Research map / Bandits and online learning / Online convex optimization and games
Near-Optimal Learning of Extensive-Form Games with Imperfect Information
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use balanced dilated KL and balanced exploration in OMD instead of the original dilated KL approach | OMD for two-player zero-sum IIEFGs · Two-player zero-sum IIEFGs, bandit feedback | sample complexity for ε-NE: \widetilde{\mathcal{O}}\left( \left( X^2A+Y^2B \right) /\varepsilon ^2 \right) -> \widetilde{\mathcal{O}}\left( \left( X | Helped |
| Integrate balanced exploration into both sampling and local Hedge regret minimization | MCCFR for two-player zero-sum IIEFGs · Two-player zero-sum IIEFGs, bandit feedback | sample complexity for ε-NE: \mc{O}(H^4(XA+YB) \iota/\eps^2) vs. \tO((X^2A+Y^2B)/\eps^2) | Helped |
| Replace Hedge with Regret Matching as the local regret minimizer (ablation) | Balanced CFR · Two-player zero-sum IIEFGs | sample complexity for ε-NE: \tO((H^4(XA^2+YB^2)/\eps^2) versus \mc{O}(H^4(XA+YB) \iota/\eps^2) | Hurt |
| Use balanced exploration policies in OMD self-play | Balanced OMD for multi-player general-sum IIEFGs · m-player general-sum IIEFGs | sample complexity for ε-NFCCE: \max_{i\in[m]}X_i^2A_i -> \max_{i\in[m]}X_iA_i | Helped |
Closest papers by meaning
- Learning Imperfect Information Extensive-form Games with Last-iterate Convergence under Bandit Feedback (ICML 2025 · similarity 0.94)
- Learning in two-player zero-sum partially observable Markov games with perfect recall (NeurIPS 2021 · similarity 0.92)
- Local and Adaptive Mirror Descents in Extensive-Form Games (NeurIPS 2024 · similarity 0.92)
- The Harder Path: Last Iterate Convergence for Uncoupled Learning in Zero-Sum Games with Bandit Feedback (ICML 2025 · similarity 0.90)
- A Policy-Gradient Approach to Solving Imperfect-Information Games with Best-Iterate Convergence (ICLR 2025 · similarity 0.90)
- Near-Optimal Last-Iterate Convergence for Zero-Sum Games with Bandit Feedback and Opponent Actions (NeurIPS 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.