AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Removing Undesirable Feature Contributions Using Out-of-Distribution Data

Saehyung Lee, Changhwa Park, Hyungyu Lee, Jihun Yi, Jonghyun Lee, Sungroh Yoon

ICLR 2021 · Label-efficient and robust learning › OOD and anomaly detection · approach: Out-of-distribution generalization theory · 1 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 OOD data from 80M-TI with uniform labels to TRADES adversarial trainingCIFAR10 TRADES · CIFAR10accuracy (AA): 52.83 -> 54.63Improved
Add OOD data from 80M-TI with uniform labels to TRADES adversarial trainingCIFAR100 TRADES · CIFAR100accuracy (AA): 26.91 -> 27.83Improved
Add ImgNet990 OOD data with uniform labels to PGD trainingImgNet10 PGD · ImgNet10 (64 x 64)accuracy (AA): 48.34 -> 53.83Improved
Add 80M-TI OOD data with uniform labelsCIFAR10 Standard · CIFAR10 (N=2,500)accuracy: 65.44 -> 72.49Improved
Add Simpson OOD data with uniform labelsCIFAR100 Standard · CIFAR100 (N=2,500)accuracy: 24.41 -> 27.04Improved
Add OOD data to pseudo-label trainingCIFAR10 Pseudo-label · CIFAR10 (Full)accuracy: 95.28 -> 95.53Improved
Add OOD data to pseudo-label trainingCIFAR100 Pseudo-label · CIFAR100 (Full)accuracy: 77.24 -> 77.36Improved
Increase OOD loss weight from alpha=1.0 to alpha=5.0 (ablation)OAT · CIFAR10accuracy (PGD20): 57.45 -> 59.48Improved
Increase OOD loss weight from alpha=1.0 to alpha=5.0 (ablation)OAT · CIFAR10accuracy (CW20): 52.65 -> 51.24Hurt
Use VisDA17 OOD data instead of no OOD augmentationImgNet10 PGD · ImgNet10 (64 x 64)accuracy (PGD20): 49.00 -> 55.34Improved
Use Places365 OOD data instead of no OOD augmentationCIFAR10 PGD · CIFAR10accuracy (Clean): 87.48 -> 82.37Hurt
Use OAT on full-data ImgNet10 (64 x 64), dog-breed class subsetStandard learning · ImgNet10 (64 x 64, Full)accuracy: 75.77 -> 75.17Hurt

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