基于协方差矩阵调整的多目标多任务优化算法

doi:10.19678/j.issn.1000-3428.0062365

摘要/Abstract

摘要： 多任务进化（EMT）是进化计算领域的一个新兴研究方向，区别于传统的单任务搜索算法，EMT通过在任务间传递有用知识，对多个任务同时实施进化搜索，以提升多个任务的收敛性能。目前，大多数进化算法只考虑了知识迁移而忽略了任务间的联系。提出一种多目标多任务优化算法，结合迁移学习的思想，采用任务间种群的协方差矩阵差异表示任务间种群分布特征差异，使用任务间种群均值的距离表示任务间种群的分布距离，并通过任务间种群的分布特征差异和分布距离表示任务间的相似度。对于某个目标任务，将其最相似任务中的解集实施K最近邻分类，以筛选出对目标任务有价值的解，并使其迁移到目标任务中。实验结果表明，与EMTSD、MaTEA、MO-MFEA-II等多目标多任务优化算法相比，所提算法具有较佳的收敛性能，平均运行效率约提高了66.62%。

关键词: 多目标多任务优化, 进化算法, 多任务进化, 迁移学习, 协方差矩阵

Abstract: Evolutionary Multi-Tasking (EMT) is an emerging research direction in the field of evolutionary computations.Differing from a traditional single-task search algorithm, EMT implements an evolutionary search for multiple tasks simultaneously by transferring useful knowledge between tasks, thereby improving the convergence of such tasks.At present, most evolutionary algorithms focus on knowledge transferring and ignore the connections between tasks.Therefore, by combining the ideas of transfer learning, a Multi-Objective Multi-Tasking Optimization(MTO) algorithm is proposed.First, the difference in the covariance matrix between tasks represents the difference in population distribution characteristics between tasks, and the distance between the means of different tasks is used to represent the distribution distance of the population between tasks.Thus, the difference in population distribution characteristics and the population distribution distance between tasks are used to express the similarity between tasks.For the target task, K-Nearest Neighbor (KNN) classification of the solutions to the most similar task is conducted, so as to screen out valuble solutions to target task and then transfer them to the target task.The experimental results show that, compared with EMTSD, MaTEA, MO-MFEA-II, and other MTO algorithms, the proposed algorithm achieves a better convergence, and the average running efficiency is increased by approximately 66.62%.

Key words: Multi-Objective Multi-Tasking Optimization(MTO), evolutionary algorithm, Evolutionary Multi-Tasking(EMT), transfer learning, covariance matrix

中图分类号:

TP391

邱鸿辉, 刘海林, 陈磊. 基于协方差矩阵调整的多目标多任务优化算法[J]. 计算机工程, 2022, 48(8): 306-312.

QIU Honghui, LIU Hailin, CHEN Lei. Multi-Objective Multi-Tasking Optimization Algorithm Based on Adjustment of Covariance Matrix[J]. Computer Engineering, 2022, 48(8): 306-312.

http://www.ecice06.com/CN/Y2022/V48/I8/306

图/表 7

20220825094939

20220825094942

20220825094946

20220825094950

20220825094954

20220825094957

20220825095001

参考文献

[1] PISARSKI S, RUGALA A, BYRSKI A, et al.Evolutionary multi-agent system in hard benchmark continuous optimisation[M].Berlin, Heidelberg:Springer, 2013:132-141.
[2] 李智翔, 李赟, 褚衍杰.基于改进平衡策略的多目标分解优化算法[J].计算机工程, 2019, 45(3):155-161. LI Z X, LI Y, CHU Y J.Decomposition multi-objective optimization algorithm based on improved balancing strategy[J].Computer Engineering, 2019, 45(3):155-161.(in Chinese)
[3] ZHU Z X, WANG F X, HE S, et al.Global path planning of mobile robots using a memetic algorithm[J].International Journal of Systems Science, 2015, 46(11):1982-1993.
[4] HIERONS R M, LI M Q, LIU X H, et al.SIP:optimal product selection from feature models using many-objective evolutionary optimization[J].ACM Transactions on Software Engineering and Methodology, 2016, 25(2):17-29.
[5] ONG Y S, GUPTA A.Evolutionary multitasking:a computer science view of cognitive multi-tasking[J].Cognitive Computation, 2016, 8(2):125-142.
[6] XU Q Z, WANG N, WANG L, et al.Multi-task optimization and multi-task evolutionary computation in the past five years:a brief review[J].Mathematics, 2021, 9(8):864-873.
[7] YANG C E, DING J L, JIN Y C, et al.Multitasking multi-objective evolutionary operational indices optimization of beneficiation processes[J].IEEE Transactions on Automation Science and Engineering, 2019, 16(3):1046-1057.
[8] BALI K K, GUPTA A, FENG L, et al.Linearized domain adaptation in evolutionary multitasking[C]//Proceedings of 2017 IEEE Congress on Evolutionary Computation.Washington D.C., USA:IEEE Press, 2017:1295-1302.
[9] GUPTA A, ONG Y S, FENG L, et al.Multi objective multifactorial optimization in evolutionary multitasking[J].IEEE Transactions on Cybernetics, 2017, 47(7):1652-1665.
[10] FENG L, ZHOU L, ZHONG J H, et al.Evolutionary multitasking via explicit autoencoding[J].IEEE Transactions on Cybernetics, 2019, 49(9):3457-3470.
[11] SHEN Y, LI Y, KANG H W, et al.Research on swarm size of multi-swarm particle swarm optimization algorithm[C]//Proceedings of the 4th International Conference on Computer and Communications.Washington D.C., USA:IEEE Press, 2018:37-43.
[12] 邱宁佳, 李娜, 胡小娟, 等.基于粒子群优化的朴素贝叶斯改进算法[J].计算机工程, 2018, 44(11):27-32, 39. QIU N J, LI N, HU X J, et al.Improved native Bayes algorithm based on particle swarm optimization[J].Computer Engineering, 2018, 44(11):27-32, 39.(in Chinese)
[13] 刘树强, 秦进.一种求解动态优化问题的改进自适应差分进化算法[J].计算机工程, 2021, 47(4):84-91, 99. LIU S Q, QIN J.An improved self-adaptive differential evolution algorithm for solving dynamic optimization problem[J].Computer Engineering, 2021, 47(4):84-91, 99.(in Chinese)
[14] 吕铭晟, 沈洪远, 李志高, 等.多变异策略差分进化算法的研究与应用[J].计算机工程, 2014, 40(12):146-150. LÜM S, SHEN H Y, LI Z G, et al.Research and application of differential evolution algorithm under multiple mutation strategy[J].Computer Engineering, 2014, 40(12):146-150.(in Chinese)
[15] CHEN Y L, ZHONG J H, TAN M K.A fast memetic multi-objective differential evolution for multi-tasking optimization[C]//Proceedings of 2018 IEEE Congress on Evolutionary Computation.Washington D.C., USA:IEEE Press, 2018:1-8.
[16] LIN J B, LIU H L, TAN K C, et al.An effective knowledge transfer approach for multi objective multitasking optimization[J].IEEE Transactions on Cybernetics, 2021, 51(6):3238-3248.
[17] SAGARNA R, ONG Y S.Concurrently searching branches in software tests generation through multitask evolution[C]//Proceedings of 2016 IEEE Symposium Series on Computational Intelligence.Washington D.C., USA:IEEE Press, 2016:46-56.
[18] ZHOU L, FENG L, ZHONG J H, et al.A study of similarity measure between tasks for multifactorial evolutionary algorithm[C]//Proceedings of the Genetic and Evolutionary Computation Conference Companion.New York, USA:ACM Press, 2018:39-46.
[19] GUPTA A, ONG Y S, DA B, et al.Landscape synergy in evolutionary multitasking[C]//Proceedings of 2016 IEEE Congress on Evolutionary Computation.Washington D.C., USA:IEEE Press, 2016:112-129.
[20] LIANG J, QIAO K J, YUAN M H, et al.Evolutionary multi-task optimization for parameters extraction of photovoltaic models[J].Energy Conversion and Management, 2020, 207:112509-112524.
[21] CHEN Y L, ZHONG J H, FENG L, et al.An adaptive archive-based evolutionary framework for many-task optimization[J].IEEE Transactions on Emerging Topics in Computational Intelligence, 2020, 4(3):369-384.
[22] BALI K K, GUPTA A, ONG Y S, et al.Cognizant multitasking in multiobjective multifactorial evolution:MO-MFEA-II[J].IEEE Transactions on Cybernetics, 2021, 51(4):1784-1796.
[23] JIA Y H, ZHOU Y R, LIN Y, et al.A distributed cooperative co-evolutionary CMA evolution strategy for global optimization of large-scale overlapping problems[J].IEEE Access, 2019, 7:19821-19834.
[24] HANSEN N.The CMA evolution strategy:a comparing review[EB/OL].[2021-07-09].https://link.springer.com/chapter/10.1007/3-540-32494-1_4.
[25] YUAN Y, ONG Y S, FENG L, et al.Evolutionary multitasking for multiobjective continuous optimization:benchmark problems, performance metrics and baseline results[EB/OL].[2021-07-09].https://arxiv.org/abs/1706.02766.

选择文件类型/文献管理软件名称

选择包含的内容