Figure 1.
Algorithmic overview of the iterative federated averaging protocol with 4 steps: 1) server broadcasts artificial intelligence model to all participants; 2) each participant individually trains model on its local data to create an updated model; 3) participants send model updates back to central server; 4) server averages all model updates to aggregate them to a new global model for the next round.