PROJECT TITLE :

Convex Sparse Spectral Clustering: Single-View to Multi-View

ABSTRACT:

Spectral clustering (SC) is one in every of the most widely used methods for data clustering. It 1st finds a coffee-dimensional embedding U of knowledge by computing the eigenvectors of the normalized Laplacian matrix, and then performs k-means that on $ text U^prime $ to urge the final clustering result. During this paper, we observe that, in the perfect case, $ text U text U ^high $ should be block diagonal and therefore sparse. Therefore, we tend to propose the sparse SC (SSC) method that extends the SC with sparse regularization on $ text U text U ^high $ . To handle the computational issue of the nonconvex SSC model, we have a tendency to propose a unique convex relaxation of SSC primarily based on the convex hull of the mounted rank projection matrices. Then, the convex SSC model will be efficiently solved by the alternating direction methodology of multipliers Furthermore, we propose the pairwise SSC that extends SSC to boost the clustering performance by using the multi-read info of knowledge. Experimental comparisons with many baselines on real-world datasets testify to the efficacy of our proposed ways.