多种群协同进化数值优化算法  

Muti-population coevolutionary algorithm for numerical optimization

作　　者：彭复明[1] 

机构地区：[1]南京工业职业技术学院计算机与软件学院,南京210046

出　　处：《计算机应用》2011年第3期660-665,共6页journal of Computer Applications

摘　　要：为了提高进化算法的抗早熟性与效率,提出了一种基于多种群的新算法。根据杂种优势理论的原理,算法让多个种群同时进化。各个种群之间既相对隔离又分工合作,目的就是保持种群的多样性。不同类型的种群采用不同的算子,并在不同的栖息地繁殖后代;不同类型的种群分别担负着广度与深度的搜索任务,以便算法能够收敛到高精度的全局最优解。多个数值实验也验证了新算法的优良性能。To improve the anti-premature ability and efficiency of evolutionary algorithm, a new algorithm based on multi-population was proposed. According to the heterotic theory, the algorithm made multi-populations concurrently evolve. The populations were both relatively isolated and cooperative to keep the diversity of populations. As a metaphor for creatures breeding in different habitats, different populations used different operators. Such a setting balances the breadth and depth of the algorithm, which made the algorithm to be converged to global optimal solutions. The good performance of the presented algorithm was validated in the experiments as well.

关 键 词：多种群 适应重心 小栖息地 大栖息地 配子 跳跃基因 杂种优势 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]