Research on Overlapping Individuals in MOEAs: Causation, Probability for Quantity and Influence on Assessment
Yunfei Zhu, Jiongliang Xie, Jinhua Zheng, Biao Luo, LI Wang-yi · 2009
The existence of overlapping individuals in the evolution population means that the MOEAs do a redundant work of searching the overlapping region in the searching space, which weakens the ability of the MOEAs to exploit new feasible regions. Hence the running efficiency of the algorithm turns to be lower. This paper focused on the overlapping individuals in MOEAs. We probed into the causation why overlapping individuals come into being and then gave a probability analysis of the quantity with the statistical results supported. We also managed to find the influence that the overlapping individuals have on the assessment of the algorithm. The experiments illustrated that MOEAs with overlapping individuals removed gained solution sets with better diversity than that obtained by the original ones. The famous NSGA-II was taken as the example MOEA and the conclusion can be extended to other MOEAs.