Research map / Architectures and efficiency / RNNs and normalization
Maximal Sparsity with Deep Networks?
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the proposed support-classification network instead of SBL | photometric stereo · q=10 images | average angular error (degrees): 4.02 -> 1.48 | Helped |
| use the proposed support-classification network instead of SBL | photometric stereo · q=20 and q=40 images | average angular error (degrees): 1.86 -> 1.95; 0.50 -> 1.20 | Hurt |
| use the proposed network instead of SBL | photometric stereo · q=10, q=20, and q=40 images | runtime (sec.): 59.1 -> 1.17; 42.5 -> 1.21; 75.2 -> 1.25 | Helped |
| add residual connections (ablation) | 20-layer support-classification network · basic sparse estimation | strict and loose accuracy | Helped |
| use ReLU instead of HELU hard-threshold activations (ablation) | 20-layer support-classification network · basic sparse estimation | strict and loose accuracy | Helped |
| use multi-label classification loss instead of quadratic loss (ablation) | 20-layer support-classification network · basic sparse estimation | strict and loose accuracy | Helped |
| increase the training-set size | proposed support-classification network · basic sparse estimation | strict and loose accuracy | Helped |
| train and test with mismatched nonzero coefficient magnitude distributions | proposed support-classification network · basic sparse estimation | strict and loose accuracy | No clear effect |
Closest papers by meaning
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 2024 · similarity 0.88)
- Straight-Through Meets Sparse Recovery: the Support Exploration Algorithm (ICML 2024 · similarity 0.87)
- Global Minimizers of $\ell^p$-Regularized Objectives Yield the Sparsest ReLU Neural Networks (NeurIPS 2025 · similarity 0.86)
- The Iterative Optimal Brain Surgeon: Faster Sparse Recovery by Leveraging Second-Order Information (NeurIPS 2024 · similarity 0.86)
- Sparse maximal update parameterization: A holistic approach to sparse training dynamics (NeurIPS 2024 · similarity 0.86)
- Preserving Deep Representations in One-Shot Pruning: A Hessian-Free Second-Order Optimization Framework (ICLR 2025 · similarity 0.86)
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.