Playing Atari with Deep Reinforcement Learning
Abstract¶
We present the first deep learning model to successfully learn control policies directly from high-dimensional sensory input using reinforcement learning. The model is a convolutional neural network, trained with a variant of Q-learning, whose input is raw pixels and whose output is a value function estimating future rewards. We apply our method to seven Atari 2600 games from the Arcade Learning Environment, with no adjustment of the architecture or learning algorithm. We find that it outperforms all previous approaches on six of the games and surpasses a human expert on three of them.
Summary¶
Presents the first DQN result: a convolutional neural network trained by Q-learning directly from Atari pixels with a single architecture reused across games. Its importance is showing that deep networks can learn control policies from high-dimensional observations without hand-engineered state, outperforming prior methods on most tested games and setting the template for later deep RL systems.