Superhuman AI for heads-up no-limit poker: Libratus beats top professionals

被引:326
作者
Brown, Noam [1 ]
Sandholm, Tuomas [1 ]
机构
[1] Carnegie Mellon Univ, Comp Sci Dept, 5000 Forbes Ave, Pittsburgh, PA 15213 USA
基金
美国安德鲁·梅隆基金会; 美国国家科学基金会;
关键词
D O I
10.1126/science.aao1733
中图分类号
O [数理科学和化学]; P [天文学、地球科学]; Q [生物科学]; N [自然科学总论];
学科分类号
07 ; 0710 ; 09 ;
摘要
No-limit Texas hold'em is the most popular form of poker. Despite artificial intelligence (AI) successes in perfect-information games, the private information and massive game tree have made no-limit poker difficult to tackle. We present Libratus, an AI that, in a 120,000-hand competition, defeated four top human specialist professionals in heads-up no-limit Texas hold'em, the leading benchmark and long-standing challenge problem in imperfect-information game solving. Our game-theoretic approach features application-independent techniques: an algorithm for computing a blueprint for the overall strategy, an algorithm that fleshes out the details of the strategy for subgames that are reached during play, and a self-improver algorithm that fixes potential weaknesses that opponents have identified in the blueprint strategy.
引用
收藏
页码:418 / +
页数:7
相关论文
共 44 条
  • [1] [Anonymous], 1950, Non-cooperative games
  • [2] [Anonymous], 2015, ACM Transactions on Economics and Computation (TEAC)
  • [3] [Anonymous], 2016, AAAI C ART INT AAAI
  • [4] [Anonymous], 1997, One Jump Ahead: Challenging Human Supremacy in Checkers
  • [5] [Anonymous], 2011, Proceedings of the Twenty-Second International Joint Conference on Artificial Intelligence IJCAI, DOI DOI 10.5591/978-1-57735-516-8/IJCAI11-054
  • [6] Billings D, 1998, FIFTEENTH NATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI-98) AND TENTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICAL INTELLIGENCE (IAAI-98) - PROCEEDINGS, P493
  • [7] Billings D., 2003, IJCAI: Proceedings of the 18th International Joint Conference on Artificial Intelligence, P661
  • [8] Heads-up limit hold'em poker is solved
    Bowling, Michael
    Burch, Neil
    Johanson, Michael
    Tammelin, Oskari
    [J]. SCIENCE, 2015, 347 (6218) : 145 - 149
  • [9] Brown N., 2017, ADV NEURAL INF PROCE, V30, P689
  • [10] Brown N., 2015, AAAI C ART INT AAAI, P4270