Research map / Architectures and efficiency / CNN design
Dynamic Channel Pruning: Feature Boosting and Suppression
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 input-dependent channel saliency to dynamically skip channels rather than pruning them statically | ResNet-18 · ImageNet ILSVRC2012 validation | MAC saving: 1.98$\times$ (FBS) vs 1.39$\times$ (Network Slimming) | Helped |
| Use FBS rather than Network Slimming for channel acceleration | VGG-16 · ImageNet ILSVRC2012 validation, 3$\times$ speed-up | top-5 error increase (%): 1.37 -> 0.04 | Helped |
| Use FBS rather than Network Slimming for channel acceleration | VGG-16 · ImageNet ILSVRC2012 validation, 4$\times$ speed-up | top-5 error increase (%): 3.26 -> 0.52 | Helped |
| Use FBS rather than Network Slimming for channel acceleration | VGG-16 · ImageNet ILSVRC2012 validation, 5$\times$ speed-up | top-5 error increase (%): 5.18 -> 0.59 | Helped |
| Add FBS after Network Slimming | M-CifarNet · CIFAR-10, 90.50% accuracy constraint | speed-up: $1.19\times$ -> $3.22\times$ | Helped |
| Add FBS auxiliary layers without suppressing channels (d = 1) | ResNet-18 · ImageNet ILSVRC2012 | top-1 accuracy: +1.73% | Helped |
| Add FBS auxiliary layers without suppressing channels (d = 1) | ResNet-18 · ImageNet ILSVRC2012 | top-5 accuracy: +0.46% | Helped |
| Use dynamic FBS instead of static Network Slimming | M-CifarNet · CIFAR-10, $2\times$ speed-up target | top-1 accuracy (%): $87.54\%$ -> $91.55\%$ | Helped |
| Set FBS channel density to d = 0.5 (ablation) | M-CifarNet · CIFAR-10 | top-1 accuracy (%): 91.37% -> 90.59% | Hurt |
Closest papers by meaning
- Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers (ICLR 2018 · similarity 0.91)
- RePr: Improved Training of Convolutional Filters (CVPR 2019 · similarity 0.89)
- BilevelPruning: Unified Dynamic and Static Channel Pruning for Convolutional Neural Networks (CVPR 2024 · similarity 0.89)
- REPrune: Channel Pruning via Kernel Representative Selection (AAAI 2024 · similarity 0.89)
- CHEX: CHannel EXploration for CNN Model Compression (CVPR 2022 · similarity 0.88)
- Rethinking the Pruning Criteria for Convolutional Neural Network (NeurIPS 2021 · 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.