Research map / Learning theory and algorithms / Matrix and tensor methods
Dictionary Learning Based on Sparse Distribution Tomography
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 SparsDT (distribution tomography with alpha-stable model) instead of l2/l1, l1/l2, l1+l2 sparse coding dictionary learning | dictionary learning on synthetic SαS signals · Synthetic SαS, m=16, n=24, α=1.2, K=500 | % dictionaries found: 100 vs 50/75/94 | Helped |
| Use SparsDT instead of baselines for image denoising dictionary | dictionary learning for image denoising · AT&T faces, Gaussian noise σ=10 | PSNR (dB): +0.6 dB or more over baselines | Helped |
| Use SparsDT instead of baselines for missing pixel recovery | dictionary learning for inpainting · AT&T face, 50% missing pixels | PSNR (dB): 28.91 vs 27.6 best baseline | Helped |
| Use SparsDT (no regularization parameters) instead of baselines with grid-searched regularization | dictionary learning runtime · Synthetic SαS, α=1.2 | avg time (s): 5.19 s vs 0.07 s for l2/l1 but faster than l1/l2 (95.75 s) and l1+l2 (19.89 s) | Mixed |
| Use exact alpha instead of estimated alpha in SparsDT (ablation) | SparsDT · Synthetic SαS, m=16, n=24, α=1 and 1.5, varying K | average correlation | No clear effect |
| Increase number of samples K (ablation) | SparsDT · Synthetic SαS, m=16, n=24, α=1 and 1.5 | average correlation | Helped |
Closest papers by meaning
- Hiding Data Helps: On the Benefits of Masking for Sparse Coding (ICML 2023 · similarity 0.89)
- Understanding l4-based Dictionary Learning: Interpretation, Stability, and Robustness (ICLR 2020 · similarity 0.88)
- Beyond the Grid: Continuous Dictionary Pursuit for Interpretable Signal Decomposition (NeurIPS 2026 · similarity 0.87)
- Learning brain regions via large-scale online structured sparse dictionary learning (NeurIPS 2016 · similarity 0.86)
- Sparse Bayesian Generative Modeling for Compressive Sensing (NeurIPS 2024 · similarity 0.85)
- The Price of Sparsity: Sufficient Conditions for Sparse Recovery using Sparse and Sparsified Measurements (NeurIPS 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.