Research on cause for overlapping solutions and on their influence in NSGA-II algorithm
Jinhua Zheng · Computer Engineering and Applications Journal · 2008
The existence of overlapping individuals in the evolution populations means overlapping regions in the searching space,which makes the algorithm much less efficiently in exploiting new feasible region.Additionally,it wastes the positions in the population and leads to the information redundancy,which reduces the diversity of the obtained solution set.This phenomenon is quite obvious in the famous NSGA-II algorithm when applied to low-variable dimension problems.In this paper,the cause of the overlapping solutions in NSGA-II is analyzed and the relation between the number of them and the coding method is diseovered,as well as and the variable dimension;then overlapping solutions have influence on the performance of the NSGA-II algorithm and on the quality of the obtained solution set is illuminated.The experimental results demonstrate that eliminating overlapping solutions make the NSGA-II algorithm more steady and gain a solution set with better diversity.