Single-objective optimization, adaptive differential evolution, CEC 2013 functions

Four adaptive variants of differential evolution algorithm, namely JADE, EPSDE, CoDE, and b6e6rl variant of competitive differential evolution, were compared experimentally in CEC 2013 benchmark suite of 28 functions at dimension $D=30$. The results show that all the algorithms in comparison are not efficient in the problems with rotated objective function, whilst their efficiency on unrotated functions is better. There are statistically significant differences in the efficiency of the algorithms. JADE was mostly the best performing, CoDE the worst. The failure of the adaptive differential evolution in the problems with rotated objective function should be studied in more details.