global optimization; differential evolution; control parameters; competitive adaption; exponential crossover; numerical comparison

Two types of crossover in differential evolution algorithms are concerned, both in standard and adaptive variants of differential evolution. Proper choice of mutation probability in exponential crossover is proposed and implemented in respective variants of differential evolution in tests. The impact of crossover type and the use of random localization in mutation were examined in benchmark tests. New variant of self-adaptive differential evolution with both type of crossover and competitive setting of control parameters outperformed the others. This variant is suggested for the real-world problems, where we need an adaptive algorithm applicable without time-wasting parameter tuning.