Research map / Architectures and efficiency / Quantization
Learnable Lookup Table for Neural Network 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 learnable lookup-table quantization instead of QIL | ResNet-20 · CIFAR-10, 4/4-bit | Top-1 accuracy (%): 91.69 -> 92.17 | Helped |
| Use learnable lookup-table quantization instead of DAQ | RDN · B100, ×4 SR, 4-bit | PSNR: 27.38 -> 27.66 | Helped |
| Add exponential formulation for scale parameters (ablation) | LLT with ResNet-20 · CIFAR-10, 2/2-bit | Top-1 accuracy (%): 89.81 -> 90.10 | Helped |
| Add gradient rescaling (ablation) | LLT with ResNet-20 and exponential scale formulation · CIFAR-10, 2/2-bit | Top-1 accuracy (%): 90.10 -> 90.63 | Helped |
| Increase lookup-table granularity from K=1 to K=9 (ablation) | LLT with ResNet-20 · CIFAR-10, 2/2-bit | Top-1 accuracy (%): 90.42 -> 90.63 | Helped |
| Increase lookup-table granularity from K=9 to K=13 (ablation) | LLT with ResNet-20 · CIFAR-10, 2/2-bit | Top-1 accuracy (%): 90.63 -> 90.62 | No clear effect |
| Use lookup-table activation quantization instead of QIL | ResNet-18 · ImageNet validation | Activation quantization time (ms): 90ms -> 20ms | Helped |
Closest papers by meaning
- Quantization Networks (CVPR 2019 · similarity 0.88)
- Network Quantization With Element-Wise Gradient Scaling (CVPR 2021 · similarity 0.88)
- Relaxed Quantization for Discretized Neural Networks (ICLR 2019 · similarity 0.88)
- Up or Down? Adaptive Rounding for Post-Training Quantization (ICML 2020 · similarity 0.86)
- Scalable methods for 8-bit training of neural networks (NeurIPS 2018 · similarity 0.86)
- Efficient Quantization for Neural Networks with Binary Weights and Low Bitwidth Activations (AAAI 2019 · 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.