AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

MeCo: Zero-Shot NAS with One Data and Single Forward Pass via Minimum Eigenvalue of Correlation

Tangyu Jiang, Haodi Wang, Rongfang Bie

NeurIPS 2023 · Architectures and efficiency › Neural architecture search · approach: One-shot and few-shot supernets · 5 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
increase fixed channel number n from 4 to 8 (ablation)MeCoopt · NATS-Bench-TSS CIFAR-10Spearman rho: 0.87 -> 0.90Improved
increase fixed channel number n from 4 to 8 (ablation)MeCoopt · NATS-Bench-SSS ImageNet16-120Spearman rho: 0.84 -> 0.89Improved
increase fixed channel number n from 4 to 8 (ablation)MeCoopt · NATS-Bench-TSS CIFAR-100Spearman rho: 0.88 -> 0.89Improved
apply MeCoopt channel optimization on top of MeCo (ablation)MeCo · NATS-Bench-SSS CIFAR-100Spearman rho: -0.87 -> 0.83Improved
use random Gaussian data instead of real data (ablation)MeCo · NATS-Bench-TSS CIFAR-10Spearman rho: 0.894 -> 0.87No clear effect
use MeCo instead of NTK under random dataNTK · NATS-Bench-TSS CIFAR-10Spearman rho: 0.78 -> 0.87Improved
vary input sample numbers from 1 to 5 (ablation)MeCo · NATS-Bench-TSS and SSSSpearman rho: fluctuates around 0.01No clear effect
increase random data size from 16 to 256 (ablation)MeCo · NATS-Bench-TSSSpearman rhoMixed
use MeCo instead of ZiCo as proxy in Zero-Cost-PTZiCo · DARTS-CNN CIFAR-10test error (%): 2.80 ± 0.03 -> 2.69± 0.05Improved
use MeCo instead of ZiCo as proxy in Zero-Cost-PTZiCo · DARTS-CNN CIFAR-100test error (%): 19.54 ± 0.28 -> 19.33 ± 0.25Improved
use MeCo instead of ZiCoZiCo · MobileNet-V3 ImageNet-1KTop-1 accuracy (%): 77.1 -> 77.8Improved
use MeCo instead of ZiCoZiCo · NAS-Bench-301 CIFAR-10Spearman rho: 0.66 -> 0.7±0.01Improved

Later papers that compared with this method

1 later accepted paper reports MeCo as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

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