Research map / Optimization / Optimizers
Proximal Alternating Direction Network: A Globally Converged Deep Unrolling Framework
Risheng Liu, Xin Fan, Shichao Cheng, Xiangyu Wang, Zhongxuan Luo
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 PADNet with a learned prior and explicit task fidelity | Non-blind deconvolution · Levin benchmark | PSNR: 32.74 -> 33.41 | Improved |
| Use PADNet with a learned prior and explicit task fidelity | Non-blind deconvolution · Sun benchmark | PSNR: 32.44 -> 32.69 | Improved |
| Use implicit PADNet to learn priors from training data | Non-blind deconvolution · Sun benchmark | PSNR: 32.44 -> 32.71 | Improved |
| Use PADNet for haze removal | Single-image haze removal · Fattal benchmark | PSNR: 27.11 -> 28.47 | Improved |
| Use PADNet for haze removal | Single-image haze removal · Fattal benchmark | Time (s): 3.73 -> 2.74 | Improved |
| Replace conventional optimization iterations with learned PADNet propagation | Gradient-domain deconvolution · Levin benchmark | Number of iterations: 29 -> 5 | Improved |
| Use the learned implicit prior instead of handcrafted ℓ1 or ℓ0 priors | Gradient-domain deconvolution · Levin benchmark | Relative reconstruction error: 1.3146 -> 0.4291; 1.7114 -> 0.4291 | Improved |
| Use the learned implicit prior rather than the explicit ℓ0.8 prior | Gradient-domain deconvolution · Levin benchmark | Relative reconstruction error: 0.4140 -> 0.4291 | Hurt |
| Choose explicit PADNet with ℓ0.8 for standard images and implicit PADNet for challenging real-world images | Non-blind deconvolution · Levin and Sun benchmarks | PSNR: 33.41 vs 33.37; 32.69 vs 32.71 | Mixed |
Closest papers by meaning
- Variational Inference for Infinitely Deep Neural Networks (ICML 2022 · similarity 0.87)
- A statistical perspective on algorithm unrolling models for inverse problems (JMLR 2025 · similarity 0.87)
- Directional convergence and alignment in deep learning (NeurIPS 2020 · similarity 0.87)
- What's in a Prior? Learned Proximal Networks for Inverse Problems (ICLR 2024 · similarity 0.85)
- Unrolled denoising networks provably learn to perform optimal Bayesian inference (NeurIPS 2024 · similarity 0.85)
- Variational Learning is Effective for Large Deep Networks (ICML 2024 · similarity 0.84)
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.
