Implementing random deep learning things in julia for fun
Basic DNN works (no autoencoder/RBM training for deep tanh/sigmoid networks, sorry), only works on L1 loss right now.
Tutorial: Install julia, install IJulia and PyPlot, open command window in folder with NNTestbed.ipynb, run "ipython notebook --profile julia" and play with the notebook.
TODO:
-more loss functions
-refactor code to make it cleaner
-sigmoid/tanh pretraining with autoencoder
-(Clockwork) RNN/LSTM/capabilities with more exotic network structure