Abstract: DQN is a classical algorithm in reinforcement learning, combining traditional Q-learning with neural network. In previous researches, DQN has been used to implement Atari Game, and other ...