A Study on Parallel Distributed Genetic Algorithms (Discussion on a Stepping Stone Population Model of Distributed Genetic Algorithms)

摘要

This paper discusses about the characteristics of distributed genetic algorithms (DGAs). Among the several types of distributed models of GAs, this paper focuses on a stepping stone population model. The effects of the number of subpopulations, subpopulation immigration rate and immigration interval on the performance of the DGA are examined. Solving standard mathematical problems and a typical structural optimization problem, we find that there are several advantages in the stepping stone population model of DGAs: this model is very suitable for parallel computation and the optimum solutions can be obtained efficiently due to the partitioning of the population.