ALBLP: Adaptive Load-Balancing Architecture Based on Link-State Prediction in Software-Defined Networking

Algorithm 1

LSTM training algorithm.

	Input:
	Link residual bandwidth matrix sequence ;
	The input time series length and the output time series length ;
	Training batch size batch_size and training times epochs;
	Output:
	model parameters after training;
(01)	for i = 1 to do
(02)
(03)	;
(04)	Data_train add;
(05)	end for
(06)	Initialize model net;
(07)	repeat
(08)	Take out Data_train from Data_batch, the size is batch_size;
(09)	Out = net(Data_batch);
(10)	Loss = MSEloss(Out,);
(11)	Back propagation to update model parameters;
(12)	epochs = epochs − 1;
(13)	until epochs = 0