EdgeML: Towards network-accelerated federated learning over wireless edge

Pinyarash Pinyoanuntapong, Prabhu Janakaraj, Ravikumar Balakrishnan, Minwoo Lee, Chen Chen, Pu Wang

Abstract: Highlights•The limited network bandwidth in Wireless multi-hop networks degraded scalability of federated learning.•Model-based optimization for multi-hop federated learning system is non-trivial.•The multi-agent reinforcement learning routing is proposed to minimize the federated learning convergence time.•The first work in the literature to reveal, formulate, and experiment on the inherent interplay between multi-hop wireless networking and federated learning.•EdgeML is the first experimental framework in the literature for federated learning over multi-hop wireless edge computing networks.