基于多模态函数优化的改进克隆选择算法

文章分析了deCastro和VonZuben在2002年提出的用于多模态函数优化的克隆选择算法(CLONALG)的不足,并且运用小生境技术、记忆方法、梯度法和相似性抑制法对该算法进行了改造,提出了小生境克隆选择算法(NCSA)。利用马尔柯夫链为数学工具,从理论上证明了NCSA的完全收敛性(CompleteConvergence)。该算法与CLONALG进行了仿真比较实验,不仅验证了NCSA理论上的完全收敛性结论,同时验证了所提算法对于求解多模态问题更为有效,且具有很好的稳定性。

An Improved Clonal Selection Algorithm for Multi-modal Function Optimization

This paper analyses the disadvantages of CLONALG which was proposed by de Castro and Von Zuben for multimodal function optimization in 2002.Using niche technology,memory method,gradient method and resemble restrain method to reform it,we propose a Niche Clonal Selection Algorithm(NCSA) for multimodal function optimization.We prove the complete convergence of NCSA in theory with Markov Chain.The contrast tests between CLONALG and NCSA validate the theory conclusion and show the stability of NCSA for multimodal function optimization.