Research map / Optimization / Nonconvex and smooth optimization
Extra-gradient with player sampling for faster convergence in n-player 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 |
|---|---|---|---|
| cyclic player sampling (extrapolate one player, update another) instead of simultaneous full extra-gradient | WGAN-GP training on CIFAR10 (ResNet) · CIFAR10 WGAN-GP | FID (50k) and Inception score: FID 19.69 -> 17.10; IS 8.26 -> 8.38 | Helped |
| random player sampling instead of cyclic sampling (ablation) | WGAN-GP training on CIFAR10 · CIFAR10 WGAN-GP | Inception score: score 6.2 IS (random) vs ~8.4 (cyclic) | Hurt |
| doubly-stochastic extra-gradient (player sampling) vs full SEG for mixtures of GANs | mixtures of GANs · mixtures of GANs | FID / training speed | Helped |
| player sampling in noisy non-smooth convex games (theoretical rate with noise factor sqrt(b/n)) | doubly-stochastic extra-gradient · non-smooth convex games with noisy gradients | convergence rate bound | Helped |
| player sampling with variance reduction in smooth games (theory) | doubly-stochastic extra-gradient · smooth convex games | convergence rate bound | Mixed |
| player sampling in the noiseless case / Corollary 1 naive bounds | doubly-stochastic extra-gradient · convex games | convergence rate bound | No clear effect |
| cyclic player sampling instead of random sampling (1/5 players) (ablation) | 5-player random quadratic games · random convex quadratic games | Nash gap | Helped |
| player sampling with high noise in skew 50-player games (ablation) | doubly-stochastic extra-gradient · 50-player skew games | Nash gap | Mixed |
| player sampling in noiseless convex games (ablation) | doubly-stochastic extra-gradient · noiseless quadratic games | Nash gap | Hurt |
| cyclic sampling instead of random sampling (spectral radius of algorithm operator) (ablation) | two-player random matrix games · random 2-player quadratic games | spectral radius | Helped |
Closest papers by meaning
- A mean-field analysis of two-player zero-sum games (NeurIPS 2020 · similarity 0.90)
- Finding Mixed Nash Equilibria of Generative Adversarial Networks (ICML 2019 · similarity 0.90)
- Discretization Drift in Two-Player Games (ICML 2021 · similarity 0.89)
- Competing Against Nash Equilibria in Adversarially Changing Zero-Sum Games (ICML 2019 · similarity 0.88)
- From Average-Iterate to Last-Iterate Convergence in Games: A Reduction and Its Applications (NeurIPS 2025 · similarity 0.87)
- Learning Equilibria in Adversarial Team Markov Games: A Nonconvex-Hidden-Concave Min-Max Optimization Problem (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.