Research map / Architectures and efficiency / CNN design
Multi-layer State Evolution Under Random Convolutional Design
Max Daniels, Cedric Gerbelot, Florent Krzakala, Lenka Zdeborova
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 random multichannel convolutional (MCC) sensing matrices instead of dense i.i.d. Gaussian matrices in AMP | AMP for sparse compressive sensing · Sparse compressive sensing, P=q=1024, k=3, rho in {0.25, 0.5, 0.75} | log(MSE) vs measurement ratio beta | No clear effect |
| Reduce convolution size q from 1024 to 10 (q much smaller than P) | MCC AMP · Sparse compressive sensing | agreement with dense state evolution | No clear effect |
| Stack multiple layers (L = 1, 2, 3) with MCC couplings and identity channels | ML-AMP · Multi-layer compressive sensing, linear activations, q=10, k=3, n(L)=5000 | log(MSE) | No clear effect |
| Use ReLU channel functions instead of identity in deeper layers | ML-AMP with MCC layers · Multi-layer compressive sensing, ReLU activations | log(MSE) | No clear effect |
| Vary measurement ratio beta near the recovery threshold (0.4 to 0.55) at rho = 0.25 | MCC AMP · Sparse compressive sensing | AMP iterate convergence in log(MSE) | No clear effect |
Closest papers by meaning
- Signal Recovery with Non-Expansive Generative Network Priors (NeurIPS 2022 · similarity 0.87)
- Constant-Expansion Suffices for Compressed Sensing with Generative Priors (NeurIPS 2020 · similarity 0.85)
- Unrolled denoising networks provably learn to perform optimal Bayesian inference (NeurIPS 2024 · similarity 0.84)
- Efficient Algorithms for Non-gaussian Single Index Models with Generative Priors (AAAI 2024 · similarity 0.84)
- Sparse Bayesian Generative Modeling for Compressive Sensing (NeurIPS 2024 · similarity 0.84)
- Consistency Posterior Sampling for Diverse Image Synthesis (CVPR 2025 · similarity 0.83)
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.
