This model makes use of Federated Averaging (FDA) for SGD. The optimization of your loss purpose is carried out with SGD, as for most federated Mastering systems. Every single MEC operates SGD on its own non-public details and periodically synchronizes with other MECs by amassing averages and sharing the optimal Answer from the many nearby MEC type