DeepBeliefNetwork

C++ implementation of the Deep Belief Network for classifying MNIST digits discussed in G. Hinton paper "A fast learning algorithm for deep belief nets".

State of the project

Until now I have implemented the associative memory layer in Python and the data handiling and the associative memory in C++.

To do

• hidden layer implementation
• try to parallelize with CUDA for gpu acceleration
• fine tuning with wake-sleep algorithm