Research map / Architectures and efficiency / Quantization
Fixed Point Quantization of Deep Convolutional Networks
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 |
|---|---|---|---|
| Optimize convolutional-layer bit-widths across layers instead of assigning equal bit-widths | CIFAR-10 DCN · CIFAR-10 | model size at the same error rate: >20% reduction | Helped |
| Apply cross-layer bit-width optimization | AlexNet-like DCN · ImageNet-1000 | convolutional-layer model size: upto 1Mbits | Helped |
| Fine-tune after quantizing weights and activations to 4-bit (ablation) | CIFAR-10 DCN · CIFAR-10 | error rate: 8.30 | Hurt |
| Use floating-point weights and 8-bit activations rather than floating-point weights and activations (ablation) | CIFAR-10 DCN · CIFAR-10 | error rate: 6.98 -> 6.78 | Helped |
| Use 4-bit activations rather than floating-point activations with floating-point weights (ablation) | CIFAR-10 DCN · CIFAR-10 | error rate: 6.98 -> 7.44 | Hurt |
| Use 4-bit rather than 8-bit weights with 8-bit activations (ablation) | CIFAR-10 DCN · CIFAR-10 | error rate: 6.95 -> 7.58 | Hurt |
Closest papers by meaning
- Fixed-Point Back-Propagation Training (CVPR 2020 · similarity 0.88)
- F8Net: Fixed-Point 8-bit Only Multiplication for Network Quantization (ICLR 2022 · similarity 0.87)
- LEARNED STEP SIZE QUANTIZATION (ICLR 2020 · similarity 0.87)
- Incremental Network Quantization: Towards Lossless CNNs with Low-precision Weights (ICLR 2017 · similarity 0.86)
- DQT: Dynamic Quantization Training via Dequantization-Free Nested Integer Arithmetic (AAAI 2026 · similarity 0.86)
- Efficient Multi-bit Quantization Network Training via Weight Bias Correction and Bit-wise Coreset Sampling (NeurIPS 2025 · similarity 0.85)
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.