多群落双向驱动协作搜索算法

发布时间：2018-06-10 06:40

本文选题：混杂多变数据 + 多群落　；参考：《计算机集成制造系统》2017年07期

【摘要】：针对复杂优化问题中数据混杂多变的特点,提出一种能够根据环境变化不断优化种群适应度的多群落双向驱动协作搜索算法。该算法在分析微粒群落特性的基础上,基于无向加权图建立了多群落协作网演化模型,该模型依据群落适应值的优劣程度对群落类型进行划分,并根据不同群落间的协作权重和群落节点响应度评估群落节点强度,由节点强度最大的群落引导整个协作网进化,改进传统群集智能算法面对复杂优化问题中环境变化的自适应性能缺陷;构建了一种多群落双向驱动的进化新模式,给出了多群落协作的异步并行搜索算法,实现了不同环境下群落内部与群落之间的并行进化,降低了数据分析中巨大的计算时空开销。实验结果表明,该方法面向混杂多变数据不断优化种群适应度,能够较快地适应环境变化,并在可接受的时间内得到精确解,为复杂优化问题的求解提供了有效手段。
[Abstract]:In order to solve the complex optimization problem, a bidirectional cooperative search algorithm for multi-community is proposed, which can optimize the population fitness according to the changing environment. Based on the analysis of particle community characteristics, a multi-community cooperative network evolution model is established based on undirected weighted graph. The model is divided into community types according to the degree of community fitness. The community node strength was evaluated according to the cooperative weight and the community node response degree among different communities, and the community with the largest node strength led the whole cooperative network to evolve. This paper improves the adaptive performance of traditional cluster intelligence algorithm for complex optimization problems, constructs a new evolutionary model of bidirectional drive of multi-community, and presents an asynchronous parallel search algorithm for multi-community collaboration. The parallel evolution within and between communities is realized in different environments, and the huge computational time and space cost in data analysis is reduced. The experimental results show that this method can adapt to the environment change quickly and get the exact solution in the acceptable time, which provides an effective method for solving the complex optimization problem.
【作者单位】：昆明理工大学机电工程学院;
【基金】：国家自然科学基金资助项目(51365022) 云南省教育厅科学研究基金资助项目(2016YJS022)~~
【分类号】：TP18

【参考文献】