电子学报 ›› 2019, Vol. 47 ›› Issue (1): 40-48.DOI: 10.3969/j.issn.0372-2112.2019.01.006

• 学术论文 • 上一篇    下一篇

基于变因子加权学习与邻代维度交叉策略的改进CSA算法

赵世杰, 高雷阜, 于冬梅, 徒君   

  1. 辽宁工程技术大学优化与决策研究所, 辽宁阜新 123000
  • 收稿日期:2017-05-02 修回日期:2018-01-26 出版日期:2019-01-25
    • 作者简介:
    • 赵世杰 男,1987年出生,山东五莲人.博士,辽宁工程技术大学讲师.主要研究领域为人工智能与数据挖掘、优化与管理决策等.E-mail:zhao2008shijie@126.com;高雷阜 男,1963年出生,辽宁阜新人.辽宁工程技术大学教授、博士生导师.主要研究领域为最优化理论与应用、非线性动力系统等.E-mail:gaoleifu@163.com;于冬梅 女,1986出生,辽宁鞍山人.博士,辽宁工程技术大学讲师.主要研究领域为最优化理论与应用;徒君 男,1982出生,安徽全椒人.博士,辽宁工程技术大学讲师.主要研究领域为智能算法、供应链管理等.
    • 基金资助:
    • 辽宁省博士启动基金 (No.20170520075); 国家自然科学基金青年基金 (No.51704140); 辽宁省教育厅基金 (No.LJ2017QL031,No.LJYL043); 辽宁省社科规划基金 (No.L17BGL004)

Improved Crow Search Algorithm Based on Variable-Factor Weighted Learning and Adjacent-Generations Dimension Crossover Strategy

ZHAO Shi-jie, GAO Lei-fu, YU Dong-mei, TU Jun   

  1. Institute of Optimization and Decision, Liaoning Technical University, Fuxin, Liaoning 123000, China
  • Received:2017-05-02 Revised:2018-01-26 Online:2019-01-25 Published:2019-01-25

摘要: 针对乌鸦搜索算法(CSA)优化高维问题时存在寻优精度低、局部极值逃逸能力弱等问题,提出一种耦合多个体变因子加权学习机制与最优个体邻代维度交叉策略的改进乌鸦搜索算法(ICSA).该算法随迭代进程动态修正模型控制参数(感知概率和飞行长度),利用多个体的变因子加权学习机制保证子代个体同时继承跟随乌鸦与上代最优个体的位置信息以避免单个体继承的过快种群同化并减小陷入局部极值的风险;同时构建历史最优个体的邻代维度交叉策略,并按维度绝对差异大的优先替换原则更新最优个体位置,以保留历代最优维度信息并提高算法的局部极值逃逸能力.数值实验结果分别验证了模型参数对CSA算法性能的一定影响,加权学习因子不同递变形式对ICSA算法性能改善的有效性与差异性以及改进算法的优越寻优性能.

关键词: 智能优化算法, 乌鸦搜索算法, 变因子加权学习机制, 邻代维度交叉策略, 基准测试函数

Abstract: Considering that crow search algorithm (CSA) has low optimization accuracy and weak local-optimum escape ability in optimizing high-dimensional problems,an improved crow search algorithm(ICSA) is proposed by coupling the variable-factors' weighted learning mechanism of multiple individuals(Mi-VWL) and the adjacent-generations dimension crossover strategy of the best individual(Bi-ADC).In the proposed algorithm,the model parameters,i.e.awareness probability and flight length,are firstly modified dynamically with increasing number of iterations.Meanwhile,the Mi-VWL is introduced to guarantee that offspring individuals of crow population can inherit position information from the followed crow and the best individual of the last generation simultaneously,which is advantageous to avoid the over-rapid population intensification of single-individual learning and reduce the algorithm's risk on dropping into local optimum.Furthermore,Bi-ADC is constructed and the priority replacement principle of larger absolute value difference of dimensions between two-generations is adopted to update position of the best individual,which is beneficial to retain the optimal dimension information of historical best crows and enhance the local extreme escape ability of algorithms.Experimental results verify the influence of modal parameters on CSA's performance,the effectiveness and differences of different-type weighted learning factor on improving ICSA's capability and the superior optimization ability of the proposed algorithm,respectively.

Key words: intelligent optimization algorithm, crow search algorithm, variable-factor weighted learning mechanism, adjacent-generations dimension crossover strategy, benchmark function

中图分类号: