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.
%0 Journal Article
%1 mnih2013playing
%A Mnih, Volodymyr
%A Kavukcuoglu, Koray
%A Silver, David
%A Graves, Alex
%A Antonoglou, Ioannis
%A Wierstra, Daan
%A Riedmiller, Martin
%D 2013
%J arXiv preprint arXiv:1312.5602
%K deeplearning neural_networks thema thema:deep_reinforcement_learning
%T Playing Atari with Deep Reinforcement Learning
%U http://arxiv.org/abs/1312.5602
%X 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.
@article{mnih2013playing,
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.},
added-at = {2016-09-28T18:46:44.000+0200},
author = {Mnih, Volodymyr and Kavukcuoglu, Koray and Silver, David and Graves, Alex and Antonoglou, Ioannis and Wierstra, Daan and Riedmiller, Martin},
biburl = {https://www.bibsonomy.org/bibtex/230bfa945486626ab74baed02ffdc9ec5/dallmann},
interhash = {78966703f649bae69a08a6a23a4e8879},
intrahash = {30bfa945486626ab74baed02ffdc9ec5},
journal = {arXiv preprint arXiv:1312.5602},
keywords = {deeplearning neural_networks thema thema:deep_reinforcement_learning},
timestamp = {2016-09-28T18:46:44.000+0200},
title = {Playing Atari with Deep Reinforcement Learning},
url = {http://arxiv.org/abs/1312.5602},
year = 2013
}