Research map / Architectures and efficiency / Quantization
Trained Ternary 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 |
|---|---|---|---|
| Use TTQ instead of full-precision weights | ResNet-20 · CIFAR-10 | error rate: 8.23 -> 8.87 | Hurt |
| Use TTQ instead of full-precision weights | ResNet-32 · CIFAR-10 | error rate: 7.67 -> 7.63 | Helped |
| Use TTQ instead of full-precision weights | ResNet-44 · CIFAR-10 | error rate: 7.18 -> 7.02 | Helped |
| Use TTQ instead of full-precision weights | ResNet-56 · CIFAR-10 | error rate: 6.80 -> 6.44 | Helped |
| Use TTQ instead of TWN | AlexNet · ImageNet validation | Top1 error: 45.5% -> 42.5% | Helped |
| Use TTQ instead of full-precision weights | AlexNet · ImageNet validation | Top1 error: 42.8% -> 42.5% | Helped |
| Use TTQ instead of TWN | AlexNet · ImageNet validation | Top5 error: 23.2% -> 20.3% | Helped |
| Use TTQ instead of TWN | ResNet-18 · ImageNet validation | Top1 error: 34.7% -> 33.4% | Helped |
| Use TTQ instead of TWN | ResNet-18 · ImageNet validation | Top5 error: 13.8% -> 12.8% | Helped |
| Increase weight sparsity from 0 toward 50% (ablation) | TTQ ResNet-20 · CIFAR-10 | error rate | Helped |
| Increase weight sparsity beyond 50% (ablation) | TTQ ResNet-20 · CIFAR-10 | error rate | Hurt |
| Use a fixed threshold rather than enforcing constant sparsity (ablation) | TTQ · CIFAR-10 | error rate | Helped |
Closest papers by meaning
- TRQ: Ternary Neural Networks With Residual Quantization (AAAI 2021 · similarity 0.92)
- Simultaneously Optimizing Weight and Quantizer of Ternary Neural Network Using Truncated Gaussian Approximation (CVPR 2019 · similarity 0.90)
- RTN: Reparameterized Ternary Network (AAAI 2020 · similarity 0.90)
- ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization (NeurIPS 2025 · similarity 0.87)
- QBB: Quantization with Binary Bases for LLMs (NeurIPS 2024 · similarity 0.86)
- Efficient Multi-bit Quantization Network Training via Weight Bias Correction and Bit-wise Coreset Sampling (NeurIPS 2025 · 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.