AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

Not All Operations Contribute Equally: Hierarchical Operation-Adaptive Predictor for Neural Architecture Search

Ziye Chen, Yibing Zhan, Baosheng Yu, Mingming Gong, Bo Du

ICCV 2021 · Architectures and efficiency › Neural architecture search · approach: Bi-level differentiable NAS · 13 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
add CGM gated fusion of layer embeddings (ablation)HOP · NAS-Bench-101 (0.05%)Kendall's Tau: 0.7773 -> 0.7819Improved
add CGM gated fusion of layer embeddings (ablation)HOP · NAS-Bench-201 (1%)Kendall's Tau: 0.7562 -> 0.7613Improved
replace CGM with mean fusion of layer embeddings (ablation)HOP · NAS-Bench-101 (0.05%)Kendall's Tau: 0.7819 -> 0.7704Hurt
replace CGM with mean fusion of layer embeddings (ablation)HOP · NAS-Bench-201 (1%)Kendall's Tau: 0.7613 -> 0.7523Hurt
increase OAM layers from 2 to 6 (ablation)HOP · NAS-Bench-101 (0.05%)Kendall's Tau: 0.6440 -> 0.7852Improved
increase OAM layers from 2 to 6 (ablation)HOP · NAS-Bench-201 (1%)Kendall's Tau: 0.5764 -> 0.7659Improved
use operation-adaptive attention (OAM) instead of GCN aggregationGATES · NAS-Bench-101 (0.05%)Kendall's Tau: 0.7634 -> 0.7773Improved
use operation-adaptive attention (OAM) instead of GCN aggregationGATES · NAS-Bench-201 (1%)Kendall's Tau: 0.7401 -> 0.7562Improved
HOP vs GATES with 0.1% training architecturesGATES · NAS-Bench-101 (0.1%)Kendall's Tau: 0.7789 -> 0.8134Improved
HOP vs GATES with 100% training dataGATES · NAS-Bench-101 (100%)Kendall's Tau: 0.9030 -> 0.9086Improved
HOP vs GATES with few training architecturesGATES · NAS-Bench-101 (0.05%)N@5: 83 -> 21Improved
HOP vs GATES with 1% training architecturesGATES · NAS-Bench-201 (1%)N@5: 19 -> 1Improved

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