软件导刊2025，Vol.24Issue(5)：87-96,10.DOI:10.11907/rjdk.241166
基于强化学习的改进差分演化算法求解Thomson问题
Reinforcement Learning-Based Improved Differential Evolutionary Algorithm for Solving Thomson Problems
摘要
Abstract
Aiming at the optimal arrangement problem of point charges with the lowest energy on a sphere proposed by physicist Thomson,an improved differential evolution algorithm based on reinforcement learning,RLHDE_ILS,is proposed.To improve the search capability of the algorithm,a strategy based on reinforcement learning adaptive control differential evolution algorithm scaling factor is designed,using rein-forcement learning agents to select the most suitable parameter values for each generation;At the same time,in order to enhance the local search capability of the algorithm,a two-stage hybrid local search operator combining stochastic gradient descent and sequential quadratic pro-gramming is proposed for fine search in the later stages of iteration.Simulation experiments were conducted on 8 different scales of Thomson problems,and the results showed that compared with 15 other representative related algorithms,the RLHDE_ILS algorithm has faster conver-gence speed and higher solution accuracy in dealing with Thomson problems.关键词
球面点分布/Thomson问题/差分演化算法/强化学习/局部搜索Key words
spherical point distribution/Thomson problem/differential evolution algorithm/reinforcement learning/local search分类
信息技术与安全科学引用本文复制引用
张婉冰,戴光明,彭雷,王茂才,宋志明,陈晓宇,袁卓铭..基于强化学习的改进差分演化算法求解Thomson问题[J].软件导刊,2025,24(5):87-96,10.基金项目
国家自然科学基金项目(42271391,62006214) （42271391,62006214）
装备预研教育部联合基金项目(8091B022148) （8091B022148）
湖北省重点研发专项(2023BIB015) （2023BIB015）
湖北省优秀中青年科技创新团队计划项目(T2021031) （T2021031）
