R Deep Learning Cookbook

Setting up a Deep Belief Network

Deep belief networks are a type of Deep Neural Network (DNN), and are composed of multiple hidden layers (or latent variables). Here, the connections are present only between the layers and not within the nodes of each layer. The DBN can be trained both as an unsupervised and supervised model.

The unsupervised model is used to reconstruct the input with noise removal and the supervised model (after pretraining) is used to perform classification. As there are no connections within the nodes in each layer, the DBNs can be considered as a set of unsupervised RBMs or autoencoders, where each hidden layer serves as a visible layer to its subsequent connected hidden layer.

This kind of stacked RBM enhances the performance ...