Research map / Optimization / Nonconvex and smooth optimization
Learning to solve TV regularised problems with unrolled algorithms
Hamza Cherkaoui, Jeremias Sulam, Thomas Moreau
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 analysis formulation instead of synthesis formulation (ablation) | unrolled PGD / FISTA for 1D TV · Simulation (n=2000, k=8, m=5), lambda 0.1 and 0.8 | risk vs iterations | Improved |
| learn parameters of unrolled PGD with prox-TV (LPGD-Taut, LPGD-LISTA) vs iterative PGD | analysis PGD · Simulation, low iteration regime | risk vs iterations | Improved |
| high regularization (lambda 0.8) vs low (0.1) for LPGD-LISTA | LPGD-LISTA with 50 inner layers · Simulation | risk vs iterations | Hurt |
| high regularization (lambda 0.8) vs low (0.1) for LPGD-Taut | LPGD-Taut · Simulation | risk vs iterations | Hurt |
| train the inner LISTA parameters of unrolled prox-TV vs untrained initialization (ablation) | LPGD-LISTA · Simulation, prox-TV error, T in [20,50] | prox-TV error | No clear effect |
| LPGD-Taut vs Accelerated PGD (analysis) | fMRI deconvolution, UK Biobank · fMRI data, lambda 0.1 | risk vs iterations | Improved |
| raise lambda from 0.1 to 0.8 lambda_max | LPGD-Taut · fMRI data | acceleration over PGD | Hurt |
Closest papers by meaning
- Modular Proximal Optimization for Multidimensional Total-Variation Regularization (JMLR 2018 · similarity 0.91)
- TV: A Sparse Optimization Method for Impulse Noise Image Restoration (TPAMI 2017 · similarity 0.86)
- Weakly Convex Regularisers for Inverse Problems: Convergence of Critical Points and Primal-Dual Optimisation (ICML 2024 · similarity 0.84)
- What's in a Prior? Learned Proximal Networks for Inverse Problems (ICLR 2024 · similarity 0.83)
- Proximal Algorithm Unrolling: Flexible and Efficient Reconstruction Networks for Single-Pixel Imaging (CVPR 2025 · similarity 0.83)
- Controlling the Flow: Stability and Convergence for Stochastic Gradient Descent with Decaying Regularization (NeurIPS 2025 · similarity 0.82)
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.
