Research map / Architectures and efficiency / Quantization
Data-Free Network Compression via Parametric Non-Uniform Mixed Precision Quantization
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 |
|---|---|---|---|
| Use a tuned non-uniform grid instead of a uniform grid (ablation) | PNMQ · ImageNet, ResNet-50 | Top-1 accuracy: 74.17% -> 75.32% | Helped |
| Use a tuned non-uniform grid instead of a uniform grid (ablation) | PNMQ · ImageNet, MobileNet-v2 | Top-1 accuracy: 69.73% -> 70.35% | Helped |
| Tune the quantization scale (ablation) | DFQ · ImageNet, ResNet-50 | Top-1 accuracy: 32.08% -> 46.60% | Helped |
| Tune the quantization scale (ablation) | DFQ · ImageNet, MobileNet-v2 | Top-1 accuracy: 55.31% -> 64.14% | Helped |
| Use a small calibration set to optimize quantization parameters rather than no data (ablation) | PNMQ · ImageNet, ResNet-50 | Top-1 accuracy: 75.32% -> 75.50% | Helped |
| Use a small calibration set to optimize quantization parameters rather than no data (ablation) | PNMQ · ImageNet, MobileNet-v2 | Top-1 accuracy: 70.35% -> 70.64% | Helped |
| Use mixed bitwidths rather than a fixed 5-bit width (ablation) | PNMQ / Lloyd-Max quantization · ImageNet, ResNet-50 | Top-1 accuracy: 73.77% -> 75.56% | Helped |
| Use mixed bitwidths rather than a fixed 5-bit width (ablation) | PNMQ / Lloyd-Max quantization · ImageNet, MobileNet-v2 | Top-1 accuracy: 68.62% -> 70.93% | Helped |
| Compress weights with data-free PNMQ rather than 6-bit DFQ | Faster R-CNN and Mask R-CNN · COCO-2017 | Bbox mAP: 0.350 -> 0.369 | Helped |
| Compress weights with data-aware PNMQ rather than 6-bit DFQ | Faster R-CNN and Mask R-CNN · COCO-2017 | Bbox mAP: 0.350 -> 0.371 | Helped |
Closest papers by meaning
- Deep Neural Network Compression With Single and Multiple Level Quantization (AAAI 2018 · similarity 0.89)
- OPQ: Compressing Deep Neural Networks with One-shot Pruning-Quantization (AAAI 2021 · similarity 0.89)
- Towards the Limit of Network Quantization (ICLR 2017 · similarity 0.88)
- Quantization without Tears (CVPR 2025 · similarity 0.86)
- Retraining-Free Model Quantization via One-Shot Weight-Coupling Learning (CVPR 2024 · similarity 0.86)
- Data-Free Quantization via Pseudo-label Filtering (CVPR 2024 · similarity 0.86)
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.