Rank-based differential evolution with eigenvector-based crossover operator

Differential evolution (DE) is an efficient algorithm for solving optimization problems in a continuous space. In recent years, many studies have reported modification and improvement for DE. Rank-based DE (RDE) is one of the modified DE algorithms, which allocates different control parameter values for each individual based on the ranking information in the current population. We attempt to improve the search ability of RDE by using an eigenvector-based (EIG) crossover. The EIG crossover is a rotationally invariant operator which provides superior performance on non-separable problems. In the EIG crossover, population is rotated to an appropriate coordinate system, and then a crossover operator is executed on the rotated population. In this paper, the performance of the RDE with EIG crossover is evaluated on the basic benchmark functions. Through the experiments, we show that the EIG crossover can enhance the search ability of RDE. © 2016 ICIC International.