Spectral clustering of high-dimensional data exploiting sparse representation vectors

Abstract

Clustering high-dimensional data has been a challenging problem in data mining and machining learning. Spectral clustering via sparse representation has been proposed for clustering high-dimensional data. A critical stepin spectral clustering is to effectively construct a weight matrix by assessing the proximity between each pair of objects. While sparse representation has proved its effectiveness for compressing high-dimensional signals, existing spectral clustering algorithms based on sparse representation use individual sparse coefficients directly. However, exploiting complete sparse representation vectors is expected to reflect more truthful similarity among data objects, since more contextual information is being considered. The intuition is that sparse representation vectors corresponding to two similar objects are expected to be similar, while those of two dissimilar objects are dissimilar. In particular, we propose two weight matrix constructions for spectral clustering based on the similarity of the sparse representation vectors. Experimental results on several real-world, high-dimensional datasets demonstrate that spectral clustering based on the proposed weight matrices outperforms existing spectral clustering algorithms, which use sparse coefficients directly.