Research map / Architectures and efficiency / RNNs and normalization
On Multi-Layer Basis Pursuit, Efficient Algorithms and Convolutional Neural Networks
Jeremias Sulam, Aviad Aberdam, Amir Beck, Michael Elad
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 |
|---|---|---|---|
| unfold ML-ISTA iterations on top of the feed-forward CNN (same parameters) (ablation) | ML-ISTA · CIFAR10 | accuracy (%): 79.00 -> 82.93 | Improved |
| unfold ML-FISTA iterations instead of feed-forward | CNN · SVHN | accuracy (%): 92.44 -> 93.79 | Improved |
| unfold ML-ISTA instead of feed-forward | CNN · MNIST | accuracy (%): 98.78 -> 99.10 | Improved |
| use global multi-layer pursuit (ML-ISTA) instead of layer-wise Layered BP | unrolled CNN · CIFAR10 | accuracy (%): 80.73 -> 82.93 | Improved |
| use global multi-layer pursuit (ML-ISTA) instead of layer-wise Layered BP | unrolled CNN · MNIST | accuracy (%): 99.19 -> 99.10 | Hurt |
| replace dictionaries with LISTA-style factors W and B (ML-LISTA) | ML-ISTA · CIFAR10 | accuracy (%): 82.93 -> 82.68 | No clear effect |
| free all filters across unfoldings (All-Free) instead of tied ML-ISTA | ML-ISTA · CIFAR10 | accuracy (%): 82.93 -> 81.48 | Hurt |
| free all filters across unfoldings (All-Free) instead of tied ML-ISTA | ML-ISTA · SVHN | accuracy (%): 93.52 -> 94.06 | No clear effect |
| add intermediate-layer penalty (lambda_1 > 0) to basis pursuit (ablation) | ISTA/ML-ISTA · synthetic two-layer model (SNR=10) | recovery error: lower recovery error in both gamma_2 and gamma_1 | Improved |
| decrease mu in ML-ISTA/ML-FISTA (ablation) | ML-ISTA/ML-FISTA · synthetic two-layer model | solution accuracy | Mixed |
Closest papers by meaning
- Convolutional Neural Networks Analyzed via Convolutional Sparse Coding (JMLR 2017 · similarity 0.93)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · similarity 0.89)
- A General Efficient Hyperparameter-Free Algorithm for Convolutional Sparse Learning (AAAI 2017 · similarity 0.89)
- Global Minimizers of $\ell^p$-Regularized Objectives Yield the Sparsest ReLU Neural Networks (NeurIPS 2025 · similarity 0.88)
- From Flat to Hierarchical: Extracting Sparse Representations with Matching Pursuit (NeurIPS 2025 · similarity 0.88)
- Sparse Cocktail: Every Sparse Pattern Every Sparse Ratio All At Once (ICML 2024 · similarity 0.87)
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.
