A Rank-Constrained Matrix Representation for Hypergraph-Based Subspace Clustering

Algorithm 2

Hypergraph-based subspace clustering.

Input: data matrix , number of classes , rank of coefficient matrix

(1) Obtain the rank-constrained matrix representation via optimization Algorithm 1.

(2) Construct a K-nearest neighbors hypergraph by using the -rank representation to define

    the hyperedge and matrix of the hypergraph.

(3) Compute the hypergraph Laplacian matrix via (18).

(4) Spectral decomposition of hypergraph Laplacian matrix and Take the first eigenvectors with

    non-zero eigenvalues as the embedded representation.

Output: Use a -means clustering algorithm on the eigenspace to partition the vertices of the graph into clusters.