The multi-view information bottleneck clustering

Abstract

In this paper, we propose a new algorithm for information bottleneck method in multi-view setting where instances have multiple independent representations. By introducing the two important conditions, conditional independence and compatibility, into the information bottleneck clustering, the compatible constraint maximizing the agreement between clustering hypotheses on different views is imposed on the individual views to cluster instances. Our algorithm is developed by the compatible constraint. Experiments on three real-world dataseis indicate that our algorithm considering the relationship among multiple views can provide solution with improved quality in multi-view setting. © Springer-Verlag Berlin Heidelberg 2007.