zvIGsceFPRJqqkHG/z6EiLxe7Pl/v4h8QELfMWT841FKRL7tpveK/a7UMUb5TxeR+9w+/w9gvP+jE0TkNvf5PSYir2ja59eK/X4yICK/EJFj3bzgu92D7m/6lTKJ3MZYXxRba/giY8wGOXAWM+Zx6yA+p/GOncH7eLdbb5eIvNHNuwp4DfDX7n3/71j7FjTUVSpsCTAf+0vqVdj/j6+78S5gBPj8mGvDOcBjwELgk8DXREaf2Cew7HeBX2IvmD+Erfk0lgOV8VtABjgZWAx8BkBEzga+CfwV0A6sB54a53WaPR84EXiRG98AHO9e4z7gO6FlPwWcCTwHu3//GvCxNb1eGywkIqcBy4CbJlEOpZRSM8wYMwLcALw+NPkVwKPGmAcBD/hL7DnuPOAi4O0H2q7YgPMa7C2qxwMvbFpkyL1mO3AF8DYR+T03b73rt7tbTX/WtO352PPLv2DPr/8M3CQiC0KLvRp4I/ZclnBlaUXPvUqpyfhz4Pew/8NLgR7g2qZlzsH+P78S+Czwt9hj4MnAK0Tk+U3L/g57jP0g8N9if0zKYo9xlxlj8tj//wfGKNMHsYHTsdhjSkPbnsaYVcaYp9zox4E1wDrgOOwx4+/cvDcBrxORC8WGY2cD73TzrgOqbp3TgUuAIMT6Q+x1zuuBAvBSYP8YZQ3KNARcBux0x/mcMWZneBkRWQN8D7gaWATcjA05w6HeK4BLgdXAqcAbWr2eiFwJ/A3wcret/3PbDnsxNhA/1W33RU3zMcZ8DXgr8DNX5g+KyIXYW8JfAXQCW4HrJ7LtsfadMeZ1wDbgJe51PtnibV0PPI39O/wD4KOuLIGXumXagR8yxrWv2B89/xt4P/bv8HfAc1stOwktP5cJ7qsXYc9552LPd1/GnutWAM8C/ii07BJX5mXYv/svi8ja0PxXA/8I5IF7gP8FHnTLXwRcLSKjPmfnWqDoyvkm19UYY8bKBsD+X2/Gfkf5rnuPz8b+/7wW+LyIBHcDjPl9SEROAv4NGwp2Am2u7IHxjkd/7JZf4crxVux3nAbu/+kH2O8784H/BH4/9D7vMsZc4JbNAre597QYeBXwb66cgVcBHwbmAU9g9z/GmOC73Wnub/o/3Ph05jbXMcZxK7TuRD+n8Y6dwfsIPp8/Aa4VkXmuqZrvAJ907/sl47wXMMZop91R2WEvpF7ohi8AykBqnOXXAT2h8buwt1WBPek8EZqXAQywZDLLYg9CVSATmv9t4NsTfE+1MmIP4j4wr8VyXwI+c6D94sY/FLw+sMqV9ZhxytDulmnDHmBHsAfi5uVS2JPI8W78U8C/zfbfhXbaaaeddgfusO0D9gbnTeCnwF+OsezVwPdD4wY4zg1fB3zEDf878PHQcmvCy7bY7meDc1no/BQLzX8DcLcbfh3wy6b1fwa8wQ3fBbw/NO/twC0T3Bd67tVOO+0aOhqvM36DrSUWzOsEKtimEIP/72Wh+fuBV4bG/wu42g2/AdgJSGj+L90xLuuOy78PpA9QvieBS0PjVwFPt1hOsAHSsaFp5wFbQuO/D2wH9gHPc9M6gFK4HNhg7U43vBF45xhlazju03ieuKC5nE3Hyw8AN4TmRYAdwAWhz+W1ofmfBL44Rjk2AH/StK1hYGWonM8Lzb8BeO8Y23oD7nzkxr+GDWyC8Zz7m1h1oG0fYN/V/u7cePD3FcMGdR6QD83/GHBdaD/+KDTvJGBkjNd5PfDzpr+Tp6lf79Y+k+ZyjFPulp/LBPfVc0PzNwHvCY1/Gvhs6O+nCmSb9u0HQn9r3wzNOwfY1lTW9wFfb/Eeoq5cJ4SmfTT8uTctfxeN+cBvQ/NOce+rIzRtP7BujG2Fvw/9HfC90LwMNueYyPHoTdgg+9RWrxNaZz2jj0P34P5Pm5Z9JfB/TdO+BHwwtM+/Gpp3ObaSwFjHgwuYptyGAx+3Jvw5cYBjp3sfIzR+b90DnBvaL6P2Z6tOa+oqVbfXGFMMRsQ+KfFL7taFfuytne0y9tNDdwcDxphhN5ib5LJLgWdC08B+SWrpAGVc4bbV02LVFdhfVA9WrUwiEhWRj7tbOPqp1zpa6LpUq9dy+/o/gNeKSAR7wPzWIZRJKaXUDDHG3I29gP89d5vc2dgaC4jIGnfr2253Xvgo9nxwIEtpPOdtDc8UkXNE5E53e10ftgbJRLYbbHtr07StNNZe2R0aHmaMc7iee5VSk7QS+L67XbcXG6p42AAh0B0aHmkxHj4e7TDuqt/ZCiw1tibrK7HHxl1ib2c+YYwyjXu8DVmEDT02hcp/i5se+F9soPWYOzcE7znuyhGs9yVsTT049OPhWBqO9cYYH/s+J32sx76Hz4XK/ww2qDmYbR2onIPYMGgi2z7YfRdcZw6Eph3oPJiS1k0mNPz9uL/HMa9ZJ2is9zuRfTWZ/58e09i28Vb3GoHw+1iJbe6jN/R38Dc0/u8GFmGD0Yn8X7XSXGaMMS3fxwG+DzV/NsM01oIf73j0LeyPBteLbULqkyISb1HWpbQ+DrWyEjinaR++BhuiBib7fzRduc2Bjlsw8c9pIsfO/aaxPeDJHENqNNRVqs40jb8bWAucY4wpUL+1c7zbJg7VLmC+iGRC01aMs/x4ZdzuttXeYr3tjN3GzxD2ABRY0mKZ8L56NXAl9haxNuwvsUEZ9mFvQRnrtb6BPahfBAybpttllVJKHda+ia2t81pgY+hL7ReAR7G1QQvYC6CJnDt30XjO62qa/13s7aArjDFtwBdD220+hzfbif2yHtaFrcE1WXruVUpNxnZskwjtoS5ljDmY4w/AsqYm3rqwxziMMRuNMRdja989CnxljG0c6Hgb2IcNKU4Olb3NhB4Kh71V+jdAp4gEt7lvx9Z4Wxhar2CMOTk0f6xj1DBjHw8ndax3+2kFB3es3w78adPnljbG3HMQ2zpQObPY27knUs7x9t14+2cn9vyUD0072PNgw99PaD8HJnJOm6hD2VetzHPbCNT+f5zwPtyOrVkZ/hvIG2Mub7HdvdhawBP5vzpU430f2gWE2/lPY/dXYMzjkTGmYoz5sDHmJGzzLS+msamtwC5aH4da2Q78uOn1csaYtx3E+w5MV25zoOPWZEzk2DmeAx3rajTUVWpseew/Yq9ri++D0/2CxpitwL3Ah0QkISLnAeO1oTJmGY0xu7C3Df2b2MbD4yISHOC+BrxRRC4S21j6stAv+Q8Ar3LLn4Vtb2k8eezBbz/25P3RUBl87O20/yy2kfmo2IfXJN38n2FvU/00WlNIKaXmmm9iQ8W3YIPCQB7oBwbduWWiX9xvAN4gIie5Hzebz7t5bC2jotj2aV8dmrcXez45Zoxt3wysEZFXi0hMRF6JvbX0xgmWrbkceu5VSk3UF4F/FPeQLRFZJLa91oO1GPgLd7z4Q2xb2zeLSIfYhzFmsceHQez/eis3AO9zx6nl2HY2R3HHk68AnxGRxa78y8S1KeqOb2/EBj9/DPyriCxzx8JbgU+LSMEd846VetvAXwWuEZEzxTpO6g8hewB4tTt2XYpt+zPQDSwQkbZx3tcV7jgbx4Y9Jeyt4ZP1Rew+Otm91za3v6fC97Dng3Xu2PxR4Bem3o7xeMbbd92McR40xmzH7oePiX0g1qnYdjxbPpD7AG4CThaRl4utyfsXNAa3DwDrRaTLfVbvO4jXCBzKvhrLh9219vnY4PI/x1jul8CA2Idypd3f5LNEZNSDBY0xHrad4Q+5mqMn0dRW9RQa7/vQ/wNeIvYBXglsUxjhcHPM45GIvEBEThFbw7Uf2yxDq2PIz7ABdnAcejn2jq1WbsR+/3qdWzYu9kGQJ46xfLMx/6ZDpiS3mcBxazLbGvfYOQETed+AhrpKjeezQBr7K8vPsdXlZ8JrsO2t7Ac+gr1NsjTGsgcq4+uwB+NHsW20XA1gjPkl9gvYZ4A+4MfUfwH9APbX3x5sg+XfPUB5v4m93WIH8IgrR9g1wK+BX2FvW/oEjceeb2LbozmYLxRKKaVmibugugfbjuMPQ7OuwV5gDGC/0P7HqJVbb28D9rx2B/ZBGXc0LfJ24O9FZADbZtwNoXWHsbXFfir2Nrdzm7a9H3vh9m7s+fWvgRcbY/ZNpGxN9NyrlJqMz2GPkbe649fPsW11HqxfYB+qtg973PsDd4yLAO/C1jp8BhuGjvWj2oexx5At2BBjvB943oM9Jv9c7K3NPwLWikgBeyx5h6vl93/YH6++LiKCDXoT2GNUDzZs6gQwxvynK/t3seeKH2AffAT2QWsvwbYP/Bo3D7feo9iQ70l3rA/fNo8x5jHs3SP/6vbPS7APDiuP8/5aMsZ8H3vsvN6974ewD2o7ZMaYH2GP+/+FrfV4LPZhURNZd7x99zHg/W7ftHrY5x9h7+zYCXwf267pjw6i/PuAP8Q+CGo/9u/xp6H5t2HP/ZuxbdwezA+owbYOel+NYTf273En9mFUb3V/V61e28N+d1iH/V/Zhw3Vx/pR4R3Y2+d3Y9tE/fohlHM8430fehj7I8312P01iP0uEuQJ4x2PlmD/T/uxte9/TItjg/t/ejm2jdlnsM2+/HergrrmPi7BfmY7sfvmE0Bygu/1Q8A33N/0K8ZYZipzmzGPWweh5bFzgut+DTjJve8fjLegNDaDoZQ63IjIf2AbC5/2msKzQUReD1xljHnebJdFKaWUOhrouVepuUlE3oB94I/+7yqlDkhEctgfSI43xmyZ7fKoqac1dZU6zLjbEY511f0vxbaZN+6vM3OV2Ntr3w58ebbLopRSSh0N9NyrlFJKHblE5CWuCYgs8CnsnTtPzW6p1HTRUFepw88S4C7srRL/ArzNGHP/rJZoGrj2ZPZi24s50G2mSimllDpEeu5VSimljnhXYps62IltGuNVRm/RP2Jp8wtKKaWUUkoppZRSSik1h2hNXaWUUkoppZRSSimllJpDNNRVSiml5hgR+XcR2SMiD40xX0TkX0TkCRHZLCJnzHQZlVJKKaWUUkpNn9hsF2CyFi5caFatWjXbxVBKKXWY2rRp0z5jzKLZLsc0uw74PPDNMeZfhm1D63jgHOALrj8uPccqpZQaz1Fyjp1yen5VSik1noM9v865UHfVqlXce++9s10MpZRShykR2TrbZZhuxpifiMiqcRa5EvimeyjCz0WkXUQ6jTG7xtuunmOVUkqN52g4x04HPb8qpZQaz8GeX7X5BaWUUurIswzYHhp/2k1TSiml1ASJSEpEfikiD4rIwyLyYTf9OhHZIiIPuG7dbJdVKaXU0WfaQl1t708ppZQ6/InIVSJyr4jcu3fv3tkujlJKKXU4KQEXGmNOA9YBl4rIuW7eXxlj1rnugdkrolJKqaPVdNbUvQ64dJz54fb+rsK296eUUkqpQ7cDWBEaX+6mjWKM+bIx5ixjzFmLFmkziUoppVTAWINuNO46M4tFUkoppWqmLdQ1xvwEeGacRWrt/Rljfg60i0jndJVHKaWUOor8EHi9uyvmXKDvQO3pKqWUUmo0EYmKyAPAHuA2Y8wv3Kx/dHecfkZEkrNYRKWUUkep2WxTV9v7U0oppQ6CiHwP+BmwVkSeFpE/EZG3ishb3SI3A08CTwBfAd4+S0VVSiml5jRjjGeMWYe96+VsEXkW8D7gBODZwHzgPc3rafNGSimlpltstgswESJyFbaJBrq6uma5NEoppdTsMsb80QHmG+DPZqg4Siml1BHPGNMrIncClxpjPuUml0Tk68A1LZb/MvBlgLPOOkubbFBKKTXlZrOmrrb3p5RSSimllFLqsCQii0Sk3Q2ngYuBR4NmA0VEgN8DWj4cXCmllJpOs1lT94fAO0TkeuActL0/pZRSSimllFKHj07gGyISxVaIusEYc6OI3CEiiwABHgDeOt5GlFJKqekwbaGua+/vAmChiDwNfBD7tFCMMV/Etvd3Oba9v2HgjdNVFqWUUkoppZRSajKMMZuB01tMv3AWiqOUUko1mLZQV9v7U0oppZRSSimllFJKqak3m23qKqWUUkoppZRSSimllJokDXWVUkoppZRSSimllFJqDpnNB6UppZSaAOP7eJ6HX624fhWvWsGrVvGrHl61YqcFy1Srdp7n+tUqnmeXNb6PMb7rG3zfDbvxSc333XzTNN9vWn+C8+d1LuPyd7x7tne3UkqpOSY4v/ieh+9V8T3fngO9Ksbza+dA37fnUNtvXtbDeJ5bx7Pn1NB8u55fX9/z8KpVjO/hufMrGFsWY8AYbM8Hg53n+26weTnXx2B847YTrGvsezzYdd3w6Ze+hJPWazOwSiml1JFEQ12l1FHF973GMDQUkgYXaL4LRRuC0VEBaRWvWg9Rg3Ubt1MZc3q9DGPMC61nLxRnmAiRSAQRQSJR148gkfp4w/yImy+N8wn1RSJAU18EiCGRCL4Xn/n3qZRSUyD8o5fv+zDGj17hc0xwXhl97K+0ON9Ux1i3UjuXjPmj3qjXav36vufONQIiYgdwh2kRxM4IJtfGRQgta+fXlxVqc0Xcdqn17XapLSv1jdfKELy+/SHRawxXPa8W5M40iUSIRmNEYlEikShSO9e59xnuI0ik/n7s+S/YX5GG/TPRdYPzarD/JRI+38qo7UXjiRnfR0oppZSaXhrqKqUOme97eJVKw8VicLHpVSpNF6QVvOBCsmGd+sWoV6k0XbCGtzn64rfhgjYIab3RF65+1bM1V6aJSIRoLEbEddFQF4nWp0With9Ppuy0FvOisdHTIs3bi0YRiYJEgSgQQSSKIYIxESCC8V3fRDBGMJ5gDPgeGCN4nsH4gu8bfM9eNHseGM/gewbP8/GrBq/q22WqPl7VzvO9YNjH89w8N92vGryyj+8ZV0PpwGLp/LR9NkqpQ2eMcTUWvVpNxXCo1jjePM/VdgxqSIbDuXHXd9PDNSSb1w/VnBz77oLG0LXVnQOTmd8c2s60SDQ66vzR+nwRJ5ZIEI1liMTibpko0WA4dL6JRKLBJ21rf9oPvaF2aCCoHRrUCq0vG9QMtfPc5NCy9XOCXTZYP/yaofHg9UPvOxykRtz7qXWRYL9EkWDZ8Pxo1J2P3bKh4WjM/sAYbdhm83AkdP6VKf5UlVJKqaPLqvfeNNtFOGRPffyKWX19DXWVOgIEF9rVcplKqUS1VKJStv1quT5cKZWolst2WjDP9YN1gqC1oaZQOEythGoCuf50BKXBBWssFncXnXEiMXchGo0SqY3HiKdSJKPRxgvahgtbu3w0uAiOxd1F7egL4IaLZPc64Yvg4DUlYkNU40XxfPCqYoPMio/ngs/6sE81NFybXnHLhadVfcrDzesZvIo3alnfM4DvusoU7HMhEosQjQqRqBCNRey0aIRoTNzFrJ0eSwQXvuH5wfrBcq3XHWubyYyekpSaCsb3qZRLVIpF25WKlF2/UhyhUipRKY7YabXp9Xnl0DLN62Mm+CvNFBGJNARpYwVskUgEcfOa7xiIRGONdxw031Hgpk3k7oPw3Qp2vpvm+pFIdPSdDpH6svXzSXBeG+sHvfr5prZctH4uk4g+FkMppZRS6minV9BKTZPgNstqpWyD0EqFaqViA9SmYLUexhZDYWy5MYytlEPBbHhdO3wwNYUkEiGeTBFPJoklEsQSth9cZCZS6VEXnvaCM04sHqsFnvWaQm483lgDKFg/1mI74eVr06bogtX3DdWyR6Xkub7fOF72qJZ8KiWP4qBHtRRM86iUfbdcsbZufZ7tH0q2IREbekbjEaIx28XiERuKuuF4MkoqG29YZvSwtFx31PLxIFBtEbLGhEhEtNaRUjPM+L4NVIMgtVgPV8u1oLVFKDtqWjiYdeHrJMQSSeLJJPFUmngySSKVJp5Kks4vtueIVIpEKkU8mSIaT4xRizHaUDtSGsYba00GtSElEiUai7p+qJZkJNp4S7tSSimllFJqFA111RHH9z28coVqtYJXLuNVK1TLlVpTAEHIWnVBq1cpu+HQ9GqFqlvXc+uOWia0La8y+vUOpX23SDRmg9ZkkrgLWoPhTFtb47RkkliiMZgda91YU4Abjc3uIcDzfKolj2rZpzRi+9XyMNWyHa6UvaZhvyl4HSeoLXt4lckF3RIR4okIsWSUeCJa68eTUdL5BDE3bOfZ0LVhWsKGqC0D1nDQ6kJVpdTcYIyxP6AVXZBaHKmHqs0h66iasCNNoW19vFoqTaocsSBwDQJYF7Rm29pr4avth8NZF8omU8RSqYb1E6kUsWQydNu9UkoppZRSaq7QUFcddowxVIojFIeGKA0NUhoaojg0SDE0XBoepDQ4SHF49DKTvUgeSyyesDVIXReLx4nGE659OttGXSqXs7VTa8skQsva5WPxuF0mEScWixNNJIgnU8QSCRe0BsFsslZTdrbDVmPsbf42ZA0CU78xYK14Y8634arvlqkHr+H51bJto3WyorGIDVRdmBoLha5xN310KBsKaoPxpvXjiSiRmNZWnW6mXMYfHq53Q0O1YUmmyD3vubNdRHWEqZSKDPf1MtTbw1BfLyN9ffWAdbxg9hCaHbDHdVe7NRSiZtvaG8br84Nlxw5l44mk1lpV6jBnjHHtz5taM0m2H2qDPmiXvurX261v7o+3bnV0O/aN2261vM+ZL1rJs56/fLZ3kVJKKaWmkIa6alp41SqlFoFrPZQdaghli4M2qA2C3AM1JZDMZElmsySzOVKZLO1LlpLMZkllcyTSGVsTNQhhG0JWF7zWHhwSI5oIj9vANhKNzclwzxhDteJTHqm6zqM8UqU0UqVcrDZMLzWMB8vUa7xO9OFWNYINSBMRYnFbazXmaq+mMjFi7Uk7LRkl3jS/tl5T6Nq8nVgiSiQy9z6XucpUKmMGsP5Q03jz/OEhu8ywnWaG7HRTGbvt3+SJJ2qoqyakUi4x3Ntrw9q+XoZ7exjq62Got5fhpn6lODLmdmLxRD00TaZGNTsQjAfBazyo6dpc87W2vtZ8VTPLGBsgVkoe5VKVSsmrddXQcK0rNk8L1rFttUfjkfqPocE52Z277XDwg2ikNh780BprMe9w/LHU83wqRa/hR2m7H/z6XT/B/mlokqlauxOoWvIolxqbZKpWbSA7XSIxsW3Wj+rX7wAK+vFk43g0JuTmp6atbEoppZSaHRrqqjEFNWZHBgYYGehnZKCf4kA/I4MDNoR14WyxFsoO1YLb8S6iwTYvkMrlaqFsulBgXudSkpmsnZ5xgW0Q3GbtsslslmXh6BwAACAASURBVGQmc0ReMBvjLsqCgNUFskEYW2oR1FaKo6dPpPZrPBUlmY6RSMdIpGKkcgnaFqWJp2Mk3IVb8wVbLF5vYmDUfNfswOF24XY0Mb6PPzziwtRQoBoM1wLXoVEBrBkexhsawgRhrZs/XgDbTNJpIpkMkWzW9jMZooUC8c7O2njD/Gxm1PKRQts07iF1uKtWKgz39TDca4Paod6eWg3b5n55ZLjlNlK5PJm2drLt8+g45niybe1k2ueRddMybe1k2tpJpG1IG4keeecSdfgyvnHNBjWFq+XmsLU6Oox16zQEjm7aZO56ibr22pu7dD5BNBahWrF305SLHsP9ZXvXTa15Ix8zyTtsas0a1e6eCf+AG205r3YXTdM8EQm1hV8NtX3v9l9D00zhwNZrCGvtQ0YnLpZoam7JDeez8fp7SNofokcFrqG26ycayDb03cNLRdu9V0oppVQLGuoeJYwxlEdGXDjbR7EW1Np+cbCfkX4b3I4MDtQCXK86druwiXSmVjs2mc3S3rGEZCZHKpclmXGBrQtog2WCfiyRPOK+nAahbGm4aruhCqXhKsXhiptWqc0L15q14awNZA90d68INoh1YWwiHSXXniTRma1NT6ZjJFLR+nLBNNfFk1rb9XBgQ1gXrjYEr6HhoAsvNzzGvJGRCd8eLqlUY5iayRDN5Yl0LAmFr5lRywTj0jCeJZJOIRqOqRa8aoXhvj5Xo7bHBrO9dri5XxoaarmNZCZbC2YXrTqGVaGAtrHfRjQWn+F3qKZbEIRWyy1qUTaNV8senmcwnnG3wdv1fd9gXFcfpuV033e1Xz03z7jpnsGYsdax22v9Wm66MZO6A0aEeuiaijWEr/kFUeKp0cFsuEukwiFkzC6fiBxSe+7Bfgk3udTQnn3oM6o11VRqNc+ODw5XGj/XsnfQNV2DGsbh5pniySiZtkS9dnG4SaamgDa87xrmxSOIfmdSSiml1GFKQ905yBhja8WGatDWatIODthwdtBN63fTBgbGfHCXSIRUPk86XyCdz9PesYQlx64hXSiQzrnphQKpXKG2TDKTPSJrOI0KZlsEsqWhCsXmacMVSkPj15IVgWQmTiIThKxRCgvTTWGsnd4QxKaCQNZeYBxpYfhsMsZApYKpVjGVyhhd87xybZjm9YL2YoeGajVfvaHmmrOuFuzI+LXZw8SFrNFMthasxhYtIrJqlavx6qZns/UwNghcG8ZtGKsBrJouA8/sY8Pn/7lWo7Y4ONByuUQ6XQtkF65YSdcpp5Ftm0emvZ1M2zyy7e12vK2dWCIxw+9CTZYxBq/i12uchh5YWWsGIBy+hm9rr93aPjqgDcLDyZKIIBGIiCBRIeJqOUoEO+y60cP1+bV+VIi4YC+oLRmJun4Eu31x88LDY7xWPHHgQDaejB6Wd7+ICNGYrXlKdnpew/f8hgel1h6GWvYwvqntv8YHl+qP1UoppZQ6Ommoe5jxqhWevP9e+nbvqtea7Q8Ftq7ve17L9SUSccGr7eZ1LrPhbN4FtIU2Url6gJvOt5HMZI6oh6/Yp5T7DYFscagppK0Fs6OD2/FuyxPBhrKZOKlMjGQmRn5+imQmRjIbJ5mJkcrYftItF8xLJKNa2+MAjDGYYhGvr892vX14fb1uuBe/vx+/VKqHqg2B6gHC2RbzmETzAhMm0liz1YWp8Y6OxnA1m53YeCZ9RP1/qiNbPJHCq1SY17mM5Sc+q16L1oW02Xbb/EE8qW07TqWgXVWvattFrVb82rBXsQ9LahxvGq6MMb/F+kHQFg5sq2VvMs+Rq9VCDT/UMpG0P2hm25K2tmUyVrutPT5urcrGW+ODwFXNTZFohEQ6QiKtlyhKKaWUUgei35gOEwP797H5RxvYfPtGhvt6AdvubNrVoE3l88xftrwhsE3l8q42baFWmzaRzhxRFzNBQDsyWKY4WKE4WGHE9YtDofGhcmi4Mv7tewLJdGPoWgtmmwPZbCik1WB2wowxtpaqC2MbQ9qgs9P9ILh180y5PPaG43EiqRQSj9e7WKxxPB4nkk5BIY/E4qPmjVovMXoeo7aZCA2Pfr2GLpXSEFYdtVK5HH/0D/8028U4rPmeT3GoWjuvjQzY89bIgB0vjVTHD1bHCGynSjRm20gPamTa4XoXS9gmAMI1JW3NydHtjgaBbeO43c6R9F1FKaWUUkqp2aCh7iwyxrD94V/zwK038sSvfo4xhmNOP4t1l1zB0rUnkUinj7iLnmrZsxevQTBbC2ndxW0oqC255bzKGBerAqlsnHQuTioXp7AwTceqAslsnFQ23hDS1sdtcwYazE6MMQZ/cLCh1qzf19cY0oZD21A3Xi1YSaWItrXVusSqVUSC8fZ2N9312+vLyRH4P6GUmruCJnvCPzg2hLWD5dB0O680XB2zbdVEKkoiEyMWjzaEqrFElGQ2HK4K0fAyo8JXIRYPHsIUIRZvsUx89Lpay1UppZRSSqm5Q0PdWVAaHuaR/7uDB2+9mf1PbyOVL3DWi1/GaRdfRtviJbNdvAnzKn4ooC1THKqOuoCt16a1F7njtY2XzMZqIW1+fopFXXnSWRvYpnJBeJuw/axtm1bbUBufMQZTKuH19+MPDuL39+MNDOIP9OP1D+APDuD1D+AN9OMPDNp+X39jODtGUx9g23q1oasNYJPHHVcPa10Y2xjWthNtKxBJ6a3XSqnDj++b0A+O5YYfHkcGQsOh89xYPzxGImLPXXl7Dlu4PNd4LssH57b6eS0a11r+SimllFJKqYnRUHcG7X96G/dvvIlHfnIHleIIHcccz6Vv/0vWnPc84onkbBevgTGG4f4y/fuK9O8bYWD/iB3eP8LA/iIjgxUqxbHDvkQ6Vrt4zbQlWLA0SzK4mM3ai9hwWJvMxA7picxHKuP79sFb/f14AwP4AzaE9Qdah7O16aHlzYHajY3FiObzRAp5ork80bYCsc4lTUFsY43ZIKyN6AOMlFKHOeMbhgfKDDxTZPCZEoM99hxWHJh8LdpU3gaw2fakC2kTtbtF0vnQcC5OIh3TWq9KKaWUUkqpaaOh7jTzqlV+d+/PeWDjTWx/5NdE43HWnnc+6150BZ3HrZ3VshWHKgzst6FtENj27yvaAHd/cVTto0whQWFhio7VbWTy9VA23ARCMB6NaUAb5pdKVPfsodrdTfWZZ0Lh7IANX/v78QYbw9igf6Cnz0g6TTSXI1IoEM3niba3k1ixwoa0+TyRfIFoIU8kl7f9fON0SaU0eFBKzVnVisfgMyUGnim64Lbohkt2uKc4qp31Sdeizel5TSmllFJKKXV40VB3mgz19rD5R7ew+UcbGOx5hsKixZz/6jfwrBdcTKbQNiNlqJQ8W7M2FNj277OB7cD+IuWRasPyyUyM/IIU8zqzrHzWAgoL0+QXpCgsTFNYkCKWiM5IuecSYwxeTw/V7m4q3d1Uu/fY8HZPaLy7G6+3t/UGRIjkci5ktWFrfOlSovlcizC20Dg9nyeayyFaW1YpdYQyxlAcqtRD2/02pB3c74LbnhIj/U0PVxTItiXJz0+yaGWeY05fRH5+ivz8FLn5KXLzkiQzWotWzSzjefhDQ/YhnYkEEtOv4DPJGAPVKsbzwPMwrgummaoHXjBctcs0TGs933ihZX0P4/vgGzB+wzDGYHwDvu/mhYdbr1Mb9n2MOfDw6O02Drdd+VIKl1wy2x+FUkoppaaQfqOcQsYYdjz2CA9svInf/uKn+J7HqtPO4IVv+TNWn34WkcjUhqKe5zP4TLExrHX9/n0jjAw03nYfi0fIL0xTWJhi6XHtFBamKCxIk1+YorAgRTITn9LyzXXh2rXhgLaypz5c3bNndPMGIkQXLCC+eDHxpUtJn76OeEcHscUdxDo6iC2YT7RQIFIoEMlmkYjW/lJKHZ08z2eopzSqhm14vLkt9lg8Qm5+ivyCFAuX58gvsGFtfp6dlm1Paq1aNW38crnxgaG1B4m6h4n294fG692oO29cuBtJJJBkEkkmiSQTSCLpxhNEasOh8dryCSLJpFveDbccT9jXSda3FUkkIB6f1A8bxvcxlUq9KwfDZRuMhue16spjTK+OnsZY6wWhbBDEelWoeqOG6/Prw/hjP9PhsCECkQhEIvazmYrhiIDYYX9gcLbfoVJKKaWmmIa6U6BSLPKbu+/igY03snfbUyQzWda96MWcdvHlzF+67KC3a3zDUF/Z1bath7VBUwlDPaWG64NIRMjNT1JYmGb1qQtrAW5hQZrCwjTp/OS+wB+pWtaubQ5ru7vtQ8KaSDpNfPFiYh0dpM84g3jH4npYu3iRDW8XLULiGpArpVRppGrD2aBmbS2stcHtUF9pVBu26bx9WOa8zixdJy8I1bJNkl+QIpXVc5k6NMYY/KFh/L5evP7+pmB2/HDWjIyMveFIhGihYNudb28jOn8eiVWr6m3R5/Pge/ilEqZUxpRKmHIZvxwaL5Vq49X+AUy5hN8wr4wp2yD1kIg0hLySTCKxmA1OW4Sv4z009ZBEo0g8PrEulSQSiyPRKMSiSDTWOByL2u3VhmNINNJiWrQ+PGqdpuGYew03LTzcON/Ni0TsvnX9iYaviOhxTSmllFKTpqHuIXhm5w4evPUmHv7x7ZSGh1i0cjUXX/UOTnzuBcRTqYPaZnGowt03/Jbup/oZ2F/Eq4ZqFrhbSgsLUyw9vt2FtfXatrn25FH/sDFTqVDZvZvq7t1UarVpu6ns2XPg2rULFxBf3EF8+XLSZ57hAlob4MY7bD+Sz+uXbqWUGsNQb4k7v/NoLcgtNz1QMxIVcvNsOLvihHnkFtjANqhlm5uX1KZ+1AEZz8MfGbHB7PAQ/vAwZngYf3gYr3+gFsx6fX2uVm3/qHB2vFBUEonaA0IjbW3Ely8ndfLJLpwtNDwwNNrWbh8kWigQyeVm7O4bU63aQNgFw7XQt1TGlINhFxaXS/Vla2FyadS4qVTsj9JjBquJxvFYDElMMJBttd0gEFVKKaWUUgdFQ91J8n2PJ++7lwc23sjWzfcTicZYc+5zWXfJFSxde+IhBX59e4e58fOb6d83wurTFrL6tIW19mwLC9Pk56eIxo/u0BbALxapbN9Oeft2ylu3Ud62lcrWbZS3b6eyc+eo2iSSydjatYsXj6pdG4S1sYULtXatUpNgjKHiVxiqDDFcHWa4MsxwdZihyhAjlZHatKHqUG3ecKW+XG2+W3+kMsIx7cfw7cu/PdtvTR2CaDzCYE+J/II0S4+fZ2vXupq2+QUpMvmEvR1YHRWMMTZcdIFrOHxt6IZCwyNNyw6NXt6UShN6/UguFwpgCyQ71tTGa6FtIQhpXTjb1kbkIH+Yn0kSiyGxGJFMZraLopRSSimlZomGuhM03N/Hr++4lc0/2kD/3j3k5i/gua94Ladc9CKy7fMOefs7f9vLhi/+GoArrz6dpce3H/I25zJvcJDKtm2Ut22nvM0Ft264unt3w7KRQoHEypWkTzmFwouvILF8BbElHbambUeHrTmjtWvVUc43PgPlgYYgNRysjlRHxg9hQ6FtEMJWzcRv/83EMmTiGbLxLJlYhnQszbzUPJblltlp8QzLc8uncQ+omZDKxnnV+8+e7WKoQ1BrmmCgH69/oN4fHHD9wdGh7MgwpkX46g8PT6otU0mniWQyjV0uR2zxYiKZDJIJz8/Wh7Oun07bh4m2txHN5/XHWqWUUkopdUTTUHccxhh2P/E4D2y8kcd+9n941SpdzzqVC173Zo496xwiU3TL2GM/38Ud33qUwsI0V/zZqbQvPvJrXRhj8Hp7bY3boLZtKMT19u9vWD66cCGJri6y555LvGsFia6VJFZ2kVixgmj70R2AKwUwUB5g19Audg/tbuyGd7NrcBfdw91U/MqBNwSkY+lRIWxbqo3OWGdtPBPPtOwHAW14WiqWIiJ6l4FSM8H4Pv7QkG0TdmAAr78ff2CgHtAODOD3D9i+C2y9gX78/gG73MDAgYPYWIxINhSqujA2CF+DkFXC4Wy6KYBt6iSd1geHKqWUUkopNQka6rZQKZd47Kc/4YFbb6L7ySdIpNOcctGLWHfJFSxY3jVlr2N8wy/+90k2bdjKsrXzuPSqZ5HKHjm1SowxVPfubQpug5q32/D7+xuWj3V2kujqIn/hhQ3BbXz5CqK57Cy9C6VmX8kr0T3U3RDa7hraxe7h3ewetMHtUGWoYZ2oRFmcWcyS7BJOWXQKF2cvZlF6Ebl4jnTchbYtQth0LK0BrFKzyHheLVy1gexgLXSt9wdsaDs4WAtva/3BQRqeotpCJJu1zQ7kckQKBeKLO4gcdxzRfIFIIU80XyBayBPJ5W0/GC8U7LqJxAztDaWUUkoppdRYNNQN6e3ezYO33cxDd95GcXCABcu7uOhNb+Ok9S8gkZ7a2rPVssft3/gNT2zaw0nP7WT9q9cSnYMPOTOeR7W72wa1zcHt9u2Y4eH6wtEo8WXLSHR10XbqqU3B7XIiyeTsvRGlZonne+wd2dtQu7YW3g7b8WeKz4xab35qPkuyS1hZWMm5S89lSWYJS7L1blF6EdGIPoBGqcOVqVYpb91K6bHHKD7+OKXHHqf0+ONUduw44LqRfJ5IPmfD13ye+NKlRNeutUFtPu+C2TyRfN4+wMv1o/m8bZIopl//lFJKKaWUmuuO+m/1xvd56sH7eODWm3jy/nsREY5/9nmse9EVLD/plGlpi3W4v8zNX9hM91P9POflx7Hu4hVzos1X43kM37uJwbvuorxlC+Vt26hs346p1G/plkSC+IoVLZtKiHd2avt26qhijKG31Ds6qHW1a3cN7WLv8F480/hwv2w8S2e2k45sByfOP5HObCdLsktq/Y5sB8mo/gii1FxR3beP4mOP1YLb4uOPUX7id5hy2S4QjZJYvYr0aafRduWVRNsK9dqxzf1sFpmi5p+UUkoppZRSc9dRG+qODA7w8J238eBtG+jt3kWmrZ1zX/5KTr3oUvILFk7b6+7fMciN1z5IcbDCZX96CsesWzRtrzUVjOcxvGkTA7fcQv+tt+Ht24ckkyRWryZ53HHkL7rQhbg2uI11dGibeOqoEYS2W/u3sn1gO9sGtrFr0DaL0D3Uze6h3RS9YsM68Ui8Vpv22R3PbqhdG4S2+UR+lt6RUupQ+MUipSd+R+mxxyg9Xq+B6z1Tr20fW7SI5Jo1ZF97Hqm1a0iuWUPi2GO1SQOllFJKKaXUpBx1oa4xhtv//Ys8fOdtVCtllp1wEs995Ws5/pznEI1Nby3SrQ/tZ+NXHyKRjPLya85kUdfhGdzUg9yN9N92K97efUg6Te75z6dw6aXk1p9PJHPkP8xNqUBvsZetA1vZ1r+NbQPb2NpfHx4oD9SWE4RF6UUsyS1hzbw1rF++vhbUBt381Hxts1apOc74PpUdO0Y1nVDeurX2kDFJpUgefzy5F1xAau1akmvWkly7hti8ebNceqWUUkoppdSR4KgLdUUEjOGk9Rdy2iWXs3jVMTPyupvvfJq7b3icBctzXPH208jNO7xunTaex8h999F/y0b6b91og9xUyga5l11Kbv16DXLVEa2v1MfW/q21Wrfh4La/XH+oX0QidGY76cp3cfnqy+nKd7GysJKuQhfLcstIRLW2nVJHEq+vzzaZ4ILb0mOPUfrtb/GDNuNFiK9YQWrtGgqXX05yzRpSa9cQX7FCm0lQSimllFJKTZujLtQFeOGb3z5jr+V7Pnf/5xP8+q6nWXXqQi5+00kkUofHbje+XwtyBzZupLp3rw1y16+3Qe7zn69Brjqi9JX62Na/ja0DW9nev72h9m1fqa+2nCA2uC10cdnqy1iRX1ELbpfnlmtwq9QRyFQqlLZscbVu6zVwq7t315aJtrWRXLuWtpe/nOTaNaTWrCF53HFEstlZLLlSSimllFLqaHR4pItHqPJIlY1ffZhtD+9n3QtXcN7LjyMSmd0HohnfZ+T+++tB7p49SDLZGOTqxamaw/rL/Ta4DdW0DYLcVsHtisIKXrTyRXQVumq1bpfnNbhV6khljKG6Z0+t1m2tBu6TT0Lw4M94nOQxx5A5+9k2uF27luSaNcQWL54TDzZVSimllFJKHfk01J0m/ftHuOnazfTsHuaC16zl5POXzVpZjO8z8sAD9G+4pSnIPZ/8pZeSv+ACDXLVnNJf7rc1bfu3jqp121vqrS0nCEuyS+jKd3HJyktYWVhZq3W7PL+cZPTwagZFKTV9jDH0Xn89e//1840PLuvsJLnmeHLr15Ncs4bk2jUkV69G4tPbzr5SSimllFJKHQoNdafB7i193PyFX+NVfF7y56ex4sT5M14GG+Q+SP8tGxjYeCvV7m4kkSD3/PXkX3QpuQsuIJrTIFcdfoYrw+wZ3sOe4T10D3ezZ3gPe0f22vGhbrYPbKen1NOwzpLsElbmV/LClS9kZX5lrdbt8vxyUrHULL0TpdThotrTw66/fT+Dd9xB5txzyV/8QlsDd80aom1ts108pZRSSimllJo0DXWn2G/v7eb2b/yGbFuC3/vL05nfOXPBqfF9Rh58kIFbbqF/461Ud+9GEgmy68+ncM015F7wAg1y1azxfI9nis80hLWtuoHKwKh1s/EsizOLWZxZzIVdF9bat+3Kd7Eiv0KDW6XUmIZ+9jN2/vV78Hp76Xjfe5n3utchkchsF0sppdQcICIp4CdAEnvt/P+MMR8UkdXA9cACYBPwOmNMefZKqpRS6mikoe4UMcawacNWfvHDJ+k8to3L3nYK6dz0t8lpjKH44IP0b7iF/ltvpbprFxKPk12/nsK73+WC3Ny0l0Md3YYrw+MGtd3D3ewb2YdnvIb1ohJlQXoBHZkOVrWt4uzOs1mcWUxHpqMW4i7OLCYb1x8jlFKTY8pl9v7rv7L/q18jsXo1K778JVInnjjbxVJKKTW3lIALjTGDIhIH7haRDcC7gM8YY64XkS8CfwJ8YTYLqpRS6uijoe4U8Co+d37nUR77+W7WnNPBha89kWh8+moBGWMobt5sg9yNG+tB7vnnU/jLq22Qm89P2+uro4fne+wv7mfv8N6G0LY5wB2sDI5aNxfP1ULZczrPGRXULs4sZkFqAdFIdBbemVLqSFZ+6il2XPNXFB96iPZXvIKO976HSCYz28VSSqlpU6wW6S310lvqpafYM6r/gq4X8Jylz5ntYs45xhgDBF90464zwIXAq930bwAfQkNdpZRSM0xD3UM0Mlhmwxd/za4n+jj7Jas56/JV0/JkbGMMxV//2gW5t1Dd6YLc5z1Pg1x1yPpKfTyw5wE27dnEtv5tteB2/8j+lrVrF6YX0pHp4Ji2Yzi389xRYW1HpoNMXAMUpdTMMsbQ94P/Yfc//AMSj7PsXz5H4ZJLZrtYSik1KWWvXAtke0u99JR66C029vtKfQ3LjFRHxtxeIVFgVdsqDXUPkohEsU0sHAdcC/wO6DXGVN0iTwOz91RspZRSRy0NdQ9Bz+4hbrx2M0M9JS5588kcf1bHlG6/FuTespGBW26hsnMnxOPknvtcCu98J7kLL9QgVx2U7qFu7ttzH5u6N3Hfnvt4oucJDIZYJEZXvovFmcWc23luQ+3aYHh+ar7WrlVKHXa8/n52f+jD9N98M5lnP5uln/wE8c7O2S6WUuooV/EqtWA2HMTW+iUXzBbrtWyHq8Njbi+fyNOebGdech6LMos4ft7xzEvOoz3VXpvenqr3C4kCsYhe8h0KY4wHrBORduD7wAkTWU9ErgKuAujq6pq+AiqllDpq6Rn+ID396DPc8uWHiESF33vX6Sw5Zuqenl18/HH6/ud/GLhlI5UdO2yQ+5znsPAv/pz8hRcSLRSm7LXUkc8Yw9b+rfUQt/s+nh58GoB0LM26Reu4ZN0lnNlxJqcsPEUfOqaUmnOG77ufnddcQ6W7m0VXX82Ct7wZieqPT0qpyTPGUPErjFRHKFaLFL1iY79aZMQbqQ0H4wPlgZa1aFs1URXIxXO0J10Ym5rHMW3HNASytXluvC3ZRjwSn8G9ocKMMb0icidwHtAuIjFXW3c5sKPF8l8Gvgxw1llnmRktrFJKqaOChroH4ZG7d/Lj7z5G+5IMV7z9VAoL01O27ZEHH+Sp174OgOxzzmPhO95B/iINctXEeb7H4z2PN4S4+4v7AZiXnMfpi0/nVSe8irM6zmLt/LVae0MpNWeZapV9X/oS+679N+JLl7LqO98mvW7dbBdLKTVNKn5lVJhaG/eKjFRHKHklO68plB2pjjQGs6Flg3WDYd/4ky5bJpZhXmpeLYjtKnTZMNYFtsH0WmibbCce1YD2cCcii4CKC3TTwMXAJ4A7gT8Argf+GPif2SulUkqpo5WmOZNgfMPPvv877r9tG10nzeeStzyLZHrqdmF1/36efufVxDs6WHX994gtXDhl21ZHrrJX5qF9D7GpexOb9mziwT0P1mqFdGY7OW/peZzRcQZnLj6T1W2rp6XNZ6WUmmmVnTvZ8Vd/zcimTRRe8hKWfPDviOZys10spdQU6h7q5panbuGWLbfw6DOPUq01YTpxsUiMdDRNKpaqdcF4LpEjGU2SjqVJRUPzm8fd8rVlg21F7bLJaFKbpjpydQLfcO3qRoAbjDE3isgjwPUi8hHgfuBrs1lIpZRSRycNdSeoUvK47d8fZsuD+zjl+ct43iuOJxKNTNn2TbXKjndfg9fTw6rvfVcDXTWmwfIgD+x9gPu6bU3ch/Y9RNkvA3Bs27FctvqyWojbmdP2JJVSR57+W25h1wf+DnyfpZ/8BG0vfelsF0kpNUV6i73ctu02NmzZwL2778VgOHH+ibz+5NeTi9sQtha8umC1VRCbiqZIxpLaXIE6JMaYzcDpLaY/CZw98yVSSiml6jTUnYDBnhI3f2Ez+7YPcP4rj+fUF6yY8tfY+7l/YfjnP6fzox8lddJJU759NXftH9nPfXvuq4W4j/U8hm98yJ1FzQAAIABJREFUohLlxPkn8qoTXsUZHWdwxuIzmJeaN9vFVUqpaeMPDbH7ox+l77/+m9Spp7LsU/9EQh8+o9ScN1wZ5s7td3Lzlpu5Z8c9VE2VVYVVvO20t3HZ6stY1bZqtouolFJKKXXY0VD3APZuG+Cmax+kXPS4/O2nsuqUqa9B23/bbez/yldof+UraX/5y6Z8+2ruMMawY3BHQ4j7VP9TACSjSU5ddCpXnXoVZyw+g9MWnUYmnpndAiul1AwZefhhdr77Gspbt7LgT/+URe/4MySuNfCUmqvKXpm7d9zNhi0buGv7XRS9Ih2ZDl570mu5fPXlnDD/BG0ySimllFJqHBrqjmPLg3u59WsPk8rGeflfncnC5VPfVl9pyxZ2vfd9pE45hY6//Zsp3746vPnG53e9v6sFuJv2bGLP8B4A8ok8Zyw+g5cd/zLOWHwGJy84WR+ooZQ66hjf55mvX8eez36W2Pz5dF13Hdlz9I5XpeYiz/f4Vfev2LBlA7dtvY2B8gDzkvO48rgruWz1ZZy++HQiMnXNmymllFJKHck01G3BGMMDP9rOPf/9BItXFrj8baeQbUtO+ev4w8Ps+Iu/QOJxln/us0QSiSl/DXV4qfpVHtn/CJu6N3Ff933cv/d++kp9ACxOL7bNKLimFI6fd7xe2CiljmqVPXvY9d73MXTPPeQvfiFL/v7vic3TZmaUmkuMMWzet5kNWzaw8amN7BvZRyaW4aKui7hs9WWcu/RcbfdWKaWUUuogaKjbxPN8fvK9x3nk7p0ce8ZiXviGE4klpv5ptsYYdn3g7yg98TtWfPUrxJcunfLXUIeHkeoI9+y8hzu23cFd2++iv9wPwMrCSi5ccSFndpzJGR1nsDy3XG8zVEpNmIhcCnwOiAJfNcZ8vGl+F/ANoN0t815jzM0zXtCDNHDnnez6m7/FHxlhyYc/TPsr/lCPkUrNIb/t+S0btmxgw5YNPD34NIlIgvXL13PZ6stYv3w9qVhqtouolFJKKTWnaagbUhyqsPErD/H0oz2ceelKznnpMUhkei4ge779HfpvuolFV19N7rnPnZbXULOnr9THj5/+MbdvvZ17dt5D0StSSBS4YMUFnL/8fM7qOIuF6alvn1kpdXQQkShwLXAx8DTwKxH5oTHmkdBi7wduMMZ8QUROAm4GVs14YSfJL5XY80+foufb3yZ5wgks+/SnSB577GwXSyk1AU8PPM2GLRu4ecvNPNH7BBGJcG7nufzpaX/KRV0XkU/kZ7uISimllFJHDA11nb69w9x07Wb69o5w0R+fyAnndU7baw3fdx/dn/gEuQsvZMFVb5m211Eza/fQbu7Ydgd3bLuDe7vvxTMeHZkOXnb8y7iwy9bI1dsLlVJT5GzgCWPMkwAicj1wJRAOdQ1QcMNtwM4ZLeFBKP32t+x49zWUHn+c+X/8eha9611EklPf/JFSaursG9nHxqc2cvOWm9m8dzMA6xat431nv49LVl2iP2IrpZRSSk2TaQ1158qtoTuf6GXDF36NwfDSd65j2Zrpa6+vuncvO955NfGlS1n68Y8hEW0zdS57svdJbt92O7dvu52H9z8MwOq21bzxWW/koq6LOHnByXq7sFJqOiwDtofGnwbOaVrmQ8CtIvLnQBZ44cwUbfKMMfR873vs+cQnieRyrPjyl8itXz/bxVJKjaG/3M/tW2/npi038avdv8I3PmvnreXqM67mstWXsTSnzYoppZRSSk23aQt158qtoY/9Yjd3fOs3FBakueLtp9LekZm21zKVCjv+8l14AwOs+upXiBYKB15JHVZ84/PQvoe4Y9sd3L7tdp7qfwqAUxaewjvPeCcXdl3IMW3HzG4hlVLK+iPgOmPMp0XkPOBbIvIsY4wfXkhErgKuAujq6prxQlZ7etj1t+9n8I47yD7veSz9+MeILdSafUodbkaqI/x4+4+5ecvN3L3jbip+hRX5Fbz5lDdz+erLObZdm0lRSimllJpJ01lT97C+NdT4hl/euIV7b36KZWvbufSqU0hlp/fW+D3//BmG772XpZ/8BKm1a6f1tdTUqfgVfrX7V9yx7Q7u3HYne0b2EJMYZy05i//P3n2HR1lmfRz/3um9EVJIgRB6D4ReQ1GK4topgqxY1vd1RRDrqqu4u/bC7usWxQbSLKwiTXpTpIUqPQRIQiqQ3iYz9/vHE6qAQMozyZzPdc2VzJOZyW+yLjNznnOfe2zrsSREJRDqHWp2TCGEY0kDoi64Hll57EITgaEAWutNSikPIBjIuvBGWusPgQ8B4uPjdU0FvpyiTZs4+fQzWHNzCX3uWQLHjZMVLELYEYvNwqaTm1iSvITVJ1ZTUlFCQ8+GjGo1iuExw2VFkhBCCCGEiWqyqGu3S0Mryq2smrmfI9uyaN0rnP5jWuLsUrMfIvOX/cDpTz8lcMwY/EeOrNHfJaqu2FLMTyd/YtWJVaxLXUdBeQGeLp70btSbgdED6RfZD393f7NjCiEc11aguVIqBqOYOwoYc8ltTgCDgM+UUq0BDyC7VlNegS4vJ/sf/+DUjI9xa9KEqP/8G482bcyOJYTAWJW0PXM7S5KXsOL4CvLK8vBz82N4zHCGxwynS2gXnJ2czY4phBBCCOHwzN4ordaXhmqbZuHfd5KelEfP22OJuym6xjsMypKSSH/+eTw7diT02Wdq9HeJG5dbmsva1LWsOrGKTSc3UWYtw9/dn4FRAxkYPZCejXri6eJpdkwhhEBrXaGUegz4AWMm/Sda61+UUtOAbVrrhcCTwEdKqckYK2MmaK1rtRP3csqPHSNt6lOU7t1LwD33EPrsMzh51dzoIyHE1WmtKako4WjeUZYmL2XZsWVkFWfh6eLJgKgBDI8ZTu9GvXF1ls1ehRBCCCHsSU0Wde1yaahyUrTuFU7HQVHExoVU5aGuibWwiNQ/Po7y8CBi+vsoN7ca/53i2qUXprM6xZiPm5iZiFVbCfMO487mdzIoehCdQzvj4mT2uQ8hhPi1yo1Fl1xy7KULvt8H9K7tXFeitSbv2+/IePVVlKsrEdOn43fzTWbHEqLeKa0oJbcslzOlZ8gty73o+wuPXXi9zFoGgIuTC30i+jA1fir9I/vj5SonXIQQQggh7FVNVqvsdmlo6161syOv1pr0F16g/Ngxoj/5BNewsFr5veLKtNYk5Sax6sQqVp1Yxf7T+wGI9Y/lgXYPMKjxINoEtZH5cEIIUY2s+flkvPwK+UuW4NW1K43efAPX8HCzYwlh98qsZeSWVhZgyyoLsqWV31/w9cJCbUlFyRUfz8/Nj0CPQALcAwjzCqNVUCsC3QPxd/cn1DuUvhF9ZbyUEEIIIUQdUWNF3bq8NLS6nP7scwqWLSNk6pN497h0nLCoLTZtY3f2blanrGb1idUczz8OQIeGHZjcZTIDowbSxL+JuSGFEKKeKk7cwcmpU7FkZtLwiUk0eOghlLPM4xSOx2K1nC/OXqYge7njxRXFV3w8XzdfAtwDCHQPpKFXQ5oHNifQPZAAj4BzxwM8zn/1c/OT1UdCCCGEEPVIjb6zq2tLQ6tT8datZL39Nr5DBhM0caLZcRyOxWphS8YWVp9YzZqUNWSXZOOiXOgW3o3xbcYzIGoAIV41P35DCCEcla6oIOc//yHnn//CNTycJrO/wLNTJ7NjCVEtLDYLeWV5VxxxkFeWd3GRtiyXIkvRFR/Px9WHAHejGBvkEUSsf+xFBdmzPzt73d/dH1cnmXErhBBCCOHI5HR9DbBkZpE6eQpuUVGEv/aaLOWvJfnl+fyU9hNrUtawIXUDBZYCPF086RPRh0HRg+gb2Rc/Nz+zYwohRL1nOXmStKeepmT7dvxuuYWwP7+Es6+v2bGEuKwKWwV5ZXlXnTl70fHSXAosBVd8PC8Xr3MjDgLcA2js3/jciIMgj6Bzx88Vbd0DZBMyIYQQQghx3aSoW820xULa5MnYiopo/OknOPv4mB2p3tJacyz/GOtT17Mudd25jc4C3QMZ1HgQg6IH0SO8Bx4uHmZHFUIIh5G/bBnpL/0ZKipo9OYb+I8caXYk4UCsNit55XkXz6H9jVEHBeVXLtB6unie75L1CCTKN+riguwFhdmzx92d3WvxGQshhBBCCEclRd1qlvnWW5QkJtLonbdxb97c7Dj1jsVqYVvmtnOF3JSCFABaBLbggXYP0C+yH+2D2+PsJPMahRCitpUePEjaE5Px6NCBiLffwi062uxIop47O25pTcoa1qeuJ6MoA83lt2dwd3a/qIM2okHEuaKsv7v/uZ9deBs5MSyEEEIIIeyVFHWrUd6ixZyZOYvA8ePwHzHC7Dj1Rk5JDhtSN7AhbQM/nfyJIksR7s7udAvrxv1t7qdfZD/CfWQXdSGEMJtHy5ZE/uuf+PTpg3KV5eSiZhSUF7AhdYMxbiltA0WWIjxdPOkZ3pPbmt12fvbsJR21ni6eZkcXQgghhBCi2khRt5qUHT5M+osv4tm5M6FPPWV2nDpNa82B0wdYl7qO9anr2ZOzB4AQrxCGxwynf2R/uoV3kw9nQghhh3wTEsyOIOqhjKIM1qSsYc2JNWzN2EqFriDII4ibm9xMQlSCjFsSQgghhBAOR4q61cBaWEjqHx/HydubiPffk+6kG1BsKWZz+mbWp61nfcp6skqyUCjaB7fnsU6P0T+qPy0DW8qmc0LUV/knIT8dIruYnUQIYQe01hw6c8go5KasYd+pfQA08WvCuDbjSIhOoENwBxm3JIQQQgghHJYUdatIa036c89RnpJC488+xTUkxOxIdcbJwpPnZuNuSd9Cua0cb1dvejXqRf/I/vSJ6EMDzwZmxxRC1ASrBVK2wOHlcGQlZO6FkDbwP5vMTiaEMEmFrYIdWTtYfWI1a1LWkFaYBkCHhh2Y1HkSA6MH0tS/qckphRBCCCGEsA9S1K2i0x9/TMGKlYQ8+wxeXbuaHceuWW1WdufsPlfIPXzmMADRvtHc0/Ie+kf1p0tIF1ydpdNZiHopP90o4B5eDkfXQlk+OLlAdE8YMg2aDTE7oRCilhVbivnp5E+sSVnDutR15JXl4ebkRvfw7jzY/kEGRA0g2DPY7JhCCCGEEELYHSnqVkHRzz+T9e57+A4bStD995sdxy7ll+fzU9pPrEtdx8a0jeSW5eKiXOgc2pmp8VPpH9mfJv5NzI4phKgJ1gpI3QKHV8CRFZBhzMfGtxG0/Z1RxG06ADz8zEwphKhlOSU5rEtZx5qUNWw6uYlyWzl+bn70i+zHwOiB9G7UGy9XL7NjCiGEEEIIYdekqHuDLBkZpE15EreYGMJf/YvMeq2ktSY5P5n1KetZn7aexMxErNpKgHsAfSP60i+qH70a9cLPTYo44vporVm2N4PM/FKzo4ir8CzLJiLnRyJO/UijUz/jXlGATTmTFdCJ1GaTSAvuzRmf5qAUnAG2nwJOARDk487Ijo1MzS+EqBnJecnnNjrblb0LjaaRdyPubnk3A6MGEhcah6uTrNQRQlQ/pVRP4D6gLxAOlAB7gcXAF1rrPBPjCSGEEDdMiro3QJeXkzbpCXRpKZH/+DvOPt5mRzKVxWphW+a2c2MVUgpSAGgR2IIH2j1Av8h+tA9uL5uZiBtms2le+f4XPt903Owo4hLOWOmkjjDAeRcJTjtp53QMgEwdwLfWLqyxdeJHWzsKSrwgHaAC2H/Zx2rbyE+KukLUEzZtY3f27nMbnSXnJQPQOqg1j3Z6lIFRA2kR2EJOigshapRSailwEvgO+CuQBXgALYAE4Dul1Lta64XmpRRCCCFujBR1b0Dm629QsmsXEe+/j3tTx9ywI6ckhw2pG9iQtoGfTv5EkaUId2d3uod3Z0LbCfSN6Eu4T7jZMUU9YLVpnv1mN19tT+XBPjH8b0IzsyM5PFWUhWvyalyTV+FybC1Opblo5UxFRFdKYsZiaToYt4ZtuUkpbrqOx3VykuKOEHVZmbWMzembWX1iNWtT1nKq9BQuyoX4sHhGtRxFQlSCvDcQQtS2cVrrnEuOFQKJlZd3lFIyuFsIIUSdJEXd65T33XecmTOHoAcewG/ozWbHqVVnSs+wJHkJS44uYXfObgBCvEIYHjOc/pH96RbeDU8XT5NTivrEYrUxef5OFu1O5/FBzZk8uLl0dZnBZoW07cZs3MPLIX2ncdwnFFrdAs0Ho5om4OoZgCsg/woI4TjyyvJYn7qeNSlr2Ji2kZKKErxcvOgb2ZeEqAT6RvaVkUtCCNOcLegqpbyBEq21TSnVAmgFLNVaWy5T9BVCCCHqBCnqXofSgwdJ//PLeHXtSsiUyWbHqRUWm4UNqRv47sh3rE9bT4WtgtZBrXms02P0j+pPy8CWUmQTNaLUYuWxOYms3J/Fs8Na8Yf+sWZHurKCDDiyElK3GZt++YSBTwj4hhmFT59QcPc15sjWFYXZkLTKKOQmrYKSM6CcILIbDHwBmt8Eoe3BycnspEKIWpZWmMaaE8ZYhe2Z27FqKyGeIdza9FYSohPoFtYNN2c3s2MKIcSF1gN9lVKBwHJgK3AvMNbUVEIIIUQVSFH3Glnz80n94+M4+/kR8d67KJf6+6fTWnPg9AEWJi1k8dHFnCk7QwOPBoxtNZaRzUbSIrCF2RFFPVdSbuXhWdvYcDiHabe1ZXzPJmZHupi1AlK3wpEVRtEzw+hcx90fLMVgs/z6Pi6e4Bt6vsjrE/rr6z6h4N0QnE3498VmhZM7jE7cwyuM79HgHQIthkHzwdA0AbyCaj+bEMJUeWV57MzayY6sHWxM28jBMwcBaBbQjAfaPUBCVAJtg9vipOQkjxDCbimtdbFSaiLwT631m0qpnWaHEkIIIaqi/lYmq5G22Tj5zLNYTp6k8cyZuATXz7FLOSU5LD66mIVJCzl05hCuTq4kRCVwW7Pb6NWoFy5O8p+LqHkFpRYmfraNbcdP89ZdHbg7PsrsSIbCLKMb9/BySFoNpXmgnCGqOwx6CZoNgbD2xm1LzkBhpnEpyDz/fWGm0dWbfRCS1xmP8SsKvIMv6fYNuUL3r0/VnlPRqfPduEdWQslpoxs3Ih4S/mQUcsM6SjeuEA5Ea01qYSo7snYYl8wdJOUlAeCiXOjQsANT46eSEJVAtF+0yWmFEOKaKaVUT4zO3ImVx2QXZyGEEHWaVOmuwakPP6JwzRpC//QnvDrHmR2nWpVby1mbspaFSQvZmLYRq7bSPrg9L3R/gaExQ/F39zc7onAgucXl3P/JFn45mc/0UXHc2rGReWFsVmOcwtlu3F/NkR1idK56Bvz6vl5BxiWk9dV/h6W0stibBYUZ578vyDh/LPuAcdxW8ev7u3pf0O17hcKvT6hRJHZyBpvN6MA9+5zStgMavIKNcQrNh0DsQOnGFcKBWGwWDp4+eL6Im7WDnBJjvKSvqy8dQzoyvOlw4kLiaBfcTmbnCyHqqieA54D/aq1/UUo1BdaYnEkIIYSoEinq/obCH38ke/p0/EaMIPC++jFySWvN3py9fJf0HUuTl5Jfnk+IZwgT2k5gZOxImgY0NTuicEDZBWWM+3gzR7OL+Pd9XRjcJrT2Q5ybI1vZjXvpHNlmQyCsQ/V1rrp6QGBj43I1Nltl92/GFbp/MyFzHyStgbL8X99fORljHawWoxsXBRFdYMBzRjdueJx04wrhIArLC9mVvetcAXdPzh5KKkoAiPCJoHt4d+IaxhEXGkezgGYyUkEIUS9ordcB65RSXpXXjwKPm5tKCCGEqBop6l6F5eRJTj45FfdmzQh/dVqd3xAssyiTRUcXsTBpIUfzjuLu7M7A6IH8LvZ3dA/vjrOTrEAS5kjPK2HsjM2k55byyYSu9GleSyNObFZIS6zsXF0OJ3dizJFtCC2Gnu/GNbtz1ckJvBsYl9C2V79teTEUZV1+7IPW0HSA0Y3r3aA2kgshTJZRlEFiZuK5Iu7h3MPYtA0n5UTLwJbc3ux24kLjiGsYR6i3CSfThBCiFlSOXvgY8AGilVIdgUe01v9jbjIhhBDixklR9wpsZWWkPj4JXVFBxN+n4+TlZXakG1JaUcrqE6tZmLSQTembsGkbcSFxvNzzZW5qchO+br5mRxQO7sSpYsbM+JncYgszJ3aja5MaLqCemyO7HI6sOt+5GtkVEp43Crl1eY6smxe4NYHAJmYnEULUMqvNypHcIyRmnS/iZhRlAODp4knHhh15pMMjxIXE0aFhB7xdvU1OLIQQteZ94GZgIYDWepdSqp+5kYQQQoiqkaLuFWT+9W+U7t1L5P/9A/eYGLPjXBetNTuzd/Ldke/44dgPFFoKCfcO58H2DzIydiSN/X5jqbcQteRIViH3zdhMicXK7Ae70zHqMvNpq+qiObLLjc5cmSMrhKgHii3F7MnZc66Auzt7N4WWQgBCPEOIC41jQtsJxIXE0SKwhWx4KoRwaFrrlEtWXlrNyiKEEEJUB3l3fxm53ywg98svafDQQ/gOHmx2nGuWXpjOwqSFfH/0e47nH8fTxZMhjYdwW+xtxIfFy1w8YVf2p+dz34zNKAXzHu5B63C/6nvw4tNGF+6RFcbX4hzOz5F91ijkyhxZIUQdk12cfdGGZgdOH8CqrSgUzQKbMTxmuDFKISSORt6N6vzYKCGEqEYpSqlegFZKuQKTgP1Xu4NSKgqYCYQCGvhQaz1dKfUy8BCQXXnT57XWS2osuRBCCHEFUtS9ROm+fWRMm4ZXjx40nGT/s/OLLcWsPLGShUcWsiVjCxpN17CuPNj+QYY0HiJLK4Vd2pWSy/hPtuDp6szsh7oT29Cnag9os0H6TjiysrIbdztoG3g1gNhBld24g2SOrBCizrBpG0dzj7Ijewc7Mo0ibmphKgDuzu60D27PA+0eIC4kjo4hHfFzq8YTY0IIUf/8AZgORABpwHLgf3/jPhXAk1rrRKWUL7BdKbWi8mfvaa3frrG0QgghxDWQou4FrLm5pD4+CefAQCLeeRvlYp9/Hpu2sT1zO98d+Y4Vx1dQXFFMpE8kj3Z6lJGxI4nwiTA7ohBXtCX5NA98tpVAb1fmPNiDqKAbnFddfBqSVhuF3CMroSgbUNAoDvo9bRRyG8WBbAAohKgjDp05xPrU9ezI2sHOrJ3kl+cDEOQRRFxIHKNajSIuJI7WQa1xdXY1Oa0QQtQdWuscYOx13icdSK/8vkAptR+jKCyEEELYBfusWppA22ykPf00lsxMmnwxC5cG9tfRl5KfwsKjC/k+6XvSCtPwdvVmaMxQRsaOpHNIZ1lmKezehsPZPDRzGxEBnsx+sAdh/h7XfufSPMg5DEfXwOEVkLrV6Mb1DKzsxr0Jmg0C7+CaewJCCFHNUgpSWJq8lKXJSzmSewSAGP8YBjceTFyIMUoh2jdaXuOFEOIGKKX+gTE64bK01te0NFMp1QSIAzYDvYHHlFLjgW0Y3bxnqhxWCCGEuE5S1K2U889/UbR+A2F/fgnPjh3NjnNOkaWI5ceW8+2Rb0nMSkSh6BHeg8fiHmNQ9CA8XTzNjijENVmxL5P/nZ1I04befPFgd4J93H99o9J8OH0UTifBqcqvp4/CqaTKubiVGsVB36lGN25EF+nGFULUKdnF2Sw7toylyUvZk7MHgM4hnflT9z8xuPFggj3l5JQQQlSTbZVfewNtgPmV1+8G9l3LAyilfIBvgCe01vlKqX8Br2IUi18F3gEeuMz9HgYeBoiOjq7CUxBCCCEuT4q6QOH69eR88AH+t91GwKhRZsfBarOyJWMLC5MWsvL4SkqtpTTxa8KkzpO4pekthHmHmR1RiOuycNdJJs/fSbsIfz4f05qA/P1w7MLibeX3RdkX39G3ETSIhVYjIKip8X1Ud/AJMeeJCCHEDcory2Pl8ZUsTV56bgZ+66DWTOkyhaFNhhLuE252RCGEqHe01p8DKKUeBfporSsqr/8b2PBb96/cVO0bYLbWekHlY2Ze8POPgEVX+N0fAh8CxMfHX7FbWAghhLhRDl/ULU9NJe2pp3Fv0YKwl/9s+vJGi9XC2CVj2X96P76uvoyMHcnIZiPpENzB9GxCXLOygspC7VH27E6kbN9OFnmfomVxNk7Tsy6+rW84BMVCy2FG4TYo1ijeBsaA2w3O2xVCCDtQbClmXeo6lhxdwsaTG6mwVdDYrzGPdHyEYTHDaOrf1OyIQgjhKAIBP+B05XWfymNXpIwPXx8D+7XW715wPLxy3i7A7cDe6o8rhBBC/DaHLuraSktJe3wS2GxE/uPvOHmaP8pgxfEV7D+9n6e7Ps09Le/B3fkyS9RF3aE1lJyBwkwoyIDCLLAUg5sPuHlXXnzA/ZLrdWEDnLLCc4XbX3XcFp5rYKA9EOkahF9YK5yCO5/vuA2KhaAY4zkLIUQ9YbFa+PHkjyxJXsLalLWUVJQQ4hXC2FZjGdZ0GG2C2shJWiGEqH2vAzuUUmsABfQDXv6N+/QGxgF7lFI7K489D4xWSnXCGL9wDHikJgILIYQQv8Vhi7paazKmvUrpvn1E/uufuNnJnKM5B+YQ7RvN2NZjcVJOZscRV1JRZhRoCzMvLtgWVn49dz0TbJbrf3xn9/MFXjfvC4q+V/jq7vPrQvGl151v4P/u5UXnC7enkipn3CYb3xdmXHxbn1CjUNt8CAQ1ZXmGD+8lWolt2Z537uuFs4vMvRVC1E9Wm5XtmdtZkryEFcdXkF+ej7+7P7c0vYVhMcPoEtpFXtOFEMJEWutPlVJLge6Vh57RWmf8xn02YhSAL7WkuvMJIYQQN8Jhi7q5X31F3oIFNHj0D/gmJJgdB4BfTv3CruxdPN31afnwZwatoTQXCjLPF2svV7AtzDS6b39FgXewUdz0CYWGrcA39Px1n1DwDQNXTygvhvLCykvR+a9ll1y/6DZFUJh98fWKkmt/fi4eVy/6uvuAq5fx3M4WcgvSL34M7xCjy7bZYKPL9sKOW3ffyj+j5r0Vh/j79iPc2rER797TEVdn+e9r9voNAAAgAElEQVRZCFG/aK3Zm7OXJclL+OHYD2SXZOPp4snA6IEMjxlOz0Y9cXWqA6suhBDCcTgD2RifgVsopVpordebnEkIIYS4YQ5Z1C3Zs4fMV/+Cd+/eNHzsMbPjnDN3/1w8XTz5XbPfmR2lfqkoh6LKYuyvCrZnv6/8ubXs1/d38ThfkA1uDk36Vl6/pGDrHVz7YxNs1ouLvOWFVygMF118u7KCCwrFWVBecL6o7OFnFGqbJkCDS2bcevhdNY7Wmr8u3s+MjcncEx/Ja3d0wNlJlhkLIeqPpNwkliQvYWnyUlIKUnB1cqVvRF+GNR1G/8j+eLqYP8pJCCHExZRSbwD3Ar8AtsrDGpCirhBCiDrL4Yq62mYj/fnncW4YTKO330I528eS8NOlp1mavJTbm9+Or5uv2XHqrsxfYNMHkJ92fgxCyenL39arAfiEgU+IUaz1Cfl1V61PCLj7gb3OP3RyBg9/42Iym03zwnd7mbP5BBN6NeGlW9rgJAVdIUQ9kFaYxtLkpSxNXsqhM4dwUk50C+vGQ+0fYlDjQfi5Xf2ElxBCCNP9Dmiptb5MB4cQQghRNzlcUVc5ORHx/vvosjJcAq+64WmtWnB4AeW2cka3Gm12lLrr2EaYOxpQ0LCFsSFXdM/Ld9X6hNSNzcjqiAqrjae/3s2CHWk8OiCWp29uKRsBCSHqtJySHJYfW87S5KXszDb2x+nYsCPPdnuWm5vcTLBnsMkJhRBCXIejgCsgRV0hhBD1hsMVdQHcY2PNjnCRClsF8w/Op3t4d2ID7CtbnbH/e/h6IgQ2gXELwD/S7EQOo7zCxqR5O1i6N4OpN7XgsYHNzY4khBA3pKC8gJXHV7I0eSmbMzZj0zaaBzZnUudJDG0ylEhfeW0RQog6qhjYqZRaxQWFXa314+ZFEkIIIarGIYu69mZNyhoyijJ4rttzZkepm7Z9CounQEQXGPMleAWZnchhlFqsPPrFdtYczObFW9owsU+M2ZGEEOK6lFaUsi51HUuTl7IhdQPltnIifSKZ2G4iw2OG0yywmdkRhRBCVN3CyosQQghRb0hR1w7M2T+HCJ8I+kf2NztK3aI1rH8L1vwVmt8Ed38Gbt5mp3IYRWUVPDRzG5uOnuJvt7dnTPdosyMJIcQ1sdgs/HzyZ5YkL2H1idUUVxQT7BnMPS3vYVjMMNoHt5cRMkIIUY9orT9XSrkBLSoPHdRaW8zMJIQQQlSVFHVNdvD0QbZlbmNKlyk4O9nHpm11gs0KS5+BrR9Bx9Ew8h8yI7cW5ZVYeOCzrew4cYZ37+nI7XGyJFkIYf+yi7P5z+7/8MOxH8gty8XXzZdhMcMYFjOM+NB4eR0WQtQ/eamwYzY0HwIRnc1OYxql1ADgc+AYoIAopdT9Wuv1ZuYSQgghqkKKuiabd3Ae7s7u3NH8DrOj1B0VZbDgYdj3LfR6HIZMA+moqjWni8oZ/8lmDmYU8MGYzgxrH252JCGE+E1bM7by1LqnKLQUMjBqIMNihtE7ojduzm5mRxNCiOpVUQYHl0DiLEhaDWhw9XDooi7wDnCT1voggFKqBTAX6GJqKiGEEKIKpKhroryyPBYfXcyIpiPwd/c3O07dUJoP88dC8nq46S/Q649mJ3IoWfml3PfxZo6fKubDcfEktAoxO5IQQlyV1prPfvmM6YnTifKNYsZNM2ROrhCifsrabxRyd8+D4lPgFwH9noK4scZmwo7N9WxBF0BrfUgpJcv8hBBC1GlS1DXRt0e+paSihDGtxpgdpW4ozIIv7oSsfXD7h9DxXrMTOZS03BLGfvQzWQVlfPr7rvSKDTY7khBCXFVBeQEv/vgiq06sYkjjIUzrNQ0fNx+zYwkhRPUpK4C93xjF3LRt4OQKrYZD3HiITQAZK3PWNqXUDOCLyutjgW0m5hFCCCGqTIq6JrHarMw9MJfOIZ1pGdTS7Dj27/RRmHUHFGbC6HnGXDBRa47lFDF2xmbySy3MmtidLo0DzY4khBBXdejMIaasnUJqQSpPxT/FuDbjZPMzIUT9oDWc+Bl2zIJf/guWYmjYCm76K3QcBd5y4v0yHgX+F3i88voG4J/mxRFCCCGqToq6JtmQtoG0wjQmd5lsdhT7l74LvrgLbBVw//cQGW92IodyKLOAsTM2Y7Vp5j7Ug3YRMipECGHfvk/6nmmbjK7cj2/+mC6hMjJRCFEPFGbBzjmw4ws4dRjcfKD9XUZXbmS87DFxdS7AdK31uwBKKWfA3dxIQgghRNVIUdckc/bPIcQrhIHRA82OYt+S18PcMeDhDxMWQ8MWZidyKHvT8hj38WZcnJ2Y/3APmof6mh1JCCGuqNxazptb32T+wfl0Ce3C2/3fJthTOtaEEHWYtQKOrDS6cg8tM5oconpAnyegze/AXUbKXKNVwGCgsPK6J7Ac6GVaIiGEEKKKpKhrgqN5R9mUvok/xv0RVyeZz39Fv3wLCx6CoFi47xvwjzA7kUPZfvwMEz7dgp+HK7Mf7E6TYG+zIwkhxBWlF6bz5Lon2ZOzhwltJzCp8yRcnORtjhCijjqVZHTk7poLBeng3RB6/A/EjZMmhxvjobU+W9BFa12olPIyM5AQQghRVfJpxwRz98/F1cmVO5vfaXYU+7V1BiyeClHdjBm6XkFmJ3Iom5JOMfHzrYT4uvPFg92JDJT3vEII+/VT2k88s+EZLDYL7w14j8GNB5sdSYi6I/uQ0QWasgWCm0N4R+MS2hbc5IRurSovhv0LjU3Pjm8E5QTNhsDwt6HFzeAszSBVUKSU6qy1TgRQSnUBSkzOJIQQQlSJFHVrWWF5IQuTFjIsZhgNPBuYHcf+aA1rX4N1b0CLoXDXp+AmBcXatOZgFn+YtZ3oIC9mP9idED8PsyMJIcRl2bSNj3Z/xAc7PyA2IJb3BrxHE/8mZscSwv6VFcK+b43iYcrP4OQC4Z3gwGKjwAtGQbHB2SJvB+NrWAfwDDA3e32jNaTvNP632PM1lOVBYAwMfBE6jQG/RmYnrC+eAL5SSp0EFBAG3GtuJCGEEKJqpKhby75L+o7iimJGtxptdhT7Y7PC4idh+6fQ6T64dTo4y3+itWnZ3nT+OHcHLUJ9mTWxO0HebmZHEkKIy8ory+O5Dc+xIW0DI5qO4KUeL+HlKicBhbgirSF1G+yYCXsXQHmhUbQdMg06jgafEOM2eamQsdvYqDZ9NxzbCHu+PP84AY0vKPR2Mgq9vqHmPa+6qvg07PnKKOZm7gEXD2hzmzFeoXFvcHIyO2G9orXeqpRqBbSsPHRQa20xM5MQQghRVVIxq0U2bWPegXl0CO5Au+B2ZsexL5ZSWPAg7P8e+kyGQX+WHXxr2X93pDL1q910jPTn0993w99TlvgJIezTvlP7mLJ2CpnFmbzQ/QXuaXkPSl4zhLi8ohzYNc/owM0+AK5e0PZ2o3gY3ePi91tKQUCUcWk14vzxwmzI2HW+0Ju+yxgTcJZP2Plu3rMdvQHR8l7uUjYbJK8z/rfYvwisZUZhfMQ70O4u6YKuQZXzc6cAjbXWDymlmiulWmqtF5mdTQghhLhRUtStRZtObuJY/jFe6/ua2VHsS2kezBsLxzbAza9Bz/8xO5HDmbP5BH/6dg89Yhow4/54vN3lnwYhhH1acHgBf/35rwR6BPL50M/p0LCD2ZGEsD82KyStMbpyDywBmwUi4o1VUG3vAA+/63s8n4bQbLBxOas0DzL2nC/ypu+CIytB24yfewRcMLahstjbIBacnKvvedYVeamwYzbs/AJyTxh/my73G4X1cPk3rJZ8CmwHelZeTwO+AqSoK4QQos6Syk0tmnNgDg08GnBz45vNjmI/CjLgi7sgez/cMQM63G12Iofz8cZkXl20j4SWDfnXfV3wcHXAD1tC1DFKqaHAdMAZmKG1fv0yt7kHeBnQwC6t9ZhaDVnNSitK+dvmv/HfI/+lZ3hPXu/3OkEesommEBc5c6yyeDgH8lPBqwF0exg6j4OQ1tX7uzz8oUkf43JWeTFk7Ttf5M3YDZv/A9Zy4+euXhDW/nw3b3hHaNgKXOrhuKeKMji4xBivkLQa0BDT31iN1uoWcJU9C2pZrNb6XqXUaACtdbGSJR5CCCHqOCnq1pKU/BQ2pG7gkY6P4Co71xpOJcGs241lgWO+hGaDzE7kcP5v9WHeXn6IYe3CmD4qDjcXmd8mhL1TSjkDHwBDgFRgq1JqodZ63wW3aQ48B/TWWp9RSoWYk7Z6pBSkMGXtFA6cPsAjHR7h0Y6P4uyI3X5CXI6lFA4sgsSZxtJ+FMQOhJv/Ci2H127B1M0LIuONy1lWC2QfvLjQu3MOlH9o/NzZzSg4h10wviG0Lbh5117u6pS13yjk7p4HxafALwL6PQVxYyGwidnpHFm5UsoT40QnSqlYoMzcSEIIIUTV1GhR1xE7ia5k7sG5OCtn7m4hnagAnNxhdOii4f7vIbKL2Ykcitaat344yD/XJnF7XARv3dUBF2cp6ApRR3QDjmitjwIopeYBtwH7LrjNQ8AHWuszAFrrrFpPWU3Wpqzl+Y3PA/DBoA/oF9nP5ERC2ImMPUYhd/eXUJprzLAd8Dx0GmPMxLUXzq4Q1s64xI01jtlscPoopO88vynbgcXGrFkA5WRs4na2yBvewSiIKjt9r2KzwtE1RjE3bRs4uUKr4RA3HmITHHPkhP35M7AMiFJKzQZ6AxNMTSSEEEJUUY0VdR2xk+hKii3FfHv4W4Y0HkKIV718itcnaQ3Mvw88g2DcAghubnYih2KzaaYt2sdnPx1jTPdo/nJbO5ycZPWZEHVIBJBywfVUoPslt2kBoJT6EePE6sta62W1E696WG1WPtj5AR/t+YjWQa15d8C7RPpGmh1LCHOV5MLer43iYfpOo8u19a3GbNaY/uBkp0XPSzk5QXAz49L+LuOY1sbs2YwLZvQe2wh7vjQ36/Vo2Apu/ht0uBe8g81OIy6gtV6hlEoEegAKmKS1zjE5lhBCCFElNdmp61CdRFez6OgiCiwFjGldL5uQr8/eBbDgYQhuAfd9A37hZidyKFab5vkFe5i/LYWJfWJ4YURr2TFeiPrJBWgODAAigfVKqfZa69xLb6iUehh4GCA6Oro2M17R6dLTPL3+aTanb+aO5nfwfPfncXd2NzuWEObQ2ihu7pgF+76DilIIbQ/D3oT2d4NXPZktrZTRYRwQBa1GnD9emG0UeAtOmpftWoS0gYguxvMQdkcp1RvYqbVerJS6D3heKTVda33c7GxCCCHEjarJoq5DdBL9Fq01cw/MpXVQazo27Gh2HHNt/hCWPg3RPWH0XPAMMDuRQ7FYbTz55S4W7jrJ4wObMXlICynoClE3pQEXrq2OrDx2oVRgs9baAiQrpQ5hFHm3XvpgWusPgQ8B4uPjdY0kvg67snfx5NonOVN6hld6vcIdze8wO5IQ5shPh52zYccXcCYZ3P2M0Qqdx0N4J8cpHvo0hOaDzU4h6r5/AR2VUh2BKcDHwEygv6mphBBCiCowe6O0a+oksscuomu1NWMrR3KPMK3XNMctoGkNa/4K69+CliPgro/B1dPsVA6lrMLKH+fsYPm+TJ4Z2opHB8SaHUkIceO2As2VUjEYxdxRwKVLQb4FRgOfKqWCMU6iHq3VlNdJa828g/N4c+ubhHqF8sXwL2jdoLXZsYSoXVYLHPrB6Mo9vBy0DRr3gQHPQuuRxkZkQogbUaG11kqp2zBWin6slJpodighhBCiKmqyqFttnUT21kV0PeYcmEOAewDDYoaZHcUc1gpYPNnYyKPzeBjxHjibfS7BsZSUW3nki+2sP5TNKyPbcn+vJmZHEkJUgda6Qin1GPADxiqXT7TWvyilpgHbtNYLK392k1JqH2AFntJanzIv9dUVW4p5ZdMrLEleQr/Ifvytz9/wd/c3O5YQtSfnsPFeadc8KMoCnzDo/QTE3QcN5ESsENWgQCn1HHAf0E8p5QS4mpxJCCGEqJKarK7Vy06i63Gy8CRrUtbw+7a/x8PFw+w4tc9SAt88CAcWQb+nIOFPjrNU0E4UllXwwGdb2XrsNG/e2YF7utrRbthCiBumtV4CLLnk2EsXfK8xlpdOqeVo1y05L5kpa6dwNO8oj8c9zsT2E3Gy1x3uhahOZYWw71tj07OUn0E5Q8thxqZnzQbLSXAhqte9GJ9FJ2qtM5RS0cBbJmcSQgghqqTG3i3Wx06i6zX/4HwA7m15r8lJTFCSC3NHw4lNxkYe3R8xO5HDySu2MP7TLexNy2P6qDhGdmxkdiQhhLjIiuMrePHHF3FzcuPfg/9Nz0Y9zY4kRM3SGlK3wY6Zxuax5YXQoBkMfgU6jgbfULMTClEvaa0zgHcvuH4CY6auEEIIUWfVaAtAfeokul6lFaV8c/gbBkYNJNwn3Ow4tSs/Hb64E3IOGfNz291pdiKHk1NYxriPt5CUVci/xnbmprZhZkcSQohzLDYL07dP5/N9n9MhuAPvDHiHMG/5d8rupW2H8iJjxquTdFNfl6JTsHue0ZWbvR9cvaDt7UZXbnQPWckkhBBCCCGum6zrqiFLk5eSV5bHmNaXTpyo53KOwKzboeQ0jP0KYhPMTuRwMvJKGTvjZ9JyS5hxfzz9WjQ0O5IQQpyTXZzN1HVTScxKZHSr0TwV/xSuzjLW0K7lnoDlLxqjAgD8oyFuLHQaCwEy1ueKbFZIWmN05R5YAjYLRMTDrdOh7R3g4Wd2QiGEEEIIUYdJUbcGaK2Zc2AOzQKaER8ab3ac2pO2HWbfDSiYsAgaxZmdyOGknC5m7IzNnC4qZ+YD3ekWE2R2JCGEOGdbxjamrptKcUUxr/V9jVua3mJ2JHE15cXw43T48X1AwYDnIbgZ7PgC1r5uXGITjG7TViPAxd3sxPbhzDHYMRt2zoH8VPAMgm4PGX+n0DZmpxNCCCGEEPWEFHVrwM7snRw4fYCXer6EcpTldEdWwfxx4N0Axn0rOzWb4Gh2IWNnbKa43MoXD3anU1SA2ZGEEAIwTnZ+/svnvJ/4PlG+UXx000c0D2xudixxJVrDLwtg+UtGUbLtHTBk2vmu3HZ3Gt27O2bDztnw9e+NwmWHe6HzOAhta25+M1hKjY1hE2dC8jpAQexAuPkv0HK4FLyFMJlSqjfwMtAY4zOwwpgG2NTMXEIIIURVSFG3BszZPwdfN19GxIwwO0rt2P0VfPsHaNgK7vsGfGUuYm07kJHPfTO2AJp5D/egdbgs6RRC2IfC8kJe/PFFVp5YyeDowbza+1V83HzMjiWuJH0XLH0WTvwEYe3hzo+gca9f3y4gGhKeg/5Pw9G1sGMWbPsYNv8LGnWGzuON4m99HzGQsceYk7t7PpTmGqMpBjwPncbIaAoh7MvHwGRgO8YG3UIIIUSdJ0XdapZVnMXK4ysZ03oMXq5eZsepeT//C5Y9a2yaMnoOePibncjh7E7NZfwnW3B3cWL2gz1pFiLFEiGEfTh85jCT104mtSCVqfFTGd9mvOOsYKlrinJg9auw/XPwCjLmvsaNAyfnq9/PyRmaDTIuRadgz5dGt+qiJ+CH56HN74zu3eie9WczsJJc2Pu1UcxN3wnObtD6VuPvFdNfNpETwj7laa2Xmh1CCCGEqE5S1K1mXx78Equ2MqrVKLOj1CytYdU02Piu8UHmjhng6mF2Koez7dhpfv/pVvy9XJnzYA+iGzjAiQQhRJ2w6Ogipm2ahrerNzNumkF8mAPNmK9LrBbY8pExH9dSBD0ehf7PgOcNjPDxbmDcv/sfIC3R2CBszzewaw40aAZx90HHMeAbWv3Po6ZpDcc2Gh3J+76DilIIbQdD34AO9xiFcCGEPVujlHoLWACUnT2otU682p2UUlHATCAU0MCHWuvpSqkgYD7QBDgG3KO1PlMz0YUQQojLk6JuNSq3lvPVoa/oF9mPKN96vOTOWgGLJhkbpXSZACPe/e1OHlHtfjySw4OfbyPc34PZD3Un3N/T7EhCCEG5tZy3tr7FvIPz6BzSmbf7v01Dr4ZmxxKXc2QVLHsOcg5C7CAY+ho0bFn1x1UKIrsYl5v/ZhRBE2fBypdh1avQ4majq7X5TeBs529F89ONonTiLDiTDO5+xmiFuHHGhrD1pftYiPqve+XXC88wamDgb9yvAnhSa52olPIFtiulVgATgFVa69eVUs8CzwLPVHNmIYQQ4qrs/J103fLDsR84XXqaMa3GmB2l5pQXw9cPwKGlRifPgOfkA40JVu3P5NHZiTQN9mbWxO409JUNWIQQ5sssymTK2insztnNhLYTeLzz47g6uZodS1zqVBIsfwEOLoHAGBg9D1oMrZnXczdvowjaaQzkHDE6XXfNNX63Tyh0HG3M37WnDVatFjj0g5H18HLQNmPM1IBnofVIcJNVMULUNVrrhBu8XzqQXvl9gVJqPxAB3AYMqLzZ58BapKgrhBCilv1mUVcpdSuwWGttq4U8ddrcA3Np4teEHo16mB2lZhRkwtxRcHIHDH8buj1kdiKHtHh3OpPm7aBtIz8+f6AbAV5uZkcSQtyg+vYa6+bsRnFFMe8OeJchjYeYHUdcqqwA1r8NP//TmAM7+BVjXIJLLZ0YDG4GQ16BgS8YxdLEWfDTP+DH96Fxb6P7tc1t5hVNcw4b84B3zYOiLPAJg95PGGMj7KnoLIS4bkqpUOBvQCOt9TClVBugp9b64+t4jCZAHLAZCK0s+AJkYIxnEEIIIWrVtXTq3gu8r5T6BvhEa32ghjPVSXuy97AnZw/PdXsOJ1UPN8jI2g+z74biUzBqDrQabnYih/T19lSe/noXXRoH8smErvh6SAecEHVcvXqNDfQI5Otbv8ZZRvLYF5sNds8zxh8UZkKnsTDoJfANMyePsyu0GmFcCjJg5xyjK/bbP8DSp6Hdncbmao061/xqoPIi+OW/RoE55WdQzkbXcudx0GyI/Y+HEEJcq8+AT4E/VV4/hDET95qKukopH+Ab4Amtdf6Fm35qrbVSSl/mPg8DDwNER0dXJbsQQghxWb/5TlVrfZ9Syg8YDXxW+YL1KTBXa11Q0wHrirkH5uLt6s1tzW4zO0r1S1oDX44HV0/4/RJjhpyodbN+Ps6L3+6lb/Ng/jOuC15u8kFTiLquPr7GSkHXzqRuMwqladshIh5GzTVm3doL3zDoOwX6TIbjP1WOZ5gH2z+FkLZGcbXDvdW7EZnWxt9lx0zYuwDKC42N3Aa/YoyDqIsbuQkhfkuw1vpLpdRzAFrrCqWU9VruqJRyxSjoztZaL6g8nKmUCtdapyulwoGsS++ntf4Q+BAgPj7+V0VfIYQQoqquqSpUeTbya8ATeAK4HXhKKfV3rfU/ajJgXXCq5BTLji3jrhZ34e3qbXac6pU4ExZNhuCWMGY+BNTjDeDs2Efrj/LXJfsZ3DqE/xvTGQ9XKZoIUV/Ia6yoEQUZRmfurrnG7Nrf/dsojjrZ6WoipaBJb+My7A3Y+43RPbvsWVjxktHV23k8xAy48edQdMroWE6cBdn7wdUL2t5ujH2I7iF7BAhRvxUppRpgbI6GUqoHkPdbd1JGS+7HwH6t9bsX/GghcD/weuXX76o9sRBCCPEbrmWm7kjg90AzYCbQTWudpZTyAvYBDv+B8+tDX2OxWRjdarTZUaqPzQarp8HG94wdse/+DDz8zE7lcLTWTF91mPdXHuaWDuG8d28nXJ3t9AO5EOK6yWusqHYVZbDpA9jwDljLjQ7Yvk+Cu6/Zya6dhz/EP2BcMvYa3bu75xtjEvyjIW6sMULiWk4026zGiqMdM+HAErBZjI7lW6dD2zvkvY0QjmMKRiE2Vin1I9AQuOsa7tcbGAfsUUrtrDz2PEYx90ul1ETgOHBP9UcWQgghru5aOnXvBN7TWq+/8KDWurjyRcyhWWwWvjz4Jb0a9SLGP8bsONXDUgL//QPs+xa6/B6Gv2XMvxO1SmvN60sP8J/1R7mrSyRv3NkBZyfpIhKinpHXWFE9tIaDS+GH5+FMMrQcATe9Wvc3+AprZ3TuDn4FDi42umzXvm5cYhOMLttWI3692duZY7BjtjGvNz8VPIOg28PGpmehbUx5KkII82itE5VS/YGWgAIOaq0t13C/jZW3v5xB1RhRCCGEuG7XUtR9GTi7sydKKU+M3T6Paa1X1VSwumLViVVklWTxUs+XzI5SPYpyYO5oSN0KQ16FXn+U5YgmsNk0f174C7N+Ps74no15+da2OElBV4j66GXkNVZUVdYB+OE5SFptjEsa91+IHWh2qurl6mFsoNbuTjhzHHbONoq2X//eKNh2uBc6jYacw0Zn79G1gDL+Djf/BVoO/3XhVwhR7yml+lQWZtFaVwC/XPJzPyBaa73XjHxCCCFEVVxLUfcroNcF162Vx7rWSKI6Zu7+uUT6RNInoo/ZUaou+xDMuduYw3fP59CmHm76VgdUWG08880evklM5ZH+TXl2aCuUFNaFqK/kNVbcuJIzsPYN2PIhuPvA0Deg68T6v7omsDEkPA/9n4Gja4zu3a0zYPO/jJ/7R8OA56HTGNkLQAhxp1LqTWAZsB3IBjwwxh4lAI2BJ82LJ4QQQty4aynqumity89e0VqXK6XcajBTnXHg9AESsxKZGj+17u/2nbwB5o8FZzeYsBgi481O5JAsVhtPzN/J4t3pTBnSgj8ObCYFXSHqN3mNFdfPZoXEz2H1X6D4NHSZAANfAO9gs5PVLidnaDbYuBSdggOLICAaYvrb74ZwQohapbWerJQKwhh3dDcQDpQA+4H/nO3iFUIIIeqiaynqZiulRmqtFwIopW4Dcmo2Vt0wZ/8cPF08ub357WZHqZqdc2Dh4xDUFMZ+ZXTAiFpXarHy2JxEVu7P4oURrXmwb1OzIwkhap68xorrc+xHWPoMZO6Bxr1h6OsQ3sHsVObzbgBd7jc7hRDCDmmtTwMfVV6EEEKIeuNairp/AGYrpf4PY0h8CjC+RpdVgAkAACAASURBVFPVAbmluSxJXsLI2JH4udXRnZO1hrWvwbo3IKYf3DMLPAPMTuWQissreHjmdjYeyeEvv2vHfT2ksC6Eg5DXWHFtck/Aipfgl/+CXyTc9Sm0vV3m3gshhBBCCOGgfrOoq7VOAnoopXwqrxfWeKo6YMGRBZRZyxjVapTZUW5MRRl897+w5ytjJ+gR74GLrPg1Q36phQc+3UriiTO8c3dH7uwSaXYkIUQtkddY8ZvKi+HH6fDj+4CCAc9Br8fBzcvsZEIIIYQQQggTXUunLkqpEUBbwOPsfE+t9bQazGXXrDYr8w/Mp2tYV1oEtjA7zvUrPg3zxsKJn2DQS9BninT6mORMUTn3f7qFfSfz+b8xnRnePtzsSEKIWiavseKytDa6cpe/CPmp0PYOGDJNNv4SQgghhBBCANdQ1FVK/RvwwtgddAZwF7ClhnPZtbWpazlZdJKnuj5ldpTrdyoJZt8Nealw1yfQ7k6zEzms7IIy7puxmeRTRXw4vgsDW4WaHUkIUcvkNVZcVvpuWPYsHP8RwtrDHR9Ck95mpxJCiDpNKdULaMIFn4G11jNNCySEEEJU0bV06vbSWndQSu3WWr+ilHoHWFrTwezZ3P1zCfMOY0DUALOjXJ/jm2DeaFBOcP/3EN3d7EQOKz2vhLEfbSYjv5TPJnSlVzMH27FcCHGWvMaK84pOwepXIfFz8AyEW96HzuPBydnsZEKY6kxROduOnzE7Rp3WPMSHJsHeZscwjVJqFhAL7ASslYc1IEVdIYQQdda1FHVLK78WK6UaAacAh10jfuTMETZnbGZS50m4OF3T9Ar7sOdr+PZRCIiGsV9BUFOzEzmsE6eKGTPjZ/KKLcya2I0ujYPMjiSEMI+8xgqwWmDrDGPz0rJC6PYIDHjGKOwKIXjl+1/4dudJs2PUac8Oa8Uf+seaHcNM8UAbrbU2O4gQQghRXa6lKvm9UioAeAtIxDij+VGNprJjcw/Mxc3JjTub15GxBVrD+rdhzV+gcW+49wvwqr9FxMKyCpb/koHVZp/v16w2zXsrD1FeYWPuwz1oF+FvdiQhhLnkNdbRHVkFy56DnIPQNAGGvg4hrcxOJYTd0Fqz8UgOg1uH8MTgOriXhZ0I8XM3O4LZ9gJhQLrZQYQQQojqctWirlLKCViltc4FvlFKLQI8tNZ5tZLOzuSX5/P90e8Z3nQ4gR51oHumohy+nwS75kCHUTDy7+BSv9/QvbnsADM3HTc7xlWF+Loz7+GetAzzNTuKEMJE8hrr4E4lwfIX4OASCIyBUXOh5TDZuFSISxzOKiSnsJyb2oTJyXBRFcHAPqX+n737jo6y2tc4/t0JkFACSBXpJVJSiCFRAlIUKRawcBSxACKgWLCinOs5il7PFRVFwIIoAscDWADFI4qAEAtNAiICBkIkUgwIoSakZ98/EsYgCQlkJu8keT5rZZH3nbc8M8b1m71nz97mByD91E5r7QDnIomIiJTMWTt1rbU5xpg3gEvyttPJVwQrmkU7F5Galcpt7W5zOkrRUo/Ah3dCwnfQ8+/Q48ly31Dcc/gk837Yzc2dmjCmV6DTcQpVP8AP/8qaH1GkolONraDST+R+g2btm+BbBa4aD53vK/cfuoqcr9U7DwEQ1bquw0mkjBvvdAARERF3K870C18bYwYCCyvyHEQ5NocPYj/gkgaX0L5ue6fjnN3hXTD3ltx/b5wOHQc5nahUvLY8DmMMj/Vpy4W1/J2OIyJSHKqxFUVODmz+AJaPh+QD0PE2uOoZCLjQ6WQiXm11fBJN61SlaZ1qTkeRMsxa+40xpiEQmbfrB2vtH05mEhERKanidOreAzwKZBlj0gADWGttTY8m8zLf7/ue3Sd288AlDzgd5ez2rId5t0JOFgxZBC26Op2oVMQdOMEnP+7l7stbqkNXRMoS1diKYG8MfPkk7IuBxp3g1rnQJMLpVCJeLzvHsvbXJK4O1vqRUjLGmFvInb8+mtxaO9UYM9ZaO9/RYCIiIiVQZKeutVYTfwJzY+dSv2p9rmp+ldNRCrf1E/jkXghoBLd/DPW8dwoCd3t12Q6qVvZldM82TkcRESk21dhy7sR+WP5s7tz2NRrCDdMgdBD4+DidTKRM2Pb7cY6nZdGljaZekBJ7Cog8NTrXGFMfWA6oU1dERMqsIjt1jTHdC9pvrf3W/XG8U8KxBFbtW8V9YfdR2aey03HOZC2sei33K51NL8tdbKV6xXnz+/PeY3y5ZT9jegVSp3oVp+OIiBSbamw5lZWeO2futxMhOwO6PgzdHwc/9eGLnIvV8Xnz6baqOO9rxWN8/jLdQhKgT9hERKRMK870C2Pz/e4PXApsAK70SCIv9MH2D6jkU4mbL77Z6Shnys6ExY/BxtkQdBPc8BZUrljTD0xcup3a1SozoltLp6OIiJyrCl9jyxVrYfuX8NX/wJFd0PYa6PM81G3tdDKRMmlVfBJtGtSgQc2K9d5WPGKJMeYrYF7e9iDgCwfziIiIlFhxpl/on3/bGNMUeM1jibxMSmYKn+78lL4t+lKvaj2n45wu7Rh8NBR+XQndHocrnqpwX+lc92sS3+w4yN+vbkdNfy8cRS0ichYVvcaWK3/Ewld/h/gVUK8t3LEQ2vRyOpVImZWRlcP6XYe5JaKJ01GkHLDWjs1bmPTUgiPTrbWfOJlJRESkpIozUvev9gLt3R3EW30W/xkpmSnc1u42p6Oc7uhumHMLJMXB9W/AJXc4najUWWuZuHQ7DQL8GBLVwuk4IiLuUKFqbLmQegSiX4QfpkOVGtBvAkSOAF990ChSEj/tPUpqZjZRrb1sUIWUWdbaBcACp3OIiIi4S3Hm1J0K2LxNHyAM2OjJUN7CWsu82HkE1w0mtH6o03H+tG8DzL01d86+OxZAq55OJ3JE9I6DrE84wv/eEEzVKr5OxxEROWcVucaWeTnZuVMfrXgeTh6GTsPgyn9AdXVAibjD6p1JGAOdW9VxOoqUYcaY7621lxtjTvBnvQUwgLXW1nQomoiISIkVZ6RuTL7fs4B51tpVHsrjVdYmrmXXsV383+X/53SUP/3yOSwYATXqw9D/QoN2TidyRE6OZeJX22lapyqDIpo6HUdE5HxV2BpbpiWsgiVPwv6foXnX3NG5jbzow1+RcmB1/CGCLqpJ7WpaBFfOn7X28rx/tVKliIiUO8Xp1J0PpFlrswGMMb7GmGrW2pOejea8ubFzqeNfh74t+jodJXfxlbVvwldPQeNwGPwB1GjgdCrHfLllP1t/P86rt3SkSqWKNY+wiJQrFbbGlklH98Cyf8LWT6BmE/jbTAi6EYxxOplIuZKakc2Pu48yrGsLp6NIOWGMed9ae2dR+0RERMqS4vSGfQ1UzbddFVjumTjeY++JvXyz5xsGBg6kiq/DIwSys+CLx3NX027fH4Z+XqE7dLOyc3hl2XYCG9Tg+rDGTscRESmJClljy5yMkxA9AV6PhO1LoOff4YH1EHyTOnRFPGDDb0fIyM4hqnVdp6NI+RGUf8MYUwno5FAWERERtyjOSF1/a23yqQ1rbbIxppoHM3mFD7d/iI/x4Za2tzgbJP0EfHwX7FwGXcbAVc+CT8Uembrwx338ejCFaXd0wtdHjWkRKdMqZI0tM6zNHZW77Gk4tgeCboLez0FtTfsj4kmr4w9RyccQ2ULz6UrJGGP+DvwPUNUYc/zUbiADmO5YMBERETcoTqduijEm3Fq7EcAY0wlI9WwsZ6VmpbIwbiG9mvXiwuoXOhfk2D6YOwj+2AbXTYKI4c5l8RLpWdlMXh5HaJNa9A1q6HQcEZGSqnA1tsxI3AxLxsFvq+DCELjxbWjR1elUIhXCqvgkOjatTQ2/4jRVRApnrX0BeMEY84K19u9O5xEREXGn4rxTehj42BjzO7mfal4IDPJoKoct/nUxxzOOc1v725wLkfhTboduejLc/hG0ucq5LF5k3rrd7DuayoSBIRh95VVEyr4KV2O9XsohWPE8bJwNVS+A616D8CHg4+t0MpEK4XhaJj/vPcoDV7RxOoqUL18aY7r/dae19lsnwoiIiLhDkZ261tr1xph2QNu8XduttZmejeUcay1zY+fS9oK2hDcIdybEjq9yp1yoegHc/RU0DCr6nArgZEYWr6/cSedWdbi8TT2n44iIlFhFq7FeLTsT1r8L0S/kfqB66T3Q88ncWiwipeaHXw+TYyGqtd7riVuNzfe7P3ApsAG40pk4IiIiJVdkp64x5n5gjrV2S972BcaYwdbaNz2ezgEbDmwg7kgcz3Z51pmRoEcS4MM7oEF7uO0jCHBw+gcvM2t1AoeSM3j7zrYapSsi5UJFq7Fea+fXsOTvcGg7tL4S+r4ADdo5nUqkQlodn4RfJR8uaVbb6ShSjlhr++ffNsY0BV5zKI6IiIhbFGfFrZHW2qOnNqy1R4CRnovkrLmxc6nlV4trWl7jTICvnwPjC4M/UIduPsdSM5kWHU+vdg3o1FyLZohIuVGhaqzXSYqHeYPhPzdBdgbcOg/uWKgOXREHrY4/RESLC/CvrClPxKP2Au2dDiEiIlISxZlT19cYY6y1FsAY4wtU8WwsZ+xP2c+K3SsYEjQE/0r+pR9g3wbYsgC6j4WaF5X+/b3YO9/+yvG0LB7r07bog0VEyo4KU2O9SvoJ+HYirH0TfKvAVeOh831Qyc/pZCIVWlJyOrH7TzC2r97viXsZY6YCNm/TBwgDNjqXSEREpOSK06m7BPjQGPN23vY9wJeei+Scj7Z/hMUyqK0Da9RYC0v/CdXrQ9eHSv/+XuzgiXTeW7WL60Ib0eGimk7HERFxpwpTY72CtfDTB7B8PCTvh463wVXP6JsxIl5i7a+HAYhqXdfhJFIOxeT7PQuYZ61d5VQYERERdyhOp+6TwCjg3rztzeSuzl2upGenM3/HfHo06UHjGo1LP8D2L+G3VXDtK+AXUPr392JvRu8kPSuHR3tf7HQUERF3qxA11mt8/2ruNEeNO8Gtc6BJhNOJRCSf1fGHqOFXidDGtZyOIuXPfCDNWpsNud+MMcZUs9aedDiXiIjIeStyTl1rbQ6wDkggd5XQK4FfPBur9C3ZtYQj6Ue4rf1tpX/z7CxY/gzUDYTwoaV/fy+272gqc9bu5m/hTWhVv4bTcURE3Kqi1FivsHstrPgXBA+Eu5erQ1fEC62OT+LSlnWo5FucZT9EzsnXQNV821WB5Q5lERERcYtCR+oaYy4GBuf9HAI+BLDWXlE60UqPtZa5sXNpXas1l114WekH2DgbDu2AW+eCb+XSv78Xm7I8DoAxVwU6nERExH0qUo31CicPw/y7oXYzuO418FGHkYi3+f1oKrsOpXD7Zc2cjiLlk7+1NvnUhrU22RhTzclAIiIiJXW2Vk0suSOGrrPWXm6tnQpkl06s0vXTwZ/YlrSNwe0GY4wp3Zunn4DoF6BZF2h7Tene28vFH0xm/sa93N65GY1rVy36BBGRsqPC1FjHWQuL7ofkA/C398Bfc7OLeKM18UkAdGldz+EkUk6lGGPCT20YYzoBqQ7mERERKbGzdereBCQCK40x7xhjegGl3ONZOubGziWgcgD9W/cv/ZuvmgIpB6HP81DaHcpebtKyHfhV8uG+nm2cjiIi4m4lrrHGmH7GmO3GmJ3GmHFnOW6gMcYaYyrmfAPr3obtX0Dv56BxeNHHi4gjVscncUG1yrS7UGtLiEc8DHxsjPnOGPM9ud+QecDhTCIiIiVSaKeutfZTa+2tQDtgJbmFsIEx5i1jTJ/SCuhpB08eZFnCMq5vcz3VKpfyN3COJ8Ka1yHoJmjSqXTv7eW2/n6MzzcnMrxrS+oH+DkdR0TErUpaY40xvsAbwNVAB2CwMaZDAccFAA+RO29vxfP7Jlj2T7j4aug82uk0IlIIay1r4g8R1bouPj4a5CDuZ61dT27NHU3u4qTtrbUbijrPGPOeMeYPY8yWfPvGG2P2GWM25f3o65YiIuKI4iyUlmKtnWut7Q80AX4kd7XuIpWFUUTzd8wny2YxuN3g0r41rPwXZGdCr6dL/95e7pWlO6jpX4mR3Vs5HUVExGNKUGMvBXZaa3+11mYAHwDXF3Dc/wIvAmnuylxmpJ+A+XdB9fpww5v6NoyIF/st6SS/H0sjSlMviIfkzZ/7JPCQtXYL0MIYc10xTp0F9Ctg/yRrbVjezxdujCoiIlJs57RSiLX2iLV2urW2V1HHloVRRJnZmXy04yMub3w5zWqW8qIMB7bBpjlw6Sio07J07+3lNvx2mBWxf3Bvz9bUqqqF40SkYjiXGgs0Bvbk296bt88lb+7AptbaxW6MWTZYC58/AkcSYOC7UK2O04lE5CxWu+bTretwEilPjDHXGWNq5G3OBDKAqLztfcDzRV3DWvstcNgzCUVERErGk8s/e/0oomW/LeNQ6iFua3dbad8alj0NfgHQ/fHSv7cXs9by0pLt1Kvhx7AuLZyOIyJSJhljfIBXgceKcewoY0yMMSbm4MGDng9XGn78D/z8MfT8H2jexek0IlKE1fGHaFjTj1b1qjsdRcqXX4Fpeb+3tta+BGQCWGtPUrL1Yh4wxmzOm57hghLmFBEROS+e7NT1+lFEc2Pn0rxmc7o27lq6N/41GnYug26Pa/TQX3y/8xDrdh3mwSvbUK1KJafjiIh4q31A03zbTfL2nRIABAPRxpgEoDPwWUHTHOWNDo6w1kbUr1/fg5FLycHt8MVYaNkduj3qdBoRKUJOjmVNfBJdWtfDaJoUcSNr7Tbg73mbGcaYqoAFMMa0BtLP89JvAa2BMHIXPX2loIPK5YemIiLiVTzZqXtWTo8i2pq0lZ8O/sStbW/Fx5Tiy5CTA0v/CbWa5U69IC7WWl7+ajuNa1fl1kubFn2CiEjFtR4INMa0NMZUAW4FPjv1oLX2mLW2nrW2hbW2BbAWGGCtjXEmbinJTIWPh0GV6nDTO+Dj63QiESnCjj9OkJSSoakXxCOstacGGY0HlgBNjTFzgK8p5joxBVzzgLU221qbA7xD7jdUCzqufH1oKiIiXseTvZlePYpo7i9zqVqpKte3KWhGCA/6+SPYvzl3cbTK/qV7by/31dYDbN57jIevCsSvkhriIiKFsdZmAQ8AXwG/AB9Za7caY54zxgxwNp2Dlvwd/tgGN74NARc6nUZEimH1ztz5dKPUqSseZK1dCtwEDAPmARHW2pXncy1jTKN8mzcCW0ocUERE5Dx48vvtrlFE5Hbm3gq4Jq+11h4DXEvcGmOigcdLYxTR4bTDLNm1hBsDbySgSoCnb/enzFT4+n+hURgEDyy9+5YB2TmWV5Zup3X96tx4SeOiTxARqeDyVtv+4i/7ni7k2J6lkclRWz+BDTOh60MQeJXTaUSkmFbHJ9G8bjWaXFDN6ShSjhlj3gceODXtnzGmuTHmw6IWJzXGzAN6AvWMMXuBZ4CexpgwcqdySADu8WR2ERGRwnisU9dam2WMOTWKyBd479QoIiDGWvvZ2a/gOQvjFpKRk1H6C6StmwbH98KN08DHsZkvvNKiTfuI+yOZN28Pp5KvXhsRETkHh3fBZ2OgSSRc+U+n04hIMWVl57Du1ySu69io6INFSuZ7YJ0x5lFy13kZSzGmAbTWDi5g9ww3ZxMRETkvHl2JyhtHEVlrWbJrCZ0bdaZV7ValcctcKUnw3atwcT9o2a307lsGZGTlMGn5DoIuqkm/IH1dVkREzkFWBswfDsbAwBngW9npRCJSTFt/P86J9CyiWtcr+mCRErDWvm2M2QqsBA4Bl1hr9zscS0REpEQ82qnrjYwxvH/N+xxOO1y6N/72JchIhqueLd37lgEfxuxhz+FUZt4VjI+PVj0WEZFz8PWz8PtGuOV9uKC502lE5Bysjs+bT7eV5tMVzzLG3An8ExgChAJfGGPustb+5GwyERGR81fhOnUBqlaqSuMapThva1I8rH8XwodAg3ald98yIDUjm6lfxxHZ4gJ6XqxVYUVE5Bzs+ArWvA6RI6BDxV0fTqSsWh1/iIsb1qB+gJ/TUaT8Gwhcbq39A5hnjPkEmA2EORtLRETk/Gny0tLw9bPg6wc9/8fpJF7n32sS+ONEOmP7tsMYjdIVEZFiOv47fHIvNAyBPv9yOo2InKP0rGzWJxymi6ZekFJgrb3BWvuHMaZa3vYPwKUOxxIRESkRdep62p4fYNsi6DoGAho6ncarHE/L5K1v4ulxcX0ubVnH6TgiIlJW5GTDgpGQlQ43z4TK/k4nEpFztGn3UdIyc+jSWlMviOcZY6KMMduA2LztjsBrzqYSEREpGXXqepK1sPQfUKMhRD3gdBqv8+53uzh6MpPH+7R1OoqIiJQl37wEv30P174C9QKdTiMi52F1fBI+Bi7TfLpSOl4D+gJJAHlz6XZ3NJGIiEgJqVPXk375L+xZB1f8D/jVcDqNV0lKTmfGd79yTciFhDSp5XQcEREpK3Z9l7v4aMfBEDbY6TQicp7WxCcR3LgWtapWdjqKVBDW2j1/2ZXtSBARERE3Uaeup2RnwvLxUL8dhN3hdBqvM+2beFIzs3m098VORxERkbIi5RAsGAF1WsE1E51OIyLn6WRGFj/uOUKUpl6Q0rPHGNMFsMaYysaYx4FfnA4lIiJSEurU9ZSYmXA4Hno/B76VnE7jVRKPpTJ7zW/cFN6ENg0CnI4jIiJlQU5O7sJoqUfgbzP1DRiRMiwm4QiZ2VaLpElpuhe4H2gM7APC8rZFRETKLPU2ekLaMfhmArToBoF9nE7jdaau2Im1lod6aR5EEREpprVvwM5luSN0G4U6nUZESmB1fBKVfAyRLS5wOopUENbaQ8DtTucQERFxJ3XqesL3r8HJJOjzv2CM02m8SsKhFD5av4fbL2tG0zrVnI4jIiJlwd4NuVMatbsOIkc4nUZESmhN/CEuaVabalXUFBHPMsY8A1gg2Vr7qtN5RERE3EnTL7jbsb2w9k0IuQUuusTpNF7nteU7qORruP/KNk5HERGRsiDtGMy/CwIugutf14elImXcsdRMft53jChNvSClIwH4DdjrcA4RERG308fj7rbiX2At9Pqn00m8Tuz+4yz66Xfu7dGaBgH+TscRERFvZy18Nib3A9PhS6CqvqotUtat+zWJHAtdtUialAJr7WynM4iIiHiKOnXdaf/P8NM86PIg1G7mdBqv88rSHdTwq8Q93Vs5HUVERMqCDTNh26dw1XhoeqnTaUTEDVbHJ+Ff2YewZrWdjiIVgDHmv+ROv1Aga+2AUowjIiLiVurUdael/4SqtaHbY04n8To/7j7Csm0HeLzPxdSuVsXpOCIi4u0ObIUlf4fWvaDLQ06nERE3WROfRGSLOvhV8nU6ilQME/P+vQm4EPhP3vZg4IAjiURERNxEnbrusnM5/LoS+r6Q27Erp5m4dDt1q1fhrq4tnY4iIiLeLiMFPh4G/rXgxrfBR0sAiJQHB0+ks/3ACa6/5CKno0gFYa39BsAY84q1NiLfQ/81xsQ4FEtERMQt1Epyh5xsWPo0XNBCq3IXYNXOQ6zamcR9V7Shup8+RxARkSJ88QQcioObpkON+k6nERE3WftrEgBdtEialL7qxhjXHHDGmJZAdQfziIiIlJh62Nzhp3nwx1b420yopKkF8rPW8vJX22lUy5/bL9M8wyIiUoTNH8Gm/0D3sdCqp9NpRMSNVscnEeBXieCLajodRSqeR4BoY8yvgAGaA6OcjSQiIlIy6tQtqYyTsOJ5aBwBQTc6ncbrLP/lDzbtOcqEm0Lwr6y500RE5CyS4uHzR6BZF+gxzuk0IuJmq+MPcVmrOlTy1ZcFpXRZa5cYYwKBdnm7Yq216U5mEhERKSl16pbU2jfgRGLuKF1jnE7jVXJyLK8s3U7LetUZ2KmJ03FERMSbZaXnzqPrWxkGvgO+eosiUp7sPXKS35JOMiSqhdNRpILK68T9yekcIiIi7qKPyUsi+SB8PxnaXQfNo5xO43X+u/l3Yvef4JHeF1NZIzJERORslj0N+zfD9W9CLX0QKFLerInPnU+3a5u6DicRERERKR/U01YS37wImSfhqvFOJ/E6mdk5TFq2g3YXBnBdSCOn44iIiDeLXQzrpsFlo6HdNU6nEREPWBOfRN3qVbi4QYDTUURERETKBXXqnq9DO2HDTIi4C+oFOp3G68zfsJeEpJOM7dsWHx9NSyEiIoU4ugc+vQ8adYTezzqdRkQ8wFrL6vgkOreuq/eF4ghjTFdjTPW83+8wxrxqjGnudC4REZGSUKfu+Vr+DFTy10IuBUjLzGby8jjCm9XmynYNnI4jIiLeKjsLFoyAnOzcuekr+TmdSEQ8YNehFPYfT6NLa029II55CzhpjOkIPAbEA/92NpKIiEjJqFP3fPy2BmI/h64PQ436TqfxOv9Z+xv7j6cxtm87jBaPExGRwkT/H+xZC/1fg7qtnU4jIh6yOm8+3S6t6zmcRCqwLGutBa4HXrfWvgFoLhARESnTtLT0ubIWlv0TAhpB1P1Op/E6yelZvBkdT7fAekRpNIaIiBQmfiV89ypccieE/M3pNCLiQWvik2hUy58Wdas5HUUqrhPGmL8DdwLdjDE+QGWHM4mIiJSIRuqeq22fwt71cMVTUEVvTP/qve93cTglg8f7tHU6ioiIeKvkP2DhKKjfFq5+yek0IuJBOTmW1fGHiGpdV9/gEicNAtKB4dba/UAT4GVnI4mIiJSMOnXPRVYGLH8WGgRB2G1Op/E6R1IyeOfbX+kb1JCOTWs7HUdERLxRTk5uh2768dx5dPUBqUi5Frv/BEdOZmrqBXFUXkfuXOACY0x/IMNaqzl1RUSkTFOn7rmImQFHdkHv58DH1+k0Xmfat/EkZ2TxmEbpiohIYVZNgl9XwtUvQsMOTqcREQ9bHX8IQIukiaOMMSOAH4CbgL8Ba40xw51NJSIiUjKaU7e4Uo/CNy9Cq57QppfTnWI9sQAAIABJREFUabzOgeNpzF6dwI1hjbm4odYcEBGRAuxeCyv+BUE3QfhQp9OISClYE59Ey3rVuah2VaejSMU2FrjEWpsEYIypC6wG3nM0lYiISAlopG5xff9qbsdu7/8FzQd2htdX7CQr2/LwVRc7HUVERLzRycMw/26o3RT6v6ZaKlIBZGXnsG7XYS2eK94gCTiRb/tE3j4REZEySyN1i+Poblg7DTreCo1CnU7jdfYcPskH63czKLIpzbx4VePMzEz27t1LWlqa01FExA38/f1p0qQJlStr8WqvZy189iAkH4C7vwL/Wk4nEjdTjZWCZGTlMKlvfepWN/zyyy9Ox5FzUA5r7E5gnTFmEWCB64HNxphHAay1rzoZTkRE5HyoU7c4VjyfO6Loyn84ncQrvbY8Dh9jePDKQKejnNXevXsJCAigRYsWWn1ZpIyz1pKUlMTevXtp2bKl03GkKD+8A7GfQ59/QeNOTqcRD1CNlYL8cTwNczyN9o1qUtlXXxAsK8ppjY3P+zllUd6/mjdORETKLHXqFuX3TbD5Q7j8EajVxOk0XifuwAk++XEvI7q14sJa/k7HOau0tDQ1NkXKCWMMdevW5eDBg05HkaIk/gRLn4LAvhB1v9NpxENUY6UgyelZ+Ff2VYduGVMea6y19lmnM4iIiLibOnXPxlpY9k+oVje3U1fO8OqyHVSrUol7e7R2OkqxqLEpUn7o/+cyIP0EfHwXVKsHN7yleXTLOf0/Kfnl5FhOZmRTp3oVp6PIeShv/z8bYyKAp4Dm5GsDW2s1t56IiJRZ+tj8bOKWwa5voceTmv+vAJv3HuXLLfsZ0a2l3rCLnIW1lpdffpkjR444HUWk9FgLix+DI7tg4LtQXQsliVQkJzOyybGWGn6eHUOiGivFNAeYCQwE+uf7ERERKbPUqVuY7CxY9jTUaQWd7nI6jVeauHQHF1SrzN2Xl5u5tjzOGMNjjz3m2p44cSLjx493y7Vr1Kjhluuc7doJCQn07NkTgJiYGMaMGeOxe7pTdHQ01113HQCzZs0679f86NGjvPnmm+d83syZM9m9e7fb/lu7S3R0NKtXrz7n84YPH06DBg0IDg4+bf/hw4fp3bs3gYGB9O7dWw3sim7T3Nzpi3qMgxZdnU4jFYBqrDMKq7HJ6VkYoLqfb7GuoxqbSzXWYw5aaz+z1u6y1v526sfpUCIiIiWhTt3CbJoDB3+Bq8ZDJY1C/au1vybx7Y6D3NezDQH+5WZVXI/z8/Nj4cKFHDp0yJH7Z2Vlue1aERERTJkyxW3XKwvO1uA822vr7+/P1KlT6d69O+np6Z6Kd87Ot8E5bNgwlixZcsb+CRMm0KtXL+Li4ujVqxcTJkxwR0wpiw5uhy8ehxbdoPvjTqeRCkI11rukpGdRtUolfH2K19xQjc2lGusxzxhj3jXGDDbG3HTqp6iTjDHvGWP+MMZsybevjjFmmTEmLu/fCzwbXUREpGDq1C1IRgqs/D9oehm0H+B0Gq9jrWXiV9tpWNOPO6OaOx2nTKlUqRKjRo1i0qRJZzyWkJDAlVdeSWhoKL169WL37t1A7pv70aNH07lzZ1q1akV0dDTDhw+nffv2DBs27LRrPPLIIwQFBdGrVy/X4hY9e/bk4YcfJiIigsmTJ7NhwwZ69OhBp06d6Nu3L4mJiWdk2bVrF1FRUYSEhPCPf/zDtd/X15c6deoAp4/MGT9+PEOHDqVbt240b96chQsX8sQTTxASEkK/fv3IzMwE4LnnniMyMpLg4GBGjRqFtRaA9evXExoaSlhYGGPHjnWNTsnOzmbs2LFERkYSGhrK22+/DUBiYiLdu3cnLCyM4OBgvvvuuzOew5IlS2jXrh3h4eEsXLjQtb9q1aquUVEHDx5k4MCBREZGEhkZyapVq1zPZ/jw4fTs2ZNWrVq5Gtbjxo0jPj7elTM6Oppu3boxYMAAOnToAMANN9xAp06dCAoKYvr06QDcdttttGjRgh49epCYmEj79u0ZOXIkQUFB9OnTh9TUVADi4+Pp168fnTp1olu3bsTGxp7T38DSpUuJiooiPDycm2++meTkZABatGjBM888Q3h4OCEhIcTGxpKQkMC0adOYNGkSYWFhBb6Ghenevbvr7yC/RYsWMXToUACGDh3Kp59+WuxrSjmSmZo7j27lanDTO+BTvFF6IiWlGus9NTY7x7I38QAPjbhDNVY11lvcBYQB/fhz6oXrinHerLxz8hsHfG2tDQS+ztsWEREpdVoorSCrX4fk/XDLv7WoSwGitx8k5rcj/OvGYPwrl83G+rP/3cq234+79ZodLqrJM/2Dijzu/vvvJzQ0lCeeeOK0/Q8++CBDhw5l6NChvPfee4wZM8b1hv3IkSOsWbOGzz77jAEDBrBq1SreffddIiMj2bRpE2FhYaSkpBAREcGkSZN47rnnePbZZ3n99dcByMjIICYmhszMTHr06MGiRYuoX78+H374IU899RTvvffeaVkeeughRo8ezZAhQ3jjjTdc+5s2bXpa4y2/+Ph4Vq5cybZt24iKimLBggW89NJL3HjjjSxevJgbbriBBx54gKeffhqAO++8k88//5z+/ftz11138c477xAVFcW4cX++L54xYwa1atVi/fr1pKen07VrV/r06cPChQvp27cvTz31FNnZ2Zw8efK0LGlpaYwcOZIVK1bQpk0bBg0a5Hos/+8PPfQQjzzyCJdffjm7d++mb9++/PLLLwDExsaycuVKTpw4Qdu2bRk9ejQTJkxgy5YtbNq0CchtdG/cuJEtW7bQsmXuNCTvvfcederUITU1lcjISAYOHEjduqfPJRoXF8e8efN45513uOWWW1iwYAF33HEHo0aNYtq0aQQGBrJu3Truu+8+VqxYUay/gSZNmvD888+zfPlyqlevzosvvsirr77qer3r1avHxo0befPNN5k4cSLvvvsu9957LzVq1ODxx88cSRkTE8O0adN49913C/zvXZADBw7QqFEjAC688EIOHDhQ7HOlHPnqf+CPrXD7AqjZyOk04gDVWNXYE2mZvPjMkzz80MP06dVTNfYvVGMdEWmtbXuuJ1lrvzXGtPjL7uuBnnm/zwaigSdLkE1EROS8qFP3r04cgFWTc0foNrvM6TReJyfH8vJX22lWpxq3RDR1Ok6ZVLNmTYYMGcKUKVOoWrWqa/+aNWtcjbk777zztAZp//79McYQEhJCw4YNCQkJASAoKIiEhATCwsLw8fFxNabuuOMObrrpz2+Undq/fft2tmzZQu/evYHcUTqnGgj5rVq1igULFriyPPlk0e9Tr776aipXrkxISAjZ2dn065c7qCEkJISEhAQAVq5cyUsvvcTJkyc5fPgwQUFBdOvWjRMnThAVFQXkjrj5/PPPgdxRMZs3b2b+/PkAHDt2jLi4OCIjIxk+fDiZmZnccMMNhIWFnZYlNjaWli1bEhgY6Ho9To3oyW/58uVs27bNtX38+HHXyJtrr70WPz8//Pz8aNCgQaGNp0svvdTV2ASYMmUKn3zyCQB79uwhLi7ujAZny5YtXZk7depEQkICycnJrF69mptvvtl1XP6vkRb1N7B37162bdtG1665c5dmZGS4XlPA9ffQqVOnQjsN8ouIiDinxuZfGWPK3crZUgz7t0DMe9BlDARe5XQaqYBUY72jxianZ7H2+294ImEnp15p1dg/qcY6YrUxpoO1dlvRhxapobX21DD8/UBDN1xTRETknKlT96+iX4Ds9Ny5dOU02TmWpz75mW2Jx5l8axiVfcvu7B3FGe3jSQ8//DDh4eHcdVfxFuHz8/MDwMfHx/X7qe3C5pnL/2a/evXqQO7UGUFBQaxZs6bIe55rYyF/xsqVK7vOP5UxLS2N++67j5iYGJo2bcr48eNJS0s76zWttUydOpW+ffue8di3337L4sWLGTZsGI8++ihDhgw5p7wAOTk5rF27Fn9//0KfD+R+Jbaw1/nUawu5o4qWL1/OmjVrqFatGj179izwOf712qmpqeTk5FC7dm3XCKXCzinsb8DX15fevXszb968s55/tudSUg0bNiQxMZFGjRqRmJhIgwYNPHIf8WIXBsPQ/0KzqKKPlXJLNVY1NjktC6xqrDupxpZYZ2CTMWYXkA4YwFprQ0tyUWutNcbYgh4zxowCRgE0a9asJLcREREpUNntlfOEg9th478h4m6o29rpNF4lMzuHRz/axAfr9zDmyjYM6HiR05HKtDp16nDLLbcwY8YM174uXbrwwQcfADBnzhy6det2TtfMyclxjbaZO3cul19++RnHtG3bloMHD7oanJmZmWzduvWM47p27XpaFnc41fCqV68eycnJrqy1a9cmICCAdevWAbjuC9C3b1/eeust13yBO3bsICUlhd9++42GDRsycuRIRowYwcaNG0+7V7t27UhISCA+Ph6g0EZYnz59mDp1qmu7sMbeKQEBAZw4caLQx48dO8YFF1xAtWrViI2NZe3atWe9Xn41a9akZcuWfPzxx0BuY/unn34q9vmdO3dm1apV7Ny5E4CUlBR27Nhx1nOKej7nasCAAcyePRuA2bNnc/3117vt2lKGtOwOvlpAU5yjGutsjc3KziE1M5ueV16lGqsa6036AYFAH/6cT7f/eV7rgDGmEUDev38UdJC1drq1NsJaG1G/fv3zvJWIiEjh1Kmb37JnoEp16KEpkfJLz8rm/jkbWbTpd57s145H+7TVV77c4LHHHjtthe6pU6cyc+ZMQkNDef/995k8efI5Xa969er88MMPBAcHs2LFCtc8b/lVqVKF+fPn8+STT9KxY0fCwsIKXJl58uTJvPHGG4SEhLBv375zf3IFqF27NiNHjiQ4OJi+ffsSGRnpemzGjBmMHDnSNW9hrVq1ABgxYgQdOnQgPDyc4OBg7rnnHrKysoiOjqZjx45ccsklfPjhhzz00EOn3cvf35/p06dz7bXXEh4eXuholilTphATE0NoaCgdOnRg2rRpZ30OdevWpWvXrgQHBzN27NgzHu/Xrx9ZWVm0b9+ecePG0blz53N6jebMmcOMGTPo2LEjQUFBLFq0qNjn1q9fn1mzZjF48GBCQ0OJiopyLQJTmP79+/PJJ58UuIhLTEwMI0aMKPC8wYMHExUVxfbt22nSpImr42TcuHEsW7aMwMBAli9fftrcjSIipUk11rkam5KRDcCk1yarxqrGeg1r7W9AU+DKvN9Pcv5t4c+AoXm/DwWK/8ckIiLiRubUyrhlRUREhI2JiXH/hRO+h1nXQq9noNuj7r9+GZWakc09/9nAtzsO8uyAIIZ2aeF0pPP2yy+/0L59e6djSAGSk5OpUaMGABMmTCAxMfGcG9xSMRX0/7UxZoO1NsKhSGWax2qslHuqsd6rtGvsvqOpHEnJoMNFNfHRIIAyrTzVWGPMM0AE0NZae7Ex5iLgY2tt1yLOm0fuomj1gAPAM8CnwEdAM+A34BZr7eGzXUf1VUTkTC3GLXY6QoklTLjWLdc53/qqOXUBcnJg6T+gZhPoPNrpNF4jOT2Lu2et54eEw7w0MJRbIrUwmnjG4sWLeeGFF8jKyqJ58+bMmjXL6UgiIiLlQmnX2JS0LKr7VVKHrnibG4FLgI0A1trfjTEBRZ1krR1cyEO93JhNRETkvKhTF2DrQvj9R7hhGlSuWvTxFcCxk5kMnfkDP+87xuRbL9EcuuJRgwYNcq0eLiIiIu5TmjU2MzuHtKxsalfXvNridTLyL2pmjKle1AkiIiLeTnPqZqXD18/ChSEQqk4lgEPJ6dz6zlq2/X6ct24PV4euiIiIiBQpJT0LgBp+GjciXucjY8zbQG1jzEhgOfCuw5lERERKRO+4fpgOR3fDnZ+Cj/q49x9L4/Z317LvaCrvDo2g+8VaqVVEREREipacnoWvj6FqZV+no4icxlo70RjTGzgOtAWettYucziWiIhIiVTsTt2Th+Hbl6HNVdD6CqfTOG7P4ZPc/u46Dqdk8O/hl3FpyzpORxIRERGRMiI5PYvqVSphNJ+ueBljzIvW2ieBZQXsExFxVHlYMAzct2iYFF/FHpr63SuQfgJ6P+d0Esf9ejCZW95ew7HUTP4zQh26Iu5kreXll1/myJEjTkcRERHxiIysbDKycqjhX7pjRlRjpZh6F7Dv6lJPISIi4kYVt1P3SELu1Atht0HDIKfTOCp2/3FueXstmdk5fDCqM2FNazsdqdwyxvDYY4+5tidOnMj48ePdcu0aNWq45Tpnu3ZCQgI9e/YEICYmhjFjxnjsnu4UHR3NddddB8CsWbPO+zU/evQob7755jmfN3PmTHbv3u22/9buEh0dzerVq8/pnD179nDFFVfQoUMHgoKCmDx5suuxw4cP07t3bwIDA+ndu7ergW2tZcyYMbRp04bQ0FA2btzo1uchIgKqsU45VWOT07NZ9NFcXnvxX+d1HdVY1VhPMMaMNsb8DLQ1xmzO97ML2Ox0PhERkZKouJ26Xz8HxheueMrpJI7avPcot05fi68PfDAqivaNajodqVzz8/Nj4cKFHDp0yJH7Z2Vlue1aERERTJkyxW3XKwvO1uA822vr7+/P1KlT6d69O+np6Z6Kd87Op8FZqVIlXnnlFbZt28batWt544032LZtGwATJkygV69exMXF0atXLyZMmADAl19+SVxcHHFxcUyfPp3Ro0e7/blUNMaYfsaY7caYncaYcQU8/qgxZltew/VrY0xzJ3KKlCbVWGclp2fhYwy+Puc39YJqrGqsh8wF+gOf5f176qeTtfYOJ4OJiIiUlEc7db220blvA2xZAF0egJoXlcotvVFMwmFuf2cdNfwq8fE9XWjTwHOjUCRXpUqVGDVqFJMmTTrjsYSEBK688kpCQ0Pp1asXu3fvBmDYsGGMHj2azp0706pVK6Kjoxk+fDjt27dn2LBhp13jkUceISgoiF69enHw4EEAevbsycMPP0xERASTJ09mw4YN9OjRg06dOtG3b18SExPPyLJr1y6ioqIICQnhH//4h2u/r68vderkTs2Rf/Tr+PHjGTp0KN26daN58+YsXLiQJ554gpCQEPr160dmZiYAzz33HJGRkQQHBzNq1CistQCsX7+e0NBQwsLCGDt2LMHBwQBkZ2czduxYIiMjCQ0N5e233wYgMTGR7t27ExYWRnBwMN99990Zz2HJkiW0a9eO8PBwFi5c6NpftWpV16iogwcPMnDgQCIjI4mMjGTVqlWu5zN8+HB69uxJq1atXA3rcePGER8f78oZHR1Nt27dGDBgAB06dADghhtuoFOnTgQFBTF9+nQAbrvtNlq0aEGPHj1ITEykffv2jBw5kqCgIPr06UNqaioA8fHx9OvXj06dOtGtWzdiY2PP6W9g6dKlREVFER4ezs0330xycjIALVq04JlnniE8PJyQkBBiY2NJSEhg2rRpTJo0ibCwsAJfw4I0atSI8PBwAAICAmjfvj379u0DYNGiRQwdOhSAoUOH8umnn7r2DxkyBGMMnTt35ujRowX+3UnxGGN8gTfI/dpoB2CwMabDXw77EYiw1oYC84GXSjelSOlTjXW2xqakZ1E7oDoBAQGAaqxqrHew1h6z1iZYawdba3/L93PY6WwiIiIl5bFOXa9tdFoLS/8J1etD14c8fjtvtWrnIe6c8QP1A/z4+N4omtWt5nSkCuP+++9nzpw5HDt27LT9Dz74IEOHDmXz5s3cfvvtp33t8siRI6xZs4ZJkyYxYMAAHnnkEbZu3crPP//Mpk2bAEhJSSEiIoKtW7fSo0cPnn32Wdf5GRkZrq9yPvjgg8yfP58NGzYwfPhwnnrqzNHqDz30EKNHj+bnn3+mUaNGrv1NmzY9rYM0v/j4eFasWMFnn33GHXfcwRVXXMHPP/9M1apVWbw4d+L3Bx54gPXr17NlyxZSU1P5/PPPAbjrrrt4++232bRpE76+f66YPWPGDGrVqsX69etZv34977zzDrt27WLu3Ln07duXTZs28dNPPxEWFnZalrS0NEaOHMl///tfNmzYwP79+12PDRo0iMcff9z1PB955BHWr1/PggULGDFihOu42NhYvvrqK3744QeeffZZMjMzmTBhAq1bt2bTpk28/PLLAGzcuJHJkyezY8cOAN577z02bNhATEwMU6ZMISkp6YzXKi4ujvvvv5+tW7dSu3ZtFixYAMCoUaOYOnUqGzZsYOLEidx3333F/hs4dOgQzz//PMuXL2fjxo1ERETw6quvus6vV68eGzduZPTo0UycOJEWLVpw77338sgjj7Bp0ya6det2WsaYmJjTXo+CJCQk8OOPP3LZZZcBcODAAdffy4UXXsiBAwcA2LdvH02bNnWd16RJE1cjVc7LpcBOa+2v1toM4APg+vwHWGtXWmtP5m2uBZqUckYRR6jGOlNjc6wlMzuHWwerxqrGioiISGnx5EoGrkYngDHmVKNz26kDrLUr8x2/Fiidr8Bcdg/kZIFfQKncztt8/csBRs/ZSKt61Xn/7suoH+DndKTS9+U42P+ze695YQhcPaHIw2rWrMmQIUOYMmUKVatWde1fs2aNqzF355138sQTT7ge69+/P8YYQkJCaNiwISEhIQAEBQWRkJBAWFgYPj4+DBo0CIA77riDm266yXX+qf3bt29ny5Yt9O6du1ZEdnb2aQ3KU1atWuVqBN155508+WTRCwNfffXVVK5cmZCQELKzs+nXrx8AISEhJCQkALBy5UpeeuklTp48yeHDhwkKCqJbt26cOHGCqKgoIHfEzamG6NKlS9m8eTPz588H4NixY8TFxREZGcnw4cPJzMzkhhtuOKPBGRsbS8uWLQkMDHS9HqdG9OS3fPly19caAY4fP+4aeXPttdfi5+eHn58fDRo0cDWe/urSSy+lZcuWru0pU6bwySefALlz48XFxVG3bt3TzmnZsqUrc6dOnUhISCA5OZnVq1dz8803u47L/zXSov4G9u7dy7Zt2+jatSuQ28lw6jUFXH8PnTp1KrTTIL+IiAjefffdQh9PTk5m4MCBvPbaa9Sseea0LcYYrX7uOY2BPfm29wKXneX4u4EvPZpIJD/VWKBi1djX35wGQA2/P5sWqrGFU40VERERd/Bkp653NjqNgQ7XF31cObV4cyIPffAjQRfVZPbwS6ldrYrTkSqkhx9+mPDwcO66665iHe/nl9vx7uPj4/r91HZh88zlf7NfvXp1IHcxjaCgINasWVPkPc+1sZA/Y+XKlV3nn8qYlpbGfffdR0xMDE2bNmX8+PGkpaWd9ZrWWqZOnUrfvn3PeOzbb79l8eLFDBs2jEcffZQhQ4acU16AnJwc1q5di7+/f6HPB3K/ElvY63zqtYXcr8suX76cNWvWUK1aNXr27Fngc/zrtVNTU8nJyaF27dquUWGFnVPY34Cvry+9e/dm3rx5Zz3/bM+luDIzMxk4cCC33377aR0bDRs2JDExkUaNGpGYmEiDBg0AaNy4MXv2/FkO9u7dS+PGjUuUQYrHGHMHEAH0OMsxo4BRAM2aNSulZCKeoxpb+jU2O8dSxdeHKr5/fglQNfb8qMaKiIhIcXnFQmn5Gp0vF/L4KGNMjDEm5tQcZnLu5m/Yy4PzNnJJs9r8Z8RlFbtD9+oJcNdi9/4UYwTRKXXq1OGWW25hxowZrn1dunThgw8+AGDOnDlnfFWvKDk5Oa7RNnPnzuXyyy8/45i2bdty8OBBV4MzMzOTrVu3nnFc165dT8viDqcaXvXq1SM5OdmVtXbt2gQEBLBu3ToA130B+vbty1tvveWaL3DHjh2kpKTw22+/0bBhQ0aOHMmIESPOWOm5Xbt2JCQkEB8fD1BoI6xPnz5MnTrVtV1YY++UgIAATpw4Uejjx44d44ILLqBatWrExsaydu3as14vv5o1a9KyZUs+/vhjILex/dNPPxX7/M6dO7Nq1Sp27twJ5H5V+NTXVQtT1PMpiLWWu+++m/bt2/Poo4+e9tiAAQOYPXs2ALNnz+b666937f/3v/+NtZa1a9dSq1atAkevSbHtA5rm226St+80xpirgKeAAdbaQlcPstZOt9ZGWGsj6tev7/awUgGpxgIVq8Zm5Viq+1U6rbNaNVY1VkRERDzLk526bmt0qsFZcu+vSeDxj3+ia5t6zB5+KQH+lZ2OVOE99thjp63QPXXqVGbOnEloaCjvv/8+kydPPqfrVa9enR9++IHg4GBWrFjB008/fcYxVapUYf78+Tz55JN07NiRsLCwAldmnjx5Mm+88QYhISFum5etdu3ajBw5kuDgYPr27UtkZKTrsRkzZjBy5EjCwsJISUmhVq1aAIwYMYIOHToQHh5OcHAw99xzD1lZWURHR9OxY0cuueQSPvzwQx566PT5sf39/Zk+fTrXXnst4eHhrtEsfzVlyhRiYmIIDQ2lQ4cOTJs27azPoW7dunTt2pXg4GDGjh17xuP9+vUjKyuL9u3bM27cODp37nxOr9GcOXOYMWMGHTt2JCgoiEWLFhX73Pr16zNr1iwGDx5MaGgoUVFRrkVgCtO/f38++eSTAhdxKWy+v1WrVvH++++zYsUKwsLCCAsL44svvgByF7lZtmwZgYGBLF++nHHjctfHvOaaa2jVqhVt2rRh5MiRha5uLsW2Hgg0xrQ0xlQBbiV3VW8XY8wlwNvk1tY/HMgo4ijV2NKrsXXq1sNiT5t6AVRjVWNFRETE08yplXHdfmFjKgE7gF7kduauB26z1m7Nd8wl5C6Q1s9aG1ec60ZERNiYmBgPJC6/pn8bz/99EctV7Rvw+m3h+Ff2LfqkcuiXX36hffv2TseQAiQnJ1OjRg0AJkyYQGJi4jk3uKViKuj/a2PMBmtthEORSoUx5hrgNcAXeM9a+y9jzHNAjLX2M2PMciAEOLUE+m5r7YCirqsaK+dLNdZ7ebrGHjyRTuKxVNpfWJPKlbziS4DiJhW1xnqC6quInE3e6twgAAAgAElEQVSLcYudjuAWCROuPafjy8PzPtfnXJjzra8em1PXWptljHkA+Io/G51b8zc6yZ1uoQbwcd7XtYrV6JTisdby2vI4Jn8dx3WhjZg0KIzKvnqzLd5n8eLFvPDCC2RlZdG8eXNmzZrldCQRr2at/QL44i/7ns73+1WlHkpEvJKna2xyehZ+lXzVoSsiIiJSyjy5UJoanQ6y1vLCl7FM//ZX/tapCS8ODMXXR6vkincaNGiQa/VwERERcR9P1tgca0lJz+KCaprWS0RERKS0ebRTV5yRk2N5+rMt/GftboZENWd8/yB81KErIiIiIm6UmpFNjj1zPl0RERER8Ty9AytnsrJzeHLBzyzYuJd7erRiXL92p61ELCIiIiLiDinpWQBUV6euiIiISKnTO7ByJCMrh0c+3MTinxN5tPfFPHhlG3XoioiIiIhHJKdn4V/Zl0pas0FERESk1KlTt5xIy8zm/jkb+Tr2D/5xbXtGdGvldCQRERERKadyciwpGdnUrV7F6SgiIiIiFZI+Vi8HTmZkcffs9Xwd+wfP3xCsDl0RL2Ot5eWXX+bIkSNORxEREXGLkxlZWC+YT1c1VkRERCoqdeqWccfTMhky4wfWxCfxys0duaNzc6cjyVkYY3jsscdc2xMnTmT8+PFuuXaNGjXccp2zXTshIYGePXsCEBMTw5gxYzx2T3eKjo7muuuuA2DWrFnn/ZofPXqUN99885zPmzlzJrt373bbf2t3iY6OZvXq1ed83vDhw2nQoAHBwcGn7T98+DC9e/cmMDCQ3r17uxrY1lrGjBlDmzZtCA0NZePGja5zZs+eTWBgIIGBgcyePbtkT0hEKjTV2NKVnJ6NwbB+zXeqsQVQjRURERFPU6duGXYkJYPb31nHpj1Hef22cAZ2auJ0JCmCn58fCxcu5NChQ47cPysry23XioiIYMqUKW67Xllwtgbn2V5bf39/pk6dSvfu3UlPT/dUvHN2vg3OYcOGsWTJkjP2T5gwgV69ehEXF0evXr2YMGECAF9++SVxcXHExf0/e/ceV2WVL37883ARFFTEWw44isEobPZmB2wFEaUcBDPNLl4qBSWh0PKSqUx2UadzovLkEdRMxUsdU09estEZM0b51VFEkSHzgiIjeRky7wIKcnl+fyDPQICCghvk+3695jXs9Vz2dy0e+u61XHutTJYtW0Z0dDRQ1kGdO3cuKSkp7N+/n7lz58pMKyHEPZMc+2DlFRbTsoUllhb1052QHFtGcqwQQgghaksGdZuoX3MLGL1sH8fP57IszIcn9V3MHZKoBSsrK6KioliwYEGVY9nZ2TzxxBMYDAYGDhzI6dOngbIP99HR0fj5+dGjRw+SkpKIiIjA3d2dcePGVbrHtGnT0Ol0DBw4kAsXLgAQFBTE1KlT8fX1ZeHChRw8eJABAwbg4+NDSEgIOTk5VWI5deoU/v7+6PV63n77ba3c0tISR0dHoPLs1zlz5hAeHk5gYCDdunVj8+bNzJw5E71eT2hoKEVFRQDMmzcPk8mEp6cnUVFRqKoKwIEDBzAYDBiNRmbMmKHNTikpKWHGjBmYTCYMBgOfffYZADk5OfTv3x+j0Yinpyc//PBDlTrs2LGDXr164e3tzebNm7Xyli1barOiLly4wHPPPYfJZMJkMrFnzx6tPhEREQQFBdGjRw+tYx0TE0NWVpYWZ1JSEoGBgQwbNgwPDw8Ahg8fjo+PDzqdjmXLlgHw4osv0r17dwYMGEBOTg7u7u5ERkai0+kYNGgQN2/eBCArK4vQ0FB8fHwIDAwkIyOjTs/Azp078ff3x9vbmxEjRpCXlwdA9+7dee+99/D29kav15ORkUF2djZLly5lwYIFGI3GatuwJv3799eeg4q2bt1KeHg4AOHh4Xz99ddaeVhYGIqi4Ofnx9WrV8nJyeHbb78lODgYR0dH2rVrR3BwcLUdWSGEqA3JsQ8ux27/698I7uvNM8H9JMdKjhVCCCGEmchGaU3Qv67e5KUVKZy/XsDqcSb6unYwd0hNzof7PyTjcka93rOXYy9m9Z511/MmTZqEwWBg5syZlcpff/11wsPDCQ8PZ+XKlUyePFn7wH7lyhWSk5P55ptvGDZsGHv27GHFihWYTCbS09MxGo3k5+fj6+vLggULmDdvHnPnzmXRokUA3Lp1i9TUVIqKihgwYABbt26lY8eObNiwgdmzZ7Ny5cpKsUyZMoXo6GjCwsJYvHixVt61a9dKnbeKsrKy2L17N0ePHsXf359Nmzbx0Ucf8cwzz7B9+3aGDx/Oa6+9xrvvvgvA2LFj2bZtG0OHDmX8+PEsX74cf39/YmJitHsmJCTQtm1bDhw4QGFhIQEBAQwaNIjNmzcTEhLC7NmzKSkp4caNG5ViKSgoIDIykl27duHq6sqoUaO0YxV/njJlCtOmTaNfv36cPn2akJAQjh07BkBGRga7d+8mNzeXnj17Eh0dTWxsLIcPHyY9PR0o63SnpaVx+PBhXFxcAFi5ciWOjo7cvHkTk8nEc889R/v27SvFl5mZybp161i+fDkjR45k06ZNjBkzhqioKJYuXYqbmxspKSlMnDiRXbt21eoZcHZ25v333ycxMRE7Ozs+/PBDPvnkE629O3ToQFpaGkuWLGH+/PmsWLGCV199FXt7e958880qv8/U1FSWLl3KihUrqv19V+f8+fN06VL2D0yPPPII58+fB+DcuXN07dpVO8/Z2Zlz587VWC6EaNokxz78OfaVqCg+Xfc1Qb56Xg4fox2THCs5VgghhBAPjgzqNjE/X8rnxeUpXL9ZxBcv98anW9V/yReNW5s2bQgLCyMuLo6WLVtq5cnJyVpnbuzYsZU6pEOHDkVRFPR6PZ07d0av1wOg0+nIzs7GaDRiYWGhdabGjBnDs88+q11fXn78+HEOHz5McHAwUDZLp7yDUNGePXvYtGmTFsusWXfvSA8ePBhra2v0ej0lJSWEhoYCoNfryc7OBmD37t189NFH3Lhxg8uXL6PT6QgMDCQ3Nxd/f3+gbMbNtm3bgLJZMYcOHWLjxo0AXLt2jczMTEwmExERERQVFTF8+HCMRmOlWDIyMnBxccHNzU1rj/IZPRUlJiZy9OhR7fX169e1mTdDhgzBxsYGGxsbOnXqpHWefqt3795aZxMgLi6OLVu2AHDmzBkyMzOrdDhdXFy0mH18fMjOziYvL4+9e/cyYsQI7byKXyO92zNw9uxZjh49SkBAAFA2yFDepoD2PPj4+NQ4aFCRr69vnTqbv6UoCoqi3PP1QghxLyTHPpgc6/z7bnTv4YqdjbXkWCTHCiGEEMI8ZFC3CTn5ay4vrUjhVnEp66L88HRqa+6QmqzazPZpSFOnTsXb25vx48fX6nwbGxsALCwstJ/LX9e0zlzFD/t2dnZA2WYaOp2O5OTku75nXTsLFWO0trbWri+PsaCggIkTJ5KamkrXrl2ZM2cOBQUFd7ynqqrEx8cTEhJS5dj333/P9u3bGTduHG+88QZhYWF1ihegtLSUffv2YWtrW2N9oOwrsTW1c3nbQtmsosTERJKTk2nVqhVBQUHV1vG397558yalpaU4ODhoM5RquqamZ8DS0pLg4GDWrVt3x+vvVJf71blzZ3JycujSpQs5OTl06tQJACcnJ86cOaOdd/bsWZycnHByciIpKalSefkmQUKIpkty7MOfY0tUlVYtLLGwqLkekmPrl+RYIYQQQvyWrKnbRBz51zVGfbaPklJYH+UvA7pNnKOjIyNHjiQhIUEr69u3L+vXrwdg7dq1BAYG1umepaWl2mybL7/8kn79+lU5p2fPnly4cEHrcBYVFXHkyJEq5wUEBFSKpT6Ud7w6dOhAXl6eFquDgwOtW7cmJSUFQHtfgJCQED799FNtvcATJ06Qn5/Pzz//TOfOnYmMjGTChAmVdnoG6NWrF9nZ2WRlZQHU2AkbNGgQ8fHx2uuaOnvlWrduTW5ubo3Hr127Rrt27WjVqhUZGRns27fvjverqE2bNri4uPDVV18BZZ3tH3/8sdbX+/n5sWfPHk6ePAlAfn4+J06cuOM1d6tPXQ0bNkzbXXvNmjU8/fTTWvnnn3+Oqqrs27ePtm3b0qVLF0JCQti5cydXrlzhypUr7Ny5s9rBBSGEqAvJsQ2bY13d/sDZ0z9z8VzZusSSY6snOVYIIYQQDU0GdZuAf5y+wgvL9mFjZcFXr/rT85HW5g5J1IPp06dX2qE7Pj6eVatWYTAY+OKLL1i4cGGd7mdnZ8f+/fvx9PRk165d2jpvFbVo0YKNGzcya9YsvLy8MBqN1e7MvHDhQhYvXoxer6+39dccHByIjIzE09OTkJAQTCaTdiwhIYHIyEht3cK2bcv+0WLChAl4eHjg7e2Np6cnr7zyCsXFxSQlJeHl5cVjjz3Ghg0bmDJlSqX3srW1ZdmyZQwZMgRvb29tNstvxcXFkZqaisFgwMPDg6VLl96xDu3btycgIABPT09mzJhR5XhoaCjFxcW4u7sTExODn59fndpo7dq1JCQk4OXlhU6nY+vWrbW+tmPHjqxevZoXXngBg8GAv7+/tglMTYYOHcqWLVuq3cQlNTWVCRMmVHvdCy+8gL+/P8ePH8fZ2VkbOImJieG7777Dzc2NxMREbe3GJ598kh49euDq6kpkZKS2u7mjoyPvvPOOtonOu+++W+3mMEIIUVeSYxsux5YoVrwb+9+MHfWs5Ng7kBwrhBBCiIamlO+M21T4+vqqqamp5g7jgdn3z0u8vPoAHVrbsHZCH5zbtTJ3SE3WsWPHcHd3N3cYohp5eXnabtmxsbHk5OTUucMtmqfq/q4VRTmoqqqvmUJq0ppbjhX1R3Js41XfOfbclZtcuXELj9+1wULWdX2oSY6tP5JfhRB30j1mu7lDqBfZsUPqdP7DUO+61rkm95pfZU3dRizp+K+88sVBujq2Yu2EPnRuU3VNMiEeBtu3b+eDDz6guLiYbt26sXr1anOHJIQQQjwU6jvH5hUWY2djJQO6QgghhBBmJoO6jdSOw7/w+ro03Dq15ouXe9Pe3ubuFwnRRI0aNUrbPVwIIYQQ9ac+c2xRSSmFxSU42lnXy/2EEEIIIcS9k0HdRmhr+jne+N8fMTi3ZfX43rRtKR+chRBCCCGEeeUVFgNgbyNdCCGEEEIIc5NPZI3M+v2n+dOWn+jj4siKcJN8aBZCCCGEEI1CfkExlhYKttaW5g5FCCGEEKLZkxHDRmTVnlPM/ctRBvyhI5+N9ZEPzEIIIYQQotHIKyzG3sYKRdbTFUIIIYQwOwtzByDKLN59krl/OUqIrjPLwmRAV4iHiaqqfPzxx1y5csXcoQghhBD35FZxCbdKSrFrZN8ikxwrhBBCiOZKBnXNTFVVPv42g4+/Pc5w4+9Y/KI3NlYyoPuwUhSF6dOna6/nz5/PnDlz6uXe9vb29XKfO907OzuboKAgAFJTU5k8eXKDvWd9SkpK4qmnngJg9erV99zmV69eZcmSJXW+btWqVZw+fbreftf1JSkpib1799b5uu7du6PX6zEajfj6+mrlly9fJjg4GDc3N4KDg7UOtqqqTJ48GVdXVwwGA2lpado1a9aswc3NDTc3N9asWXP/lRJCNFuSYxtWTevpSo6tnuRYIYQQQjQ0GdQ1I1VVmbftKIt3Z/FC767810gjVpbyK3mY2djYsHnzZi5evGiW9y8uLq63e/n6+hIXF1dv92sK7tThvFPb2traEh8fT//+/SksLGyo8OrsXjucALt37yY9PZ3U1FStLDY2loEDB5KZmcnAgQOJjY0F4G9/+xuZmZlkZmaybNkyoqOjgbIO6ty5c0lJSWH//v3MnTtXZloJIe6Z5NiGlVdQgpWFBTZWDfNZVXLsv0mObRoURclWFOUnRVHSFUVJvfsVQgghRP2SEUQzKSlV+dPmn1i1J5vxAd35z2f0WFrI+mQPOysrK6KioliwYEGVY9nZ2TzxxBMYDAYGDhzI6dOnARg3bhzR0dH4+fnRo0cPkpKSiIiIwN3dnXHjxlW6x7Rp09DpdAwcOJALFy4AEBQUxNSpU/H19WXhwoUcPHiQAQMG4OPjQ0hICDk5OVViOXXqFP7+/uj1et5++22t3NLSEkdHR6DyzJw5c+YQHh5OYGAg3bp1Y/PmzcycORO9Xk9oaChFRUUAzJs3D5PJhKenJ1FRUaiqCsCBAwcwGAwYjUZmzJiBp6cnACUlJcyYMQOTyYTBYOCzzz4DICcnh/79+2M0GvH09OSHH36oUocdO3bQq1cvvL292bx5s1besmVLbVbUhQsXeO655zCZTJhMJvbs2aPVJyIigqCgIHr06KF1rGNiYsjKytLiTEpKIjAwkGHDhuHh4QHA8OHD8fHxQafTsWzZMgBefPFFunfvzoABA8jJycHd3Z3IyEh0Oh2DBg3i5s2bAGRlZREaGoqPjw+BgYFkZGTU6RnYuXMn/v7+eHt7M2LECPLy8oCyWT/vvfce3t7e6PV6MjIyyM7OZunSpSxYsACj0VhtG9bV1q1bCQ8PByA8PJyvv/5aKw8LC0NRFPz8/Lh69So5OTl8++23BAcH4+joSLt27QgODmbHjh33HYcQonmSHNtwOVZV1Urr6UqOlRwrNI+rqmpUVdX37qcKIYQQ9UsGdc2guKSUN/43nfUHzvDa4668+5SHbDjRjEyaNIm1a9dy7dq1SuWvv/464eHhHDp0iJdeeqnS1y6vXLlCcnIyCxYsYNiwYUybNo0jR47w008/kZ6eDkB+fj6+vr4cOXKEAQMGMHfuXO36W7duaV/lfP3119m4cSMHDx4kIiKC2bNnV4lxypQpREdH89NPP9GlSxetvGvXrpU6bxVlZWWxa9cuvvnmG8aMGcPjjz/OTz/9RMuWLdm+fTsAr732GgcOHODw4cPcvHmTbdu2ATB+/Hg+++wz0tPTsbT89/IjCQkJtG3blgMHDnDgwAGWL1/OqVOn+PLLLwkJCSE9PZ0ff/wRo9FYKZaCggIiIyP5y1/+wsGDB/nll1+0Y6NGjeLNN9/U6jlt2jQOHDjApk2bmDBhgnZeRkYG3377rTa7paioiNjYWB599FHS09P5+OOPAUhLS2PhwoWcOHECgJUrV3Lw4EFSU1OJi4vj0qVLVdoqMzOTSZMmceTIERwcHNi0aRMAUVFRxMfHc/DgQebPn8/EiRNr/QxcvHiR999/n8TERNLS0vD19eWTTz7Rru/QoQNpaWlER0czf/58unfvzquvvsq0adNIT08nMDCwUoypqamV2qMiRVEYNGgQPj4+Wqca4Pz589rz8sgjj3D+/HkAzp07R9euXbXznJ2dOXfuXI3lQghxryTHNkyOLSwupbi0FHtbS8mxkmOFEEII0Ug0rp0OmoHC4hImr/sH3x45z4yQnkx63NXcITVLv/znf1J4LKNe72nj3otH3nrrrue1adOGsLAw4uLiaNmypVaenJysdebGjh3LzJkztWNDhw5FURT0ej2dO3dGr9cDoNPpyM7Oxmg0YmFhwahRowAYM2YMzz77rHZ9efnx48c5fPgwwcHBQNksnYodynJ79uzROkFjx45l1qxZd63X4MGDsba2Rq/XU1JSQmhoKAB6vZ7s7Gyg7OuEH330ETdu3ODy5cvodDoCAwPJzc3F398fKJtxU94R3blzJ4cOHWLjxo0AXLt2jczMTEwmExERERQVFTF8+PAqg7oZGRm4uLjg5uamtUfFjlG5xMREjh49qr2+fv26NvNmyJAh2NjYYGNjQ6dOnbTO02/17t0bFxcX7XVcXBxbtmwB4MyZM2RmZtK+fftK17i4uGgx+/j4kJ2dTV5eHnv37mXEiBHaeRW/Rnq3Z+Ds2bMcPXqUgIAAoGyQobxNAe158PHxqXHQoCJfX19WrFhR7bH/+7//w8nJiV9//ZXg4GB69epF//79K52jKIr8Y5UQzZTk2Icvx17MK8tH9jZWHD38k+RYybGijArsVBRFBT5TVbXqH4IQQgjRgGRQ9wEqKCrhlS8O8v9OXOC9oR6MD3C5+0XioTR16lS8vb0ZP358rc63sbEBwMLCQvu5/HVN68xV/LBvZ2cHlK3jrNPpSE5Ovut71rWzUDFGa2tr7fryGAsKCpg4cSKpqal07dqVOXPmUFBQcMd7qqpKfHw8ISEhVY59//33bN++nXHjxvHGG28QFhZWp3gBSktL2bdvH7a2tjXWB8q+EltTO5e3LZR9XTYxMZHk5GRatWpFUFBQtXX87b1v3rxJaWkpDg4O2qywmq6p6RmwtLQkODiYdevW3fH6O9WltpycnADo1KkTzzzzDPv376d///507tyZnJwcunTpQk5ODp06ddLOP3PmjHb92bNncXJywsnJiaSkpErl5ZsECSHEvZIcW/85dsCQ52hhaUGLOmzmKzn23kiObVL6qap6TlGUTsB3iqJkqKr6fflBRVGigCiA3//+9+aKUQghxENMBnUfkLzCYiasOUDKqcvEPqtndG9J7OZUm9k+DcnR0ZGRI0eSkJBAREQEAH379mX9+vWMHTuWtWvXVvmq3t2UlpayceNGRo8ezZdffkm/fv2qnNOzZ08uXLhAcnIy/v7+FBUVceLECXQ6XaXzAgICWL9+PWPGjGHt2rX3XtEKyjteHTp0IC8vj40bN/L888/j4OBA69atSUlJoU+fPqxfv167JiQkhE8//ZQnnngCa2trTpw4gZOTExcvXsTZ2ZnIyEgKCwtJS0urNKjbq1cvsrOzycrK4tFHH62xEzZo0CDi4+OZMWMGAOnp6VVm/VbUunVrcnNzazx+7do12rVrR6tWrcjIyGDfvn21bp82bdrg4uLCV199xYgRI1BVlUOHDuHl5VWr6/38/Jg0aRInT57E1dWV/Px8zp07xx/+8Ic71uf69eu1jhHKvoJcWlpK69atyc/PZ+fOnbz77rsADBs2jDVr1hATE8OaNWt4+umntfJFixYxevRoUlJSaNu2LV26dCEkJIS33npL27hl586dfPDBB3WKRwjR+EiOfbhy7MGDB/H549O0tbUGJMdKjhXlVFU9d/v/f1UUZQvQG/i+wvFlwDIAX19f1SxBCiGEeKjJmroPwLUbRYxNSOFA9hX+e5RRBnQFANOnT6+0Q3d8fDyrVq3CYDDwxRdfsHDhwjrdz87Ojv379+Pp6cmuXbu0TkBFLVq0YOPGjcyaNQsvLy+MRmO1OzMvXLiQxYsXo9fr6239NQcHByIjI/H09CQkJASTyaQdS0hIIDIyEqPRSH5+Pm3btgVgwoQJeHh44O3tjaenJ6+88grFxcUkJSXh5eXFY489xoYNG5gyZUql97K1tWXZsmUMGTIEb29vbTbLb8XFxZGamorBYMDDw4OlS5fesQ7t27cnICAAT09PrZNaUWhoKMXFxbi7uxMTE4Ofn1+d2mjt2rUkJCTg5eWFTqdj69attb62Y8eOrF69mhdeeAGDwYC/v7+2CUxNhg4dypYtW6rdxKWm9f7Onz9Pv3798PLyonfv3gwZMkT7GnBMTAzfffcdbm5uJCYmEhMTA8CTTz5Jjx49cHV1JTIyUtvd3NHRkXfeeUfbROfdd9/VNgkSQoj7ITm2/nLsqxNfo6RUxc62bC6I5FjJsQIURbFTFKV1+c/AIOCweaMSQgjR3CjlO+M2Fb6+vmpqaqq5w6i1S3mFjE3Yz8lf84h/8TFCdI+YO6Rm69ixY7i7u5s7DFGNvLw8bbfs2NhYcnJy6tzhFs1TdX/XiqIclF2o701Ty7Gi8ZAc23jdb469kFtAzrUC3Lu0wdpS5oM0J5Jja6YoSg9gy+2XVsCXqqr+R03nS34VQtxJ95jt5g6hXmTHDqnT+Q9Dveta55rca36V5Rca0PnrBby0IoUzl2+wPNyXAX/oaO6QhGiUtm/fzgcffEBxcTHdunVj9erV5g5JCCGEeCjcb47NKyzBxspSBnSFqEBV1X8CtVu/QwghhGggMqjbQM5cvsFLK1K4lFfImoje+PVof/eLhGimRo0ape0eLoQQQoj6cz85tlRVyS8spl2rFvUclRBCCCGEuF8yqNsATl3M56Xl+8grLOZ/JvThsd+3M3dIQgghhBBC1MnNWyWUqir2tpbmDkUIIYQQQvyGDOrWs+O/5PLSihRKVZV1UX7oftfW3CEJIYQQQghRZ3mFxQDYtZAugxBCCCFEYyOf0OrRT2evEbYyBWtLC9ZH+eHaqbW5QxJCCCGEEOKe5BUW09LaEitZT1cIIYQQotGRT2j1JDX7Mi8u30erFlZ89aq/DOgKIYQQQogmq7RU5catEuxtZA6IEEIIIURjJIO69WDPyYuMTdhPh9Y2fPWqP93a25k7JCFEI6KqKh9//DFXrlwxdyhCCCFErdy4VYyqqtg18kFdybFCCCGEaK5kUPc+7co4z/jVB/i9Yys2vOLH7xxamjsk0YgpisL06dO11/Pnz2fOnDn1cm97e/t6uc+d7p2dnU1QUBAAqampTJ48ucHesz4lJSXx1FNPAbB69ep7bvOrV6+yZMmSOl+3atUqTp8+XW+/6/qSlJTE3r1763xdREQEnTp1wtPTs1L5nDlzcHJywmg0YjQa+etf/6od++CDD3B1daVnz558++23WvmOHTvo2bMnrq6uxMbGauWnTp2iT58+uLq6MmrUKG7dunUPNRRCNCeSY+tXXmExCspdB3Ulx1ZPcqwQQgghGlrj/qf3Rm77oRymrP8H7l3a8HlEb9rZtTB3SKKRs7GxYfPmzfzpT3+iQ4cOD/z9i4uLsbKqnz97X19ffH196+VeTUV5h3PixIlVjt2pbW1tbYmPj2fTpk0UFhZiY2PT0KHWSlJSEvb29vTt27dO140bN47XXnuNsLCwKsemTZvGm2++Wans6NGjrF+/niNHjvCvf/2LP/7xj5w4cQKASZMm8d133+Hs7IzJZGLYsGF4eHgwa9Yspk2bxujRoxCpleYAACAASURBVHn11VdJSEggOjr63isrhHjoSY6tX3mFJbRsYYmlhfJA3k9ybBnJsUKI5q57zHZzh3DfsmOHmDsE0UzITN17tOngWV5fl4axqwNrI/vIgK6oFSsrK6KioliwYEGVY9nZ2TzxxBMYDAYGDhzI6dOngbIP99HR0fj5+dGjRw+SkpKIiIjA3d2dcePGVbrHtGnT0Ol0DBw4kAsXLgAQFBTE1KlT8fX1ZeHChRw8eJABAwbg4+NDSEgIOTk5VWI5deoU/v7+6PV63n77ba3c0tISR0dHoPLMnDlz5hAeHk5gYCDdunVj8+bNzJw5E71eT2hoKEVFRQDMmzcPk8mEp6cnUVFRqKoKwIEDBzAYDBiNRmbMmKHNTikpKWHGjBmYTCYMBgOfffYZADk5OfTv3x+j0Yinpyc//PBDlTrs2LGDXr164e3tzebNm7Xyli1barOiLly4wHPPPYfJZMJkMrFnzx6tPhEREQQFBdGjRw/i4uIAiImJISsrS4szKSmJwMBArZMEMHz4cHx8fNDpdCxbtgyAF198ke7duzNgwABycnJwd3cnMjISnU7HoEGDuHnzJgBZWVmEhobi4+NDYGAgGRkZdXoGdu7cib+/P97e3owYMYK8vDwAunfvznvvvYe3tzd6vZ6MjAyys7NZunQpCxYswGg0VtuGNenfv7/2HNTG1q1bGT16NDY2Nri4uODq6sr+/fvZv38/rq6u9OjRgxYtWjB69Gi2bt2Kqqrs2rWL559/HoDw8HC+/vrrWr+fEKJ5khxbfzk2sH9/nnrcn6eC+kiOvU1yrBBCCCEaG5mpew/+Z9/PvP31YQJc27M8zJdWLaQZm5of/vcEF8/k1es9O3S1J3DkH+563qRJkzAYDMycObNS+euvv054eDjh4eGsXLmSyZMnax+yr1y5QnJyMt988w3Dhg1jz549rFixApPJRHp6Okajkfz8fHx9fVmwYAHz5s1j7ty5LFq0CIBbt26RmppKUVERAwYMYOvWrXTs2JENGzYwe/ZsVq5cWSmWKVOmEB0dTVhYGIsXL9bKu3btWqnzVlFWVha7d+/m6NGj+Pv7s2nTJj766COeeeYZtm/fzvDhw3nttdd49913ARg7dizbtm1j6NChjB8/nuXLl+Pv709MTIx2z4SEBNq2bcuBAwcoLCwkICCAQYMGsXnzZkJCQpg9ezYlJSXcuHGjUiwFBQVERkaya9cu7auF5Sr+PGXKFKZNm0a/fv04ffo0ISEhHDt2DICMjAx2795Nbm4uPXv2JDo6mtjYWA4fPkx6ejpQ1ulOS0vj8OHDuLi4ALBy5UocHR25efMmJpOJ5557jvbt21eKLzMzk3Xr1rF8+XJGjhzJpk2bGDNmDFFRUSxduhQ3NzdSUlKYOHEiu3btqtUz4OzszPvvv09iYiJ2dnZ8+OGHfPLJJ1p7d+jQgbS0NJYsWcL8+fNZsWIFr776Kvb29lVm/UDZV3+XLl3KihUrqv1912TRokV8/vnn+Pr68l//9V+0a9eOc+fO4efnp53j7OzMuXPngLJnqmJ5SkoKly5dwsHBQZuVVfF8IUTjJzm26efYoCf+yIgJk+nWzhaL0qJKsUiOlRwrhBBCiMZBRiPraMUP/+T97cd4olcnlrzkja21pblDEk1MmzZtCAsLIy4ujpYt/70Gc3JystaZGzt2bKUO6dChQ1EUBb1eT+fOndHr9QDodDqys7MxGo1YWFhonakxY8bw7LPPateXlx8/fpzDhw8THBwMlM3S6dKlS5UY9+zZw6ZNm7RYZs2addd6DR48GGtra/R6PSUlJYSGhgKg1+vJzs4GYPfu3Xz00UfcuHGDy5cvo9PpCAwMJDc3F39/f6Bsxs22bduAslkxhw4dYuPGjQBcu3aNzMxMTCYTERERFBUVMXz4cIxGY6VYMjIycHFxwc3NTWuP8hk9FSUmJnL06FHt9fXr17WZN0OGDMHGxgYbGxs6derE+fPnq6137969tc4mQFxcHFu2bAHgzJkzZGZmVulwuri4aDH7+PiQnZ1NXl4ee/fuZcSIEdp5hYWF2s93ewbOnj3L0aNHCQgIAMoGGcrbFNCeBx8fnxoHDSry9fWtc2czOjqad955B0VReOedd5g+fXqVwQwhhGhIkmPrJ8eGjxvPxes3iBwzCm/vxyrFIjlWcqwQQgghGgcZ1K0lVVWJ+/tJFiSeYIi+CwtGGWlhJatXNFW1me3TkKZOnYq3tzfjx4+v1fnl68NZWFhUWivOwsKC4uLiaq9RlH+vgWdnZweUPcc6nY7k5OS7vmfF6+sao7W1tXZ9eYwFBQVMnDiR1NRUunbtypw5cygoKLjjPVVVJT4+npCQkCrHvv/+e7Zv3864ceN44403ql177m5KS0vZt28ftra2NdYHyr4SW1M7l7ctlM0qSkxMJDk5mVatWhEUFFRtHX9775s3b1JaWoqDg4M2Q6mma2p6BiwtLQkODmbdunV3vP5OdblfnTt31n6OjIzUvjrs5OTEmTNntGNnz57FyckJoNry9u3bc/XqVW0NxYrnCyEaP8mxTT/Hfr7lb/zw92+JiBgvOVZyrBBCCCEaKRmVrAVVVYndkcGCxBM85+3MwtEyoCvuj6OjIyNHjiQhIUEr69u3L+vXrwdg7dq1BAYG1umepaWl2mybL7/8kn79+lU5p2fPnly4cEHrcBYVFXHkyJEq5wUEBFSKpT6Ud7w6dOhAXl6eFquDgwOtW7cmJSUFQHtfgJCQED799FNtvcATJ06Qn5/Pzz//TOfOnYmMjGTChAmkpaVVeq9evXqRnZ1NVlYWQI2dsEGDBhEfH6+9rqmzV65169bk5ubWePzatWu0a9eOVq1akZGRwb59++54v4ratGmDi4sLX331FVD2350ff/yx1tf7+fmxZ88eTp48CUB+fr62UUpN7lafuqq4duSWLVu0dRuHDRvG+vXrKSws5NSpU2RmZtK7d29MJhOZmZmcOnWKW7dusX79eoYNG4aiKDz++OPaM7JmzRqefvrpeotTCPFwkxx7fzk265+nsHdoz4QJkmPLSY4VQgghRGMkI5N3UVqq8t43R/js//2TMX6/5+PnDVhZSrOJ+zd9+nQuXryovY6Pj2fVqlUYDAa++OILFi5cWKf72dnZsX//fjw9Pdm1a5e2zltFLVq0YOPGjcyaNQsvLy+MRiN79+6tct7ChQtZvHgxer2+3tZZc3BwIDIyEk9PT0JCQjCZTNqxhIQEIiMjtXUL27ZtC8CECRPw8PDA29sbT09PXnnlFYqLi0lKSsLLy4vHHnuMDRs2MGXKlErvZWtry7JlyxgyZAje3t506tSp2pji4uJITU3FYDDg4eHB0qVL71iH9u3bExAQgKenJzNmzKhyPDQ0lOLiYtzd3YmJiam0xl1trF27loSEBLy8vNDpdGzdurXW13bs2JHVq1fzwgsvYDAY8Pf31zaBqcnQoUPZsmVLtZu4pKamMmHChGqve+GFF/D39+f48eM4OztrAyflG/cYDAZ2796tbVak0+kYOXIkHh4ehIaGsnjxYiwtLbGysmLRokWEhITg7u7OyJEj0el0ANp6ha6urly6dImXX3651m0hhBCSY+89x+5M3MWIQf0I7u8nOfY2ybFCCCGEaIyU8p1xmwpfX181NTX1gbxXSanKrE2H2HjwLFH9e/Cnwb3q/HU50XgcO3YMd3d3c4chqpGXl6ftlh0bG0tOTk6dO9yiearu71pRlIOqqvqaKaQm7UHmWPFwkRzbeNU1x567coMrN4rw+F0bLORzb7MmObb+SH4Vova6x2w3dwj3LTt2SJ3OfxjqDM2z3nWtc03uNb/Kmro1KCopZdqGdLYdymHqH92YMtBNBnSFaCDbt2/ngw8+oLi4mG7durF69WpzhySEEEI8FOqaY/MKS7CzsZIBXSGEEEKIRk4GdatRUFTCa1+mkXjsV956shdR/R81d0hCPNRGjRql7R4uhBBCiPpTlxx7q7iUwuISHO1aNHBUQgghhBDifsmg7m/cuFXMK18c5IfMi/z5aR1j/bubOyQhhBBCCCEaXH5hMQD2NpZmjkQIIYQQQtyNDOpWkFtQRMTqAxz8+QrzR3jxvI+zuUMSQgghhBDigcgrLMbSQsHWWgZ1hRBCCCEaOxnUve3qjVuErdzP0X9dJ+6Fx3jK8DtzhySEEEIIIcQDoaoq+YXF2NtYyT4SQgghhBBNgIW5A2gMLuQWMnrZPjJ+yeWzsT4yoCuEqFeqqvLxxx9z5coVc4cihBBCVOtWSSm3Skqxt2lacz4kxwohhBCiuWr2g7o5124y6rNkfr50g5XhJga6dzZ3SOIhpigK06dP117Pnz+fOXPm1Mu97e3t6+U+d7p3dnY2QUFBAKSmpjJ58uQGe8/6lJSUxFNPPQXA6tWr77nNr169ypIlS+p83apVqzh9+nS9/a7rS1JSEnv37q3zdVevXuX555+nV69euLu7k5ycDMDly5cJDg7Gzc2N4OBgrYOtqiqTJ0/G1dUVg8FAWlqadq81a9bg5uaGm5sba9as0coPHjyIXq/H1dWVyZMno6rqfdZWCPGwkxx7f/Jur6drV8dBXcmx1ZMcK4QQQoiG1qCDuoqihCqKclxRlJOKosRUc9xGUZQNt4+nKIrSvSHj+a3Tl24wYmkyF3IL+fzl3vRz6/Ag3140QzY2NmzevJmLFy+a5f2Li4vr7V6+vr7ExcXV2/2agjt1OO/Utra2tsTHx9O/f38KCwsbKrw6u9cO55QpUwgNDSUjI4Mff/wRd3d3AGJjYxk4cCCZmZkMHDiQ2NhYAP72t7+RmZlJZmYmy5YtIzo6GijroM6dO5eUlBT279/P3LlztU5qdHQ0y5cv167bsWNHPdX64dHYc6wQD5rk2PuTX1CMtaUFNlbmmfMhObaM5FghhBBC1FaDfWpTFMUSWAwMBjyAFxRF8fjNaS8DV1RVdQUWAB82VDy/dfLXPEZ+lkxeYTFrI/tg6u74oN5aNGNWVlZERUWxYMGCKseys7N54oknMBgMDBw4kNOnTwMwbtw4oqOj8fPzo0ePHiQlJREREYG7uzvjxo2rdI9p06ah0+kYOHAgFy5cACAoKIipU6fi6+vLwoULOXjwIAMGDMDHx4eQkBBycnKqxHLq1Cn8/f3R6/W8/fbbWrmlpSWOjmV/KxVn5syZM4fw8HACAwPp1q0bmzdvZubMmej1ekJDQykqKgJg3rx5mEwmPD09iYqK0maGHDhwAIPBgNFoZMaMGXh6egJQUlLCjBkzMJlMGAwGPvvsMwBycnLo378/RqMRT09Pfvjhhyp12LFjB7169cLb25vNmzdr5S1bttRmRV24cIHnnnsOk8mEyWRiz549Wn0iIiIICgqiR48eWsc6JiaGrKwsLc6kpCQCAwMZNmwYHh5l/3kbPnw4Pj4+6HQ6li1bBsCLL75I9+7dGTBgADk5Obi7uxMZGYlOp2PQoEHcvHkTgKysLEJDQ/Hx8SEwMJCMjIw6PQM7d+7E398fb29vRowYQV5eHgDdu3fnvffew9vbG71eT0ZGBtnZ2SxdupQFCxZgNBqrbcPqXLt2je+//56XX34ZgBYtWuDg4ADA1q1bCQ8PByA8PJyvv/5aKw8LC0NRFPz8/Lh69So5OTl8++23BAcH4+joSLt27QgODmbHjh3k5ORw/fp1/Pz8UBSFsLAw7V6iTGPPsUKYg+TY+8uxK1csx87Gil9++UVyrORYIYQQQjQBDbloVm/gpKqq/wRQFGU98DRwtMI5TwNzbv+8EVikKIqiNvB3gI7lXGfMihQUBdZH+dHrkTYN+XZCVDJp0iQMBgMzZ86sVP76668THh5OeHg4K1euZPLkydqH7CtXrpCcnMw333zDsGHD2LNnDytWrMBkMpGeno7RaCQ/Px9fX18WLFjAvHnzmDt3LosWLQLg1q1bpKamUlRUxIABA9i6dSsdO3Zkw4YNzJ49m5UrV1aKZcqUKURHRxMWFsbixYu18q5du1bqvFWUlZXF7t27OXr0KP7+/mzatImPPvqIZ555hu3btzN8+HBee+013n33XQDGjh3Ltm3bGDp0KOPHj2f58uX4+/sTE/PvCYcJCQm0bduWAwcOUFhYSEBAAIMGDWLz5s2EhIQwe/ZsSkpKuHHjRqVYCgoKiIyMZNeuXbi6ujJq1CjtWMWfp0yZwrRp0+jXrx+nT58mJCSEY8eOAZCRkcHu3bvJzc2lZ8+eREdHExsby+HDh0lPTwfKOt1paWkcPnwYFxcXAFauXImjoyM3b97EZDLx3HPP0b59+0rxZWZmsm7dOpYvX87IkSPZtGkTY8aMISoqiqVLl+Lm5kZKSgoTJ05k165dtXoGnJ2def/990lMTMTOzo4PP/yQTz75RGvvDh06kJaWxpIlS5g/fz4rVqzg1Vdfxd7enjfffLPK7zM1NZWlS5eyYsWKSuWnTp2iY8eOjB8/nh9//BEfHx8WLlyInZ0d58+fp0uXLgA88sgjnD9/HoBz587RtWtX7R7Ozs6cO3fujuXOzs5VykUljTLHFhSVkHLqckPdXjRibYpKyC0oMncYhL0cRV+TD9GTp1FYVEJhcVlc0ZMmMfLFl3hpTBhfrFnNxEmvse6rTRSVlJJ78RI7d3/PX7f9hWHDhrFzVxILFn1KUD9/9qQcwOBVlmN1BiPzPviI2P98n9nvvMd//fdCSkpV8m4WsPv/kikqKmJw8EDWf7WJDh07sumr/2VmzJ9Y8tnySjFOev11xk2I5MWXxrJs6acA5BYU4dDxEdZ8uYHcgiJu3CqmuFQlt6CIwuISTmSeZPu335Fx7Ch/DOrPF+s28M68/+DFkc+zcctWnhr2NOETXmHazD8BEBkxjq82f83gIU8RPm4ccYuX0sfPj/fefotStey+qxJWYGtnz64f9pJ3o4A/BvXn6acGs2vHNsmxkmOFEEII0QQ05KCuE3CmwuuzQJ+azlFVtVhRlGtAe6DS9+YURYkCogB+//vf31dQpaUqU9b/gxZWFqyd0IceHRtujTTReO1evYxff/5nvd6zU7cePD4u6q7ntWnThrCwMOLi4mjZsqVWnpycrA2Yjh07ttKg79ChQ1EUBb1eT+fOndHr9QDodDqys7MxGo1YWFhonakxY8bw7LPPateXlx8/fpzDhw8THBwMlM3SKe8gVLRnzx42bdqkxTJr1qy71mvw4MFYW1uj1+spKSkhNDQUAL1eT3Z2NgC7d+/mo48+4saNG1y+fBmdTkdgYCC5ubn4+/sDZTNutm3bBpTNijl06BAbN24EymawZGZmYjKZiIiIoKioiOHDh2M0GivFkpGRgYuLC25ublp7lM/oqSgxMZGjR/89Bnb9+nVt5s2QIUOwsbHBxsaGTp06aZ2n3+rdu7fW2QSIi4tjy5YtAJw5c4bMzMwqHU4XFxctZh8fH7Kzs8nLy2Pv3r2MGDFCO6/i10jv9gycPXuWo0ePEhAQAJQN5Je3KaA9Dz4+PjUOzFfk6+tbpbMJZV+BTUtLIz4+nj59+jBlyhRiY2P585//XOk8RVFk9/SG1Shz7OX8W4Sv3H9f9xBN0/JhXbC8mA/A8a+/IPdfP9fr/Vv/rhs9h4+94zmlKly6ZUnoMyP54OMF2NracuNGEacu5rMveR/vL1rNqYv59Bk0nNl/iuHUxXzyCorx6/9Hsi/doM3vHqVd+460eqQHP1++iXOPP3DgpxO0dnLDwsIC78eHcOpiPgEhz/BG1FhOXcynoKiEvn8cyqmL+WRmHOXIkSMMDg0BynJsh06PcOp2u5Tbu2cvf45fxamL+fiHDKd09ltVzsm5VsCNW8WcupjP1RtF+PZ7nLPXbtHqkR4Ul5Tg6t2PUxfz+V2PnvzjaCa6vvkk/vVbVn0aR8HNG1y7epXO3Vz5XS9vrl67TidXPacu5tM35Gm++cs2Tl3M5y9/3cGJY0f46quyHJube51/nT4lOVZyrBBCCCGaiCaxva2qqsuAZQC+vr73NcPIwkJh8Yve2Fpb0tWxVb3EJ0RdTZ06FW9vb8aPH1+r821sbACwsLDQfi5/XdM6cxU/7NvZ2QFlm2nodDpt0407qWtnoWKM1tbW2vXlMRYUFDBx4kRSU1Pp2rUrc+bMoaCg4I73VFWV+Ph4QkJCqhz7/vvv2b59O+PGjeONN94gLCysTvEClJaWsm/fPmxtbWusD5R9Jbamdi5vWyibVZSYmEhycjKtWrUiKCio2jr+9t43b96ktLQUBwcHbYZSTdfU9AxYWloSHBzMunXr7nj9nepSG87Ozjg7O9OnT9n44fPPP6+t69e5c2dycnLo0qULOTk5dOrUCQAnJyfOnPn3+OPZs2dxcnLCycmJpKSkSuVBQUE4OTlx9uzZKueLhlGfOba9fQs2Rfetl7hE02J1/V88evsfyn9paU2RtWW93r9tS2vt/jWxUODRjva8PetN+vn1ZkxYODatWvBoR3ssLBQe7WiPtbU1RUVF2uvWtlZ07dCWRzvaY5Vvj12rltr7tG3Zgg52VtrrRzvaY2VlhUVuK2ysLXm0oz221pa4ObXn0Y723DzfEp3Og13/7//uGKelhYJrp9ZYWVlx3aZUi7uic21bYtei7L3btWqBvb29dk4La2tcO7UGoL29Lfa2lji1tiL2nRn8sGcfzl278h9/ngeouHSww8rCQrv25i+taGFVFnurFpbExcXxx+BBAFhZKNjc/r1JjkV7LTlWCGEu3WO2mzuE+5YdO8TcIQjxUGvIQd1zQNcKr51vl1V3zllFUayAtsClBowJALfOrRv6LUQjV5sZtQ3J0dGRkSNHkpCQQEREBAB9+/Zl/fr1jB07lrVr1xIYGFine5aWlrJx40ZGjx7Nl19+Sb9+/aqc07NnTy5cuEBycjL+/v4UFRVx4sQJdDpdpfMCAgJYv349Y8aMYe3atfde0QrKO14dOnQgLy+PjRs38vzzz+Pg4EDr1q1JSUmhT58+rF+/XrsmJCSETz/9lCeeeAJra2tOnDiBk5MTFy9exNnZmcjISAoLC0lLS6vU4ezVqxfZ2dlkZWXx6KOP1tgJGzRoEPHx8cyYMQNAW8qiJq1btyY3N7fG49euXaNdu3a0atWKjIwM9u3bV+v2adOmDS4uLnz11VeMGDECVVU5dOgQXl5etbrez8+PSZMmcfLkSVxdXcnPz+fcuXP84Q9/uGN9rl+/XusYoewrn127duX48eP07NmTv//979pah8OGDWPNmjXExMSwZs0ann76aa180aJFjB49mpSUFNq2bUuXLl0ICQnhrbfe0jZu2blzJx988AGOjo60adOGffv20adPHz7//HNef/31OsXZDDTKHGtjZYlPt3YN+RaikTp27BfsbMo+Vg56+VWzxWFnY4Vdl06MGjWSL9asIiIiAjsbKwL69uUvWzYyduxYVq/7H/oHBmJnY4WVpQW21pbY2VjRysYKCwWtHhWPlZaW8re/fM3o0aP5etP/atdbWii0bGGFnY0Vj+l1XLp4kUNpB+6aY/+yZSNjxozh840btLgratnCEksLBTsbK1pYWdDCyqLSOeU/lx+zVItRgN87PUJJUQHffL2Z559/HqfOHWjTpjWH0w/Sp08ftm7eqNVxyOBQVq1YxpMhwZJj70JyrBBCCCEao4bc3vYA4KYoiouiKC2A0cA3vznnGyD89s/PA7saej1dIRqL6dOnV9qhOz4+nlWrVmEwGPjiiy9YuHBhne5nZ2fH/v378fT0ZNeuXdo6bxW1aNGCjRs3MmvWLLy8vDAajdXuzLxw4UIWL16MXq+vt3XWHBwciIyMxNPTk5CQEEwmk3YsISGByMhIbW3gtm3bAjBhwgQ8PDzw9vbG09OTV155heLiYpKSkvDy8uKxxx5jw4YNTJkypdJ72drasmzZMoYMGYK3t7c2m+W34uLiSE1NxWAw4OHhwdKlS+9Yh/bt2xMQEICnp6fWSa0oNDSU4uJi3N3diYmJwc/Pr05ttHbtWhISEvDy8kKn07F169ZaX9uxY0dWr17NCy+8gMFgwN/fX9sEpiZDhw5ly5Yt1W7ikpqayoQJE6q9Lj4+npdeegmDwUB6ejpvvfUWULbJzXfffYebmxuJiYna+shPPvkkPXr0wNXVlcjISG13c0dHR9555x1tE513331X2yRoyZIlTJgwAVdXVx599FEGDx5c67ZoJiTHCnEHkmMlx/6W5FjJsUIIIcTDRmnI/p2iKE8C/w1YAitVVf0PRVHmAamqqn6jKIot8AXwGHAZGF2+6UtNfH191dTU1AaLWTy8jh07hru7u7nDENXIy8vTdsuOjY0lJyenzh1u0TxV93etKMpBVVV9zRTSAyM5VjQmkmMbL8mx4l415xxb3+orvz4MX8cH+Up+bT0Mv+97+V03x3o/DHWG5lnv+vrv2b3m1wZdU1dV1b8Cf/1N2bsVfi4ARvz2OiFE87J9+3Y++OADiouL6datG6tXrzZ3SEI0epJjhRC1ITlWCNHUycCPEEJUr0lslCaEeLiNGjWKUaNGmTsMIYQQ4qEjOVYIIYQQ4uHUkGvqCiGEEEIIIYQQQgghhKhnMqgrmhXZI0iIh4f8PQvRuMjfpBAPD/l7FkIIIRo/GdQVzYatrS2XLl2SD6lCPARUVeXSpUvY2tqaOxQhBJJjhXiYSI4VQgghmgZZU1c0G87Ozpw9e5YLFy6YOxQhRD2wtbXF2dnZ3GEIIZAcK8TDRnKsEEII0fjJoK5oNqytrXFxcTF3GEIIIcRDR3KsEKK5URQlFFgIWAIrVFWNNXNIQgghmhlZfkEIIYQQQgghhKglRVEsgcXAYMADeEFRFA/zRiWEEKK5kUFdIYQQQgghhBCi9noDJ1VV/aeqqreA9cDTZo5JCCFEMyODukIIIYQQQgghRO05AWcqvD57u0wIIYR4YJSmtkuxoigXgJ/NHccD9n/UqwAAEJpJREFU0AG4aO4gmihpu/sj7XfvpO3uT321XzdVVTvWw32anWaSY+Xv9P5I+90fab97J213fyTH1iNFUZ4HQlVVnXD79Vigj6qqr1U4JwqIuv2yJ3D8gQd6b5rr31pzrHdzrDNIvZubplLve8qvTW6jtObyIUJRlFRVVX3NHUdTJG13f6T97p203f2R9jO/5pBj5Tm7P9J+90fa795J290fab96dw7oWuG18+0yjaqqy4BlDzKo+tBcn5XmWO/mWGeQeps7jgftYa+3LL8ghBBCCCGEEELU3gHATVEUF0VRWgCjgW/MHJMQQohmpsnN1BVCCCGEEEIIIcxFVdViRVFeA74FLIGVqqoeMXNYQgghmhkZ1G28mtxXdRoRabv7I+1376Tt7o+0n3gQ5Dm7P9J+90fa795J290fab96pqrqX4G/mjuOBtBcn5XmWO/mWGeQejc3D3W9m9xGaUIIIYQQQgghhBBCCNGcyZq6QgghhBBCCCGEEEII0YTIoG4joihKV0VRdiuKclRRlCOKokwxd0xNkaIoloqi/ENRlG3mjqUpURTFQVGUjYqiZCiKckxRFH9zx9SUKIoy7fbf7WFFUdYpimJr7pgaM0VRViqK8quiKIcrlDkqivKdoiiZt/+/nTljFA8XybH3T/LrvZMce38kx9aN5FhRW83lWanpM0BNdVXKxCmKclJRlEOKonibtwa1V5ff6Z3qqShK+O3zMxVFCTdHXepCUZRsRVF+UhQlXVGU1NtlD129a/j9zlEU5dztuqcrivJkhWN/ul3P44qihFQoD71ddlJRlJgK5S6KoqTcLt+glG0E2ShU91mqPn/HiqL43H6GTt6+VjFHPetKBnUbl2JguqqqHoAfMElRFA8zx9QUTQGOmTuIJmghsENV1V6AF9KGtaYoihMwGfBVVdWTsg0zRps3qkZvNRD6m7IY4O+qqroBf7/9Woj6Ijn2/kl+vXeSY++R5Nh7shrJsaJ2VtM8npWaPgPUVNfBgNvt/0UBnz74kO/Zamr/O622noqiOALvAX2A3sB7TWRw/3FVVY2qqvrefv0w1ns1VX+/AAtu1914e71vbj/jowHd7WuWKGX/QG8J/P/27j/ojqq+4/j7kzygISBQZdIUsMlEEYFICD+aEMpEZGKnWqOIBUURmhH8QZLSAcGWqba1M6ggFMV0CIRfIURE0FgEwiASRCIhIZI8YFtLgkADCYQABQyQfPvHOZdnud77/MpN9u59Pq+ZZ569Z/fsc87u5n4355w9eynpOBwAfLJwT/yNvK93Ac8BM7ZrbQam0b1UK8/xHOBzhXyNjnPbcaNuG4mIdRGxIi+/SLpI9y63VNUiaR/gQ8DlZZelSiTtDhwNXAEQEa9GxKZyS1U5XcAISV3ALsD/llyethYRS4CNdcnTgavz8tXAR3dooayjOcZuG8fXwXOMbQnH2AFwjLX+GirXSi/3AM3qOh24JpKlwB6SRu/gYg/KAM9ps3p+ELgjIjZGxHPAHVSkgatOx9W7yfltZjqwMCI2R8Qa4LekhswjgN9GxKMR8SqwEJieR6YeA9yY87fNv/9e7qVaco7zurdFxNJILx67hjape1/cqNumJI0BDgF+VW5JKudi4MvA1rILUjFjgQ3AlUqP1l4uaWTZhaqKiHgSuAD4HbAOeD4iFpdbqkoaFRHr8vJTwKgyC2OdyzF2UBxfB88xdhs4xraMY6z1V0dfK3X3AM3qujfweCHbE1S7I3ig9axi/QNYLGm5pNNy2lCod80ZeZqBeYWRpwOt59uBTRHxel16O2h2L9Wqc7x3Xq5Pb3tu1G1DknYFfgj8bUS8UHZ5qkLSh4H1EbG87LJUUBcwEZgTEYcAL9EZj1rtEDlwTicFmz8BRkr6dLmlqrbcQxpll8M6j2PswDm+bjPH2G3gGNt6jrHWX512rfR2D9BpdW2mg+t5VERMJD12/yVJRxdXdnC9IU0bMA6YQOr8vLDc4mwXfd5Ldfg5bsqNum1G0k6kQHNdRNxUdnkqZgrwEUlrSY8QHCNpfrlFqowngCciojZq7UbSl6b1z7HAmojYEBGvATcBR5Zcpip6uvZoW/69vuTyWIdxjB00x9dt4xi7bRxjW8Mx1vqrI6+VJvcAzer6JLBvIfs+Oa2qBlrPytU/P9VBRKwHbiZNMdDx9QaIiKcjYktEbAXmkuoOA6/ns6RpCrrq0ttBs3upVp3jJ/NyfXrbc6NuG8lzmFwBPBIR3y67PFUTEV+JiH0iYgxpQvCfRYRHcvRDRDwFPC7pPTnpA8DDJRapan4HTJK0S/53/AH8EpzBWATU3kD6WeDHJZbFOoxj7OA5vm4bx9ht5hjbGo6x1l8dd630cg/QrK6LgJOVTCJN+7KO6hpoPW8HpknaMz8tMS2ntSVJIyXtVlsmlXc1HV7vmrr5nj9Gqjukep4o6S2SxpJe/nU/sAx4t6SxknYm3dstyiNd7wKOz/nb5t9/L/dSLTnHed0Lkibl74uTaZO696Wr701sB5oCfAZYJWllTvv7yG8vNNvOZgLX5S/2R4FTSy5PZUTEryTdCKwgvV33QeCyckvV3iRdD0wF3iHpCdJbSM8HbpA0A3gM+OvySmgdyDHWyuQYO0iOsQPnGGv9NYSulYb3ADSv60+BvyS9WOplKvSdPcBz2rCeEbFR0r+QGv8A/jki+vtyrjKMAm5ObXF0AQsi4jZJy+iwejc5v1MlTSBNPbAWOB0gIrol3UBq/Hwd+FJEbMn7OYPUwDkcmBcR3flPnAMslPR1Ury9YgdVrT8a3UsNo3Xn+IvAVcAI4Nb80/aUGuPNzMzMzMzMzMzMrAo8/YKZmZmZmZmZmZlZhbhR18zMzMzMzMzMzKxC3KhrZmZmZmZmZmZmViFu1DUzMzMzMzMzMzOrEDfqmpmZmZmZmZmZmVWIG3XNSiRpi6SVhZ9zW7jvMZJWt2p/ZmZmVeIYa2ZmZttKydmS9iy7LGb13KhrVq5XImJC4ef8sgtkZmbWIRxjzcysZSSFpAsLn8+S9LUW7fv/WrGf3vadOyR/npcPk3TJ9vqbrSRpqqT/yMunDPaYS9pD0hcHkfVU4J3AoP7u9pKPy5Fll8PK5UZdszYkaa2kb0paJel+Se/K6WMk/UzSQ5LulPTOnD5K0s2Sfp1/al/uwyXNldQtabGkEXn7WZIezvtZWFI1zczMdjjHWDMzG6TNwHGS3lHGH5fU1ap9RcQDETGrVfuriD2Aho26fRzb30fETGCJpLdsl5INzlTAjbpDnBt1zco1ou7R0BMK656PiPHAd4GLc9p3gKsj4n3AdUCtd/US4O6IOBiYCHTn9HcDl0bEgcAm4OM5/VzgkLyfz2+vypmZmZXIMdbMzFrpdeAy4Mz6Fb10DF4laY6kpZIezaMr50l6RNJVdfu4KHcU3ilpr5z2c0kXS3oAmC3pUEl3S1ou6XZJoxuUZayk+3Ln5dcLq7YAG/M2xdGvX5N0taR7JD0m6bhC5+dtknbK2/2jpGWSVku6TJJy+uG53islfUt5eiJJw/PnZXn96Tl9tKQlefvVkv68QR3+QtJvJK0AjiusegWojTzeS9IP8/6XSZpSqM+8fOwelVRrvD4fGFco59Rc50XAwznvj/Kx7ZZ0GkBELJC0FrgbGJ3PXaNO3XH5eC3P+91/INeApGn5vK2Q9ANJu+b0tZL+KaevkrS/pDGke4wzc33+4Bja0OBGXbNy1T8a+v3CuusLvyfn5cnAgrx8LXBUXj4GmAMQEVsi4vmcviYiVubl5cCYvPwQcJ2kT5NuTszMzDqNY6yZmbXapcBJknavS2/WMQiwJynGnAksAi4CDgTGS5qQtxkJPJA7Cu8GvlrIv3NEHJb3+R3g+Ig4FJgH/GuDMv4bMCd3Xq6rJUbE4xFxXIPtAcaR4t1HgPnAXTn/K8CH8jbfjYjDI+IgYATw4Zx+JXB6REwgNRzXzCB1oh4OHA58TtJY4FPA7Xn7g4GVhTxIeiswF/gr4FDgjwt1+H5EXFCo50V5/x8HLi/sZn/gg8ARwFdzw/S5wP/ke4Kz83YTgdkRsV/+/Df52B4GzJL09gbHqlmn7mXAzJz/LOB7hTy9XgNKo7/PA46NiInAA8DfFfI/k9PnAGdFxFrg33P9J0TEPQ3KaUNAy4bvm1nLRZPlgdhcWN5CCr6QAvPRpED5D5LGR4T/42lmZkOFY6yZmQ1YRLwg6RpgFqnBs2YyPSNKrwW+WVj3k4gISauApyNiFYCkblKH4EpgK1DrfJwP3FTIX0t/D3AQcEceJDucQqNtwRR6GhqvBb7Rj6rdGhGv5TIOB27L6avo6bR8v6QvA7sAfwR0S7oH2C0i7svbLKCnsXca8D5Jx+fPu5MaRJcB83JD648KHaQ1+5M6Tv8bQNJ84LQGZT4WOCAfC4C31Ua3ArdExGZgs6T1wKgm9b4/ItYUPs+S9LG8vG8u77N1ef6gUzf/3SOBHxTKU5yqoa9rYB/gAODenH9n4L5C/tr1sJw3j1y2Ic4jdc3a1wmF37Uv9F8CJ+blk4Baj9ydwBfgjcdc6nuO3yBpGLBvRNwFnEMKrrs2297MzKwDOcaamdlgXUwahTqyn9vXOgG38uYOwa00H2hX7HB8Kf8W0F14AmV8REzrR/5+lzEitgKvRUQt/1agK4+e/R5plPB40kjat/axT5FGrtbKOzYiFkfEElLn55PAVZJOHmBZa4YBkwr73zsiai+cq+94bXaca8cWSVNJDcWT85RLD9K4jo32PQzYVPeE0Hsb5Gl2DQi4o5D3gIiY0SB/b3WxIciNumblqp/vr/hm7j0lPQTMpmfeppnAqTn9M3kd+ff7c8/fclIvXzPDgfl52weBSyJiUwvrZGZm1g4cY83MrOUiYiNwA6lht6ZZx2B/DQNqI1o/BfyiwTb/CewlaTKApJ0kHdhgu3vrytIKtcbNZ/Ko1OMBcox7UdKf5fUnFvLcDnxBPXPy7idppKQ/JY1WnUuaMmFi3d/6DWn067j8+ZNNyrSYFLvJ+5/QZLuaF4Hdelm/O/BcRLyc58Od1Mf+3hARLwBrJH0il0WSDu5vfmApMEU9L28dKWm/PvL0VR8bAtzCb1aiiBjey+pvRcQ5dds/RprrqH4/TwPTG+zjoMI2FxTSj2qwrZmZWcdwjDUzs+3oQuCMwueZwJWSzgY2AKcOcH8vAUdIOg9YT88TJW+IiFfzVAaX5KdGukijhrvrNp0NLJB0DvDjAZajoYjYJGkusBp4ijSFQs0MYK6kraT5gGtzz19OmlpghdKcAhuAjwJTgbMlvUZ66dmbRupGxO/zS8pukfQyqYG8UePlLODS3BnbBSyhlxeURsSzku5VepHbrcAtdZvcBnxe0iOkBvSlzY9IQycBc/I53AlYCPy6PxkjYoOkU4DrJdWmbTgP+K9esv0EuFHSdNKIaM+rOwSpZ1S9mbULpbdrHhYRz5RdFjMzs07iGGtmZtY6knatTXsg6VxgdETM7iObmbWAG3XNzMzMzMzMzGzAJJ0AfIU0WvYx4JSI2FBuqcyGBjfqmpmZmZmZmZmZmVWIX5RmZmZmZmZmZmZmViFu1DUzMzMzMzMzMzOrEDfqmpmZmZmZmZmZmVWIG3XNzMzMzMzMzMzMKsSNumZmZmZmZmZmZmYV4kZdMzMzMzMzMzMzswr5fx5zJrYt4Ty6AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 1728x432 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.figure(figsize=(24,6))\n",
        "\n",
        "plt.subplot(131)\n",
        "plt.title('Training accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "for i in range(len(train_sizes)):\n",
        "  plt.plot(np.arange(1,11,1), training_histories[i].history['accuracy'], label=\"Nombre d'images d'entraînement : {}\".format(train_sizes[i]))\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(132)\n",
        "plt.title('Validation accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "for i in range(len(train_sizes)):\n",
        "  plt.plot(np.arange(1,11,1), training_histories[i].history['val_accuracy'], label=\"Nombre d'images d'entraînement : {}\".format(train_sizes[i]))\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(133)\n",
        "plt.title(\"Temps d'exécution en fonction du nombre d'images d'entraînement\")\n",
        "plt.xlabel(\"Nombre d'images d'entraînement\")\n",
        "plt.ylabel(\"temps d'exécution (en secondes)\")\n",
        "plt.bar(np.arange(1,7,1), times)\n",
        "plt.xticks(ticks=np.arange(1,7,1), labels=train_sizes);"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eb-ObwSwiu_G"
      },
      "source": [
        "# <a name=\"architecture\"></a> V - Influence de l'architecture du réseau de neurones"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2UtR_OVSp4Rs"
      },
      "source": [
        "## <a name=\"dense\"></a> A - Architecture très basique VS architecture plus complexe"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "As-P23AWmeeM"
      },
      "source": [
        "Nous allons maintenant évaluer l'influence de l'architecture d'un modèle sur ses performances.\n",
        "- (a) Exécuter la cellule suivante pour instancier le modèle très simple que nous avons utilisé dans les parties précédentes."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "EZx-S4kvYrMb"
      },
      "outputs": [],
      "source": [
        "model1 = Sequential()\n",
        "model1.add(Dense(units = 20, input_dim = num_pixels, kernel_initializer ='normal', activation ='tanh'))\n",
        "model1.add(Dense(units = num_classes, kernel_initializer ='normal', activation ='softmax'))\n",
        "\n",
        "model1.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QiJ7EcWynYbr"
      },
      "source": [
        "- (b) Instancier un modèle appelé ```model2``` avec au moins trois couches denses (en comptant la couche finale). Vous êtes libres de choisir les paramètres des couches."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "frJPj3l_Yz3b"
      },
      "outputs": [],
      "source": [
        "model2 = Sequential()\n",
        "model2.add(Dense(units = 1024, input_dim = num_pixels, kernel_initializer ='normal', activation ='tanh'))\n",
        "model2.add(Dense(units = 256, kernel_initializer ='normal', activation ='tanh'))\n",
        "model2.add(Dense(units = num_classes, kernel_initializer ='normal', activation ='softmax'))\n",
        "\n",
        "model2.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yg5a88Tjn-6F"
      },
      "source": [
        "- (c) Entraîner les modèles grâce à la méthode ```fit``` :\n",
        ">  - L'entraînement devra se faire sur 10 epochs (paramètre ```epochs```).\n",
        ">  \n",
        ">  - Les batchs devront avoir une taille de 20 (paramètre ```batch_size```)\n",
        ">\n",
        ">  - La perfomance du modèle devra être évaluée sur un échantillon de validation contenant 20% des données (paramètre ```validation_split```).\n",
        ">\n",
        ">  - La sortie des entraînements devra être stockée dans des variables appelées ```training_history_1``` et ```training_history_2```."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ldwcGd23ZBhe",
        "outputId": "f8f34810-5a9b-46e8-eaf3-9b93df528842"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "2400/2400 [==============================] - 4s 1ms/step - loss: 0.8607 - accuracy: 0.8022 - val_loss: 0.2612 - val_accuracy: 0.9262\n",
            "Epoch 2/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.2528 - accuracy: 0.9281 - val_loss: 0.2171 - val_accuracy: 0.9392\n",
            "Epoch 3/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.2041 - accuracy: 0.9411 - val_loss: 0.1994 - val_accuracy: 0.9442\n",
            "Epoch 4/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1821 - accuracy: 0.9482 - val_loss: 0.1873 - val_accuracy: 0.9482\n",
            "Epoch 5/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1633 - accuracy: 0.9531 - val_loss: 0.1864 - val_accuracy: 0.9479\n",
            "Epoch 6/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1520 - accuracy: 0.9555 - val_loss: 0.1763 - val_accuracy: 0.9503\n",
            "Epoch 7/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1414 - accuracy: 0.9588 - val_loss: 0.1783 - val_accuracy: 0.9501\n",
            "Epoch 8/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1368 - accuracy: 0.9588 - val_loss: 0.1777 - val_accuracy: 0.9492\n",
            "Epoch 9/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1268 - accuracy: 0.9619 - val_loss: 0.1721 - val_accuracy: 0.9527\n",
            "Epoch 10/10\n",
            "2400/2400 [==============================] - 3s 1ms/step - loss: 0.1164 - accuracy: 0.9654 - val_loss: 0.1754 - val_accuracy: 0.9508\n"
          ]
        }
      ],
      "source": [
        "training_history_1 = model1.fit(X_train, y_train, epochs = 10, batch_size = 20, validation_split = 0.2)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UTeIHqr0ZWZh",
        "outputId": "0ddcc594-87db-4543-a7d2-d52972593ab1"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "2400/2400 [==============================] - 20s 8ms/step - loss: 0.3480 - accuracy: 0.8935 - val_loss: 0.1760 - val_accuracy: 0.9486\n",
            "Epoch 2/10\n",
            "2400/2400 [==============================] - 19s 8ms/step - loss: 0.1341 - accuracy: 0.9579 - val_loss: 0.1336 - val_accuracy: 0.9594\n",
            "Epoch 3/10\n",
            "2400/2400 [==============================] - 19s 8ms/step - loss: 0.1041 - accuracy: 0.9678 - val_loss: 0.1304 - val_accuracy: 0.9617\n",
            "Epoch 4/10\n",
            "2400/2400 [==============================] - 19s 8ms/step - loss: 0.0905 - accuracy: 0.9713 - val_loss: 0.1210 - val_accuracy: 0.9650\n",
            "Epoch 5/10\n",
            "2400/2400 [==============================] - 20s 8ms/step - loss: 0.0707 - accuracy: 0.9785 - val_loss: 0.1209 - val_accuracy: 0.9647\n",
            "Epoch 6/10\n",
            "2400/2400 [==============================] - 20s 8ms/step - loss: 0.0603 - accuracy: 0.9801 - val_loss: 0.1153 - val_accuracy: 0.9683\n",
            "Epoch 7/10\n",
            "2400/2400 [==============================] - 20s 8ms/step - loss: 0.0560 - accuracy: 0.9809 - val_loss: 0.1166 - val_accuracy: 0.9659\n",
            "Epoch 8/10\n",
            "2400/2400 [==============================] - 20s 8ms/step - loss: 0.0516 - accuracy: 0.9829 - val_loss: 0.1246 - val_accuracy: 0.9655\n",
            "Epoch 9/10\n",
            "2400/2400 [==============================] - 20s 8ms/step - loss: 0.0495 - accuracy: 0.9839 - val_loss: 0.1540 - val_accuracy: 0.9607\n",
            "Epoch 10/10\n",
            "2400/2400 [==============================] - 20s 8ms/step - loss: 0.0449 - accuracy: 0.9843 - val_loss: 0.1149 - val_accuracy: 0.9678\n"
          ]
        }
      ],
      "source": [
        "training_history_2 = model2.fit(X_train, y_train, epochs = 10, batch_size = 20, validation_split = 0.2)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sRcSI0XXogWq"
      },
      "source": [
        "- (d) Tracer sur un même graphe l'évolution de la précision d'entraînement en fonction des epochs pour les deux modèles.\n",
        "\n",
        "- (e) Sur un deuxième graphe, tracer l'évolution de la précision de validation en fonction des epochs pour les deux modèles.\n",
        "\n",
        "Que constatez-vous?"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 513
        },
        "id": "JArsgqg_cXqv",
        "outputId": "6720cee2-07c3-42c0-bdf1-ef7e7046dc49"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1728x576 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.figure(figsize=(24,8))\n",
        "\n",
        "plt.subplot(121)\n",
        "plt.title('Training Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_1.history['accuracy'], label = 'model1', color='blue')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_2.history['accuracy'], label = 'model2', color='red')\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(122)\n",
        "plt.title('Validation Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_1.history['val_accuracy'], label = 'model1', color='blue')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_2.history['val_accuracy'], label = 'model2', color='red')\n",
        "plt.legend();"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8ujuPijbi2Ho"
      },
      "source": [
        "## <a name=\"CNN\"></a> B - Introduction aux réseaux de neurones convolutionnels\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UyJbmG8spTLP"
      },
      "source": [
        "Les couches denses ne sont pas les seules couches que l'on peut créer sur python. Par exemple, on peut définir le réseau de convolution suivant (architecture **LeNet**), qui est bien plus adapté au traitement des images que les modèles définis précédemment.\n",
        "- (f) Exécuter les cellules suivantes pour constater l'efficacité de l'architecture **LeNet**."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "vJ9bwTyPdCto"
      },
      "outputs": [],
      "source": [
        "X_train_conv = X_train.reshape((-1, 28, 28, 1))\n",
        "X_test_conv = X_test.reshape((-1, 28, 28, 1))\n",
        "\n",
        "model_conv = Sequential()\n",
        "model_conv.add(Conv2D(filters=30, kernel_size=(5,5), padding='valid', input_shape=(28,28,1), activation='relu'))\n",
        "model_conv.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model_conv.add(Conv2D(filters=16, kernel_size=(3,3), padding='valid', activation='relu'))\n",
        "model_conv.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model_conv.add(Dropout(rate=0.2))\n",
        "model_conv.add(Flatten())\n",
        "model_conv.add(Dense(units = 128, activation ='relu'))\n",
        "model_conv.add(Dense(units = num_classes, activation ='softmax'))\n",
        "\n",
        "model_conv.compile(loss = 'categorical_crossentropy', optimizer = 'adam', metrics = ['accuracy'])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sKPtVldjfgw7",
        "outputId": "e56b9fa7-e51c-4b33-83f5-542a7cd52d73"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "2400/2400 [==============================] - 37s 15ms/step - loss: 0.4091 - accuracy: 0.8727 - val_loss: 0.0632 - val_accuracy: 0.9823\n",
            "Epoch 2/10\n",
            "2400/2400 [==============================] - 36s 15ms/step - loss: 0.0695 - accuracy: 0.9782 - val_loss: 0.0442 - val_accuracy: 0.9871\n",
            "Epoch 3/10\n",
            "2400/2400 [==============================] - 36s 15ms/step - loss: 0.0500 - accuracy: 0.9842 - val_loss: 0.0486 - val_accuracy: 0.9860\n",
            "Epoch 4/10\n",
            "2400/2400 [==============================] - 36s 15ms/step - loss: 0.0405 - accuracy: 0.9867 - val_loss: 0.0375 - val_accuracy: 0.9893\n",
            "Epoch 5/10\n",
            "2400/2400 [==============================] - 37s 15ms/step - loss: 0.0312 - accuracy: 0.9902 - val_loss: 0.0414 - val_accuracy: 0.9882\n",
            "Epoch 6/10\n",
            "2400/2400 [==============================] - 36s 15ms/step - loss: 0.0295 - accuracy: 0.9909 - val_loss: 0.0337 - val_accuracy: 0.9902\n",
            "Epoch 7/10\n",
            "2400/2400 [==============================] - 37s 15ms/step - loss: 0.0245 - accuracy: 0.9922 - val_loss: 0.0358 - val_accuracy: 0.9903\n",
            "Epoch 8/10\n",
            "2400/2400 [==============================] - 36s 15ms/step - loss: 0.0257 - accuracy: 0.9916 - val_loss: 0.0432 - val_accuracy: 0.9889\n",
            "Epoch 9/10\n",
            "2400/2400 [==============================] - 37s 15ms/step - loss: 0.0197 - accuracy: 0.9937 - val_loss: 0.0360 - val_accuracy: 0.9908\n",
            "Epoch 10/10\n",
            "2400/2400 [==============================] - 37s 15ms/step - loss: 0.0167 - accuracy: 0.9943 - val_loss: 0.0371 - val_accuracy: 0.9908\n"
          ]
        }
      ],
      "source": [
        "training_history_conv = model_conv.fit(X_train_conv, y_train, epochs = 10, batch_size = 20, validation_split = 0.2)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 513
        },
        "id": "RgAMhjq7hyxc",
        "outputId": "0c85fcfa-14a7-471e-ec36-36df9f186366"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1728x576 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "plt.figure(figsize=(24,8))\n",
        "\n",
        "plt.subplot(121)\n",
        "plt.title('Training Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_1.history['accuracy'], label = 'model1', color='blue')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_2.history['accuracy'], label = 'model2', color='red')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_conv.history['accuracy'], label = 'model_conv', color='green')\n",
        "plt.legend()\n",
        "\n",
        "plt.subplot(122)\n",
        "plt.title('Validation Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_1.history['val_accuracy'], label = 'model1', color='blue')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_2.history['val_accuracy'], label = 'model2', color='red')\n",
        "plt.plot(np.arange(1 , 11, 1), training_history_conv.history['val_accuracy'], label = 'model_conv', color='green')\n",
        "plt.legend();"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0QjhXvIZSJZS"
      },
      "source": []
    }
  ],
  "metadata": {
    "colab": {
      "toc_visible": true,
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}