基于自适应邻域选择的多目标免疫进化算法

在分析以往求解多目标进化算法中个体选择方法的基础上,给出了一种基于个体邻域的选择方法,分析表明这种选择方法可有效地维持群体的多样性,且个体的适应度在选择过程中将随着该个体邻域中所包含个体数目作自适应调整,文中称之为基于个体邻域的自适应校正选择方法。此外,由于每一个待求问题本身或多或少都有自身一些基本的、显见的特征信息或知识。因此,在求解过程中忽视问题本身的特征信息或舍弃可供应用的信息,有时并不是一个明智之举。基于以上考虑,在传统进化算法的基础上又引入免疫算子,其中免疫算子依次通过疫苗提取、接种疫苗和免疫选择3个步骤来完成,进而设计了一种基于邻域选择的多目标免疫进化算法。最后,用算法分别对2个变量和30个变量的双目标优化问题进行数值模拟的结果表明,算法都能够找到所给问题的分布较均匀且涵盖范围较宽广的Pareto最优解集,显示了算法的有效性及可行性。

Adaptive neighborhood-based selection for multi-objective immunity genetic algorithm

A neighborhood-based selection for multi-objective immunity genetic algorithm is presented on the basis of analyzing the multi-objective genetic algorithm in literatures, a further analysis indicates the given selection method can keep diversity of the population effectively, and the fitness of each individual will be adjusted along with the number of individuals containing in its neighborhood, which we call an adapted selection method based on its neighborhood. In addition, each pending problem has some essential and apparent characteristic information more or less, hence, it is not a wise choose to ignore or discard these information. Considering the above reasons, an immune operator into classical genetic algorithm is given and a neighborhood-based selection for multi-objective immunity genetic algorithm is designed, where the immune operator is realized by vaccine extraction, vaccination and immune selection in turn. Finally, the numerical simulations show the proposed method can find the pareto-solutions with maximum possible converge and uniformity along the Pareto front.