Table 3. Working of Federated clients.

Steps	Description
Step 1	Clients train on a batch locally
Step 2	In every iteration, client will update its weights as if isolated training is being performed
Step 3	All clients would transmit weights according to their configuration
Step 4	Federated averaging algorithm will collect local data models and calculate average and update the shared model to CS
Step 5	On receiving the updated average. CS sends a message to rest of the clients to pull the latest model before training on future batches