Table 7.

Hyperparameter settings.

Hyperparameters	Explain	Default Value
learning_rate	Optimizer learning rate	0.01
num_rounds	Communication rounds	1 00
clients_per_round	Number of clients per training round	1 0
num _epochs	Number of local training rounds	1 0
num_iters	Number of iterations per round	2 0
eval_every	Evaluate every few rounds	1
batch_size	Batch sample size	1 0
see	random seed	0
mu	Constraint difference parameters $\mu$	1
stepsize(ss)	Constrained attention aggregation weight parameters $\lambda$	1.2