Research map / Architectures and efficiency / Pruning and efficient tuning
Learning sparse neural networks via sensitivity-driven regularization
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 unspecific sensitivity instead of specific sensitivity (ablation) | Sensitivity-based regularization on LeNet300 · MNIST (LeNet300, lower-error results) | compression ratio: 80x -> 103x | Helped |
| Use unspecific sensitivity instead of specific sensitivity (ablation) | Sensitivity-based regularization on LeNet5 · MNIST (LeNet5) | compression ratio: 41.9x -> 51.1x | Helped |
| Use specific sensitivity instead of unspecific sensitivity (ablation) | Sensitivity-based regularization on VGG-16 · ImageNet | compression ratio: 12.17x -> 14.12x | Helped |
| Use sensitivity-based regularization rather than no regularizer (ablation) | LeNet300 · MNIST | test error | Helped |
| Use sensitivity-based regularization rather than L1 regularization (ablation) | LeNet300 · MNIST | test error | Helped |
| Use sensitivity-based regularization rather than L2 regularization (ablation) | LeNet300 · MNIST | test error | Helped |
| Apply sensitivity-based sparsification | LeNet300 · MNIST (Top-1 error 1.6% target) | compression ratio: 12.2x vs. 27.8x | Helped |
| Apply sensitivity-based sparsification | LeNet300 · MNIST (lower-error results) | compression ratio: 103x vs. 68x | Helped |
| Apply sensitivity-based sparsification | LeNet5 · MNIST | compression ratio: 51.07x vs. 11.87x | Helped |
| Apply sensitivity-based sparsification | VGG-16 · ImageNet | Top-1 error: 9.80% vs 10.88% | Helped |
| Use sensitivity-based sparsification rather than Sparse VD | LeNet5 · MNIST | compression ratio: 51.1x vs. 280x | Hurt |
Closest papers by meaning
- Learning Sparse Neural Networks through L_0 Regularization (ICLR 2018 · similarity 0.91)
- Global Minimizers of $\ell^p$-Regularized Objectives Yield the Sparsest ReLU Neural Networks (NeurIPS 2025 · similarity 0.90)
- Sparsifying Networks via Subdifferential Inclusion (ICML 2021 · similarity 0.89)
- DeepHoyer: Learning Sparser Neural Network with Differentiable Scale-Invariant Sparsity Measures (ICLR 2020 · similarity 0.89)
- A Single-Step, Sharpness-Aware Minimization is All You Need to Achieve Efficient and Accurate Sparse Training (NeurIPS 2024 · similarity 0.88)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · similarity 0.88)
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.