A new parameter-less evolution strategy for solving unconstrained global optimization problems

Abstract

Several evolutionary approaches have been applied to unconstrained global optimization problems with significant success. These algorithms mimic the natural evolution of the species in biological systems and do not require any continuity or convexity properties of the problem being solved. Moreover, unlike conventional algorithms, only information regarding the objective function is required to perform the search. Evolution strategies proved to be one of the most efficient evolutionary approach to global optimization. However, these algorithms have several parameters which the setting is not simple. Thus, it is crucial to investigate how to set dynamically these parameters during the search. In this paper, a new parameter-less evolution strategy, which has only one single parameter to set, is proposed. The influence of this parameter is also investigated. The new algorithm is compared with the traditional evolution strategies considering a set of difficult test problems. The statistical analysis of the results obtained indicates a promising performance of the new approach.