Research map / Optimization / Nonconvex and smooth optimization
Beating Price of Anarchy and Gradient Descent without Regret in Potential Games
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Studies equilibrium selection in potential games via a class of continuous-time no-regret dynamics called Q-Replicator Dynamics (QRD), which includes gradient descent, log-barrier, and replicator dynamics.
Problem
Equilibrium selection in potential games: whether self-interested no-regret learning dynamics converge to socially optimal equilibria when multiple equilibria exist.
Approach
Introduces Q-Replicator Dynamics, a rich class of continuous-time no-regret dynamics for potential games that contains gradient descent, log-barrier, and replicator dynamics as special cases, and analyzes which equilibria they select.
Experimental setup
Builds on: gradient descent replicator dynamics log-barrier dynamics
Closest papers by meaning
- Convergence of No-Swap-Regret Dynamics in Self-Play (NeurIPS 2024 · similarity 0.91)
- On Last-Iterate Convergence Beyond Zero-Sum Games (ICML 2022 · similarity 0.91)
- Paradoxes of Game Theoretic Equilibria and Price of Anarchy (NeurIPS 2026 · similarity 0.91)
- No-Regret Learning and Mixed Nash Equilibria: They Do Not Mix (NeurIPS 2020 · similarity 0.91)
- No-regret Learning in Harmonic Games: Extrapolation in the Face of Conflicting Interests (NeurIPS 2024 · similarity 0.89)
- No Coin Left Behind: Maximizing Strategic Surplus Against No-Regret Dynamics (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.