计算机工程与应用 ›› 2019, Vol. 55 ›› Issue (3): 147-153.DOI: 10.3778/j.issn.1002-8331.1602-0058

• 模式识别与人工智能 • 上一篇    下一篇

基于混沌和自适应搜索策略的GSO算法分析与优化

黄宇达1,王迤冉2,牛四杰3   

  1. 1.周口职业技术学院 信息工程学院,河南 周口 466000
    2.周口师范学院 计算机科学与技术学院,河南 周口 466000
    3.南京理工大学 计算机科学与工程学院,南京 210094
  • 出版日期:2019-02-01 发布日期:2019-01-24

Analysis and Optimization of GSO Algorithm Based on Chaos and Self-Adaptive Search Strategy

HUANG Yuda1, WANG Yiran2, NIU Sijie3   

  1. 1.Institute of Information & Engineering, Zhoukou Vocational and Technical College, Zhoukou, Henan 466000, China
    2.College of Computer Science & Technology, Zhoukou Normal University, Zhoukou, Henan 466000, China
    3.College of Computer Science & Engineering, Nanjing University of Science & Technology, Nanjing 210094, China
  • Online:2019-02-01 Published:2019-01-24

摘要: 针对基本萤火虫群算法在全局优化问题求解过程中存在的求解精度偏低、易陷入局部最优、收敛速度较慢等问题,提出一种基于混沌和自适应搜索策略的萤火虫优化算法(CSAGSO)。利用混沌搜索技术对萤火虫种群进行初始化以得到分布更为均匀、合理的较优初始解;运用混沌扰动优化策略对每一代适应度较差的部分萤火虫个体进行混沌扰动以增强种群多样性和提高全局搜索能力。采用动态步长的自适应搜索策略,并对寻优过程中静止不动的萤火虫个体位置进行更新,加快了算法前期收敛速度,减少了后期震荡现象发生。仿真实验结果表明,优化后的萤火虫算法参数较少并具有较好稳定性,同时在求解精度和收敛速度上都明显优于基本萤火虫群算法。

关键词: 萤火虫群优化, Chebyshev混沌映射, 优化, 混沌扰动, 动态步长, 自适应搜索

Abstract: According to basic Glowworm Swarm Optimization(GSO) algorithm has lower precision defects, easily falling into local optimum value and slow convergence speed in solving global optimization problems, an improved Chaos and Self-Adaptive Search Glowworm Swarm Optimization(CSAGSO) algorithm based on chaos and self-adaptive search strategy is proposed. By using chaotic search technology to initialize the glowworm population in order to achieve more reasonable and uniformly distributed initial solutions; meanwhile, CSAGSO applies chaos perturbation optimization strategy to disturb some individuals with low fitness values each generation so as to increase the diversity of the population and the global search ability. CSAGSO applies the self-adaptive search strategy with dynamic step size, and updates the static glowworm location in the process of searching the best value, and so accelerates the convergence speed of the algorithm’s early stage, reduces the shock phenomenon of the algorithm’s late stage. Simulation experiment results show that CSAGSO has less parameters and better stability, and is significantly superior to GSO in computational precision and convergence rate.

Key words: Glowworm Swarm Optimization(GSO), Chebyshev chaotic map, optimization, chaos perturbation, dynamic step, self-adaptive search