Research map / Architectures and efficiency / Pruning and efficient tuning
Automatic Neural Network Compression by Sparsity-Quantization Joint Learning: A Constrained Optimization-Based Approach
Check 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 |
|---|---|---|---|
| jointly prune and quantize instead of quantizing alone (ablation) | Ours on CIFAR-10 ResNet-20 · CIFAR-10, ResNet-20 | compression rate: 16× -> 35.4× | Helped |
| jointly prune and quantize instead of quantizing alone (ablation) | Ours on CIFAR-10 ResNet-20 · CIFAR-10, ResNet-20 | accuracy drop: 0.00% -> 0.14% | Hurt |
| use quantization-only compression rather than ReLeQ | ResNet-20 · CIFAR-10 | accuracy drop: 0.12% -> 0.00% | Helped |
| increase compression from 462× to 836× using joint pruning and quantization (ablation) | Ours on CIFAR-10 ResNet-50 · CIFAR-10, ResNet-50 | accuracy drop: -1.25% -> 0.00% | Mixed |
| add pruning to quantization (ablation) | Ours on ImageNet MobileNet · ImageNet, MobileNet | compression rate: 16× -> 26.7× | Helped |
| add pruning to quantization (ablation) | Ours on ImageNet MobileNet · ImageNet, MobileNet | top-1 accuracy drop: 1.19% -> 4.41% | Hurt |
| automatically allocate quantization bits rather than use fixed 4-bit widths | MnasNet compression · ImageNet, MnasNet | top-1 accuracy drop: 3.14% -> 1.66% | Helped |
| automatically allocate quantization bits rather than use fixed 4-bit widths | ProxylessNAS-mobile compression · ImageNet, ProxylessNAS-mobile | top-1 accuracy drop: 3.17% -> 2.13% | Helped |
| compress to 2-bit quantization instead of using HAQ | MobileNet compression · ImageNet, MobileNet | top-1 accuracy drop: 13.76% -> 7.10% | Helped |
| compress to 3-bit quantization instead of using HAQ | MobileNet compression · ImageNet, MobileNet | top-1 accuracy drop: 3.24% -> 1.19% | Helped |
| compress AlexNet at a similar rate using Ours instead of CLIP-Q | AlexNet compression · ImageNet, AlexNet | top-1 accuracy drop: -0.70% -> -1.00% | Helped |
| compress AlexNet at a similar rate using Ours instead of Ye et al. | AlexNet compression · ImageNet, AlexNet | top-1 accuracy drop: 0.10% -> -0.08% | Helped |
Closest papers by meaning
- OPQ: Compressing Deep Neural Networks with One-shot Pruning-Quantization (AAAI 2021 · similarity 0.91)
- Effective Interplay between Sparsity and Quantization: From Theory to Practice (ICLR 2025 · similarity 0.91)
- CLIP-Q: Deep Network Compression Learning by In-Parallel Pruning-Quantization (CVPR 2018 · similarity 0.91)
- Optimal Brain Compression: A Framework for Accurate Post-Training Quantization and Pruning (NeurIPS 2022 · similarity 0.90)
- Focused Quantization for Sparse CNNs (NeurIPS 2019 · similarity 0.90)
- Automatic Joint Structured Pruning and Quantization for Efficient Neural Network Training and Compression (CVPR 2025 · similarity 0.90)
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.