Playing Atari with Deep Reinforcement Learning

TL;DR

This paper introduces a deep convolutional neural network trained with reinforcement learning to play Atari games directly from raw pixel input, achieving superhuman performance on several titles.

Contribution

It is the first to successfully learn control policies directly from high-dimensional sensory input using deep reinforcement learning without task-specific adjustments.

Findings

Outperforms previous methods on six Atari games

Surpasses human expert performance on three games

Demonstrates effectiveness of deep learning in 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.