基于信息共享策略的加权逐点预测动态多目标优化  

WEIGHTED POINT-WISE PREDICTION DYNAMIC MULTIPLE OBJECTIVE OPTIMIZATION BASED ON INFORMATION SHARING STRATEGY

作　　者：包全磊[1] 陈红星[2] Bao Quanlei;Chen Hongxing(Taiyuan University,Taiyuan 030032,Shanxi,China;College of Computer and Information Technology,Shanxi University,Taiyuan 030006,Shanxi,China)

机构地区：[1]太原学院,山西太原030032 [2]山西大学计算机与信息技术学院,山西太原030006

出　　处：《计算机应用与软件》2024年第6期273-281,共9页Computer Applications and Software

基　　金：山西省专利推广项目计划(201005-201105)。

摘　　要：为了有效实现Pareto最优解,并且提升对于输入误差的鲁棒性,提出一种基于信息共享策略的加权逐点预测动态多目标优化算法。引入一种信息共享策略,该策略允许每个点利用其相邻解决方案中的信息有效提升模型的鲁棒性。引入一个相似性度量,并对其与一些常用的相似性度量进行对比分析,进一步提出一种加权逐点预测方法,逐点特性极大增强了捕捉各种模式的能力。在EC2018 DMO测试套件上的实验结果验证了该方法的有效性。In order to achieve the Pareto optimal solution effectively and improve the robustness to the input error,a weighted point-wise prediction dynamic multiple objective optimization algorithm based on information sharing strategy is proposed.An information sharing strategy was introduced,which allowed each point to make use of the information in its adjacent solutions to improve the robustness of the model.A similarity measure was introduced,and by comparing with some common similarity measures,a weighted point-wise prediction method was proposed,which greatly enhanced the ability to capture various patterns.Experimental results on EC2018 DMO test suite show the effectiveness of the proposed method.

关 键 词：信息共享 鲁棒性 多目标优化 加权逐点预测 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]