Figure 2.

The weighted aggregation process of Federated Learning (such as FedAvg) obtains the average of the $w_t$ value received from the client of K from the server and sends the updated $w_t+1$ back to the client.