Fast Constrained Spectral Clustering and Cluster Ensemble with Random Projection

Algorithm 1

Fast constrained spectral clustering.

Input: data set , the number of landmark points , constraint matrix , cluster number ,

      confidence parameter , sample rate ;

Output: the grouping result.

() Compute the sparse representation in Equation (5);

() Compute Laplacian and , where

   is the Laplacian matrix of , is the Laplacian matrix of ;

() Solve the first non-trivial generalized eigenvectors of Equation (12);

() Compute ;

() Embed into a -dimensional sphere using the embedding process in [17];

() Sample row vectors of randomly and run -means clustering on the sampled row vectors;

() Get the clustering result utilizing distances between centers of -means clustering and row vectors of .