Deep Q Learning

[martin]

http://cs.stanford.edu/people/karpathy/convnetjs/demo/rldemo.html