Abstract: The objective of self-expression based spectral clustering is to learn an affinity matrix which accurately reflects the similarity among data, and the Laplacian constraint is usually ...