Device Selection for Passive RIS-Assisted Over-the-Air Federated Learning

In this paper, we consider an AirComp FL system with beyond-diagonal RIS. It is widely recognized that increasing the number of participating devices in federated learning generally leads to improved learning performance. Therefore, we propose a beamforming vector and scattering matrix design scheme that maximizes the number of participating edge devices with MSE constraint. Simulation results show that fully connected beyond-diagonal RIS can accommodate more devices in the AirComp FL.