The AlphaGo, AlphaGo Zero, and AlphaZero series of algorithms are a
remarkable demonstration of deep reinforcement learning's capabilities,
achieving superhuman performance in the complex game of Go with progressively
increasing autonomy. However, many obstacles remain in the understanding of and
usability of these promising approaches by the research community. Toward
elucidating unresolved mysteries and facilitating future research, we propose
ELF OpenGo, an open-source reimplementation of the AlphaZero algorithm. ELF
OpenGo is the first open-source Go AI to convincingly demonstrate superhuman
performance with a perfect (20:0) record against global top professionals. We
apply ELF OpenGo to conduct extensive ablation studies, and to identify and
analyze numerous interesting phenomena in both the model training and in the
gameplay inference procedures. Our code, models, selfplay datasets, and
auxiliary data are publicly available.
Description
[1902.04522] ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero
%0 Generic
%1 tian2019opengo
%A Tian, Yuandong
%A Ma, Jerry
%A Gong, Qucheng
%A Sengupta, Shubho
%A Chen, Zhuoyuan
%A Pinkerton, James
%A Zitnick, C. Lawrence
%D 2019
%K 2019 arxiv facebook games go paper reinforcement-learning
%T ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero
%U http://arxiv.org/abs/1902.04522
%X The AlphaGo, AlphaGo Zero, and AlphaZero series of algorithms are a
remarkable demonstration of deep reinforcement learning's capabilities,
achieving superhuman performance in the complex game of Go with progressively
increasing autonomy. However, many obstacles remain in the understanding of and
usability of these promising approaches by the research community. Toward
elucidating unresolved mysteries and facilitating future research, we propose
ELF OpenGo, an open-source reimplementation of the AlphaZero algorithm. ELF
OpenGo is the first open-source Go AI to convincingly demonstrate superhuman
performance with a perfect (20:0) record against global top professionals. We
apply ELF OpenGo to conduct extensive ablation studies, and to identify and
analyze numerous interesting phenomena in both the model training and in the
gameplay inference procedures. Our code, models, selfplay datasets, and
auxiliary data are publicly available.
@misc{tian2019opengo,
abstract = {The AlphaGo, AlphaGo Zero, and AlphaZero series of algorithms are a
remarkable demonstration of deep reinforcement learning's capabilities,
achieving superhuman performance in the complex game of Go with progressively
increasing autonomy. However, many obstacles remain in the understanding of and
usability of these promising approaches by the research community. Toward
elucidating unresolved mysteries and facilitating future research, we propose
ELF OpenGo, an open-source reimplementation of the AlphaZero algorithm. ELF
OpenGo is the first open-source Go AI to convincingly demonstrate superhuman
performance with a perfect (20:0) record against global top professionals. We
apply ELF OpenGo to conduct extensive ablation studies, and to identify and
analyze numerous interesting phenomena in both the model training and in the
gameplay inference procedures. Our code, models, selfplay datasets, and
auxiliary data are publicly available.},
added-at = {2019-02-13T12:31:07.000+0100},
author = {Tian, Yuandong and Ma, Jerry and Gong, Qucheng and Sengupta, Shubho and Chen, Zhuoyuan and Pinkerton, James and Zitnick, C. Lawrence},
biburl = {https://www.bibsonomy.org/bibtex/25709fb4bb7d14326fbb484cbeb4f726f/analyst},
description = {[1902.04522] ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero},
interhash = {bd8603037b367fd8222851e43b47f872},
intrahash = {5709fb4bb7d14326fbb484cbeb4f726f},
keywords = {2019 arxiv facebook games go paper reinforcement-learning},
note = {cite arxiv:1902.04522},
timestamp = {2019-02-13T12:31:07.000+0100},
title = {ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero},
url = {http://arxiv.org/abs/1902.04522},
year = 2019
}