Research map / Learning theory and algorithms / Matrix and tensor methods
Predictive Coding Machine for Compressed Sensing and Image Denoising
Jun Li, Hongfu Liu, Yun Fu
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 |
|---|---|---|---|
| train DNN encoder with APG guidance instead of ISTA (GDgAPG vs LISTA) | LISTA · compressed sensing | average REE: 0.1465 -> 0.1004 | Improved |
| replace iterative l1 solvers with DNN encoder at inference | ISTA/APG · compressed sensing | test time (s): 0.1164 (APG) -> 0.00599 | Improved |
| add predictive DNN encoder (GDgAPG) to sparse coding | APG · image denoising, sigma=e-3 | average PSNR (dB): 27.19 -> 31.24 | Improved |
| GDgAPG vs LISTA | LISTA · image denoising, sigma=e-1 | average PSNR (dB): 24.65 -> 26.32 | Improved |
| use DNN encoder trained on low-noise data | APG and ISTA · image denoising, sigma=e-1 | average PSNR (dB): 17.81 -> 26.32 (GDgAPG) | Improved |
| use DNN encoder at inference | APG · image denoising | test time (s): 149.9066 -> 0.2534 | Improved |
| train DNN encoder (GDgAPG vs APG) | APG · image denoising | training time (s): 0 -> 1137.1975 (needs training) | Hurt |
| vary regularization parameter lambda (ablation) | GDgAPG · compressed sensing | RRE/REE | Mixed |
| vary number of hidden units h (ablation) | GDgAPG · compressed sensing | RRE/REE | No clear effect |
| vary DNN layers L (2 to 5) (ablation) | GDgAPG · compressed sensing | REE | Mixed |
Closest papers by meaning
- Practical Compact Deep Compressed Sensing (TPAMI 2024 · similarity 0.85)
- Deep Predictive Coding Network for Object Recognition (ICML 2018 · similarity 0.85)
- Variational Sparse Coding with Learned Thresholding (ICML 2022 · similarity 0.85)
- Learning Deep ℓ0 Encoders (AAAI 2016 · similarity 0.85)
- Benchmarking Predictive Coding Networks -- Made Simple (ICLR 2025 · similarity 0.85)
- Understanding and Improving Optimization in Predictive Coding Networks (AAAI 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.
