Algoritmo genético acoplado a um método multi-grid e a teoria dos grafos para determinação da estrutura de equilíbrio de aglomerados atômicos

Abstract

In this work we present a proposal to improve Genetic Algorithm method by coupling it to the techniques of discretization of the configurational space via the multi-grid methodology, and by employing a topological selection of the offsprings via graph theory. The best performance for clusters of 13 and 19 aluminum atoms shows that the multi-grid tecniques can increase the efficiency of the genetic algorithm, mainly when a more extensive search is performed in an initially sparse grid of points. We also show that a greater improvement in the efficiency of the genetic algorithm can be obtained when we select the offsprings of the sucessive generations in order to be topologically distinct from each other.