Deep Q learning by DeepMind implemented in Google's tensor flow. Multithreaded gameplay speeds up training! - View it on GitHub