Research map / Architectures and efficiency / CNN design
Learning Efficient Convolutional Networks Through Network Slimming
Zhuang Liu, Jianguo Li, Zhiqiang Shen, Gao Huang, Shoumeng Yan, Changshui Zhang
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 |
|---|---|---|---|
| prune 40% channels after sparsity training | DenseNet-40 · CIFAR-10 | test error (%): 6.11 -> 5.19 | Improved |
| prune 70% channels after sparsity training | VGGNet · CIFAR-10 | test error (%): 6.34 -> 6.20 | Improved |
| prune 40% channels | ResNet-164 · CIFAR-100 | test error (%): 23.37 -> 22.87 | Improved |
| prune 60% channels instead of 40% (ablation) | ResNet-164 · CIFAR-100 | test error (%): 22.87 -> 23.91 | Hurt |
| prune 40% channels | ResNet-164 · SVHN | test error (%): 1.78 -> 1.85 | No clear effect |
| prune 50% channels of VGG-A | VGG-A · ImageNet | validation error (%): 36.69 -> 36.66 | No clear effect |
| replace per-layer pruning with global threshold slimming (ablation) | VGGNet · CIFAR-10 | test error (%): 6.88 -> 6.20 | Improved |
| replace per-layer pruning with global threshold slimming (ablation) | VGGNet · CIFAR-100 | test error (%): 28.36 -> 26.52 | Improved |
| replace SSL with network slimming (ablation) | 3-layer FC · MNIST | test error (%): 1.53 -> 1.49 | Improved |
| add multi-pass pruning (iteration 5) to single pass | VGGNet · CIFAR-10 | test error (%): 6.34 -> 5.96 | Improved |
| add more multi-pass iterations beyond 3 (ablation) | VGGNet · CIFAR-100 | test error (%): 26.52 -> 29.08 | Hurt |
| prune over 80% of channels (ablation) | DenseNet-40 · CIFAR-10 | test error (%) | Hurt |
Closest papers by meaning
- Dynamic Slimmable Network (CVPR 2021 · similarity 0.89)
- Slimmable Neural Networks (ICLR 2019 · similarity 0.89)
- RePr: Improved Training of Convolutional Filters (CVPR 2019 · similarity 0.89)
- ThiNet: Pruning CNN Filters for a Thinner Net (TPAMI 2018 · similarity 0.89)
- ChannelNets: Compact and Efficient Convolutional Neural Networks via Channel-Wise Convolutions (NeurIPS 2018 · similarity 0.88)
- ThiNet: A Filter Level Pruning Method for Deep Neural Network Compression (ICCV 2017 · 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.
