Genetic CFL: Hyperparameter Optimization in Clustered Federated Learning

Table 3

Performance of FL against genetic CFL on MNIST dataset for various hyperparameters.

Client ratio	Rounds	FL		Genetic CFL
		Accuracy	Loss	Accuracy	Loss
0.1	3	0.9133	0.3136	0.9679	0.1203
	6	0.9265	0.2493	0.9730	0.1343
	10	0.9367	0.2115	0.9777	0.1923
0.15	3	0.9176	0.2878	0.9665	0.1049
	6	0.9740	0.0876	0.9740	0.0876
	10	0.9443	0.1828	0.9763	0.0910
0.3	3	0.9178	0.2989	0.9698	0.0964
	6	0.9326	0.2359	0.9780	0.0804
	10	0.9450	0.1946	0.9799	0.0849