Research map / Architectures and efficiency / CNN design
Rethinking the Pruning Criteria for Convolutional Neural Network
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 |
|---|---|---|---|
| choose filters with l1 vs l2 vs Fermat vs GM criteria (layer-wise) | ResNet56 · CIFAR10/100 | pruned/finetuned accuracy | No clear effect |
| use l1 vs l2 criterion in global pruning | VGG16 · ImageNet-style pruning similarity | Spearman rank correlation | Mixed |
| randomly prune among filters with near-identical l2 scores | VGG16 · CIFAR100 | accuracy gap max-min: 1.24 | Hurt |
| randomly prune among filters with near-identical l2 scores (20%) | VGG19 · CIFAR100 | accuracy gap max-min: 1.91 | Hurt |
| increase number of filters in second conv layer from 64 to 256 (ablation) | VGG16 · CIFAR10 | Spearman rank correlation | Helped |
Closest papers by meaning
- Convolutional Neural Network Pruning With Structural Redundancy Reduction (CVPR 2021 · similarity 0.93)
- Learning Filter Pruning Criteria for Deep Convolutional Neural Networks Acceleration (CVPR 2020 · similarity 0.92)
- Pruning Filters for Efficient ConvNets (ICLR 2017 · similarity 0.92)
- Reborn Filters: Pruning Convolutional Neural Networks with Limited Data (AAAI 2020 · similarity 0.91)
- Entropy Induced Pruning Framework for Convolutional Neural Networks (AAAI 2024 · similarity 0.89)
- BilevelPruning: Unified Dynamic and Static Channel Pruning for Convolutional Neural Networks (CVPR 2024 · 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.