AI Paper Atlas

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

ICCV 2023 · Architectures and efficiency › Neural architecture search · approach: One-shot and few-shot supernets · 9 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use xi-GSNR proxy instead of ZiCo or GradSign for ranking in NAS-Bench-201Zero-Shot NAS proxy · NAS-Bench-201 CIFAR-10Spearman rho: 0.808 -> 0.845Improved
Use xi-GSNR instead of parameter count in NAS-Bench-101Zero-Shot NAS proxy · NAS-Bench-101 CIFAR-10Spearman rho: 0.423 -> 0.615Improved
Search with xi-GSNR instead of TE-NAS in NAS-Bench-201Pruning-based zero-shot search · NAS-Bench-201 ImageNet-16-120test accuracy (%): 42.38 -> 46.24Improved
Search in DARTS space with xi-GSNR instead of TE-NASPruning-based zero-shot search · DARTS ImageNettop-1 accuracy (%): 75.5 -> 75.5No clear effect
Add xi to the GSNR denominator (ablation)GSNR proxy · NAS-Bench-201 ImageNet-16-120Spearman rho: 0.764 -> 0.828Improved
Use gradient mean over variance (GSNR) instead of squared mean only (ablation)GSNR proxy · NAS-Bench-201 ImageNet-16-120Spearman rho: 0.556 -> 0.764Improved
Use inverse gradient variance only as proxy (ablation)GSNR proxy · NAS-Bench-201 ImageNet-16-120Spearman rho: 0.828 -> 0.200Hurt
Set batch size to 64 for proxy computation (ablation)xi-GSNR · NAS-Bench-201 (200 archs)ranking consistencyImproved
Use 8 batches (N=8) to compute the proxy (ablation)xi-GSNR · NAS-Bench-201 (200 archs)ranking consistencyImproved
Search ProxylessNAS space with xi-GSNREvolutionary search · ImageNet ProxylessNAS spacetop-1 accuracy (%): 77.5 -> 78.2Improved

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.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy