Kernelized Graph-based Multi-view Clustering on High Dimensional Data

Abstract

Kernelized graph-based learning methods have gained popularity because of its better performances in the clustering task. But in high dimensional data, there exist many redundant features which may degrade the clustering performances. To solve this issue, we propose a novel multi-view kernelized graph-based clustering (MVKGC) framework for high dimensional data that performs the clustering task while reducing the dimensionality of the data. The proposed method also uses multiple views which help to improve the clustering performances by providing different partial information of a given data set. The extensive experiments of the proposed method on different real-world benchmark data sets show a better and efficient performance of the proposed method than other existing methods.