Dimensions overall score 7.0

• NeiGAD is a plug-and-play module that enhances existing graph anomaly detection methods by leveraging neighbor information encoded in spectral components.method 95%
• Eigenvectors of the adjacency matrix inherently encode neighbor information and linearly amplify anomaly signals.technical 90%
• NeiGAD consistently improves detection accuracy over state-of-the-art GAD methods across eight real-world datasets.result 85%
• Recent GNN-based methods often fail to explicitly model neighbor information or its interaction with attributes, limiting detection performance.limitation 80%
• The neighbor information anomaly score for a node can be computed as the Frobenius norm between an eigenvector and its reconstructed version.technical 90%
• The combined loss when incorporating eigenvectors with node features decomposes additively into feature reconstruction loss and individual neighbor information anomaly scores.technical 85%
• NeiGAD can be integrated into existing methods like DOMINANT with minimal overhead.method 80%
• Spectral analysis provides an effective approach for explicit neighbor modeling in graph anomaly detection.result 85%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card