http://www.wildml.com/2016/10/learning-reinforcement-learning/
https://github.com/dennybritz/reinforcement-learning