Hierarchical Federated Learning
Overview
Hierarchical federated learning studies how devices can train models locally and combine updates through intermediate edge or fog nodes before reaching a central server. Its methods include clustered and personalized aggregation, update compression, adaptive client selection, and resource-aware scheduling to manage non-identical data, limited bandwidth, energy, and intermittent connections. Applications include IoT, vehicles, satellites, wearables, and infrastructure monitoring, where centralized data collection may be impractical or sensitive. A central challenge is balancing accuracy and fairness against communication, delay, and energy costs, while establishing stronger theoretical guarantees and validating methods on real hardware.
Latest papers
19 papers1 paper in the last four weeks, against 2 the four weeks before. 0.0% of all new papers.