基于目标相关性的一种高维目标演化算法

A Many-Objective Evolutionary Algorithm Based on Objective Correlation

针对高维目标问题中非支配解数量随目标数量增加而剧增的问题,提出一种基于目标相关性信息的降维方法.该方法利用非支配解的目标值分析目标之间的相关性,对正相关较强的目标进行合并,从而降低目标数量,使部分非支配解之间产生支配关系,达到减少非支配解数量的目的.该方法可与基于Pareto支配的演化算法结合.实验结果表明,结合该目标降维方法的演化算法可以取得收敛性更好的结果.

Aiming at the problem caused by ever-increasing non-dominated solutions with the increasing in the number of objectives,an objective reduction method based on objective correlation is proposed.The method employs the objective values of non-dominated solutions to combine the high-positively correlative objectives,and accordingly reduces the number of objectives.The reduced objective set makes some former non-dominated solutions dominated by some others and thus the number of non-dominated solutions decreases.Then,the objective reduction method can be integrated into Pareto-based evolutionary algorithms.The experiments show that evolutionary algorithms combined with the objective reduction method can converge better.