面向多峰函数优化的PASFLI算法及应用

PASFLI algorithm for multimodal function optimization and its application

针对基本混合蛙跳算法在高维多峰函数优化时早熟及难以找到所有全局极值的问题,提出了一种具有混合智能的多态子种群自适应混合蛙跳免疫算法,证明了算法以概率1收敛于全局最优解。该算法采用双层进化模式,融合了混合蛙跳、免疫克隆选择技术。在低层混合蛙跳操作中,加入了多态自适应子种群机制,提高了子种群多样性,有效抑制了早熟现象;在算法进化后期,提出了全局极值筛选策略,将子种群极值点提升到高层免疫克隆选择操作,进一步提高了全局寻优能力。通过复杂多峰函数仿真实验,表明该算法能够快速有效地给出全部全局最优解。

As basic Shuffled Frog Leaping Algorithm （SFLA） has many problems in the high dimensional multimodal function optimization such as premature and difficult to find all global extremes, a hybrid intelligent Polymorphic Adaptive Shuffled Frog Leaping Immune Algorithm（PASFLIA） is presented. PASFLIA which combines the SFLA and immune Clonal Selection Algorithm （CSA） technology is convergent to the global optimal solution with probability 1. PASFLIA adopts a double-layer model evolu- tion： in the lower layer, the polymorphic adaptive population mechanism is joined to SFLA, which improves the diversity of the population, and effectively suppresses premature behavior of the convergence progress; in the higher layer, the global extreme screening strategy is proposed, which puts the population extreme points to high-level immune clonal selection operation, further enhancing the global optimization ability. The complex multimodal function simulation experiments results show that the PASFLIA can quickly and efficiently give all the global optimal solution.