AI Paper Atlas

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

On the Strong Correlation Between Model Invariance and Generalization

Weijian Deng, Stephen Gould, Liang Zheng

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 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
measure invariance with EI instead of JSinvariance measure · ImageNet-RPearson r with accuracy: -0.173 -> 0.928Improved
measure invariance with EI instead of JSinvariance measure · ImageNet-SPearson r with accuracy: 0.131 -> 0.915Improved
measure invariance with EI instead of JSinvariance measure · ImageNet-APearson r with accuracy: -0.297 -> 0.910Improved
measure invariance with EI instead of JSinvariance measure · ImageNet-ValPearson r with accuracy: -0.861 -> 0.929Improved
use (pre)training with more data instead of heavy augmentationResNet-50 · ImageNet-Aaccuracy-invariance trendImproved
use l2 distance or accuracy difference as invariance measure instead of EIinvariance measure · OOD ImageNet setscorrelation with accuracyHurt
use average confidence as invariance score instead of EIinvariance measure · ImageNetcorrelation with accuracyHurt
use prediction consistency only instead of EIinvariance measure · OOD ImageNet setscorrelation with accuracyHurt
pretrain or train semi-supervised on more datastandard models · ImageNetrotation and grayscale invariance and accuracyImproved
evaluate on style-shift ImageNet-R vs ImageNet-Val vs ImageNet-V2-Agrayscale EI · ImageNetcorrelation: 0.940 vs 0.900 vs 0.868Improved

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