求解多目标优化问题的分级变异量子进化算法  

Quantum-inspired immune evolutionary algorithm with classified mutation for solving multi-objective optimization problems

作　　者：杨海东[1] 杨春[1] 

机构地区：[1]华南理工大学自动化科学与工程学院,广州510640

出　　处：《控制与决策》2009年第6期894-898,902,共6页Control and Decision

基　　金：广东省自然科学基金项目(x2zdB6080490);教育部高校博士点基金项目(20070561081);广东省工业科技攻关项目(2007B010200046)

摘　　要：分析量子进化算法和免疫算子的特点,提出一种分级变异的量子进化算法,用于求解多目标优化问题.算法主要基于两个策略:首先,利用快速非受控排序和密度距离计算种群抗原-抗体的亲和度;然后,基于亲和度排序将个体进行分级,最优分级中的个体作为算法中的最优个体,大部分实施量子旋转更新和免疫操作,而剩余分级中的个体实施免疫交叉操作以获得新的个体补充种群.求解多目标0/1背包问题的实验结果表明了该算法的有效性.Based on quatum-inspired evolutionary algorithm and immune operator, a quantum-inspired immune evolutionary algorithm with classified mutation is proposed. The following two main strategies are introduced. First, the fast nondominated sorting and crowding distance calculation are used in the algorithm to compute the affinity between antigens and antibodies of the population. Then, the sorted population is classified into several groups. The individuals in the best group are utilized to update individuals in other groups. Most of the individuals perform immune operator and Q-gate update. And the left individuals perform quantam crossover to gain new individuals to the new generation population. Finally, the proposed algorithm is used for solving multiobjective 0/1 knapsack problem and the results show its effectiveness.

关 键 词：量子进化算法 多目标优化 分级变异 快速非受控排序 背包问题 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]