PARTITION TRIPLES - A TOOL FOR REDUCTION OF DATA SETS

Data sets discussed in this paper are presented as tables with rows co rresponding to examples (entities, objects) and columns to attributes. A partition triple is defined for such a table as a triple of partiti ons on the set of examples, the set of attributes, and the set of attr ibute values, respectively, preserving the structure of a table. The i dea of a partition triple is an extension of the idea of a partition p air, introduced by J. Hartmanis and J. Steams in automata theory. Resu lts characterizing partition triples and algorithms for computing part ition triples are presented. The theory is illustrated by an example o f an application in machine learning from examples. (C) 1996 Academic Press, Inc.