Performance Analysis for Channel-Weighted Federated Learning in OMA Wireless Networks

To alleviate the negative impact of noise on wireless federated learning (FL), we propose a channel-weighted aggregation scheme FL (CWA-FL), in which parameter server (PS) makes gradients according to channel conditions devices. In proposed scheme, are transmitted PS an uncoded way through orthogonal multiple access (OMA) channel, can avoid synchronization issue among devices faced by over-the-air FL. The convergence analysis CWA-FL is conducted and theoretical results show that converge with rate $\mathcal {O} (\frac{1}{T})$ . Simulation performs better than equal-weighted (EWA-FL) more robust noise.