Research map / Optimization / Nonconvex and smooth optimization
On Gradient Descent Ascent for Nonconvex-Concave 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 |
|---|---|---|---|
| Use two-time-scale stepsizes (ηx ≪ ηy) in GDA instead of nested-loop GDmax-style inner ascent | WRM training with GDA vs GDmA · MNIST, Fashion-MNIST, CIFAR-10 (γ=0.4, 1.3) | test accuracy vs time | Helped |
| Two-time-scale GDA with ηx = O(1/κ²ℓ), ηy = O(1/ℓ) in nonconvex-strongly-concave setting | Two-time-scale GDA · Nonconvex-strongly-concave | gradient complexity to ε-stationary point | Helped |
| Two-time-scale GDA in nonconvex-concave setting (single loop, ηy/ηx = Θ(1/ϵ⁴)) | Two-time-scale GDA · Nonconvex-concave | gradient complexity to ε-stationary point (Moreau envelope) | Mixed |
| Minibatch size M = Θ(ϵ⁻²) in SGDA versus purely stochastic M=1 in nonconvex-strongly-concave setting (ablation) | Two-time-scale SGDA · Nonconvex-strongly-concave | stochastic gradient complexity: O(κ3ϵ−4) vs O(κ3ϵ−5) for M=1 | Helped |
| Use batch size M=1 in SGDA for nonconvex-concave setting | Two-time-scale SGDA · Nonconvex-concave | stochastic gradient complexity | No clear effect |
Closest papers by meaning
- Global Convergence and Variance Reduction for a Class of Nonconvex-Nonconcave Minimax Problems (NeurIPS 2020 · similarity 0.94)
- A Single-Loop Smoothed Gradient Descent-Ascent Algorithm for Nonconvex-Concave Min-Max Problems (NeurIPS 2020 · similarity 0.94)
- Universal Gradient Descent Ascent Method for Nonconvex-Nonconcave Minimax Optimization (NeurIPS 2023 · similarity 0.93)
- Tight Analysis of Extra-gradient and Optimistic Gradient Methods For Nonconvex Minimax Problems (NeurIPS 2022 · similarity 0.93)
- Fundamental Benefit of Alternating Updates in Minimax Optimization (ICML 2024 · similarity 0.91)
- High-probability complexity bounds for stochastic non-convex minimax optimization (NeurIPS 2024 · similarity 0.90)
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.