More about HKUST
Explainability of Graph Neural Networks
PhD Qualifying Examination Title: "Explainability of Graph Neural Networks" by Miss Ge LV Abstract: In recent years, Graph Neural Networks (GNNs) have received significant attention from both industry and academic world owing to their outstanding performance in many graphbased tasks. However, black-box nature of these models hinders them from being trustworthy tools with transparent decision making mechanism. As a result, extensive efforts have been devoted to promoting explainability of GNNs and the area is experiencing rapid developments. In this survey, we aim to provide intuitive understanding and inspiring insights of various techniques by introducing the problem setting of explaining a GNN and reviewing the state-of-the-art (SOTA) GNN explainability methods. We also revisit the evaluation acting for the task, including the commonly used datasets and quantitative metrics. Date: Monday, 29 May 2023 Time: 4:00pm - 6:00pm Venue: Committee Members: Prof. Lei Chen (Supervisor) Prof. Nevin Zhang (Chairperson) Dr. Minhao Cheng Dr. Dan Xu **** ALL are Welcome ****