Next-Generation Energy-Based Models for Graph Anomaly Detection

EasyChair Preprint 15620

Abstract

Detecting anomalies in complex networks has become a pivotal focus in graph analysis. This study introduces a novel and robust framework leveraging the capabilities of Energy-Based Models (EBMs) for efficient anomaly detection in graph-structured data. By combining the strengths of graph neural networks (GNNs) with EBMs, the proposed approach effectively captures intricate structural, relational, and feature-level patterns, ensuring high accuracy and reliability in anomaly detection. Comprehensive experiments on widely recognized benchmark datasets demonstrate the superior performance of this method, outperforming state-of-the-art techniques while showcasing its robustness, scalability, and practical applicability in real-world scenarios.

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

@booklet{EasyChair:15620,
  author    = {Camlia Costa and Pedro Silva and Chi Zhang},
  title     = {Next-Generation Energy-Based Models for Graph Anomaly Detection},
  howpublished = {EasyChair Preprint 15620},
  year      = {EasyChair, 2024}}