AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

CLOSE: Curriculum Learning on the Sharing Extent towards Better One-Shot NAS

Zixuan Zhou, Xuefei Ning, Yi Cai, Jiashu Han, Yiping Deng, Yuhan Dong, Huazhong Yang, Yu Wang

ECCV 2022 · Architectures and efficiency › Neural architecture search · approach: One-shot and few-shot supernets

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
add WIT and SRT together (ablation)CLOSE · NAS-Bench-301Kendall's Tau: 0.1104 -> 0.5168Improved
add WIT and SRT together (ablation)CLOSE · NDS ResNetKendall's Tau: 0.6339 -> 0.6786Improved
add WIT alone (without SRT) (ablation)CLOSE · NAS-Bench-301Kendall's Tau: 0.1104 -> 0.1047No clear effect
add SRT alone (without WIT) (ablation)CLOSE · NAS-Bench-301Kendall's Tau: 0.1104 -> 0.2004Improved
use learnable GATE instead of random GLOW assignment (ablation)CLOSE · NAS-Bench-201Kendall's Tau: 0.3627 -> 0.7622Improved
use learnable GATE instead of random GLOW assignment (ablation)CLOSE · NAS-Bench-301Kendall's Tau: 0.2236 -> 0.5168Improved
gradually add GLOW blocks instead of fixed 2 blocks (ablation)CLOSE · NAS-Bench-301Kendall's Tau: 0.4533 -> 0.5168Improved
gradually add GLOW blocks instead of fixed 2 blocks (ablation)CLOSE · NAS-Bench-201Kendall's Tau: 0.7320 -> 0.7622Improved
add all blocks instead of stopping after 1 (ablation)CLOSE · NAS-Bench-301Kendall's Tau: 0.3990 -> 0.5168Improved
use more curriculums (ablation)CLOSE · NAS-Bench-301 and NDS ResNetKendall's TauImproved
use CLOSENet instead of vanilla supernet (ablation)one-shot supernet · four NAS benchmarksKendall's TauImproved
use CLOSE-S (two-stage sharing, no GATE) instead of vanilla supernet (ablation)one-shot supernet · NAS-Bench-201/301Kendall's TauImproved

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