My take on playing atari using reinforcement learning: https://www.cs.toronto.edu/~vmnih/docs/dqn.pdf

Work still in progress (it's bugged). My goal is to implement version with double q-learning and prioritized experience replay.