Research map / Architectures and efficiency / Pruning and efficient tuning
Learning to Prune Deep Neural Networks via Layer-wise Optimal Brain Surgeon
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 layer-wise OBS sensitivity pruning instead of magnitude-based LWC at comparable compression | LeNet-300-100 · MNIST | error after pruning: 81.32% -> 3.10% | Helped |
| Use layer-wise OBS sensitivity pruning instead of magnitude-based LWC | LeNet-5 · MNIST | error after pruning: 89.55% -> 3.21% | Helped |
| Use layer-wise OBS sensitivity pruning instead of magnitude-based LWC | CIFAR-Net · CIFAR-10 | error after pruning: 87.65% -> 21.32% | Helped |
| Use layer-wise OBS sensitivity pruning instead of magnitude-based LWC | AlexNet · ImageNet ILSVRC-2012 | Top-1 error after pruning: 76.14 -> 50.04 | Helped |
| Use layer-wise OBS sensitivity pruning instead of magnitude-based LWC | VGG-16 · ImageNet ILSVRC-2012 | Top-1 error after pruning: 73.61 -> 37.32 | Helped |
| Iteratively prune and lightly retrain to go beyond the single-pass pruning level | L-OBS on LeNet-300-100 · MNIST | CR: 7% -> 1.5% | Helped |
| Iteratively prune and lightly retrain to go beyond the single-pass pruning level | L-OBS on LeNet-5 · MNIST | CR: 7% -> 0.9% | Helped |
| Use L-OBS instead of Net-Trim at the same reported pruning error setting | LeNet-300-100 second layer · LeNet-300-100 | pruned error: 28.45% -> 4.56% | Helped |
Closest papers by meaning
- Structured Pruning of Neural Networks With Budget-Aware Regularization (CVPR 2019 · similarity 0.93)
- Learning Pruning-Friendly Networks via Frank-Wolfe: One-Shot, Any-Sparsity, And No Retraining (ICLR 2022 · similarity 0.91)
- Pruning from Scratch (AAAI 2020 · similarity 0.91)
- Accelerate CNNs from Three Dimensions: A Comprehensive Pruning Framework (ICML 2021 · similarity 0.91)
- NISP: Pruning Networks Using Neuron Importance Score Propagation (CVPR 2018 · similarity 0.90)
- “Learning-Compression” Algorithms for Neural Net Pruning (CVPR 2018 · similarity 0.90)
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.