Research map / Architectures and efficiency / CNN design
Binarizing MobileNet via Evolution-Based Searching
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 |
|---|---|---|---|
| search heterogeneous group counts at each convolutional layer rather than use randomized groups (ablation) | MobileNet binary modules M1 and M2 · ImageNet validation | Top-1 accuracy: about 2% | Helped |
| search heterogeneous group counts rather than use randomized groups (ablation) | MobileNet binary modules M1, M2, and M3 · ImageNet validation | FLOPs | Helped |
| use heterogeneous searched groups rather than uniform group count 4 (ablation) | MobileNet M1 · ImageNet validation | Top-1 accuracy: 60.89 -> 60.90 | No clear effect |
| use heterogeneous searched groups rather than uniform group count 16 (ablation) | MobileNet M1 · ImageNet validation | Top-1 accuracy: 58.49 -> 60.90 | Helped |
| use heterogeneous searched groups rather than uniform group count 4 (ablation) | MobileNet M2 · ImageNet validation | Top-1 accuracy: 59.59 -> 59.30 | Hurt |
| use heterogeneous searched groups rather than uniform group count 16 (ablation) | MobileNet M2 · ImageNet validation | Top-1 accuracy: 54.23 -> 59.30 | Helped |
| use heterogeneous searched groups rather than uniform group count 4 (ablation) | MobileNet M3 · ImageNet validation | Top-1 accuracy: 49.90 -> 51.06 | Helped |
| use heterogeneous searched groups rather than uniform group count 16 (ablation) | MobileNet M3 · ImageNet validation | Top-1 accuracy: 45.29 -> 51.06 | Helped |
| replace depth-wise convolution with searched group convolution | Binary MobileNet · ImageNet | Top-1 accuracy | Helped |
| retain full-precision 1x1 convolution at spatial-dimension reduction | Binary MobileNet · ImageNet | Top-1 accuracy | Helped |
Closest papers by meaning
- BD-Net: Has Depth-Wise Convolution Ever Been Applied in Binary Neural Networks? (AAAI 2026 · similarity 0.88)
- MobileODE: An Extra Lightweight Network (NeurIPS 2025 · similarity 0.87)
- Structured Binary Neural Networks for Accurate Image Classification and Semantic Segmentation (CVPR 2019 · similarity 0.87)
- MnasNet: Platform-Aware Neural Architecture Search for Mobile (CVPR 2019 · similarity 0.86)
- Binarized Neural Architecture Search (AAAI 2020 · similarity 0.86)
- Fast and Slow Gradient Approximation for Binary Neural Network Optimization (AAAI 2025 · similarity 0.84)
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.