一种新的免疫算法及其在多模态函数优化中的应用

提取免疫应答的部分简化机制并结合小生境技术,提出一种用于多峰值或非连续函数优化的免疫算法.该算法由记忆细胞获取、克隆选择、亲和突变及群体更新这四种算子模块构成.这些算子的有机组合不仅为最优化问题的解决提供了实用新方法,而且反映了抗体应答抗原的简化运行机制.算法设计的重点是借鉴小生境共享实现方法的思想建立有助于增强群体多样性及保留优良抗体的记忆细胞获取算子,以及利用亲和成熟机理设计抗体突变算子.所获算法具有整体和局部搜索能力及并行搜索特点.理论证明了其收敛性.仿真事例比较表明此算法不仅是有效的,而且能快速搜索到多个最优解(针对于多解最优化问题).

Novel immune algorithm and its application to multi-modal function optimization

A novel immune algorithm, suitable for function optimization problems with either multimodal or noncontinuous objective functions, was proposed based on four immune operators of memory cell keeping, clone selection, somatic mutation and population updating from a niching technique and several immune mechanisms. The algorithm, of which had capability of global and local searching and characteristics of parallel searching, not only provided a practical method for solving optimization problems, but also reflected on some simple performance characteristics of antibodies to antigens. The key of its design was to construct a memory cell-keeping operator based on a niching method that could maintain diversity of population and keep good antibodies of current populations, and to design a mutation operator by utilizing the metaphor of affinity maturation. Further, its convergence was proved in theory. Simulation comparison showed that the algorithm can accomplish effectively optimization tasks, and specially search rapidly more than one of optimal solutions for certain optimization problems.