Project Grant 2145031

This $500,000 National Science Foundation Project Grant through the Integrative Activities program (CFDA 47.083) funds research at Auburn University towards efficient and fast hierarchical federated learning in heterogeneous wireless edge networks. The university will conduct research over five years to develop fundamental understandings and adaptive, efficient algorithms and schemes for wireless hierarchical federated learning. This addresses challenges from non-uniform device configurations, inefficient bandwidth sharing, and impacts of hierarchical structures on heterogeneous devices. The research is organized into three interdependent thrusts: 1) adaptive, cost-aware device selection and computation configuration; 2) co-design of computation configuration and communication scheduling; and 3) global model communication and aggregation. The university will implement proposed schemes and evaluate practical performance. This supports NSF's mission to enhance STEM competitiveness through integrative activities building research capacity and infrastructure.