Research map / Architectures and efficiency / Neural architecture search
Learning Architectures for Binary Networks
Dahyun Kim, Kunal Pratap Singh, Jonghyun Choi
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 |
|---|---|---|---|
| keep Zeroise layers in final architecture (ablation) | BNAS · CIFAR10 | test accuracy: 89.47 -> 92.70 | Improved |
| keep Zeroise layers in final architecture (ablation) | DARTS (floating point) · CIFAR10 | test accuracy: 97.45 -> 94.43 | Hurt |
| add diversity regularizer to search (ablation) | BNAS · CIFAR10 | test accuracy: 90.95 -> 92.70 | Improved |
| add diversity regularizer to search (ablation) | DARTS (floating point) · CIFAR10 | test accuracy: 96.53 -> 96.73 | No clear effect |
| add inter-cell skip connections (ablation) | BNAS-A · CIFAR10 | test accuracy: 61.23 -> 92.70 | Improved |
| add inter-cell skip connections (ablation) | BNAS-C · CIFAR10 | test accuracy: 70.58 -> 94.43 | Improved |
| add Zeroise layers (ablation) | BNAS-C · CIFAR10 | test accuracy: 88.74 -> 94.43 | Improved |
| add diversity regularizer (ablation) | BNAS-C · CIFAR10 | test accuracy: 92.66 -> 94.43 | Improved |
| add dilated convolutions to search space (ablation) | BNAS-A · CIFAR10 | test accuracy: 92.22 -> 92.70 | No clear effect |
| add separable convolutions to search space (ablation) | BNAS-A · CIFAR10 | test accuracy: 92.70 -> 89.66 | Hurt |
| binarize separable convolution layers (ablation) | small CNN · CIFAR10 | accuracy: 57.00 -> 10.00 | Hurt |
| keep Zeroise layers in final architecture (ablation) | BNAS-A · CIFAR10 | FLOPs: 0.36 -> 0.14 | Improved |
Closest papers by meaning
- BATS: Binary ArchitecTure Search (ECCV 2020 · similarity 0.88)
- Binarized Neural Architecture Search (AAAI 2020 · similarity 0.87)
- BN-NAS: Neural Architecture Search With Batch Normalization (ICCV 2021 · similarity 0.86)
- MathNAS: If Blocks Have a Role in Mathematical Architecture Design (NeurIPS 2023 · similarity 0.86)
- Boosting Order-Preserving and Transferability for Neural Architecture Search: a Joint Architecture Refined Search and Fine-tuning Approach (CVPR 2024 · similarity 0.84)
- A&B BNN: Add&Bit-Operation-Only Hardware-Friendly Binary Neural Network (CVPR 2024 · 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.
