Research map / Learning theory and algorithms / Matrix and tensor methods
TFPnP: Tuning-free Plug-and-Play Proximal Algorithms with Applications to Inverse Imaging Problems
Kaixuan Wei, Angelica Aviles-Rivero, Jingwei Liang, Ying Fu, Hua Huang, Carola-Bibiane Schönlieb
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 |
|---|---|---|---|
| add noise level and step info to policy input (ablation) | TFPnP policy · CS-MRI 8x noise 15 | PSNR / iterations: 26.27 -> 26.52 (iterations 30.0 -> 15.0) | Improved |
| add noise level and step info to policy input (ablation) | TFPnP policy · CS-MRI 2x noise 15 | PSNR / iterations: 30.15 -> 30.33 (iterations 27.4 -> 5.0) | Improved |
| add early-stop penalty eta=0.05 instead of eta=0 (ablation) | TFPnP reward · CS-MRI 2x noise 15 | iterations: 27.1 -> 5.0 | Improved |
| raise penalty eta from 0.05 to 0.25 (ablation) | TFPnP reward · CS-MRI 8x noise 15 | PSNR: 26.52 -> 25.53 | Hurt |
| vary iterations per step m and horizon N with m x N fixed at 30 (ablation) | TFPnP policy · CS-MRI 8x noise 15 | PSNR: 26.27, 26.39, 26.52 | No clear effect |
| optimize pi2 with model-based RL instead of model-free DDPG (ablation) | TFPnP policy · CS-MRI Medical7 8x noise 15 | PSNR: 26.15 -> 26.52 | Improved |
| use learned policy instead of oracle fixed parameters | PnP-ADMM · CS-MRI Medical7 8x noise 15 | PSNR: 26.41 -> 26.52 | Improved |
| use learned policy instead of handcrafted policy | PnP-ADMM · CS-MRI Medical7 8x noise 15 | PSNR: 25.76 -> 26.52 | Improved |
| add optimal early stopping to fixed parameters | PnP-ADMM fixed policy · CS-MRI Medical7 2x noise 15 | PSNR: 23.94 -> 28.45 | Improved |
| use U-net instead of DnCNN as denoiser prior (ablation) | PnP-ADMM CS-MRI · CS-MRI noise 15 | PSNR: 25.43 -> 25.76 | Improved |
| apply learned policy to PnP-PGM | PnP-PGM · CS-MRI 8x noise 15 | PSNR: 25.85 -> 26.28 | Improved |
| apply learned policy to PnP-HQS | PnP-HQS · CS-MRI 8x noise 15 | PSNR: 25.50 -> 25.96 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- prDeep — prDeep: Robust Phase Retrieval with a Flexible Deep Network (ICML 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PnP-Flow: Plug-and-Play Image Restoration with Flow Matching (ICLR 2025 · similarity 0.91)
- Provably Robust Score-Based Diffusion Posterior Sampling for Plug-and-Play Image Reconstruction (NeurIPS 2024 · similarity 0.91)
- Plug-and-Play image restoration with Stochastic deNOising REgularization (ICML 2024 · similarity 0.90)
- Prior Mismatch and Adaptation in PnP-ADMM with a Nonconvex Convergence Analysis (ICML 2024 · similarity 0.90)
- Proximal Algorithm Unrolling: Flexible and Efficient Reconstruction Networks for Single-Pixel Imaging (CVPR 2025 · similarity 0.89)
- Learning Cocoercive Conservative Denoisers via Helmholtz Decomposition for Poisson Imaging Inverse Problems (NeurIPS 2025 · similarity 0.89)
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.
