Research map / Architectures and efficiency / Pruning and efficient tuning
An Information Theory-Inspired Strategy for Automated Network Pruning
Xiawu Zheng, Yuexiao Ma, Teng Xi, Gang Zhang, Errui Ding, Yuchao Li, Jie Chen, Yonghong Tian, Rongrong Ji
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 |
|---|---|---|---|
| integrate FPGM with ITPruner (ablation) | ITPruner ResNet-56 · CIFAR-10 | top-1 accuracy: 93.43 -> 94.05 | Improved |
| use ITPruner layer-wise allocation instead of uniform allocation (ablation) | VGG with L1 metric · CIFAR-10 | top-1 accuracy: 92.92 -> 94.00 | Improved |
| use ITPruner allocation instead of uniform with Snip metric (ablation) | VGG · CIFAR-10 | top-1 accuracy: 92.67 -> 93.93 | Improved |
| swap local metric from L1 to Synflow (ablation) | ITPruner VGG · CIFAR-10 | top-1 accuracy: 94.00 -> 93.71 | No clear effect |
| swap local metric from L1 to Snip (ablation) | ITPruner VGG · CIFAR-10 | top-1 accuracy: 94.00 -> 93.93 | No clear effect |
| add mutual-information regularization to ResNet-20 training (ablation) | ResNet-20 · CIFAR-10 | top-1 accuracy: 89.65 -> 92.01 | Improved |
| add MI regularization on top of weight decay (ablation) | ResNet-20 · CIFAR-10 | top-1 accuracy: 92.13 -> 92.15 | No clear effect |
| replace evolutionary-search architecture with ITPruner architecture (ablation) | VGG · CIFAR-10 | top-1 accuracy: 0.19 | No clear effect |
| compare ITPruner with HRank at similar FLOPs reduction | VGG · CIFAR-10 | top-1 accuracy: 92.34 -> 94.00 | Improved |
| use ITPruner instead of AMC | MobileNetV1 · ImageNet | top-1 accuracy drop: 1.3 -> 0.8 | Improved |
| add distillation and extra tricks to ITPruner (ablation) | ITPruner ResNet-50 · ImageNet | top-1 accuracy: 75.75 -> 78.05 | Improved |
| use ITPruner instead of NViT | DeiT-S · ImageNet | top-1 accuracy: 82.19 -> 82.50 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HRank — HRank: Filter Pruning Using High-Rank Feature Map (CVPR 2020) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Automatic Network Pruning via Hilbert-Schmidt Independence Criterion Lasso under Information Bottleneck Principle (ICCV 2023 · similarity 0.92)
- Exploration and Estimation for Model Compression (ICCV 2021 · similarity 0.90)
- Discrimination-aware Network Pruning for Deep Model Compression (TPAMI 2021 · similarity 0.90)
- A Probabilistic Approach to Neural Network Pruning (ICML 2021 · similarity 0.90)
- Pruning from Scratch (AAAI 2020 · similarity 0.89)
- 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.
