Research map / Bandits and online learning / Online convex optimization and games
Deep Counterfactual Regret Minimization
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 Deep CFR instead of NFSP | imperfect-information poker solving · FHP | exploitability (mbb/g): 37 mbb/g vs 47 mbb/g | Helped |
| Use Deep CFR instead of NFSP | poker strategy · HULH vs NFSP | expected value (mbb/g): +43 ± 2 mbb/g | Helped |
| Use Deep CFR instead of a domain-specific abstraction | poker strategy · HULH vs 3.3 · 10^6-bucket abstraction | expected value (mbb/g): +6 ± 2 mbb/g | Helped |
| Use Deep CFR instead of a domain-specific abstraction | poker strategy · HULH vs 3.3 · 10^7-bucket abstraction | expected value (mbb/g): -6 ± 2 mbb/g | Mixed |
| Use Deep CFR instead of a domain-specific abstraction | poker strategy · HULH vs 3.3 · 10^8-bucket abstraction | expected value (mbb/g): -11 ± 2 mbb/g | Mixed |
| Retrain the regret model from scratch at each CFR iteration (ablation) | Deep CFR · FHP | exploitability | Helped |
| Replace reservoir sampling with a sliding-window memory (ablation) | Deep CFR · FHP | exploitability | Hurt |
| Reduce the number of game traversals per iteration (ablation) | Deep CFR · FHP | exploitability | Mixed |
| Reduce the number of SGD steps per iteration (ablation) | Deep CFR · FHP | exploitability | Hurt |
| Increase the model size (ablation) | Deep CFR · FHP | exploitability | Helped |
| Use Linear CFR weighting (ablation) | Deep CFR · FHP | convergence speed | Helped |
| Choose the highest-regret action when all positive regrets are zero, instead of choosing uniformly (ablation) | regret matching in Deep CFR · FHP | convergence | Helped |
Closest papers by meaning
- Double Neural Counterfactual Regret Minimization (ICLR 2020 · similarity 0.87)
- AutoCFR: Learning to Design Counterfactual Regret Minimization Algorithms (AAAI 2022 · similarity 0.87)
- Solving Imperfect-Information Games via Discounted Regret Minimization (AAAI 2019 · similarity 0.87)
- Dynamic Discounted Counterfactual Regret Minimization (ICLR 2024 · similarity 0.86)
- Deep (Predictive) Discounted Counterfactual Regret Minimization (AAAI 2026 · similarity 0.86)
- Accelerating Nash Equilibrium Convergence in Monte Carlo Settings Through Counterfactual Value Based Fictitious Play (NeurIPS 2024 · similarity 0.84)
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.