Research map / Architectures and efficiency / CNN design
From Hashing to CNNs: Training Binary Weight Networks via Hashing
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 BWNH instead of Full-Precision | VGG-9 · CIFAR10 | test error rate: 9.01 -> 9.21 | Hurt |
| Use BWNH instead of SQ-BWN | VGG-9 · CIFAR10 | test error rate: 9.40 -> 9.21 | Helped |
| Use BWNH instead of Full-Precision | VGG-9 · CIFAR100 | test error rate: 30.45 -> 34.35 | Hurt |
| Use BWNH instead of SQ-BWN | VGG-9 · CIFAR100 | test error rate: 35.25 -> 34.35 | Helped |
| Use BWNH instead of BWN | AlexNet · ImageNet | Top1 accuracy: 56.8 -> 58.5 | Helped |
| Use BWNH instead of BWN | AlexNet · ImageNet | Top5 accuracy: 79.4 -> 80.9 | Helped |
| Use BWNH instead of HWGQ-BWN | ResNet-18 · ImageNet | Top1 accuracy: 61.3 -> 64.3 | Helped |
| Use BWNH instead of HWGQ-BWN | ResNet-18 · ImageNet | Top5 accuracy: 83.9 -> 85.9 | Helped |
| Add a scaling factor to binary hash codes (ablation) | BWNH · CIFAR10 and CIFAR100 | classification accuracy | Helped |
| Fine-tune the binarized network after hashing-based initialization (ablation) | BWNH · VGG-9, AlexNet, and ResNet-18 | Top1 accuracy | Helped |
| Adapt each layer's binary codes to featuremaps produced by previously binarized layers (ablation) | BWNH · CIFAR10 and CIFAR100 | classification accuracy | Helped |
Closest papers by meaning
- Deep Supervised Hashing for Fast Image Retrieval (CVPR 2016 · similarity 0.90)
- Greedy Hash: Towards Fast Optimization for Accurate Hash Coding in CNN (NeurIPS 2018 · similarity 0.89)
- Towards Accurate Binary Convolutional Neural Network (NeurIPS 2017 · similarity 0.89)
- Network Sketching: Exploiting Binary Structure in Deep CNNs (CVPR 2017 · similarity 0.88)
- Deep Hashing Network for Efficient Similarity Retrieval (AAAI 2016 · similarity 0.88)
- Deep Supervised Discrete Hashing (NeurIPS 2017 · 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.