Multiview approach to spectral clustering.

In this paper we propose a generic approach to the multiview clustering problem that can be applied to any number of data views and with different topologies, either continuous, discrete, graphs, or other. The proposed method is an extension of the well-established spectral clustering algorithm to integrate the information from several data views in the partition solution. The algorithm, therefore, resolves a joint cluster structure which could be present in all views, which enables researchers to better resolve data structures in data fusion problems The application of this novel clustering approach covers an extended number of machine learning unsupervised clustering problems including biomedical analysis or machine vision.