<ul><li>STOOD-X is a two-stage methodology for Out-of-Distribution (OOD) detection in machine learning.</li><li>The first stage of STOOD-X uses feature-space distances and a nonparametric test (Wilcoxon-Mann-Whitney) to identify OOD samples without assuming a specific feature distribution.</li><li>The second stage of STOOD-X generates user-friendly, concept-based visual explanations to reveal the features driving each decision.</li><li>STOOD-X achieves competitive performance in high-dimensional and complex settings and enhances human oversight, bias detection, and model debugging.</li></ul>

STOOD-X methodology: using statistical nonparametric test for OOD Detection Large-Scale datasets enhanced with explainability

Discover more