Research map / Optimization / Nonconvex and smooth optimization
Gradient Descent Ascent for Minimax Problems on Riemannian Manifolds
Feihu Huang, Shangqian Gao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 Riemannian retraction-based update (RSGDA) instead of SGDA with retraction appended | SGDA · MNIST PGD40 eps=0.4 | test accuracy (%): 61.95 -> 64.92 | Improved |
| add STORM momentum variance reduction on the manifold (Acc-RSGDA over RSGDA) | RSGDA · MNIST PGD40 eps=0.4 | test accuracy (%): 64.92 -> 73.66 | Improved |
| add STORM momentum variance reduction on the manifold (Acc-RSGDA over RSGDA) | RSGDA · FashionMNIST PGD40 eps=0.2 | test accuracy (%): 63.48 -> 64.35 | Improved |
| use Acc-RSGDA instead of Acc-MDA | Acc-MDA · CIFAR10 PGD40 eps=0.02 | test accuracy (%): 25.76 -> 29.16 | Improved |
| use Acc-RSGDA instead of Acc-MDA | Acc-MDA · STL10 PGD40 eps=0.02 | test accuracy (%): 16.04 -> 16.76 | Improved |
| use Acc-RSGDA instead of Acc-MDA | Acc-MDA · CIFAR100 (first 10 classes) PGD40 eps=0.02 | test accuracy (%): 31.90 -> 33.50 | Improved |
| add momentum without Riemannian structure (Acc-MDA over SGDA) | SGDA · MNIST PGD40 eps=0.4 | test accuracy (%): 61.95 -> 58.06 | Hurt |
| add momentum without Riemannian structure (Acc-MDA over SGDA) | SGDA · FashionMNIST PGD40 eps=0.2 | test accuracy (%): 62.55 -> 60.76 | Hurt |
| use Acc-RSGDA instead of Acc-MDA | Acc-MDA · CIFAR100 (first 10 classes) natural images | test accuracy (%): 70.70 -> 70.10 | Hurt |
| use Acc-RSGDA instead of Acc-MDA | Acc-MDA · MNIST natural images | test accuracy (%): 99.23 -> 99.37 | Improved |
| use Acc-RSGDA instead of RCG | RCG · CIFAR10 FGSM eps=0.02 | test accuracy (%): 31.11 -> 35.94 | Improved |
| use Acc-RSGDA instead of Acc-MDA on distributionally robust optimization | Acc-MDA · CIFAR-10 / STL-10 DRO | test accuracy (%) | Mixed |
Closest papers by meaning
- Learning-Rate-Free Stochastic Optimization over Riemannian Manifolds (ICML 2024 · similarity 0.89)
- Riemannian coordinate descent algorithms on matrix manifolds (ICML 2024 · similarity 0.89)
- Convergence and Complexity Guarantee for Inexact First-order Riemannian Optimization Algorithms (ICML 2024 · similarity 0.89)
- SERENA: A Unified Stochastic Recursive Variance Reduced Gradient Framework for Riemannian Non-Convex Optimization (ICML 2025 · similarity 0.89)
- Implicit Riemannian Optimism with Applications to Min-Max Problems (ICML 2025 · similarity 0.88)
- Convergence and Trade-Offs in Riemannian Gradient Descent and Riemannian Proximal Point (ICML 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.
