Research map / Architectures and efficiency / Quantization
DBQ: A Differentiable Branch Quantizer for Lightweight Deep Neural Networks
Hassan Dbouk, Hetul Sanghvi, Mahesh Mehendale, Naresh Shanbhag
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add a second ternary branch to the pointwise layers (ablation) | DBQ on MobileNetV1 · ImageNet | top-1 accuracy: 66.45 -> 71.09 | Improved |
| replace ReLU6 clipping with BN-based clipping (8b fixed point) (ablation) | MobileNetV1 FX8 · ImageNet | top-1 accuracy: 71.60 -> 71.86 | Improved |
| replace ReLU6 clipping with BN-based clipping (ablation) | DBQ-2T · ImageNet | top-1 accuracy: 70.25 -> 70.80 | Improved |
| quantize first layer, depthwise layers, FC and activations to 8b (ablation) | DBQ-2T MobileNetV1 · ImageNet | top-1 accuracy: 70.80 -> 70.92 | No clear effect |
| ternarize pointwise layers with one branch (ablation) | MobileNetV1 · ImageNet | top-1 accuracy: 72.12 -> 66.45 | Hurt |
| use two ternary branches instead of two binary branches | ResNet-20 DBQ vs LQNet · CIFAR-10 | accuracy: 90.10 -> 91.06 | Improved |
| use two ternary branches instead of three binary branches | ResNet-20 DBQ-2T vs LQNet-3B · CIFAR-10 | accuracy: 92.00 -> 91.93 | No clear effect |
| compare DBQ-2T against IAO 8b quantization | MobileNetV1 · ImageNet | top-1 accuracy: 69.00 -> 70.92 | Improved |
| compare DBQ-2T against RQ 5b quantization | MobileNetV1 · ImageNet | top-1 accuracy: 61.50 -> 70.92 | Improved |
| apply 2T quantization of PW layers to MobileNetV2 | MobileNetV2 · ImageNet | top-1 accuracy: 71.88 -> 70.54 | Hurt |
| apply 2T quantization of PW layers to ShuffleNetV2 | ShuffleNetV2 · ImageNet | top-1 accuracy: 69.36 -> 66.74 | Hurt |
Closest papers by meaning
- Differentiable Soft Quantization: Bridging Full-Precision and Low-Bit Neural Networks (ICCV 2019 · similarity 0.89)
- LQ-Nets: Learned Quantization for Highly Accurate and Compact Deep Neural Networks (ECCV 2018 · similarity 0.89)
- InfoQ: Mixed-Precision Quantization via Global Information Flow (AAAI 2026 · similarity 0.89)
- AutoQ: Automated Kernel-Wise Neural Network Quantization (ICLR 2020 · similarity 0.88)
- Deep Neural Network Quantization via Layer-Wise Optimization Using Limited Training Data (AAAI 2019 · similarity 0.88)
- ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization (NeurIPS 2025 · similarity 0.87)
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.
