The deep reinforcement learning community has made several independent
improvements to the DQN algorithm. However, it is unclear which of these
extensions are complementary and can be fruitfully combined. This paper
examines six extensions to the DQN algorithm and empirically studies their
combination. Our experiments show that the combination provides
state-of-the-art performance on the Atari 2600 benchmark, both in terms of data
efficiency and final performance. We also provide results from a detailed
ablation study that shows the contribution of each component to overall
performance.
Description
[1710.02298] Rainbow: Combining Improvements in Deep Reinforcement Learning
%0 Journal Article
%1 hessel2017rainbow
%A Hessel, Matteo
%A Modayil, Joseph
%A van Hasselt, Hado
%A Schaul, Tom
%A Ostrovski, Georg
%A Dabney, Will
%A Horgan, Dan
%A Piot, Bilal
%A Azar, Mohammad
%A Silver, David
%D 2017
%K reinforcement-learning
%T Rainbow: Combining Improvements in Deep Reinforcement Learning
%U http://arxiv.org/abs/1710.02298
%X The deep reinforcement learning community has made several independent
improvements to the DQN algorithm. However, it is unclear which of these
extensions are complementary and can be fruitfully combined. This paper
examines six extensions to the DQN algorithm and empirically studies their
combination. Our experiments show that the combination provides
state-of-the-art performance on the Atari 2600 benchmark, both in terms of data
efficiency and final performance. We also provide results from a detailed
ablation study that shows the contribution of each component to overall
performance.
@article{hessel2017rainbow,
abstract = {The deep reinforcement learning community has made several independent
improvements to the DQN algorithm. However, it is unclear which of these
extensions are complementary and can be fruitfully combined. This paper
examines six extensions to the DQN algorithm and empirically studies their
combination. Our experiments show that the combination provides
state-of-the-art performance on the Atari 2600 benchmark, both in terms of data
efficiency and final performance. We also provide results from a detailed
ablation study that shows the contribution of each component to overall
performance.},
added-at = {2020-03-03T23:40:35.000+0100},
author = {Hessel, Matteo and Modayil, Joseph and van Hasselt, Hado and Schaul, Tom and Ostrovski, Georg and Dabney, Will and Horgan, Dan and Piot, Bilal and Azar, Mohammad and Silver, David},
biburl = {https://www.bibsonomy.org/bibtex/223e587e36693531fa1a5bd1f8d2a1bdd/kirk86},
description = {[1710.02298] Rainbow: Combining Improvements in Deep Reinforcement Learning},
interhash = {f4fb4d30fac6e6290d70a94e7420777a},
intrahash = {23e587e36693531fa1a5bd1f8d2a1bdd},
keywords = {reinforcement-learning},
note = {cite arxiv:1710.02298Comment: Under review as a conference paper at AAAI 2018},
timestamp = {2020-03-03T23:40:35.000+0100},
title = {Rainbow: Combining Improvements in Deep Reinforcement Learning},
url = {http://arxiv.org/abs/1710.02298},
year = 2017
}