In this paper, we study a problem of the control parameter settings in DifferentialEvolution algorithm and test novel variant of the algorithm called Co-BiDE. Although Differential Evolution with basic setting (i.e. CR = 0:5; F =0:5) works quite well, it starts to fail on rotated functions. In general, wewant to improve the convergence of algorithm primarily on rotated functions.It is done by adapting crossover parameter CR whereas parameter F is fixedto 0.5. There is a recommendation to set CR = 1 for rotated functions. Itmeans that trial vectors are essentially composed from mutant. However, itis not easy task to set the parameters appropriately for solving optimizationproblem but it is crucial for obtaining good results. Moreover, the quality ofpoints produced in evolution is highly aected by the coordinate system. InCoBiDE, the authors proposed new coordinate system based on the currentdistribution of points in the population. We test these two approaches byrunning both algorithms on six pairs of rotated and non-rotated functionsfrom CEC 2013 benchmark set in two levels of dimension space. This experimentalstudy aims to reveal if such algorithm's setting is invariant under arotation.
@article{616,
title = {Is differential evolution rotationally invariant?},
journal = {Tatra Mountains Mathematical Publications},
volume = {72},
year = {2019},
doi = {10.2478/tmmp-2018-27},
language = {EN},
url = {http://dml.mathdoc.fr/item/616}
}
Zámečníková, H.; Einšpiglová, D.; Poláková, Radka; Bujok, Petr. Is differential evolution rotationally invariant?. Tatra Mountains Mathematical Publications, Tome 72 (2019) . doi : 10.2478/tmmp-2018-27. http://gdmltest.u-ga.fr/item/616/