Neuro-PLS: A Generalizable Local Search Framework for Multiobjective Combinatorial Optimization

IF 15.9 1区 计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE
Haotian Zhang;Jialong Shi;Jianyong Sun;Qingfu Zhang;Zongben Xu
{"title":"Neuro-PLS: A Generalizable Local Search Framework for Multiobjective Combinatorial Optimization","authors":"Haotian Zhang;Jialong Shi;Jianyong Sun;Qingfu Zhang;Zongben Xu","doi":"10.1109/TEVC.2025.3589640","DOIUrl":null,"url":null,"abstract":"In this article, we present a novel neural multiobjective combinatorial optimization framework with generalization ability across problems. The main idea is to integrate the framework of Pareto local search (PLS), a well-known problem independent multiobjective heuristic, with the concept of learning-to-optimize (L2O). In the proposed framework, namely, Neuro-PLS, the backbone of PLS is guided by two neural networks to utilize cross-problem knowledge. Specifically, in Neuro-PLS, the solution-selection component is guided by a graph neural network (GNN) and the neighborhood-exploration component is guided by a multilayer perception. In the experimental studies, the training set of Neuro-PLS contains only one 200-dimensional multiobjective unconstrained binary quadratic programming problem (mUBQP) instance. The trained Neuro-PLS shows remarkable efficiency on optimizing large-size mUBQP instances and middle-size multiobjective TSP (mTSP) instances with different number of objectives. Extensive experiments on a variety of mUBQP and mTSP instances show that the trained Neuro-PLS significantly outperforms some recently proposed reinforcement learning-based methods.","PeriodicalId":13206,"journal":{"name":"IEEE Transactions on Evolutionary Computation","volume":"30 4","pages":"1564-1578"},"PeriodicalIF":15.9000,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Evolutionary Computation","FirstCategoryId":"94","ListUrlMain":"https://ieeexplore.ieee.org/document/11082360/","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2025/7/16 0:00:00","PubModel":"Epub","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0

Abstract

In this article, we present a novel neural multiobjective combinatorial optimization framework with generalization ability across problems. The main idea is to integrate the framework of Pareto local search (PLS), a well-known problem independent multiobjective heuristic, with the concept of learning-to-optimize (L2O). In the proposed framework, namely, Neuro-PLS, the backbone of PLS is guided by two neural networks to utilize cross-problem knowledge. Specifically, in Neuro-PLS, the solution-selection component is guided by a graph neural network (GNN) and the neighborhood-exploration component is guided by a multilayer perception. In the experimental studies, the training set of Neuro-PLS contains only one 200-dimensional multiobjective unconstrained binary quadratic programming problem (mUBQP) instance. The trained Neuro-PLS shows remarkable efficiency on optimizing large-size mUBQP instances and middle-size multiobjective TSP (mTSP) instances with different number of objectives. Extensive experiments on a variety of mUBQP and mTSP instances show that the trained Neuro-PLS significantly outperforms some recently proposed reinforcement learning-based methods.
神经pls:一种多目标组合优化的可推广局部搜索框架
本文提出了一种具有跨问题泛化能力的神经网络多目标组合优化框架。其主要思想是将著名的问题独立多目标启发式算法Pareto局部搜索(PLS)的框架与学习优化(L2O)的概念相结合。在本文提出的神经-PLS框架中,PLS的主干由两个神经网络引导,以利用交叉问题知识。具体而言,在神经pls中,解决方案选择组件由图神经网络(GNN)引导,邻域探索组件由多层感知引导。在实验研究中,神经pls的训练集只包含一个200维多目标无约束二元二次规划问题实例。训练后的神经pls在优化大型mUBQP实例和具有不同目标数的中型多目标TSP (mTSP)实例方面表现出显著的效率。在各种mUBQP和mTSP实例上进行的大量实验表明,训练后的神经pls显著优于最近提出的一些基于强化学习的方法。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
IEEE Transactions on Evolutionary Computation
IEEE Transactions on Evolutionary Computation 工程技术-计算机:理论方法
CiteScore
21.90
自引率
9.80%
发文量
196
审稿时长
3.6 months
期刊介绍: The IEEE Transactions on Evolutionary Computation is published by the IEEE Computational Intelligence Society on behalf of 13 societies: Circuits and Systems; Computer; Control Systems; Engineering in Medicine and Biology; Industrial Electronics; Industry Applications; Lasers and Electro-Optics; Oceanic Engineering; Power Engineering; Robotics and Automation; Signal Processing; Social Implications of Technology; and Systems, Man, and Cybernetics. The journal publishes original papers in evolutionary computation and related areas such as nature-inspired algorithms, population-based methods, optimization, and hybrid systems. It welcomes both purely theoretical papers and application papers that provide general insights into these areas of computation.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信
小红书