I've got derivatives in for the network so it does learn with linear, tanh, softmax, and relu nonlinearities. I'd like to add a simple convolutional example (like LeNet learning to do MNIST handwritten digit recognition -- common benchmark for computer vision) to show this is actually useful.
What do you all think? Do you think this is a useful direction for Gneural Network to move? I'm really excited about the project and the possibilities. I'm also really enjoying working with C - Python and its libraries are painful to work with.