《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (7): 2237-2247.DOI: 10.11772/j.issn.1001-9081.2022060896

• 先进计算 • 上一篇    

信息迁移多任务优化共生生物搜索算法

程美英1,2(), 钱乾2,3, 熊伟清4   

  1. 1.湖州师范学院 经济管理学院, 浙江 湖州 313000
    2.浙江省教育信息化评价与应用研究中心(湖州师范学院), 浙江 湖州 313000
    3.湖州师范学院 教师教育学院, 浙江 湖州 313000
    4.宁波大学 商学院, 浙江 宁波 315211
  • 收稿日期:2022-06-21 修回日期:2022-08-22 接受日期:2022-08-24 发布日期:2022-09-22 出版日期:2023-07-10
  • 通讯作者: 程美英
  • 作者简介:程美英(1983—),女,安徽黄山人,副教授,博士,主要研究方向:群体智能优化;
    钱乾(1983—),男,安徽芜湖人,讲师,硕士,主要研究方向:进化计算;
    熊伟清(1966—),男,安徽合肥人,教授,硕士,主要研究方向:人工智能。
  • 基金资助:
    浙江省高等教育“十三五”第二批教学改革项目(jg20190652);国家自然科学基金资助项目(62102148)

Symbiotic organisms search algorithm for information transfer multi-task optimization

Meiying CHENG1,2(), Qian QIAN2,3, Weiqing XIONG4   

  1. 1.School of Economics and Management,Huzhou University,Huzhou Zhejiang 313000,China
    2.Research Center of Education Information Evaluation and Application of Zhejiang Province (Huzhou University),Huzhou Zhejiang 313000,China
    3.School of Teacher Education,Huzhou University,Huzhou Zhejiang 313000,China
    4.Business School,Ningbo University,Ningbo Zhejiang 315211,China
  • Received:2022-06-21 Revised:2022-08-22 Accepted:2022-08-24 Online:2022-09-22 Published:2023-07-10
  • Contact: Meiying CHENG
  • About author:CHENG Meiying, born in 1983, Ph. D., associate professor. Her research interests include swarm intelligence optimization.
    QIAN Qian, born in 1983, M. S., lecturer. His research interests include evolutionary computation.
    XIONG Weiqing, born in 1966, M. S., professor. His research interests include artificial intelligence.
  • Supported by:
    the Second Batch of Teaching Reform Project of Higher Education during “the 13th Five Year Plan” in Zhejiang Province(jg20190652);National Natural Science Foundation of China(62102148)

摘要:

针对现有共生生物搜索(SOS)算法只能求解单个任务,以及信息负迁移影响多任务优化(MTO)性能这两个难题,提出一个信息迁移多任务优化共生生物搜索(ITMTSOS)算法。首先基于多种群演化MTO框架,根据任务个数设置相应数量种群;然后各种群独立运行基本SOS算法,当某一种群连续若干代停滞进化时,引入个体自身最优经验和邻域最优个体以形成知识模块并将该模块迁移至该种群个体进化过程中;最后对ITMTSOS算法时间和空间复杂度进行分析。仿真实验结果表明,ITMTSOS算法同时求解多个不同形态高维函数时均能快速收敛至全局极值解0,与单任务SOS算法相比,平均运行时间最多缩短约25.25%;而在同时求解多维0/1背包问题和师生匹配问题时,所提算法在测试集weing1和weing7上的最优适应值与目前测试集公布的最优结果相比分别提高了22 767和22 602,师生最优匹配差和平均匹配差的绝对值分别下降了26和33,平均运行时间约缩短了7.69%。

关键词: 共生生物搜索算法, 多任务优化, 信息迁移, 多任务高维函数优化, 多任务二元离散优化

Abstract:

Aiming at the problems that Symbiotic Organisms Search (SOS) algorithm only can solve single tasks and negative information transfer affects Multi-Task Optimization (MTO) performance, an Information Transfer Multi-Task SOS (ITMTSOS) algorithm was proposed. Firstly, based on multi-population evolution framework MTO, multiple populations were set according to the number of tasks. Secondly, each population ran basic SOS algorithm independently, and by introducing individual itself optimal experience and neighborhood optimal individuals, the knowledge module containing the above two was formed and transferred to the process of individual evolution when a population stagnated for several consecutive generations. Finally, the time and space complexity of ITMTSOS was analyzed. Simulation results show that ITMTSOS converges rapidly to the global optimal solution 0 when resolving a batch of different shape high-dimensional functions, and the average running time is reduced around 25.25% when compared with single task SOS; when solving the multi-dimensional 0/1 knapsack problems and the teacher-student matching problems concurrently, the optimal fitnesses on weing1 and weing7 test sets are increased by 22 767 and 22 602 respectively compared with the current published optimal results, the absolute values of the optimal and the average matching difference of teacher-student matching problem are decreased by 26 and 33 respectively, and the average running time is reduced around 7.69%.

Key words: Symbiotic Organisms Search (SOS) algorithm, Multi-Task Optimization (MTO), information transfer, multi-task high-dimensional function optimization, multi-task binary discrete optimization

中图分类号: