Weighted Nuclear Norm Minimization on Multimodality Clustering

Algorithm 1

Spectral clustering.

	Input:
(1)	Construct the similarity matrix by Gaussian kernel, where represents the similarity of the th sample and the th sample.
(2)	Compute the normalized symmetrical Laplacian , where D is a diagonal matrix with .
(3)	Let U be a matrix with columns representing the top eigenvectors of .
(4)	Normalize each row of U.
(5)	Run the k-means algorithm on U.
	Output: the result of k-means.