Over-the-Air Computation for Distributed Machine Learning and Consensus in Large Wireless Networks

We give an overview of the problem of Over-the-Air (OTA) computation and the currently available research, and of two important applications: Distributed consensus algorithms and distributed Machine Learning. For the latter, we specifically consider an application of OTA computation in Horizontal and Vertical Federated Learning. We then survey some of our own contributions to these fields, which include a scheme for analog Over-the-Air computation that works with a large class of fast-fading wireless channels, as well as applications to distributed consensus and Vertical Federated Learning. Moreover, we give an overview of our initial research on how Physical Layer Security can be applied to secure Over-the-Air computation against eavesdroppers.