Unsupervised Hypergraph Neural Network for Hyperedge Anomaly Detection

A new end-to-end hypergraph neural network detects anomalous higher-order associations in hypergraphs without requiring labeled data. The unsupervised approach allows for identifying unusual multi-entity relationships that conventional pairwise graph structures cannot capture.