CF-mMIMO中联邦学习的前传压缩与波束成形方法OA

Federated Learning(FL)is a distributed machine learning paradigm that allows multiple participants to collaboratively train a shared global model while keeping their data localized.To reduce the communication overhead in wireless FL,we explore the application of over-the-air computation to aggregate local gradients from distributed devices within a cell-free massive multi-input multi-output(CF-mMIMO)-enabled FL framework.To mitigate the aggregation errors introduced during this process,a joint optimization approach is proposed that integrates fronthaul compression and beamforming design.Specifically,we first investigate the CF-mMIMO-enabled FL architecture and establish an uplink transmission model leveraging over-the-air computation techniques to enhance communication efficiency,then we derive a closed-form expression for the aggregation error by accounting for the limited fronthaul capacity.Subsequently,the convergence behavior of the system is analyzed,and based on the derived optimal convergence interval,an optimization problem is formulated to jointly design the device transmission power,receive beamforming vectors,and fronthaul quantization parameters.The resulting problem is solved using an alternating optimization algorithm.Simulation results demonstrate that the proposed scheme improves learning performance by more than 10%compared to baseline methods,thereby validating the effectiveness and potential of the FL in supporting CF-mMIMO.