Local Epochs Inefficiency Caused by Device Heterogeneity in Federated Learning

Table 2

Hyperparameter settings.

Hyperparameter	Value
Initial learning rate	0.1
The decay period of learning rate	40
The decay ratio of learning rate
L2 regularization	0.00001
Dropout	0.2
Batchsize	128
Epoch	300
Activation	ReLu
Optimizer	Adam
Loss	Mean squared error