Research map / Optimization / Implicit bias and generalization
Subgradient Descent Learns Orthogonal Dictionaries
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 GRANSO instead of Riemannian subgradient descent | Orthogonal dictionary learning · Simulated orthogonal dictionary, n=400, m=10n² | CPU time to identify a basis: approximately 10× faster | Helped |
| Use Riemannian subgradient descent with random initialization | Orthogonal dictionary learning · Simulated data across different sparsity levels | dictionary recovery success | Helped |
| Learn a complete dictionary from preconditioned image patches | Natural images · Natural image 1, 8×8 blocks | mean coefficient sparsity (L1/L2): 5.9135 | Helped |
| Learn a complete dictionary from preconditioned image patches | Natural images · Natural image 2, 8×8 blocks | mean coefficient sparsity (L1/L2): 6.4339 | Helped |
| Increase the sparsity parameter θ | Riemannian subgradient descent for dictionary recovery · Simulated data | sample threshold for successful recovery | Hurt |
Closest papers by meaning
- Efficient Dictionary Learning with Gradient Descent (ICML 2019 · similarity 0.91)
- NOODL: Provable Online Dictionary Learning and Sparse Coding (ICLR 2019 · similarity 0.88)
- Stacey: Promoting Stochastic Steepest Descent via Accelerated $\ell_p$-Smooth Nonconvex Optimization (ICML 2025 · similarity 0.85)
- Generalized Gradient Norm Clipping & Non-Euclidean $(L_0,L_1)$-Smoothness (NeurIPS 2025 · similarity 0.85)
- On the Sparsity-Storage-Accuracy Tradeoff in Parsimoniously Activated Dictionary Learning (NeurIPS 2026 · similarity 0.85)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · 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.