Reproduce “Optimistic Hedge in General-Sum Games: Fundamental Limits and Regret Bounds via Temporal Approximation” 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
Yuval Dagan, Nadav Eisen, Liad Erez, Tomer Koren
Optimistic Hedge updates players’ action probabilities using recent losses as predictions of future losses. In full-information repeated games, the authors prove that every two-player game has regret at most log n/η + η per player when η ≤ c/n, at every horizon. For three or more players, they construct games with exponential worst-case regret and, for eligible rates, linear growth.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.