Drs. Jiliang Tang and Guan-Hua Tu, faculty members in Computer Science and Engineering, have received the Amazon Faculty Resear

Abstract: 

Graphs are universal data structures that describe the relationships between entities in various application domains, such as social media, biology, transportation, financial system, and cyber networks. Due to the rich information described by graphs, machine learning on graph data is a prominent research area with broad and promising applications. However, in real-world graphs, there might generally exist anomalous entities or relations that cause undesired consequences such as decreasing performance and even severe security concerns. Our preliminary study suggests that it is promising to detect anomalies in graph data by designing specialized graph neural networks, and anomaly detection also helps improve the prediction performance of graph neural networks. Therefore, in this proposal, we propose to comprehensively study new methods for anomaly detection with graph neural networks