Feature Transformation Enhanced Jacobi Polynomial Graph Filtering for Graph Anomaly Detection

The paper introduces JPGFN, a graph anomaly detection method that enhances frequency-domain filtering with a Feature Separation Transformation Network to capture fine-grained node features, an adaptive Jacobi polynomial graph filtering module to better model complex frequency-domain characteristics, and a node label constraint module to leverage label information. These components address limitations of static filters, attribute importance neglect, and insufficient label use found in existing approaches. Experiments on real-world datasets show that JPGFN outperforms mainstream methods.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.