Feature norm-based regularization for robust out-of-distribution detection in graph neural networks

Jiaqi Zhang

doi:10.1016/j.neucom.2026.134743

Feature norm-based regularization for robust out-of-distribution detection in graph neural networks

2026-08-07

0

PRE

AI

J

Jiaqi Zhang

DOI：10.1016/j.neucom.2026.134743

Original

Original request for help

Save

Abstract

En 中文

• We introduce feature norm regularization to enhance GNN robustness against OOD samples without altering the architecture or adding data. • We identify norm anomalies as key signals for OOD detection, improving GNN reliability in safety-critical domains. • We demonstrate through experiments that our method boosts detection accuracy and model stability across various benchmarks.

Feature norm-based regularization for robust out-of-distribution detection in graph neural networks

Abstract

Journal

Organization

Cited Papers

Citing Papers