Research map / Architectures and efficiency / Pruning and efficient tuning
Channel Pruning Guided by Classification Loss and Feature Importance
Jinyang Guo, Wanli Ouyang, Dong Xu
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 |
|---|---|---|---|
| add classification-loss guidance during channel selection (ablation) | CPLI · CIFAR-10, VGGNet, 2.86× CR | Accuracy change (%): +0.04 -> +0.96 | Improved |
| add feature-importance guidance to account for features removed at the next pruning stage (ablation) | CPLI · CIFAR-10, VGGNet, 2.86× CR | Accuracy change (%): +0.47 -> +0.96 | Improved |
| increase sampled spatial locations from 1 to 10 (ablation) | CPLI · CIFAR-10, VGGNet, 2.86× CR | Accuracy change (%): +0.12 -> +0.96 | Improved |
| increase sampled spatial locations from 10 to 20 (ablation) | CPLI · CIFAR-10, VGGNet, 2.86× CR | Accuracy change (%): +0.96 -> +0.97 | No clear effect |
| use CPLI rather than DCP | VGGNet · CIFAR-10, 2.86× CR | Top-1 accuracy change (%): +0.58 -> +0.96 | Improved |
| use CPLI rather than DCP | ResNet-56 · CIFAR-10, 1.99× CR | Top-1 accuracy change (%): -0.31 -> +0.07 | Improved |
| use CPLI rather than DCP | MobileNet-V2 · CIFAR-10, 1.36× CR | Top-1 accuracy change (%): +0.22 -> +0.35 | Improved |
| use CPLI rather than DCP | ResNet-50 · ImageNet, 2.25× CR | Top-5 accuracy change (%): -0.61 -> -0.38 | Improved |
| use CPLI rather than DCP | MobileNet-V2 · ImageNet, 1.81× CR | Top-1 accuracy change (%): -5.89 -> -4.84 | Improved |
| use CPLI rather than RBP | C3D · UCF-101, 2× CR | Clip-level accuracy change (%): -3.56 -> -2.18 | Improved |
Closest papers by meaning
- Rethinking the Pruning Criteria for Convolutional Neural Network (NeurIPS 2021 · similarity 0.89)
- Influence Function Based Second-Order Channel Pruning: Evaluating True Loss Changes for Pruning is Possible Without Retraining (TPAMI 2024 · similarity 0.89)
- Discrimination-aware Channel Pruning for Deep Neural Networks (NeurIPS 2018 · similarity 0.87)
- Network Pruning via Transformable Architecture Search (NeurIPS 2019 · similarity 0.86)
- BilevelPruning: Unified Dynamic and Static Channel Pruning for Convolutional Neural Networks (CVPR 2024 · similarity 0.85)
- REPrune: Channel Pruning via Kernel Representative Selection (AAAI 2024 · similarity 0.85)
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.
