Efficient Predictive Counterfactual Regret Minimization$^+$ Algorithm in Solving Extensive-Form Games

ICLR 2025 Conference Submission10720 Authors

27 Sept 2024 (modified: 13 Oct 2024)ICLR 2025 Conference SubmissionEveryoneRevisionsBibTeXCC BY 4.0
Keywords: Imperfect-Information Extensive-Form Games, Nash Equilibrium, Counterfactual Regret Minimization
TL;DR: We propose P2PCFR$^+$, a novel and effective varaint of PCFR$^+$, achiving faster empirical convergence rate than existing PCFR$^+$ variants.
Abstract: Imperfect-information extensive-form games (IIGs) serve as a foundational model for capturing interactions among multiple agents in sequential settings with hidden information. A common objective of IIGs is to calculate a Nash equilibrium (NE). Counterfactual Regret Minimization (CFR) algorithms have been widely developed to learn an NE in two-player zero-sum IIGs. Among CFR algorithms, Predictive CFR$^+$ (PCFR$^+$) is powerful, usually achieving an extremely fast empirical convergence rate. However, PCFR$^+$ suffers from the significant discrepancy between strategies represented by explicit accumulated counterfactual regrets across two consecutive iterations, which decreases the empirical convergence rate of PCFR$^+$ in practice. To mitigate this significant discrepancy, we introduce a novel and effective variant of PCFR$^+$, termed Pessimistic PCFR$^+$ (P2PCFR$^+$), minimizing the discrepancy between strategies represented by implicit and explicit accumulated regrets within the same iteration. We provide theoretical proof to show that P2PCFR$^+$ exhibits a faster theoretical convergence rate than PCFR$^+$. Experimental results demonstrate that P2PCFR$^+$ outperforms other tested CFR variants.
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 10720
Loading