deeplearning: An Implementation of Deep Neural Network for Regression and Classification

An implementation of deep neural network with rectifier linear units trained with stochastic gradient descent method and batch normalization. A combination of these methods have achieved state-of-the-art performance in ImageNet classification by overcoming the gradient saturation problem experienced by many deep architecture neural network models in the past. In addition, batch normalization and dropout are implemented as a means of regularization. The deeplearning package is inspired by the darch package and uses its class DArch.

Downloads:

Linking:

Please use the canonical form https://CRAN.R-project.org/package=deeplearning to link to this page.