AI Paper Atlas

Research map / Optimization / Nonconvex and smooth optimization

Gradient Descent Ascent for Minimax Problems on Riemannian Manifolds

Feihu Huang, Shangqian Gao

TPAMI 2023 · Optimization › Nonconvex and smooth optimization · approach: Saddle-point escape algorithms · 31 citations (OpenAlex)

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 changedWhereMeasuredEffect
use Riemannian retraction-based update (RSGDA) instead of SGDA with retraction appendedSGDA · MNIST PGD40 eps=0.4test accuracy (%): 61.95 -> 64.92Improved
add STORM momentum variance reduction on the manifold (Acc-RSGDA over RSGDA)RSGDA · MNIST PGD40 eps=0.4test accuracy (%): 64.92 -> 73.66Improved
add STORM momentum variance reduction on the manifold (Acc-RSGDA over RSGDA)RSGDA · FashionMNIST PGD40 eps=0.2test accuracy (%): 63.48 -> 64.35Improved
use Acc-RSGDA instead of Acc-MDAAcc-MDA · CIFAR10 PGD40 eps=0.02test accuracy (%): 25.76 -> 29.16Improved
use Acc-RSGDA instead of Acc-MDAAcc-MDA · STL10 PGD40 eps=0.02test accuracy (%): 16.04 -> 16.76Improved
use Acc-RSGDA instead of Acc-MDAAcc-MDA · CIFAR100 (first 10 classes) PGD40 eps=0.02test accuracy (%): 31.90 -> 33.50Improved
add momentum without Riemannian structure (Acc-MDA over SGDA)SGDA · MNIST PGD40 eps=0.4test accuracy (%): 61.95 -> 58.06Hurt
add momentum without Riemannian structure (Acc-MDA over SGDA)SGDA · FashionMNIST PGD40 eps=0.2test accuracy (%): 62.55 -> 60.76Hurt
use Acc-RSGDA instead of Acc-MDAAcc-MDA · CIFAR100 (first 10 classes) natural imagestest accuracy (%): 70.70 -> 70.10Hurt
use Acc-RSGDA instead of Acc-MDAAcc-MDA · MNIST natural imagestest accuracy (%): 99.23 -> 99.37Improved
use Acc-RSGDA instead of RCGRCG · CIFAR10 FGSM eps=0.02test accuracy (%): 31.11 -> 35.94Improved
use Acc-RSGDA instead of Acc-MDA on distributionally robust optimizationAcc-MDA · CIFAR-10 / STL-10 DROtest accuracy (%)Mixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy