An analysis of the XOR dynamic problem generator based on the dynamical system

Abstract

In this paper, we use the exact model (or dynamical system approach) to describe the standard evolutionary algorithm (EA) as a discrete dynamical system for dynamic optimization problems (DOPs). Based on this dynamical system model, we analyse the properties of the XOR DOP Generator, which has been widely used by researchers to create DOPs from any binary encoded problem. DOPs generated by this generator are described as DOPs with permutation, where the fitness vector is changed according to a permutation matrix. Some properties of DOPs with permutation are analyzed, which allows explaining some behaviors observed in experimental results. The analysis of the properties of problems created by the XOR DOP Generator is important to understand the results obtained in experiments with this generator and to analyze the similarity of such problems to real world DOPs.