Research map / Optimization / Nonconvex and smooth optimization
Infeasible Deterministic, Stochastic, and Variance-Reduction Algorithms for Optimization under Orthogonality Constraints
Pierre Ablin, Simon Vary, Bin Gao, Pierre-Antoine Absil
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 landing instead of Riemannian gradient descent with QR retraction | online PCA SGD · Online PCA (p=200, p=1000) | objective vs time: N(X) <= 10^-6 | Improved |
| use landing instead of l2 penalty with small lambda=10^2 | Euclidean SGD · Online PCA | distance to constraint: lambda=10^2 | Improved |
| use landing instead of l2 penalty with large lambda=10^4 | Euclidean SGD · Online PCA | convergence rate: lambda=10^4 | Improved |
| add variance reduction (SAGA) to landing SGD | Landing · ICA (n=10, N=10^4) | loss plateau | Improved |
| use landing instead of Riemannian methods | ICA · ICA (n=10, N=10^4) | speed | Improved |
| use landing instead of l2 penalty lambda=1 | VGG16/Resnet18 CIFAR-10 · orthogonal conv kernels | distance to constraint: 10^{-3} -> 10^{-8} | Improved |
| use landing instead of l2 penalty lambda=10^3 | VGG16/Resnet18 CIFAR-10 · orthogonal conv kernels | test accuracy | Improved |
| use landing instead of Riemannian GD with QR retraction | VGG16/Resnet18 CIFAR-10 · orthogonal conv kernels | time to ~90% test accuracy | Improved |
| use landing instead of GeoTorch trivialization | VGG16 CIFAR-10 · orthogonal conv kernels | training time: approximately 14 times as long | Improved |
Closest papers by meaning
- Distributed Retraction-Free and Communication-Efficient Optimization on the Stiefel Manifold (ICML 2025 · similarity 0.89)
- Efficient Optimization with Orthogonality Constraint: a Randomized Riemannian Submanifold Method (ICML 2025 · similarity 0.89)
- Optimization without Retraction on the Random Generalized Stiefel Manifold (ICML 2024 · similarity 0.88)
- An Embarrassingly Simple Way to Optimize Orthogonal Matrices at Scale (ICML 2026 · similarity 0.87)
- Riemannian coordinate descent algorithms on matrix manifolds (ICML 2024 · similarity 0.86)
- SERENA: A Unified Stochastic Recursive Variance Reduced Gradient Framework for Riemannian Non-Convex Optimization (ICML 2025 · similarity 0.85)
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.
