Abstract: In order to solve the existed problems that are the population size required to be determined by the experience, weaker multi-peak search capability and longer training time for Castro clone selection algorithm. It proposes a new immune clone selection algorithm based on real coding and adaptive zoom mutation method, which is able to dynamically determine the population size, owns strong global and local search capabilities and can search the global optimal points and possibly the greatest number of local extreme points. Simulation results show the improved algorithm to find the average running time and average number of peak points is obviously better than Castro clone selection algorithm, multimodal function optimization results are significantly improved.

Key words: artificial immune system, clone selection, real coding, adaptive zoom mutation

摘要： 为解决Castro克隆选择算法中存在的种群规模需根据经验确定、多峰搜索能力弱、训练时间长等问题，提出一种新的免疫克隆选择算法，该算法基于实数编码和自适应变焦变异方法，能够动态确定种群大小，具有很强的全局和局部搜索能力，可以搜索到全局最优点和尽可能多的局部极值点。仿真实验结果表明，该算法平均运行时间和平均找到的峰值点个数都明显优于Castro克隆选择算法，且多峰值函数的优化效果得到显著改善。

关键词: 人工免疫系统, 克隆选择, 实数编码, 自适应变焦变异

CLC Number: 

• TP311.52