Local Epochs Inefficiency Caused by Device Heterogeneity in Federated Learning

Table 4

Layer features in the convolutional layer.

Features	Description
Batchsize	The size of a batch.
Elements_matrix	The number of elements of input.
Elements_kernel	The number of elements of convolutional kernel.
Channels_in	The number of input channels.
Channels_out	The number of output channels.
Padding	The number of edge padding.
Strides	The step size of convolution.
Use_bias	Whether to use bias, 0: not use 1: use.
Opt_SGD	Whether to use SGD optimizer, 0: not use 1: use.
Opt_Adadelta	Whether to use Adadelta optimizer, 0: not use 1: use.
Opt_Adagrad	Whether to use Adagrad optimizer, 0: not use 1: use.
Opt_Momentum	Whether to use momentum optimizer, 0: not use 1: use.
Opt_Adam	Whether to use Adam optimizer, 0: not use 1: use.
Opt_RMSProp	Whether to use RMSProp optimizer, 0: not use 1: use.
Act_relu	Whether to use ReLu activation, 0: not use 1: use.
Act_tanh	Whether to use tanh activation, 0: not use 1: use.
Act_sigmoid	Whether to use sigmoid activation, 0: not use 1: use.