Research map / Architectures and efficiency / Quantization
Deep Neural Network Compression With Single and Multiple Level Quantization
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 |
|---|---|---|---|
| Apply 5-bit SLQ | Light CNN · CIFAR-10 | accuracy: 78.66$\%$ -> 81.11$\%$ | Helped |
| Apply 5-bit SLQ | ResNet20 · CIFAR-10 | accuracy: 91.70$\%$ -> 91.75$\%$ | Helped |
| Use 5-bit SLQ instead of full-precision weights | AlexNet · ImageNet | Top-1 accuracy: 57.10$\%$ -> 57.56$\%$ | Helped |
| Use 5-bit SLQ instead of 5-bit INQ | AlexNet · ImageNet | Top-1 accuracy: 57.39$\%$ -> 57.56$\%$ | Helped |
| Use 5-bit SLQ instead of 5-bit INQ | VGG16 · ImageNet | Top-1 accuracy: 70.82$\%$ -> 72.23$\%$ | Helped |
| Use 5-bit SLQ instead of full-precision weights | VGG16 · ImageNet | Top-1 accuracy: 68.54$\%$ -> 72.23$\%$ | Helped |
| Use 5-bit SLQ instead of full-precision weights | GoogleNet · ImageNet | Top-1 accuracy: 68.89$\%$ -> 69.10$\%$ | Helped |
| Use 5-bit SLQ instead of full-precision weights | ResNet18 · ImageNet | Top-1 accuracy: 68.27$\%$ -> 69.09$\%$ | Helped |
| Reduce SLQ bit-width from 5 to 3 (ablation) | VGG16 SLQ · ImageNet | Top-1 accuracy: 72.23$\%$ -> 68.38$\%$ | Hurt |
| Reduce SLQ bit-width from 5 to 4 (ablation) | VGG16 SLQ · ImageNet | Top-1 accuracy: 72.23$\%$ -> 71.18$\%$ | Hurt |
| Use linear rather than non-linear centroid initialization (ablation) | VGG16 SLQ · ImageNet | Top-1 accuracy: 72.23$\%$ -> 71.85$\%$ | Hurt |
| Constrain centroids to powers of two or zero with ESLQ | VGG16 · ImageNet | Top-1 accuracy: 68.54$\%$ -> 71.95$\%$ | Helped |
Closest papers by meaning
- Towards the Limit of Network Quantization (ICLR 2017 · similarity 0.90)
- OPQ: Compressing Deep Neural Networks with One-shot Pruning-Quantization (AAAI 2021 · similarity 0.90)
- Adaptive Loss-Aware Quantization for Multi-Bit Networks (CVPR 2020 · similarity 0.90)
- Automatic Neural Network Compression by Sparsity-Quantization Joint Learning: A Constrained Optimization-Based Approach (CVPR 2020 · similarity 0.89)
- Bit-Shrinking: Limiting Instantaneous Sharpness for Improving Post-Training Quantization (CVPR 2023 · similarity 0.89)
- Quantization without Tears (CVPR 2025 · 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.