Research map / Architectures and efficiency / CNN design
Filter Pruning via Feature Discrimination in Deep Neural Networks
Zhiqiang He, Yaguan Qian, Yuqi Wang, Bin Wang, Xiaohui Guan, Zhaoquan Gu, Xiang Ling, Shaoning Zeng, Haijiang Wang, Wujie Zhou
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 distinguishing layer pruning instead of global threshold pruning (ablation) | RFC criterion (VGG16) · CIFAR-10 | accuracy (%): 92.25 -> 93.38 | Improved |
| add distinguishing layer pruning instead of global threshold pruning (ablation) | RFC criterion (VGG16) · CIFAR-100 | accuracy (%): 71.25 -> 73.82 | Improved |
| add distinguishing layer pruning instead of global threshold pruning (ablation) | L1 criterion (VGG16) · CIFAR-10 | accuracy (%): 90.56 -> 93.12 | Improved |
| add distinguishing layer pruning instead of global threshold pruning (ablation) | L1 criterion (VGG16) · CIFAR-100 | accuracy (%): 70.90 -> 73.49 | Improved |
| add distinguishing layer pruning instead of global threshold pruning (ablation) | BN scale (NS) criterion (VGG16) · CIFAR-10 | accuracy (%): 92.64 -> 93.25 | Improved |
| add distinguishing layer pruning instead of global threshold pruning (ablation) | BN scale (NS) criterion (VGG16) · CIFAR-100 | accuracy (%): 71.12 -> 73.55 | Improved |
| use RFC criterion instead of L1 within distinguishing layer pruning (ablation) | DLP (VGG16) · CIFAR-10 | accuracy (%): 93.12 -> 93.38 | Improved |
| use RFC criterion instead of NS within distinguishing layer pruning (ablation) | DLP (VGG16) · CIFAR-10 | accuracy (%): 93.25 -> 93.38 | Improved |
| use RFC criterion instead of L1 within distinguishing layer pruning (ablation) | DLP (VGG16) · CIFAR-100 | accuracy (%): 73.49 -> 73.82 | Improved |
| use RFC criterion instead of NS within distinguishing layer pruning (ablation) | DLP (VGG16) · CIFAR-100 | accuracy (%): 73.55 -> 73.82 | Improved |
| use RFC instead of NS under global threshold pruning (ablation) | GTP (VGG16) · CIFAR-10 | accuracy (%): 92.64 -> 92.25 | Hurt |
| use RFC instead of L1 under global threshold pruning (ablation) | GTP (VGG16) · CIFAR-10 | accuracy (%): 90.56 -> 92.25 | Improved |
Closest papers by meaning
- Learning Filter Pruning Criteria for Deep Convolutional Neural Networks Acceleration (CVPR 2020 · similarity 0.92)
- Rethinking the Pruning Criteria for Convolutional Neural Network (NeurIPS 2021 · similarity 0.91)
- Provable Filter Pruning for Efficient Neural Networks (ICLR 2020 · similarity 0.90)
- Pruning Filters for Efficient ConvNets (ICLR 2017 · similarity 0.90)
- Shallowing Deep Networks: Layer-Wise Pruning Based on Feature Representations (TPAMI 2018 · similarity 0.90)
- Convolutional Neural Network Pruning With Structural Redundancy Reduction (CVPR 2021 · similarity 0.89)
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.
