Research map / Architectures and efficiency / Neural architecture search
Unleashing the Power of Gradient Signal-to-Noise Ratio for Zero-Shot NAS
Zihao Sun, Yu Sun, Longxing Yang, Shun Lu, Jilin Mei, Wenxiao Zhao, Yu Hu
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 |
|---|---|---|---|
| Use xi-GSNR proxy instead of ZiCo or GradSign for ranking in NAS-Bench-201 | Zero-Shot NAS proxy · NAS-Bench-201 CIFAR-10 | Spearman rho: 0.808 -> 0.845 | Improved |
| Use xi-GSNR instead of parameter count in NAS-Bench-101 | Zero-Shot NAS proxy · NAS-Bench-101 CIFAR-10 | Spearman rho: 0.423 -> 0.615 | Improved |
| Search with xi-GSNR instead of TE-NAS in NAS-Bench-201 | Pruning-based zero-shot search · NAS-Bench-201 ImageNet-16-120 | test accuracy (%): 42.38 -> 46.24 | Improved |
| Search in DARTS space with xi-GSNR instead of TE-NAS | Pruning-based zero-shot search · DARTS ImageNet | top-1 accuracy (%): 75.5 -> 75.5 | No clear effect |
| Add xi to the GSNR denominator (ablation) | GSNR proxy · NAS-Bench-201 ImageNet-16-120 | Spearman rho: 0.764 -> 0.828 | Improved |
| Use gradient mean over variance (GSNR) instead of squared mean only (ablation) | GSNR proxy · NAS-Bench-201 ImageNet-16-120 | Spearman rho: 0.556 -> 0.764 | Improved |
| Use inverse gradient variance only as proxy (ablation) | GSNR proxy · NAS-Bench-201 ImageNet-16-120 | Spearman rho: 0.828 -> 0.200 | Hurt |
| Set batch size to 64 for proxy computation (ablation) | xi-GSNR · NAS-Bench-201 (200 archs) | ranking consistency | Improved |
| Use 8 batches (N=8) to compute the proxy (ablation) | xi-GSNR · NAS-Bench-201 (200 archs) | ranking consistency | Improved |
| Search ProxylessNAS space with xi-GSNR | Evolutionary search · ImageNet ProxylessNAS space | top-1 accuracy (%): 77.5 -> 78.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ZiCo — ZiCo: Zero-shot NAS via inverse Coefficient of Variation on Gradients (ICLR 2023) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Zero-Shot Neural Architecture Search: Challenges, Solutions, and Opportunities (TPAMI 2024 · similarity 0.92)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.91)
- Generalizing Few-Shot NAS with Gradient Matching (ICLR 2022 · similarity 0.91)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 2025 · similarity 0.89)
- ParZC: Parametric Zero-Cost Proxies for Efficient NAS (AAAI 2025 · similarity 0.89)
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · similarity 0.88)
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.
