An Implementation of Deep Neural Network for Regression and Classification

The deeplearning package is an R package that implements deep neural networks in R. It employes Rectifier Linear Unit functions as its building blocks and trains a neural network with stochastic gradient descent method with batch normalization to speed up the training and promote regularization. Neural networks of such kind of architecture and training methods are state of the art and even achieved suplassing human-level performance in ImageNet competition. The deeplearning package is inspired by another R package darch which implements layerwise Restricted Boltzmann Machine pretraining and dropout and uses its class DArch as the default class.