Research map / Multi-agent systems and games / Games and equilibria
Policy Mirror Ascent for Efficient and Independent Learning in Mean Field Games
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 policy mirror ascent updates with TD value estimates instead of recomputing a best response at each iteration | Stationary MFG learning · N-player SAGS | sample complexity: \widetilde{\mathcal{O}}(\varepsilon^{-2}) compared with \mathcal{O}(\varepsilon^{-4}) and \mathcal{O}(\varepsilon^{-4|\s | Helped |
| Learn from a single N-agent sample path without population manipulation or a generative model | Centralized MFG-PMA · N-player SAGS | sample complexity: \widetilde{\mathcal{O}}(\varepsilon^{-2}) | Helped |
| Allow each agent to learn a separate policy from its own observations | Independent MFG-PMA · N-player SAGS | sample complexity: \widetilde{\mathcal{O}}(\varepsilon^{-2-a}); a is 0 for smooth enough problems | Mixed |
| Require sufficient regularization so the policy mirror ascent operator is contractive | Policy mirror ascent · Stationary MFG | convergence: L_{\Gamma_\eta} < 1 | Helped |
Closest papers by meaning
- Local and Adaptive Mirror Descents in Extensive-Form Games (NeurIPS 2024 · similarity 0.88)
- Adaptively Perturbed Mirror Descent for Learning in Games (ICML 2024 · similarity 0.87)
- Model-Based RL for Mean-Field Games is not Statistically Harder than Single-Agent RL (ICML 2024 · similarity 0.87)
- Last Iterate Convergence in Monotone Mean Field Games (NeurIPS 2025 · similarity 0.87)
- Learning mirror maps in policy mirror descent (ICLR 2025 · similarity 0.87)
- Bounded Rationality Equilibrium Learning in Mean Field Games (AAAI 2025 · similarity 0.87)
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.