This $266,000 National Science Foundation (NSF) CAREER project grant, awarded on August 15, 2024 under the CISE (Computer and Information Science and Engineering, CFDA 47.070) program, aims to develop novel graph neural network (GNN) models that create smaller, compressed graphs while preserving the structural information of large graphs. This will enable GNNs to be more effectively deployed in real-world applications across transportation, biomedical, social, and security domains. The research goals are to: 1) create novel graph compression techniques to address GNN oversmoothing, 2) develop compression methods that capture local, global, and higher-order structural information for improved pooling, and 3) devise compression methods to effectively capture and compress shared structures in multidimensional networks. The project will integrate research and education through outreach activities targeting high school, undergraduate, and underrepresented students. The deliverables include new GNN architectures with improved compression capabilities, evaluated on drug-disease matching and event detection tasks in collaboration with domain experts.
Generated 2/25/25, 3:41 AM