A Network Fault Prediction-Based Service Migration Approach for Unstable Mobile Edge Environment

Table 3

NFP-WD parameters.

Parameter	Values
FFM	Field size : 3
	Feature sizes : (64, 64, 64)
	Embedding size : 4
	Dropout shallow : (0.5, 0.5)
LSTM	Hidden size : 3, 512, 256, 128
	Num of layers : 3
	Epochs : 64
	Batchsize : 256
	learning rate : 0.003
Joint training	Activation : Sigmoid
Loss function	: 0.9