A Multiview Deep Learning Method for Brain Functional Connectivity Classification

Algorithm 1

The proposed MVDL-BFCC method.

	Input: a training set , where is the set of the BFCof all subjects, and is the set of all the category labels
	Output: the optimal multiview feature selection weight , the optimal network weight , and the optimal weight bias
(1)	Randomly initialize , , and
(2)	while the stop condition is met do
(3)	compute according to (8)
(4)	end while
(5)
(6)	while the stop condition is not met do
(7)	The forward propagation process. input into DNN, extract multiview features, and obtain classification result by the prototype learning
(8)	Calculate loss. According to and , calculate the function loss by (18)
(9)	The back propagation process. the gradient descent method is used to update and
(10)	end while
(11)	,
(12)	Return , , and .