• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

协同进化混合蛙跳算法

戴月明,张明明,王艳   

  1. (江南大学物联网工程学院,江苏 无锡 214122)
     
  • 收稿日期:2016-05-16 修回日期:2016-08-30 出版日期:2018-01-25 发布日期:2018-01-25
  • 基金资助:

    国家863计划(2014AA041505);国家自然科学基金(61572238)

A coevolutionary shuffled frog leaping algorithm

DAI Yue-ming,ZHANG Ming-ming,WANG Yan   

  1. (School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China)

     
  • Received:2016-05-16 Revised:2016-08-30 Online:2018-01-25 Published:2018-01-25

摘要:

针对基本混合蛙跳算法收敛速度慢、求解精度低且易陷入局部最优的问题,提出了一种新的协同进化混合蛙跳算法。该算法在局部搜索策略中,对子群内最差个体的更新引入平均值的同时充分利用最优个体的优秀基因,可有效扩大搜索空间,增加种群的多样性;同时对子群内少量的较差青蛙采取交互学习策略向邻近子群的最优个体交流学习,增加子群间交互的频繁性,提高信息共享程度,有利于进化。在全局迭代过程中采取精英群自学习进化机制,以对精英空间进行精细搜索,获得更优解,进一步提升算法的全局寻优能力,正确导向算法的进化。实验结果表明,所提算法在七个测试函数中均能收敛到最优解0,成功率为100%,优于其他对比算法。所提算法可有效避免陷入早熟收敛,极大地提高了算法的收敛速度和优化精度。

关键词:

Abstract:

To overcome the shortcomings of the basic shuffled frog leaping algorithm (SFLA), such as slow convergence speed, low optimization precision and falling into local optimum easily, etc., we propose a novel coevolutionary shuffled frog leaping algorithm (CSFLA). As for the local search strategy of the proposed algorithm, we introduce the average value and make full use of the best individuals to update the worst frog individuals in the subgroup, which expands the searching space effectively and increases the diversity of population. Meanwhile, the interactive learning strategy for a small number of worse frog individuals in the subgroup is adopted to learn from the best frog individuals in neighboring subgroups, which increases the frequency of interaction between subgroups and improves the degree of information sharing, thus conducive to algorithm evolution. In the process of global iteration, we adopt the elite group self-learning mechanism to search through the elite space for a better solution, which can further improve the ability of global optimization and lead the evolution of algorithm for the better. Experimental results show that the proposed algorithm can converge to the optimal solution 0 in the seven test functions, and the success rate is 100%, which is better than the other comparative algorithms. The proposed algorithm can effectively avoid premature convergence, and greatly improve the convergence speed and the accuracy of the algorithm.
 

Key words: shuffled frog leaping algorithm(SFLA), interaction learning, elites, self-learning, coevolution