Research map / Architectures and efficiency / Neural architecture search
HourNAS: Extremely Fast Neural Architecture Search Through an Hourglass Lens
Zhaohui Yang, Yunhe Wang, Xinghao Chen, Jianyuan Guo, Wei Zhang, Chao Xu, Chunjing Xu, Dacheng Tao, Chang Xu
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 |
|---|---|---|---|
| Prioritize vital blocks during search (ablation) | HourNAS on FBNet search space · ImageNet validation | Top-1 accuracy: 73.2 -> 74.1 | Improved |
| Prioritize vital blocks during search (ablation) | HourNAS on enlarged FBNet search space · ImageNet validation | Top-1 accuracy: 73.5 -> 74.1 | Improved |
| Use Gumbel softmax instead of Gumbel-Max (ablation) | HourNAS-FBNetSS-A · ImageNet | Top-1 accuracy: 74.1 -> 74.2 | Improved |
| Use Gumbel softmax instead of Gumbel-Max (ablation) | HourNAS-FBNetSS-A · ImageNet search | Search cost (GPU days): 0.1 -> 1.0 | Hurt |
| Increase the number of space proposals from one to eight (ablation) | HourNAS · CIFAR-10 test | Test error: 3.86 -> 3.39 | Improved |
| Increase the number of space proposals from eight to sixteen (ablation) | HourNAS · CIFAR-10 test | Test error: 3.39 -> 3.37 | Improved |
| Enlarge the FBNet search space by adding kernel size 7 and removing group 2 blocks | HourNAS · ImageNet | Top-1 accuracy: 75.0 -> 75.3 | Improved |
| Use HourNAS search instead of MnasNet-A1 search | MnasNet search space · ImageNet | Top-1 accuracy: 75.2 -> 75.7 | Improved |
| Use HourNAS search instead of EfficientNet-B0 search | EfficientNet search space · ImageNet | Top-1 accuracy: 76.8 -> 77.0 | Improved |
Closest papers by meaning
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.90)
- MathNAS: If Blocks Have a Role in Mathematical Architecture Design (NeurIPS 2023 · similarity 0.90)
- AtomNAS: Fine-Grained End-to-End Neural Architecture Search (ICLR 2020 · similarity 0.90)
- Fast and Practical Neural Architecture Search (ICCV 2019 · similarity 0.89)
- DDPNAS: Efficient Neural Architecture Search via Dynamic Distribution Pruning (IJCV 2023 · similarity 0.89)
- DA-NAS: Data Adapted Pruning for Efficient Neural Architecture Search (ECCV 2020 · similarity 0.89)
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.
