Reproduce “On the Convergence of Alternating Regret Matching Algorithms” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Anh Duc Nguyen, Ioannis Anagnostides, Brian Hu Zhang, Tuomas Sandholm
The paper analyzes alternating regret-matching updates in finite two-player zero-sum normal-form games. It halves the prediction in predictive regret matching+ and introduces an analysis bound for alternating play. Uniformly averaged strategies achieve an O(T^-1) duality-gap rate; the authors also prove a best-iterate O(T^-1/2) rate and convergence to a single Nash equilibrium. These are theoretical guarantees, not established for extensive-form benchmarks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.