Research map / Optimization / Optimizers
Trivializations for Gradient-Based Optimization on Manifolds
Mario Lezcano Casado
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 dynamic rather than fixed-basis Lie trivialization | orthogonal RNN · MNIST and P-MNIST | test accuracy: +0.4% out of the 1.3% possible in the largest size | Improved |
| Use DTriv 100 instead of the static ExpRNN trivialization | orthogonal RNN · TIMIT | test MSE: 5.30 -> 4.77 | Improved |
| Use DTriv ∞ instead of the static ExpRNN trivialization | orthogonal RNN · TIMIT | test MSE: 5.30 -> 4.71 | Improved |
| Change the basis every 1 step rather than every 100 steps (ablation) | dynamic trivialization · TIMIT (n=224) | test MSE: 4.80 -> 6.55 | Hurt |
| Change the basis every 1 step rather than every 100 steps (ablation) | dynamic trivialization · TIMIT (n=322) | test MSE: 3.80 -> 4.56 | Hurt |
| Change the basis every 1 step rather than every 100 steps (ablation) | dynamic trivialization · TIMIT (n=425) | test MSE: 2.02 -> 4.21 | Hurt |
| Change the basis every 100 steps rather than never changing it (ablation) | dynamic trivialization · TIMIT (n=224) | test MSE: 4.71 -> 4.77 | Mixed |
| Change the basis every 100 steps rather than never changing it (ablation) | dynamic trivialization · TIMIT (n=425) | test MSE: 1.97 -> 1.99 | Hurt |
Closest papers by meaning
- Riemannian coordinate descent algorithms on matrix manifolds (ICML 2024 · similarity 0.88)
- Riemannian adaptive stochastic gradient algorithms on matrix manifolds (ICML 2019 · similarity 0.88)
- Acceleration via silver step-size on Riemannian manifolds with applications to Wasserstein space (NeurIPS 2025 · similarity 0.87)
- Efficient Optimization with Orthogonality Constraint: a Randomized Riemannian Submanifold Method (ICML 2025 · similarity 0.87)
- Learning-Rate-Free Stochastic Optimization over Riemannian Manifolds (ICML 2024 · similarity 0.86)
- Convergence and Trade-Offs in Riemannian Gradient Descent and Riemannian Proximal Point (ICML 2024 · similarity 0.86)
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.
