Research map / Architectures and efficiency / Pruning and efficient tuning
Effective Sparsification of Neural Networks With Global Sparsity Constraint
Check 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 |
|---|---|---|---|
| allocate sparsity globally rather than uniformly across layers (ablation) | ProbMask · CIFAR-10, ResNet32, 99.9% pruning | accuracy: 19.12 -> 76.87 | Helped |
| allocate sparsity globally rather than uniformly across layers (ablation) | ProbMask · CIFAR-10, ResNet32, 99.5% pruning | accuracy: 74.6 -> 89.34 | Helped |
| allocate sparsity globally rather than uniformly across layers (ablation) | ProbMask · CIFAR-10, ResNet32, 90% pruning | accuracy: 94.89 -> 94.96 | No clear effect |
| use ProbMask instead of the strongest listed baseline at 99.9% pruning | VGG19 · CIFAR-10 | accuracy: 65.38 -> 89.79 | Helped |
| use ProbMask instead of the strongest listed baseline at 99.9% pruning | VGG19 · CIFAR-100 | accuracy: 5.58 -> 60.41 | Helped |
| use ProbMask instead of the strongest listed baseline at 99.9% pruning | ResNet32 · CIFAR-10 | accuracy: 38.64 -> 76.87 | Helped |
| use ProbMask instead of the strongest listed baseline at 99.9% pruning | ResNet32 · CIFAR-100 | accuracy: 5.96 -> 26.72 | Helped |
| use ProbMask instead of STR | ResNet50 · ImageNet-1K, 99% pruning | accuracy: 50.35 -> 61.07 | Helped |
| use ProbMask instead of STR | MobileNetV1 · ImageNet-1K, 94.1% sparsity | Top-1 accuracy: 23.61 -> 60.10 | Helped |
| use ProbMask to find a supermask with weights fixed at initialization | ResNet32 · CIFAR-100, 2% remaining weights | accuracy | Helped |
Closest papers by meaning
- Shaving Weights with Occam's Razor: Bayesian Sparsification for Neural Networks using the Marginal Likelihood (NeurIPS 2024 · similarity 0.90)
- Neural Pruning via Growing Regularization (ICLR 2021 · similarity 0.89)
- Powerpropagation: A sparsity inducing weight reparameterisation (NeurIPS 2021 · similarity 0.89)
- Probabilistic Neural Pruning via Sparsity Evolutionary Fokker-Planck-Kolmogorov Equation (ICLR 2025 · similarity 0.89)
- Effective Model Sparsification by Scheduled Grow-and-Prune Methods (ICLR 2022 · similarity 0.88)
- Targeted Low-rank Refinement: Enhancing Sparse Language Models with Precision (ICML 2025 · 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.