XIII Krajowa Konferencja
Warszawa, 21  22 września 2011


Archiwum / Archives



Jarosław Stańczak, Krzysztof Trojanowski Properties of Selection Methods Applied to Non–Stationary Optimization Tasks A large group of real–world problems are non–stationary optimization tasks, i.e. problems where the environment is varying in time. In this case, a recently found optimal solution can stop being an optimum in the next moment because of the changes that appear in the environment. Therefore in non–stationary environment the optimum has to be searched continuously and this way non–stationary optimization process becomes quite similar to adaptation process. It has already been showed that simple genetic algorithm is not a very efficient tool for such a non–stationary optimization because of specific requirements of this process and thus the algorithm should be extended by some additional traits. This paper presents discussion of several selection methods and their influence on behavior and efficiency of evolutionary algorithm when optimizing in the non–stationary environment. Three methods were compared: tournament selection and two other methods of selection
proposed by the authors. Some experiments were performed and the results obtained in these experiments are presented and discussed. Inne artukuły z wybranej konferencji.

 
