Research map / Optimization / Nonconvex and smooth optimization
A mean-field analysis of two-player zero-sum 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 |
|---|---|---|---|
| Include transport updates of particle positions | WFR-DA and L-DA · Synthetic polynomial games on spheres | Nikaido-Isoda error | Helped |
| Use WFR-DA instead of mirror descent | Synthetic polynomial games on spheres · High-dimensional synthetic games | Nikaido-Isoda error | Helped |
| Add dynamic mixture-weight updates to parameter updates (ablation) | WFR-DA · CIFAR10 | FID | Helped |
| Train a mixture rather than a single generator (ablation) | WFR-DA · 2D Gaussian mixture, under-parameterized single-layer MLP | log likelihood | Helped |
| Train a mixture rather than a single generator (ablation) | WFR-DA · 2D Gaussian mixture, over-parameterized 4-layer MLP | log likelihood | No clear effect |
| Increase the relative importance of transport compared with weight updates (ablation) | WFR-DA · Synthetic games | Nikaido-Isoda error | Helped |
| Use more particles, from n=50 to n=100 (ablation) | WFR-DA and L-DA · Synthetic polynomial games on spheres | Nikaido-Isoda error | No clear effect |
| Use enough generators and discriminators (G > 5, D > 2) (ablation) | WFR-DA GAN mixtures · CIFAR10 and MNIST | FID | Helped |
Closest papers by meaning
- Provably convergent quasistatic dynamics for mean-field two-player zero-sum games (ICLR 2022 · similarity 0.93)
- Two-Scale Gradient Descent Ascent Dynamics Finds Mixed Nash Equilibria of Continuous Games: A Mean-Field Perspective (ICML 2023 · similarity 0.92)
- Implicit Learning Dynamics in Stackelberg Games: Equilibria Characterization, Convergence Analysis, and Empirical Study (ICML 2020 · similarity 0.90)
- Extra-gradient with player sampling for faster convergence in n-player games (ICML 2020 · similarity 0.90)
- Towards convergence to Nash equilibria in two-team zero-sum games (ICLR 2023 · similarity 0.90)
- Fast Last-Iterate Convergence of Learning in Games Requires Forgetful Algorithms (NeurIPS 2024 · 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.