Inferring Group-wise Consistent Multimodal Brain Networks via Multi-view Spectral Clustering

. Author manuscript; available in PMC: 2014 Apr 6.

Published in final edited form as: IEEE Trans Med Imaging. 2013 May 2;32(9):1576–1586. doi: 10.1109/TMI.2013.2259248


Algorithm 2

Input: Connectivity matrices of two views $W_{1}^{0}$ , $W_{2}^{0}$ , and the number of eigenvectors to consider k.
Output: Co-trained connectivity matrices $W_{1}^{}$ , $W_{2}^{}$ .
1.	Compute the initial normalized Laplacian $L_{1}^{0}$ , $L_{2}^{0}$ of each connectivity matrix, and the first k eigenvectors $U_{1}^{0}$ , $U_{2}^{0}$ with the k smallest eigenvalues of $L_{1}^{0}$ , $L_{2}^{0}$ .
2.	for i = 1 to iter
3.	$W_{1}^{i} = proj (W_{1}^{i - 1}, U_{2}^{i - 1})$
4.	$W_{2}^{i} = proj (W_{2}^{i - 1}, U_{1}^{i - 1})$
5.	Compute Laplacian and the corresponding first k eigenvectors $U_{1}^{i}$ , $U_{2}^{i}$ of $W_{1}^{i}$ , $W_{2}^{i}$ .
6.	If converge, return $W_{1}^{} = W_{1}^{i}, W_{2}^{} = W_{2}^{i}$ .