Learning Automata-Based Object Partitioning with Pre-Specified Cardinalities

Abstract

The Object Migrating Automata (OMA) has been used as a powerful AI-based tool to resolve real-life partitioning problems. Apart from its original version, variants and enhancements that invoke the pursuit concept of Learning Automata, and the phenomena of transitivity, have more recently been used to improve its power. The single major handicap that it possesses is the fact that the number of the objects in each partition must be equal. This thesis deals with the task of relaxing this constraint. Thus, in this thesis, we will consider the problem of designing OMA-based schemes when the number of the objects can be unequal, but prespecified. By opening ourselves to this less-constrained version, we encounter a few problems that deal with the implementation of the inter-partition migration of the objects. This thesis considers how these problems can be solved, and in essence, presents the design, implementation and testing of two OMA-based methods and all its variants, that include the pursuit and transitivity phenomena.