01 August 2015
Recently I am interested in reinforcement learning and I also wanted to play with recurent neural networks for some time, after a bunch of image captioning papers were published. For both these models I found nice reference implementation on javascript from Andrej Karpathy: recurrentjs & reinforcejs and his useful article: The Unreasonable Effectiveness of Recurrent Neural Networks.
Initial code in general follows the code from two libraries above and its main purpose is to implement reference algorithms on C++.
8 November 2015
Added Convolutional Neural Network model.