改进免疫算法在函数优化中的应用

在对已有克隆选择算法的抗体行为特征分析的基础上,提出了一种新的偏心动态免疫克隆算法(EDICA,Eccentric Dynamic Immune Clone Algorithm).利用进化过程中子代抗体比父代抗体更靠近最优解的启发性信息,提出偏心变异策略,使抗体更快地靠近最优解域.引入控制因子,通过动态调整变异搜索半径的方法,在进化初期加大步长以加快搜索速度,而在后期减小搜索粒度以提高优化精度.采用超球体混沌变异策略以克服各向异性的不利影响并提高全局搜索能力.实验结果表明:EDICA不仅能够准确地找到静态函数的多个最优点,而且还能以较高的精度锁定和跟踪动态函数的最优点.

Improved immune algorithm and applications on function optimization

A novel eccentric dynamic immune clone algorithm (EDICA) was proposed based on the analysis of antibody behavior features in existed clone selection algorithm (CSA). Heuristic information implicates that descendant antibodies are always better than their parents during evolution, which derive an eccentric mutation strategy, and let the mutation center shift a proper distance along the direction which is from parent to descendant, antibodies may search towards optima more quickly. A dynamic mutation radial adjustment method was proposed with some introduced control factors. The search speed was accelerated by enlarged mutation radial at initial stage. Then the search granularity was gradually diminished so as to improve optimization precision at later stage. A hyper sphere chaos mutation strategy was adopted to avoid the adverse effects of anisotropy and ensure the ability to successfully find global optima. Experiment results show that the EDICA could not only accurately discover most optima of static function but also hit and follow optima of dynamic function with high precision.