Exploring Advanced Energy-Based Modeling Techniques for Comprehensive and Effective Graph Anomaly Detection

EasyChair Preprint 15617

Abstract

Detecting anomalies in complex networks has emerged as a critical area of focus within the realm of graph analysis. This research presents an innovative and comprehensive framework that harnesses the power of Energy-Based Models (EBMs) to efficiently detect anomalies within graph-structured data. By integrating the strengths of graph neural networks (GNNs) with EBMs, the proposed methodology captures intricate structural, relational, and feature-level patterns, enabling highly accurate and reliable anomaly detection. Extensive experiments conducted on well-established benchmark datasets highlight the exceptional performance of the proposed approach, surpassing state-of-the-art techniques, and underscoring its robustness, scalability, and overall effectiveness in real-world applications.

Keyphrases: Graph Neural Networks, energy-based models, graph anomaly detection, machine learning

@booklet{EasyChair:15617,
  author    = {James Walker and Carlos Lopez and Camila Costa and Pedro Silva and Kin Elvard},
  title     = {Exploring Advanced Energy-Based Modeling Techniques for Comprehensive and Effective Graph Anomaly Detection},
  howpublished = {EasyChair Preprint 15617},
  year      = {EasyChair, 2024}}