| Popis: |
In multimodal optimization, it is important to obtain all the optima in a single run, for an insightful description of the potential design configurations. Genetic algorithms, as population-based approaches, are very suitable for this purpose. In this paper, the genetic algorithm is designed to work on subpopulations, which use special mechanisms for the insertion of offspring and immigrants. Mainly, the subpopulations are designed to allow speciation, each species being potentially formed around a global optimum. To this end, the insertion mechanism uses a combination of fitness-based and similarity-based insertion techniques, which can help both for preserving the exploration in diverse areas and for improving the accuracy of the solutions. A highly nonlinear problem is considered as a case study – that of finding the propagation constants of a planar dielectric microwave guide. The experimental investigations carried out for different waveguide setups show that using migration with similitude-based insertion of migrants helps achieve accurate locations of all optima, with reasonable computational costs. |
