Research map / Reinforcement learning / Policy gradient and actor-critic
Independent Policy Gradient Methods for Competitive Reinforcement Learning
Constantinos Daskalakis, Dylan J Foster, Noah Golowich
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 two-timescale learning rates, with one player updating much more slowly | Independent policy gradient in two-player zero-sum stochastic games · Two-player zero-sum stochastic games | average min-player Nash gap | Improved |
| Use equal learning rates instead of two-timescale updates | Independent gradient descent-ascent · Zero-sum matrix games | distance of individual iterates from equilibrium | Hurt |
| Use ε-greedy exploration with direct policy parameterization | Independent policy gradient · Two-player zero-sum stochastic games | REINFORCE gradient-estimator variance: 24\frac{A^{2}}{\varepsilon_x\zeta^{4}} | Improved |
| Apply extragradient updates | Ratio games and randomly generated multi-state stochastic games · Stochastic games | last-iterate convergence | Mixed |
Closest papers by meaning
- Independent Policy Gradient for Large-Scale Markov Potential Games: Sharper Rates, Function Approximation, and Game-Agnostic Convergence (ICML 2022 · similarity 0.93)
- Faster Last-iterate Convergence of Policy Optimization in Zero-Sum Markov Games (ICLR 2023 · similarity 0.91)
- Optimistic Policy Gradient in Multi-Player Markov Games with a Single Controller: Convergence beyond the Minty Property (AAAI 2024 · similarity 0.91)
- Provably Fast Convergence of Independent Natural Policy Gradient for Markov Potential Games (NeurIPS 2023 · similarity 0.91)
- A Policy-Gradient Approach to Solving Imperfect-Information Games with Best-Iterate Convergence (ICLR 2025 · similarity 0.89)
- Independent Learning of Nash Equilibria in Partially Observable Markov Potential Games with Decoupled 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.
