计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (22): 35-41.DOI: 10.3778/j.issn.1002-8331.1710-0115

• 理论与研发 • 上一篇    下一篇

基于邻域引力学习的生物地理学优化算法

马  萍1,2,刘思含3,孙根云1,2,张爱竹1,2,郝艳玲1,2   

  1. 1.中国石油大学(华东) 地球科学与技术学院,山东 青岛 266580
    2.青岛海洋国家实验室 海洋矿产资源评价与探测技术功能实验室,山东 青岛 266071
    3.环境保护部卫星环境应用中心 国家环境保护卫星遥感重点实验室,北京 100094
  • 出版日期:2018-11-15 发布日期:2018-11-13

Neighbor force learning biogeography-based optimization

MA Ping1,2, LIU Sihan3, SUN Genyun1,2, ZHANG Aizhu1,2, HAO Yanling1,2   

  1. 1.School of Geosciences, China University of Petroleum(East China), Qingdao, Shandong 266580, China
    2.Laboratory for Marine Mineral Resources, Qingdao National Laboratory for Marine Science and Technology, Qingdao, Shandong 266071, China
    3.Satellite Environmental Center, Ministry of Environmental Protection, State Environmental Protection Key Laboratory of Satellite Remote Sensing, Beijing 100094, China
  • Online:2018-11-15 Published:2018-11-13

摘要: 针对生物地理学优化算法(Biogeography-Based Optimization,BBO)易发生早熟收敛、陷入局部最优的问题,提出一种基于邻域引力学习的生物地理学优化算法(Neighbor Force Learning Biogeography-Based Optimization,NFBBO)。该算法采用邻域选择的方法确定迁出栖息地,以充分利用栖息地的邻域信息,增加算法的种群多样性。同时采用引力学习策略对栖息地进行更新,拓展搜索空间,提高算法的搜索能力,避免早熟收敛问题。为使种群能够自适应地跳出局部最优,引入一种自适应高斯变异机制。基于高维标准测试函数的对比实验表明,NFBBO算法具有更快的收敛速度和更高的收敛精度。

关键词: 生物地理学优化算法, 邻域选择, 引力学习, 自适应高斯变异机制

Abstract: Biogeography-Based Optimization(BBO) easily suffers from the premature convergence and local optima trapping problems. In order to solve these issues, a new algorithm, named Neighbor Force Learning Biogeography-Based Optimization(NFBBO), is proposed in this paper. NFBBO presents a neighbor selection strategy, in which an emigrating solution is selected from the neighbors of the immigrating solution based on its suitability and distance. This operation can exploit the neighborhood information of swarm and improve the population diversity. Then, a force learning strategy is integrated with the migration operator to update the immigrating solutions. This strategy can expand the solutions search space and enhance the searching ability of BBO to avoid the prematurity. Furthermore, in order to escape from the local optima, an adaptive Gaussian mutation mechanism is further introduced, which is an effective jump-out mechanism. Experimental study is conducted on 10 well-known high-dimensional benchmark functions. The experimental results indicate that NFBBO has better search performance compared with other competing algorithms in terms of the convergence rate and the quality of the final solutions.

Key words: Biogeography-Based Optimization(BBO), neighbor selection strategy, force learning, adaptive Gaussian mutation mechanism