Visualization of clusters in very large rectangular dissimilarity data

Abstract

D is an m×n matrix of pairwise dissimilarities between m row objects Or and n column objects Oc, which, taken together, comprise m+n objects O = [o1,…om,om+1,…om+n]. There are four clustering problems associated with O: (P1) amongst the row objects Or; (P2) amongst the column objects Oc; (P3) amongst the union of the row and column objects O=Or∪Oc; and (P4) amongst the union of the row and column objects that contain at least one object of each type (co-clusters). The coVAT algorithm, which builds images for visual assessment of clustering tendency for these problems, is limited to m×n ≈ O(10⁴×10⁴). We develop a scalable version of coVAT that approximates coVAT images when D is very large. Two examples are given to illustrate and evaluate the new method.