Multiview spectral clustering via ensemble
Yong Jun Cheng, Ruilian Zhao · 2009
Clustering on multiple views is witnessing increasing interests in both real-world application and machine learning community. A typical application is to discover communities of joint interests in social network, such as Facebook and Twitter. The network can be simply modeled as a graph in which the nodes are the people while the links show relationship between the people. There may exist many relationships between a pair of nodes, such as classmates, collaborators, playmates and so on. It is important to consider how to use these graphs together rather than a single graph if we want to understand the network and their participants effectively. Motivated by the fact, we present a clustering algorithm using spectral analysis in which multiple graphs are considered to get the clusters. Our study can also be considered as an instance of multi-views learning. The experimental results on UCI data set and Corel image data demonstrate the promising results that validate our proposed algorithm.