PSA Approach to Population Models for Parallel Genetic Algorithms

Abstract

A universal approach for describing the population model of genetic algorithms is developed which is based on the Parallel Substi- tution Algorithm (PSA) theory. Genetic algorithms (GA) are a suitable method if good approximations for problems are required which were otherwise not solvable in practical environments. Optimisation of GAs can be done on several levels, in this work we concentrate on the popu- lation model. Most prominent population models are the classical global model, the island model and it’s extreme variant, the cellular model. The PSA theory supports us with a general approach which is essential for systematically studying convergence behaviour of GA population model variants and consequently for their optimisation.