Research map / Optimization / Nonconvex and smooth optimization
Train simultaneously, generalize better: Stability of gradient-based minimax learners
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 PPM instead of GDA | Bilinear convex-concave minimax learning · Bilinear convex-concave Gaussian | generalization risk: 6.2; below 0.5 | Helped |
| Use simultaneous 1,1 Adam updates instead of 1,100 updates | SN-GAN · CIFAR-10 | generalization gap | Helped |
| Use simultaneous 1,1 Adam updates instead of 1,100 updates | SN-GAN · CelebA | generalization gap | Helped |
| Use full-batch instead of stochastic updates | GDA · Strongly-convex strongly-concave Gaussian | generalization risk: below an amplitude of 0.2; vanishing | Helped |
| Use full-batch instead of stochastic updates | GDA · Bilinear convex-concave Gaussian | generalization risk: 6.2; 7.8 | Hurt |
| Use full-batch instead of stochastic updates | PPM · Convex-concave Gaussian | generalization risk | Helped |
| Fully optimize the maximization player at each minimization step | SN-GAN · CIFAR-10 and CelebA | generalization gap | Hurt |
Closest papers by meaning
- Stability and Generalization of Stochastic Gradient Methods for Minimax Problems (ICML 2021 · similarity 0.94)
- Stability and Generalization of Learning Algorithms that Converge to Global Optima (ICML 2018 · similarity 0.91)
- GDA-AM: ON THE EFFECTIVENESS OF SOLVING MIN-IMAX OPTIMIZATION VIA ANDERSON MIXING (ICLR 2022 · similarity 0.90)
- On Convergence of Gradient Descent Ascent: A Tight Local Analysis (ICML 2022 · similarity 0.90)
- Improving Generalization and Stability of Generative Adversarial Networks (ICLR 2019 · similarity 0.90)
- Stability and Generalization of Adversarial Training for Shallow Neural Networks with Smooth Activation (NeurIPS 2024 · similarity 0.88)
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.