This $439,539 Project Grant award from the National Science Foundation (CFDA 47.041 - Engineering) supports research to develop computationally efficient and trustworthy graph neural network (GNN) models with theoretical performance guarantees. The primary objectives are to systematically analyze how graph topology and network architecture influence GNN performance, optimize computational and memory resources through techniques like graph data aggregation and network pruning, and introduce novel Graph Mixture of Experts (GMoE) architectures to enhance efficiency. The research aims to advance the development of reliable AI systems applicable across domains like social networks and power grids, while also contributing to the emerging field of green AI to reduce the economic and environmental impacts of large-scale AI models. The award is being led by Rensselaer Polytechnic Institute and does not involve any sub-awards.
Generated 3/25/25, 3:19 AM