AI Paper Atlas

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

Neural Relation Graph: A Unified Framework for Identifying Label Noise and Outlier Data

Jang-Hyun Kim, Sangdoo Yun, Hyun Oh Song

NeurIPS 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Uncertainty and misclassification detection · 8 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 relational graph scores instead of the strongest baseline, MarginLabel error detection · ImageNet, 8% label noise, MAE-LargeAP: 0.484 -> 0.526Improved
Use relational graph scores instead of the strongest baseline, TracInLabel error detection · ImageNet, 8% label noise, MAE-LargeTNR95: 0.521 -> 0.695Improved
Use relational graph scores instead of the strongest baselineLabel error detection · ESC-50AP: 0.739 -> 0.779Improved
Use relational graph scores instead of the strongest baselineLabel error detection · SST2AP: 0.861 -> 0.881Improved
Use relational graph scores instead of the strongest baselineLabel error detection · MNLIAP: 0.764 -> 0.758Hurt
Use relational graph scores instead of the strongest baselineLabel error detection · MNLITNR95: 0.514 -> 0.660Improved
Use relational graph scores instead of the strongest baselineRealistic label error detection · ImageNet validation with ConvNeXt-22kAP: 0.724 -> 0.744Improved
Use relational graph scores instead of KNNOOD detection · ImageNet OOD, ALL, MAE-LargeAUROC: 0.901 -> 0.911Improved
Use relational graph scores instead of KNNOOD detection · ImageNet OOD, ALL, MAE-LargeAP: 0.854 -> 0.874Improved
Use relational scores instead of KNNOutlier detection · ImageNet-100 with SUN outliersAUROC: 0.990 -> 0.993Improved
Use relational scores instead of KNNOutlier detection · ImageNet-100 with iNaturalist outliersAP: 0.923 -> 0.940Improved
Add max-cut global edge aggregation rather than simply summing node edges (ablation)Relation label-error scoring · Label error detectionTNR95: 0.628 -> 0.647Improved

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