An Extended Extremal Optimisation Model for Parallel Architectures

Abstract

A relatively new meta-heuristic, known as extremal optimisation (EO), is based on the evolutionary science notion that poorly performing genes of an individual are replaced by random mutation over time. In combinatorial optimisation, the genes correspond to solution components. Using a generalised model of a parallel architecture, the EO model can readily be extended to a number of individuals using evolutionary population dynamics and concepts of self-organising criticality. These solutions are treated in a manner consistent with the EO model. That is, poorly performing solutions can be replaced by random ones. The performance of standard EO and the new system shows that it is capable of finding near optimal solutions efficiently to most of the test problems.