Research map / Learning theory and algorithms / Matrix and tensor methods
Simplifying Momentum-based Positive-definite Submanifold Optimization with Applications to Deep Learning
Wu Lin, Valentin Duruisseaux, Melvin Leok, Frank Nielsen, Mohammad Emtiyaz Khan, Mark Schmidt
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 the proposed optimizer instead of SGD | VGG-16 · CIFAR-100 | test error rate: 26.45 -> 26.06 | Improved |
| Use the proposed optimizer instead of SGD | PyramidNet-65 · CIFAR-100 | test error rate: 25.65 -> 25.39 | Improved |
| Use the proposed optimizer instead of KFAC | ConvMixer256-12 · CIFAR-100 | test error rate: 26.14 -> 25.79 | Improved |
| Use the proposed optimizer instead of SGD | RegNetX-1.6GF · CIFAR-100 | test error rate: 21.62 -> 21.89 | Hurt |
| Use the proposed optimizer instead of SGD | VGG-16 · TinyImageNet-200 | test error rate: 40.18 -> 40.01 | Improved |
| Use the proposed optimizer instead of KFAC | PyramidNet-65 · TinyImageNet-200 | test error rate: 41.23 -> 40.82 | Improved |
| Use the proposed optimizer instead of SGD | ConvMixer256-12 · TinyImageNet-200 | test error rate: 40.03 -> 39.09 | Improved |
| Use the proposed optimizer instead of SGD | RegNetX-1.6GF · TinyImageNet-200 | test error rate: 34.38 -> 34.85 | Hurt |
| Use the proposed optimizer instead of SGD | ConvMixer180-16 · ImageNet-100 | test error rate: 20.37 -> 18.84 | Improved |
| Use the proposed optimizer instead of KFAC | RegNetZ-500MF · ImageNet-100 | test error rate: 20.07 -> 18.30 | Improved |
| Use the proposed optimizer instead of SGD | ReXNetr-100 · ImageNet-100 | test error rate: 17.07 -> 16.73 | Improved |
| Use the proposed optimizer instead of SGD | RepVGG-B1G4 · ImageNet-100 | test error rate: 16.15 -> 15.81 | Improved |
Closest papers by meaning
- Efficient Optimization with Orthogonality Constraint: a Randomized Riemannian Submanifold Method (ICML 2025 · similarity 0.90)
- Riemannian coordinate descent algorithms on matrix manifolds (ICML 2024 · similarity 0.89)
- Intrinsic Muon: Spectral Optimization on Riemannian Matrix Manifolds (NeurIPS 2026 · similarity 0.87)
- Adaptive Riemannian ADMM for Nonsmooth Optimization: Optimal Complexity without Smoothing (NeurIPS 2025 · similarity 0.87)
- Decentralized Projected Riemannian Stochastic Recursive Momentum Method for Nonconvex Optimization (AAAI 2025 · similarity 0.87)
- Learning-Rate-Free Stochastic Optimization over Riemannian Manifolds (ICML 2024 · similarity 0.87)
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.
