On Boolean Representation of Continuous Data Biclustering

Abstract

Biclustering is considered as the method of finding two–dimensional subgroups in a matrix of scalars. The paper introduces a new approach to biclustering continuous matrices on the basis of boolean function analysis. We draw the strong relation between inclusion–maximal (maximal with respect to inclusion) biclusters of the assumed maximal difference between the data in a bicluster and prime implicants of a boolean function describing the data. These biclusters are called similarity biclusters. In the opposition to them, a new notion of dissimilarity biclusters was also introduced in the paper.