Centroid-based Initialized JADE for global optimization

JADE is a recent adaptive version of nature inspired algorithm, Differential Evolution (DE). It has shown considerable performance improvement on a set of well studied benchmark functions. This paper studies whether the performance of JADE can be improved further by changing its random population initialization with centroid based population initialization. The new algorithm is known as Centroid-based Initialized JADE (ciJADE). ciJADE not only initialize JADE from centroids of the random parents, it also reflects the worst ten percent of the centroids population. ciJADE is tested on fourteen CEC2005 competition test problems. Experimental results show that ciJADE has significantly improved the performance of JADE on various problems.