A New Multichannel Parallel Network Framework for the Special Structure of Multilead ECG

Algorithm 3

Training of MLCNN-BiLSTM model based classifier.

	Input: Training set X = [x1, x2, x3…]; Base learning algorithm MLCNN-BiLSTM model
	Output: The parameters of the model and result
	Process:
	1 begin
	2 Build the MLCNN-BiLSTM Model with a softmax output layer
	3 while training do
	4 begin
	5 Calculate the loss on the training set according to (1)
	6 Train the MLCNN-BiLSTM Model using the Adam back propagation method
	7 Evaluate the training loss on training set
	8 Record the prediction of classifier on training set
	9 if training loss stop decreasing then
	10 Store the model and break
	11 end
	12 end