model update

Author: Suwon Shon

scripts/siamese_model_phone_cz.py

@@ -0,0 +1,62 @@
+import tensorflow as tf 
+import numpy as np
+class siamese:
+
+    # Create model
+    def __init__(self,input_dim):
+        self.x1 = tf.placeholder(tf.float32, [None, input_dim])
+        self.x2 = tf.placeholder(tf.float32, [None, input_dim])
+
+        with tf.variable_scope("siamese") as scope:
+            self.a1,self.b1,self.o1 = self.network(self.x1)
+            scope.reuse_variables()
+            self.a1,self.b2,self.o2 = self.network(self.x2)
+            
+        # Create loss
+        self.y_ = tf.placeholder(tf.float32, [None])
+        self.loss = self.loss_with_cds()
+
+    def network(self, x):
+        weights = []
+        kernel_size =180
+        stride = 22
+        depth=40
+        conv1 = self.conv_layer(x, kernel_size,stride,depth,'conv1')
+        conv1r = tf.nn.relu(conv1)
+        n_prev_weight = int(x.get_shape()[1])
+        conv1_d = tf.reshape(conv1r,[-1, int(round(n_prev_weight/stride+0.5)*depth)])
+        
+        fc1 = self.fc_layer(conv1_d, 1500, "fc1")
+        ac1 = tf.nn.relu(fc1)
+        fc2 = self.fc_layer(ac1, 600, "fc2")   
+        ac2 = tf.nn.relu(fc2)
+        fc3 = self.fc_layer(ac2, 200, "fc3")
+        return fc1,fc2,fc3
+
+    def fc_layer(self, bottom, n_weight, name):
+        print( bottom.get_shape())
+        n_prev_weight = bottom.get_shape()[1]
+        W = tf.get_variable(name+'W', dtype=tf.float32, shape=[n_prev_weight, n_weight], initializer=tf.contrib.layers.xavier_initializer())
+        b = tf.get_variable(name+'b', dtype=tf.float32, initializer=tf.random_uniform([n_weight],-0.001,0.001, dtype=tf.float32))
+        fc = tf.nn.bias_add(tf.matmul(bottom, W), b)
+        return fc
+
+    def conv_layer(self, bottom, kernel_size, stride, depth, name):        
+        n_prev_weight = int(bottom.get_shape()[1])
+        num_channels = 1 # for 1 dimension
+        inputlayer = tf.reshape(bottom, [-1,n_prev_weight,1])
+        initer = tf.truncated_normal_initializer(stddev=0.1)
+        W = tf.get_variable(name+'W', dtype=tf.float32, shape=[kernel_size, num_channels, depth], initializer=tf.contrib.layers.xavier_initializer())
+        b = tf.get_variable(name+'b', dtype=tf.float32, initializer=tf.constant(0.001, shape=[depth*num_channels], dtype=tf.float32))
+        
+        conv = tf.nn.bias_add( tf.nn.conv1d(inputlayer, W, stride, padding='SAME'), b)
+        return conv
+
+    def loss_with_cds(self):
+        labels_t = self.y_
+        cds = tf.reduce_sum(tf.multiply(self.o1,self.o2),1)
+        eucd2 = tf.reduce_mean(tf.pow(tf.subtract(labels_t,cds),2))
+        eucd = tf.sqrt(eucd2, name="eucd")
+        return eucd
+    
+

scripts/siamese_model_phone_ru.py

@@ -0,0 +1,62 @@
+import tensorflow as tf 
+import numpy as np
+class siamese:
+
+    # Create model
+    def __init__(self,input_dim):
+        self.x1 = tf.placeholder(tf.float32, [None, input_dim])
+        self.x2 = tf.placeholder(tf.float32, [None, input_dim])
+
+        with tf.variable_scope("siamese") as scope:
+            self.a1,self.b1,self.o1 = self.network(self.x1)
+            scope.reuse_variables()
+            self.a1,self.b2,self.o2 = self.network(self.x2)
+            
+        # Create loss
+        self.y_ = tf.placeholder(tf.float32, [None])
+        self.loss = self.loss_with_cds()
+
+    def network(self, x):
+        weights = []
+        kernel_size =180
+        stride = 22
+        depth=40
+        conv1 = self.conv_layer(x, kernel_size,stride,depth,'conv1')
+        conv1r = tf.nn.relu(conv1)
+        n_prev_weight = int(x.get_shape()[1])
+        conv1_d = tf.reshape(conv1r,[-1, int(round(n_prev_weight/stride+0.5)*depth)])
+        
+        fc1 = self.fc_layer(conv1_d, 1500, "fc1")
+        ac1 = tf.nn.relu(fc1)
+        fc2 = self.fc_layer(ac1, 600, "fc2")   
+        ac2 = tf.nn.relu(fc2)
+        fc3 = self.fc_layer(ac2, 200, "fc3")
+        return fc1,fc2,fc3
+
+    def fc_layer(self, bottom, n_weight, name):
+        print( bottom.get_shape())
+        n_prev_weight = bottom.get_shape()[1]
+        W = tf.get_variable(name+'W', dtype=tf.float32, shape=[n_prev_weight, n_weight], initializer=tf.contrib.layers.xavier_initializer())
+        b = tf.get_variable(name+'b', dtype=tf.float32, initializer=tf.random_uniform([n_weight],-0.001,0.001, dtype=tf.float32))
+        fc = tf.nn.bias_add(tf.matmul(bottom, W), b)
+        return fc
+
+    def conv_layer(self, bottom, kernel_size, stride, depth, name):        
+        n_prev_weight = int(bottom.get_shape()[1])
+        num_channels = 1 # for 1 dimension
+        inputlayer = tf.reshape(bottom, [-1,n_prev_weight,1])
+        initer = tf.truncated_normal_initializer(stddev=0.1)
+        W = tf.get_variable(name+'W', dtype=tf.float32, shape=[kernel_size, num_channels, depth], initializer=tf.contrib.layers.xavier_initializer())
+        b = tf.get_variable(name+'b', dtype=tf.float32, initializer=tf.constant(0.001, shape=[depth*num_channels], dtype=tf.float32))
+        
+        conv = tf.nn.bias_add( tf.nn.conv1d(inputlayer, W, stride, padding='SAME'), b)
+        return conv
+
+    def loss_with_cds(self):
+        labels_t = self.y_
+        cds = tf.reduce_sum(tf.multiply(self.o1,self.o2),1)
+        eucd2 = tf.reduce_mean(tf.pow(tf.subtract(labels_t,cds),2))
+        eucd = tf.sqrt(eucd2, name="eucd")
+        return eucd
+    
+