Multi-Carrier NOMA-Empowered Wireless Federated Learning with Optimal Power and Bandwidth Allocation

Wireless federated learning (WFL) undergoes a communication bottleneck in uplink, limiting the number of users that can upload their local models each global aggregation round. This paper presents new multi-carrier non-orthogonal multiple-access (MC-NOMA)-empowered WFL system under an adaptive setting Flexible Aggregation. Since round accommodates both model training and uploading for user, use Aggregation allows to train different numbers iterations per round, adapting channel conditions computing resources. The key idea is MC-NOMA concurrently users, thereby extending times increasing participating users. A metric, namely, Weighted Global Proportion Trained Mini-batches (WGPTM), analytically established measure convergence system. Another important aspect we maximize WGPTM harness by jointly optimizing transmit powers subchannel bandwidths. nonconvex problem converted equivalently tractable convex solved efficiently using variable substitution Cauchy’s inequality. As corroborated experimentally convolutional neural network 18-layer residential network, proposed reduce delay, increase times, accelerate over 40%, compared its existing alternative.