Research map / Optimization / Nonconvex and smooth optimization
Stability and Generalization of Stochastic Gradient Methods for Minimax Problems
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 |
|---|---|---|---|
| Increase the step size | SOLAM (SGDA) · svmguide3 and w5a | distance between models trained on neighboring datasets | Hurt |
| Continue training for more passes | SOLAM (SGDA) · svmguide3 and w5a | distance between models trained on neighboring datasets | Hurt |
| Use SGDA with early stopping and an appropriate iteration count and step size | Convex-concave minimax optimization · Convex-concave problems | weak PD population risk: O(1/\sqrt{n}) | Helped |
| Use SGDA without requiring smoothness | Strongly-convex-strongly-concave minimax optimization · Strongly-convex-strongly-concave problems | weak PD population risk: O(\sqrt{\log n}/(n\rho)) | Helped |
| Allow weak-convexity-weak-concavity to decay along optimization | SGDA · Decaying weak-convexity-weak-concavity problems | weak PD generalization: O\big(1/\sqrt{n}+\sqrt{T}/n\big) | Helped |
Closest papers by meaning
- Train simultaneously, generalize better: Stability of gradient-based minimax learners (ICML 2021 · similarity 0.94)
- Train faster, generalize better: Stability of stochastic gradient descent (ICML 2016 · similarity 0.94)
- Generalization Analysis of Biased Stochastic Gradient Methods for Minimax Problems (NeurIPS 2026 · similarity 0.93)
- High-probability complexity bounds for stochastic non-convex minimax optimization (NeurIPS 2024 · similarity 0.90)
- Error Analysis Affected by Heavy-Tailed Gradients for Non-Convex Pairwise Stochastic Gradient Descent (AAAI 2025 · similarity 0.90)
- A Unified Stability Analysis of SAM vs SGD: Role of Data Coherence and Emergence of Simplicity Bias (NeurIPS 2025 · 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.