AI Paper Atlas

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

Supervision Adaptation Balancing In-Distribution Generalization and Out-of-Distribution Detection

Zhilin Zhao, Longbing Cao, Kun-Yu Lin

TPAMI 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Provable and feature-based OOD detection · 11 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 supervision adaptation (SA) with OOD training samplesBaseline ResNet18 · CIFAR10AUROC (%): 91.4 -> 98.9Improved
add SABaseline ResNet18 · CIFAR100AUROC (%): 80.1 -> 96.3Improved
use SA instead of FIGFIG · CIFAR10AUROC (%): 97.1 -> 98.9Improved
add SA to MobileNetV2Baseline MobileNetV2 · CIFAR10AUROC (%): 87.9 -> 97.6Improved
add SA (ID accuracy)Baseline ResNet18 · CIFAR100accuracy (%): 77.7 -> 78.6Improved
add outlier exposure (OE) with manual labelsBaseline ResNet18 · CIFAR10accuracy (%): 95.0 -> 94.9Hurt
add domain adaptation method DANNBaseline ResNet18 · CIFAR10AUROC (%): 91.4 -> 87.1Hurt
add label smoothing (LS)Baseline ResNet18 · CIFAR10AUROC (%): 91.4 -> 86.7Hurt
train SA with LSUN instead of TinyImageNet as OOD dataSA ResNet18 · CIFAR10AUROC (%): 98.1 -> 98.9Improved
train SA with CelebA instead of TinyImageNet as OOD dataSA ResNet18 · CIFAR10AUROC (%): 98.1 -> 96.1Hurt
increase component parameter epsilon (OOD proportion) (ablation)SA ResNet18 · CIFAR10AUROC (%)Improved
add MSMI to MBCE-only training (full SA vs MBCE) (ablation)SA ResNet18 · CIFAR100accuracy (%)Improved

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