Cell-Free Massive MIMO for Wireless Federated Learning
share
This paper proposes using cell-free massive multiple-input multiple-output (MIMO) systems to support federated learning (FL) in wireless networks. Thanks to their channel hardening property, these systems provide stable channel conditions for the FL processes to operate on. With a high probability of coverage, they also make the FL processes less prone to the users with unfavorable links. Here, we formulate an FL training time minimization problem where the local accuracy, transmit power, data rate, and users' processing frequency are jointly optimized. This mixed-timescale stochastic nonconvex problem captures the complex interactions among the training time, and transmission and computation of training updates of one FL process. By employing the online successive convex approximation approach, we develop a new algorithm to solve the formulated problem with proven convergence to the neighbourhood of its stationary points. The proposed algorithm only requires channel stability in each iteration rather than its entire run time. Our numerical results confirm that the presented joint design reduces the training time by up to 55% over baseline approaches. They also show that cell-free massive MIMO here requires the lowest training time for FL processes compared with cell-free time-division multiple access massive MIMO and collocated massive MIMO.
READ FULL TEXT
