PROJECT TITLE :

Spectral Ensemble Clustering via Weighted K-Means: Theoretical and Practical Evidence - 2017

ABSTRACT:

As a promising approach for heterogeneous data analytics, consensus clustering has attracted increasing attention in recent decades. Among numerous glorious solutions, the co-association matrix primarily based methods type a landmark, that redefines consensus clustering as a graph partition problem. Nevertheless, the relatively high time and area complexities preclude it from wide real-life applications. We tend to, therefore, propose Spectral Ensemble Clustering (SEC) to leverage the advantages of co-association matrix in information integration however run more efficiently. We tend to disclose the theoretical equivalence between SEC and weighted K-means clustering, that dramatically reduces the algorithmic complexity. We tend to conjointly derive the latent consensus function of SEC, that to our greatest data is the first to bridge co-association matrix based ways to the ways with express international objective functions. Any, we tend to prove in theory that SEC holds the robustness, generalizability, and convergence properties. We tend to finally extend SEC to fulfill the challenge arising from incomplete basic partitions, primarily based on which a row-segmentation scheme for big information clustering is proposed. Experiments on numerous real-world knowledge sets in each ensemble and multi-read clustering scenarios demonstrate the superiority of SEC to some state-of-the-art ways. In particular, SEC looks to be a promising candidate for large information clustering.