Research map / Image generation / Image restoration
DiCoDiLe: Distributed Convolutional Dictionary Learning
Thomas Moreau, Alexandre Gramfort
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 |
|---|---|---|---|
| split the domain on a 2D grid of workers instead of a line (ablation) | DiCoDiLe-Z · Mandrill 512x512 images | worker scaling limit: W = T_1/2L_1 = 32 -> up to 1024 workers | Improved |
| add soft-locks between neighboring workers (ablation) | DiCoDiLe-Z · Mandrill 512x512 images (49 workers) | convergence/reconstruction artifacts | Improved |
| use locally greedy coordinate selection instead of greedy selection (ablation) | coordinate descent for CSC · 1D synthetic signals (T=150L and T=750L, K=25, L=250) | running time | Improved |
| use locally greedy selection instead of randomized coordinate descent (ablation) | coordinate descent for CSC · 1D synthetic signals (T=150L and T=750L) | running time | Improved |
| use DiCoDiLe-Z with LGCD instead of DICOD with GCD per worker | distributed CSC · 1D synthetic signals | running time | Improved |
| use DiCoDiLe-Z with LGCD instead of greedy selection on 2D images (ablation) | DiCoDiLe-Z · 2D images with different lambda | running time | Improved |
| increase regularization lambda (ablation) | DiCoDiLe-Z · 2D images | convergence speed | Improved |
| use DiCoDiLe instead of Consensus ADMM | convolutional dictionary learning · Hubble GOODS South 512x512 patch (36 workers) | objective cost vs time | Improved |
| use DiCoDiLe instead of Consensus ADMM on full image | convolutional dictionary learning · Hubble image 6000x3600 | memory feasibility: failed to fit in 250 GB of RAM for ADMM | Improved |
Closest papers by meaning
- Convolutional dictionary learning based auto-encoders for natural exponential-family distributions (ICML 2020 · similarity 0.85)
- Multiscale Dictionary Learning: Non-Asymptotic Bounds and Robustness (JMLR 2016 · similarity 0.85)
- Dictionary Learning Based on Sparse Distribution Tomography (ICML 2017 · similarity 0.84)
- DICOD: Distributed Convolutional Coordinate Descent for Convolutional Sparse Coding (ICML 2018 · similarity 0.84)
- Online Convolutional Sparse Coding with Sample-Dependent Dictionary (ICML 2018 · similarity 0.84)
- On the Sparsity-Storage-Accuracy Tradeoff in Parsimoniously Activated Dictionary Learning (NeurIPS 2026 · similarity 0.81)
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.
