Associative clustering

Abstract

Clustering by maximizing the dependency between two paired, continuous-valued multivariate data sets is studied. The new method, associative clustering (AC), maximizes a Bayes factor between two clustering models differing only in one respect: whether the clusterings of the two data sets are dependent or independent. The model both extends Information Bottleneck (IB)-type dependency modeling to continuous-valued data and offers it a well-founded and asymptotically well-behaving criterion for small data sets: With suitable prior assumptions the Bayes factor becomes equivalent to the hypergeometric probability of a contingency table, while for large data sets it becomes the standard mutual information. An optimization algorithm is introduced, with empirical comparisons to a combination of IB and K-means, and to plain K-means. Two case studies cluster genes 1) to find dependencies between gene expression and transcription factor binding, and 2) to find dependencies between expression in different organisms. © Springer-Verlag Berlin Heidelberg 2004.