Steering of balance between exploration and exploitation properties of evolutionary algorithms - Mix selection

Abstract

In this paper the novel selection method which can be used in any evolutionary algorithm is presented. Proposed method is based on steering between exploration and exploitation properties of evolutionary algorithms. In presented approach, at the start of the algorithm operation the probability of selection of individuals for new population is equal for all individuals. In such a case the algorithm possesses maximal value of pressure on global search of a solution space (exploration of solution space). As number of generations increases, the algorithm searches the solution space in more locally manner (exploitation of solution space) at expense of global search property. The results obtained using proposed method are compared with the results obtained using other selection methods like: roulette selection, elitist selection, fan selection, tournament selection, deterministic selection, and truncation selection. The comparison is performed using test functions chosen from literature. The results obtained using proposed selection method are better in many cases than results obtained using other selection techniques. © 2010 Springer-Verlag.