Research map / Optimization / Optimizers
Understanding approximate and unrolled dictionary learning for pattern recovery
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 |
|---|---|---|---|
| Disable steps learning (ablation) | Stochastic Deep CDL · MEG, minibatch 5 and 20 s window | mean correlation: 0.845 -> 0.83 | Hurt |
| Disable steps learning (ablation) | Stochastic Deep CDL · MEG, minibatch 5 and 20 s window | time: 110 s -> 57 s | Helped |
| Shorten the time window from 20 s to 10 s (ablation) | Stochastic Deep CDL with steps learning enabled and minibatch 5 · MEG | mean correlation: 0.845 -> 0.825 | Hurt |
| Shorten the time window from 20 s to 10 s (ablation) | Stochastic Deep CDL with steps learning enabled and minibatch 5 · MEG | time: 110 s -> 56 s | Helped |
| Increase minibatch size from 5 to 20 (ablation) | Stochastic Deep CDL with steps learning enabled and 10 s window · MEG | mean correlation: 0.825 -> 0.80 | Hurt |
| Increase minibatch size from 5 to 20 (ablation) | Stochastic Deep CDL with steps learning enabled and 10 s window · MEG | time: 56 s -> 163 s | Hurt |
| Use unrolled dictionary learning instead of alternating minimization at small depths | Dictionary learning · Synthetic data and image patches | gradient direction estimation: up to 20% | Helped |
| Use too many unrolled iterations or back-propagation too deeply | Deep Dictionary Learning · Synthetic data and image patches | gradient direction estimation | Hurt |
| Use 20 FISTA iterations for approximate dictionary learning instead of standard dictionary learning | Image denoising dictionary learning · Noisy image | time: 600 seconds -> 80 seconds | Helped |
| Use stochastic DDL instead of Online dictionary learning | Dictionary recovery · Synthetic data, smaller lambda | time to reach recovery score 0.95 | Helped |
| Use stochastic Deep CDL for MEG pattern learning instead of alphacsc | MEG pattern learning · MEG | mean correlation: 80% -- 90% average correlation; around 100 sec. compared to around 1400 sec. | Mixed |
Closest papers by meaning
- Alternating minimization for dictionary learning with random initialization (NeurIPS 2017 · similarity 0.89)
- NOODL: Provable Online Dictionary Learning and Sparse Coding (ICLR 2019 · similarity 0.87)
- Dictionary Learning Based on Sparse Distribution Tomography (ICML 2017 · similarity 0.85)
- A Provable Approach for Double-Sparse Coding (AAAI 2018 · similarity 0.85)
- Unrolled denoising networks provably learn to perform optimal Bayesian inference (NeurIPS 2024 · similarity 0.85)
- Time-Embedded Algorithm Unrolling for Computational MRI (NeurIPS 2025 · similarity 0.83)
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.