Advancing Graph Anomaly Detection with Energy-Based Models

EasyChair Preprint 15606

Abstract

Graph anomaly detection is pivotal for analyzing complex networks. This study introduces a novel framework combining Energy-Based Models (EBMs) with Graph Neural Networks (GNNs) to efficiently detect anomalies in graph-structured data. By leveraging structural, relational, and feature-level insights, our approach achieves high accuracy. Experiments on benchmark datasets show superior performance over state-of-the-art methods, underscoring its robustness.

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

@booklet{EasyChair:15606,
  author    = {Chi Zhang and Ye Zhou and Ken Yamada},
  title     = {Advancing Graph Anomaly Detection with Energy-Based Models},
  howpublished = {EasyChair Preprint 15606},
  year      = {EasyChair, 2024}}