An Overview of Enabling Federated Learning over Wireless Networks

In this paper, we provide an overview of enabling federated learning (FL) techniques over wireless networks. More specifically, we present key techniques such as model compression, quantization and sparsification that increase the training accuracy of the distributed learning over the wireless medium. Next, the joint FL, resource allocation and scheduling approach is presented, which is identified in two types: a) both user and network assisted, and b) network assisted only. More specifically, the proposed FL-driven resource allocation and scheduling result in a joint optimization problem, where resource allocation and scheduling are jointly optimized. Finally, the simulation setup is described and the obtained simulation results are discussed, while several key enabling techniques are employed that further highlight the achievable performance of enabling FL over wireless networks in terms of training accuracy and loss.