Reevaluating Exponential Crossover in Differential Evolution

Exponential crossover in Differential Evolution (DE), which is similar to 1-point crossover in genetic algorithms, continues to be used today as a default crossover operator for DE. We demonstrate that exponential crossover exploits an unnatural feature of some widely used synthetic benchmarks such as the Rosenbrock function – dependencies between adjacent variables. We show that for standard DE as well as stateof-the-art adaptive DE, exponential crossover performs quite poorly on benchmarks without this artificial feature. We also show that shuffled exponential crossover, which removes this kind of search bias, significantly outperforms exponential crossover.