AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

LA3: Efficient Label-Aware AutoAugment

Mingjun Zhao, Shan Lu, Zixuan Wang, Xiaoli Wang, Di Niu

ECCV 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory

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 label-aware policies (ablation)LA3 · CIFAR-100 WRN-40-2top-1 accuracy (%): 80.08 -> 81.09Improved
add label-aware policies (ablation)LA3 · CIFAR-10 WRN-28-10top-1 accuracy (%): 97.11 -> 97.80Improved
add mRMR policy construction instead of top-100 (ablation)LA3 · CIFAR-100 WRN-40-2top-1 accuracy (%): 78.57 -> 81.09Improved
add mRMR policy construction instead of top-100 (ablation)LA3 · CIFAR-10 WRN-28-10top-1 accuracy (%): 97.49 -> 97.80Improved
add mRMR policy construction instead of top-500 (ablation)LA3 · CIFAR-100 WRN-28-10top-1 accuracy (%): 84.04 -> 84.54Improved
increase top-k from 100 to 500 (ablation)LA3 top-k variant · CIFAR-100 WRN-40-2top-1 accuracy (%): 78.57 -> 79.85Improved
increase top-k from 100 to 500 (ablation)LA3 top-k variant · CIFAR-10 WRN-28-10top-1 accuracy (%): 97.49 -> 97.26Hurt
add batch augment (x4)LA3 · ImageNet ResNet-50top-1 accuracy (%): 78.71 -> 79.97Improved
use LA3 policy instead of FastAAResNet-50 · ImageNet search costsearch cost (GPU hours): 450 -> 29.3 GPU hoursImproved
use LA3 policy instead of DADAResNet-50 · ImageNettop-1 accuracy (%): 77.5 -> 78.71Improved
use LA3 searched on WRN-40-2 instead of AutoAugmentShake-Shake / PyramidNet · CIFAR-100 PyramidNet+ShakeDroptop-1 accuracy (%): 89.3 -> 89.02Hurt

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