Differential evolution, rotationally invariant crossover, Eigenvector crossover, experimental comparison

Differential Evolution (DE) is population-based optimization algorithm which has only few control parameters. However, it is not easy task to set the parameters appropriately for solved optimization problem but it is crucial for obtaining good results. Moreover, the quality of points produced in evolution is highly affected by the coordinate system. Some types of problems could not be solved satisfactory with basic version of DE algorithm, e.g. rotated problems. In this paper, we focus on a new approach published. Coordinates of points from which the new trial point is created are transformed into the coordinate system of principal components. New coordinate system is based on the current distribution of points in the population. Does the using of such mechanism lead to better solution for rotated problems? Experiments are conducted on benchmark set developed for CEC2013 competition containing pairs of rotated and non-rotated functions.