Research map / Architectures and efficiency / Pruning and efficient tuning
AutoPrune: Automatic Network Pruning by Regularizing Auxiliary Parameters
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 |
|---|---|---|---|
| prune via regularized auxiliary parameters (AutoPrune) | MobileNetV2 · ImageNet ~100M FLOPs | top-1 accuracy: 65.40% -> 66.83% | Helped |
| AutoPrune vs Yu and Huang 2019a | MobileNetV2 · ImageNet ~300M FLOPs | top-1 accuracy: 74.20% -> 74.0% | Hurt |
| AutoPrune weight pruning | VGG-like · CIFAR-10 | compression rate: 65 -> 75 | Helped |
| AutoPrune weight pruning | LeNet5 · MNIST | compression rate: 298 -> 310 | Helped |
| AutoPrune weight pruning | AlexNet · ImageNet | compression rate: 17.7 -> 18.5 | Helped |
| proposed auxiliary-parameter update rule instead of regular BNN update (ablation) | AutoPrune · CIFAR-10 VGG-like | accuracy at high compression | Helped |
| vary auxiliary learning rate 1e-2 to 5e-1 (ablation) | AutoPrune · CIFAR-10 VGG-like | accuracy: <0.1% | No clear effect |
| Softplus STE instead of linear STE (ablation) | AutoPrune · CIFAR-10 VGG-like | compression rate | Helped |
Closest papers by meaning
- Pruning from Scratch (AAAI 2020 · similarity 0.91)
- AutoCompress: An Automatic DNN Structured Pruning Framework for Ultra-High Compression Rates (AAAI 2020 · similarity 0.91)
- Rethinking the Value of Network Pruning (ICLR 2019 · similarity 0.91)
- Auto-Train-Once: Controller Network Guided Automatic Network Pruning from Scratch (CVPR 2024 · similarity 0.90)
- GDeR: Safeguarding Efficiency, Balancing, and Robustness via Prototypical Graph Pruning (NeurIPS 2024 · similarity 0.89)
- Device-Wise Federated Network Pruning (CVPR 2024 · 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.