摘要
提出一种新的基于Pareto多目标进化免疫算法(PMEIA)。算法在每一代进化群体中选取最优非支配抗体保存到记忆细胞文档中;同时引入Parzen窗估计法计算记忆细胞的熵值,根据熵值对记忆细胞文档进行动态更新,使算法向着理想Pareto最优边界搜索。此外,算法基于点在目标空间分布情况进行克隆选择,有利于得到分布较广的Pareto最优边界,且加快了收敛速度。与已有算法相比,PMEIA在收敛性、多样性,以及解的分布性方面都得到很好的提高。
This paper proposed a new pareto-based muhi-object evolutionary immune algorithm (PMEIA). PMEIA selected optimal non-dominated antibodies which were then reserved in memory cell archive, and introduced Parzen window to calculate entropy of memory cells. Updated the memory cell archive according to entropy of memory cells. This guarantees the convergence to the true Pareto front. Moreover, the performance of clone selection was dependent on distribution in the objective space, which was favorable for getting a widely spread Pareto front and improving convergence speed. Compared with the existed algorithms , the obtained solutions of PMEIA have much better performance in the convergence, diversity and distribution.
出处
《计算机应用研究》
CSCD
北大核心
2009年第5期1687-1690,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(50778109)
上海市重点学科建设项目(J50103)