AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

NAS-OoD: Neural Architecture Search for Out-of-Distribution Generalization

Haoyue Bai, Fengwei Zhou, Lanqing Hong, Nanyang Ye, S.-H. Gary Chan, Zhenguo Li

ICCV 2021 · Architectures and efficiency › Neural architecture search · approach: Efficient NAS search strategies · 30 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 searched NAS-OoD architecture instead of DecAug ResNet-18ResNet-18 OoD baselines · NICO (average)accuracy: 82.67 -> 85.16Improved
use NAS-OoD instead of best OoD baselineL2A-OT / DecAug · PACS (ImageNet-pretrained)accuracy: 82.80 -> 84.23Improved
use NAS-OoD instead of best baselineMMD · OoD-FPaccuracy: 95.83 -> 98.77Improved
use NAS-OoD instead of best baseline when trained from scratchDecAug · PACS (from scratch)accuracy: 50.96 -> 53.80Improved
search architecture with conditional generator (add NAS-OoD) instead of plain NAS (ablation)DARTS · NICO (average)accuracy: 79.61 -> 85.16Improved
add IRM regularizer to NAS methods (ablation)DARTS · NICO (average)accuracy: 79.61 -> 77.26Hurt
add IRM regularizer to NAS methods (ablation)SNAS · NICO (average)accuracy: 83.00 -> 79.73Hurt
add IRM regularizer to NAS methods (ablation)ISTA-NAS · NICO (average)accuracy: 83.63 -> 80.09Hurt
add REx regularizer to NAS methods (ablation)SNAS · NICO (average)accuracy: 83.00 -> 71.23Hurt
add NAS-OoD search instead of random architecture sampling (ablation)random architecture · NICO (average)accuracy: 78.68 -> 85.16Improved
add generator cycle-consistency loss (ablation)NAS-OoD · NICO (average)accuracy: 83.86 -> 85.16Improved
add NAS-OoD architecture search instead of NAS+IRM (ablation)NAS + IRM · PACS (ImageNet-pretrained)accuracy: 78.93 -> 83.89Improved

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