高级检索

    王学武, 夏泽龙, 顾幸生. 基于事件触发的自适应邻域多目标进化算法[J]. 华东理工大学学报(自然科学版), 2020, 46(1): 48-57. DOI: 10.14135/j.cnki.1006-3080.20181120005
    引用本文: 王学武, 夏泽龙, 顾幸生. 基于事件触发的自适应邻域多目标进化算法[J]. 华东理工大学学报(自然科学版), 2020, 46(1): 48-57. DOI: 10.14135/j.cnki.1006-3080.20181120005
    WANG Xuewu, XIA Zelong, GU Xingsheng. Multi-objective Evolutionary Algorithm with Adaptive Neighborhood Based on Event Triggering[J]. Journal of East China University of Science and Technology, 2020, 46(1): 48-57. DOI: 10.14135/j.cnki.1006-3080.20181120005
    Citation: WANG Xuewu, XIA Zelong, GU Xingsheng. Multi-objective Evolutionary Algorithm with Adaptive Neighborhood Based on Event Triggering[J]. Journal of East China University of Science and Technology, 2020, 46(1): 48-57. DOI: 10.14135/j.cnki.1006-3080.20181120005

    基于事件触发的自适应邻域多目标进化算法

    Multi-objective Evolutionary Algorithm with Adaptive Neighborhood Based on Event Triggering

    • 摘要: 为提高多目标算法的多样性与分布性,提出了一种基于事件触发的自适应邻域多目标进化算法(MOEA/D-ET)。采用事件触发策略协调全局与局部搜索,利用网格法进行全局寻优,利用自适应邻域MOEA/D进行局部寻优。对固定邻域与自适应邻域进行了对比,结果表明采用自适应邻域能有效地改善解分布不均的问题。通过对ZDT、WFG、DTLZ测试函数的求解,并与4个经典的多目标算法和2个最新的多目标算法进行对比,结果表明本文算法在收敛性和多样性方面具有一定的优越性。

       

      Abstract: In order to improve the diversity and distribution of multi-objective algorithms, this paper proposes an adaptive neighborhood multi-objective evolutionary algorithm based on event triggering (MOEA/D-ET). By using the event triggering strategy to coordinate the global search and the local search, the proposed algorithm can converge to the Pareto frontier and attain better diversity. The grid method is adopted to achieve global optimization such that the grid coordinate system can quickly constrain the solution to the vicinity of the frontier surface and attain a good diversity, which can provide an effective guidance for local search. The adaptive neighborhood MOEA/D is used for local optimization and the search direction is optimized by the adaptive neighborhood. The number of weight vectors is enlarged for the small coverage area for increasing the search intensity such that it can jump out of the local minimum, improve the diversity of the solution, and make the solution more widely distributed in space. Three different levels of fixed neighborhood method and adaptive neighborhood method are compared via the ZDT1 problem, by which it is shown that the proposed adaptive neighborhood method can effectively improve the density of solution in the central region and make the distribution of the solution more uniform. Finally, the simulation via ZDT1—ZDT4, WFG1—WFG4, DTLZ1—DTLZ4, and DTLZ7 test functions is undergone, whose results are compared with four classic multi-objective algorithms and two new multi-objective algorithms including NSGA-III. It is shown from these simulation results that the proposed algorithm can attain better distribution in most tests than other test algorithms and has greater advantages in convergence and diversity.

       

    /

    返回文章
    返回