Research map / Architectures and efficiency / Pruning and efficient tuning
Towards Efficient Model Compression via Learned Global Ranking
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 |
|---|---|---|---|
| Learn one global filter ranking and reuse it across FLOP targets | LeGR · ResNet-56 on CIFAR-100 | pruning search time: 7× faster than AMC; 5× faster than MorphNet | Helped |
| Use the learned global filter ranking instead of uniform pruning | LeGR · CIFAR-100 | accuracy | Helped |
| Learn the ranking at different lowest FLOP targets (ablation) | LeGR on ResNet-56 · CIFAR-100 | accuracy | No clear effect |
| Increase ranking-search fine-tuning steps from 0 to 50, 200, or 500 (ablation) | LeGR on ResNet-56 · CIFAR-100 | test accuracy | Helped |
| Use more ranking-search fine-tuning steps at lower FLOP targets (ablation) | LeGR on ResNet-56 · CIFAR-100 | test accuracy | Helped |
| Use DDPG instead of evolutionary search to learn layer-wise affine transformations (ablation) | LeGR on ResNet-56 · CIFAR-100 | final network accuracy variance | Hurt |
| Search layer-wise affine filter scores rather than layer-wise filter percentages (ablation) | LeGR on ResNet-56 · CIFAR-100 | search efficiency and final performance | Helped |
Closest papers by meaning
- Pruning Filters for Efficient ConvNets (ICLR 2017 · similarity 0.89)
- Compressing Convolutional Neural Networks via Factorized Convolutional Filters (CVPR 2019 · similarity 0.89)
- RePr: Improved Training of Convolutional Filters (CVPR 2019 · similarity 0.89)
- Provable Filter Pruning for Efficient Neural Networks (ICLR 2020 · similarity 0.88)
- DPFPS: Dynamic and Progressive Filter Pruning for Compressing Convolutional Neural Networks from Scratch (AAAI 2021 · similarity 0.88)
- Entropy Induced Pruning Framework for Convolutional Neural Networks (AAAI 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.