| 具有动态学习能力的分层进化粒子群优化算法 |
| |
| 引用本文： | 徐 超,单志勇,徐好好.具有动态学习能力的分层进化粒子群优化算法[J].教育技术导刊,2021,20(1):128-131. |
| |
| 作者姓名： | 徐 超 单志勇 徐好好 |
| |
| 作者单位： | 1. 东华大学 信息科学与技术学院；2. 数字化纺织服装技术教育部工程研究中心，上海 201620 |
| |
| 基金项目： | 国家自然科学基金项目（61602110） |
| |
| 摘 要： | 针对粒子群算法（PSO）在优化高维多极值问题时容易陷入局部极值的问题，结合分层进化与动态学习策略，提出一种具有动态学习能力的分层进化粒子群优化算法（DHEPSO）。该算法首先根据粒子适应度值将粒子划分为不同层级，对不同层级粒子采取不同的进化策略，避免迭代后期种群多样性快速消失；然后根据粒子所属层级的不同动态调整粒子学习能力，在保证算法收敛精度情况下提高算法收敛速度；最后将算法在4个典型函数进行测试，结果表明DHEPSO与传统粒子群算法相比，除病态函数外均能快速达到全局最优。同时，问题维数提升对算法的全局收敛能力影响较小，证明该算法具有良好的稳定性。
|
| 关 键 词： | 粒子群算法 高维多极值 学习能力 分层进化 |
| 收稿时间： | 2020-04-23 |
A Hierarchical Evolutionary Particle Swarm Optimization Algorithm with Dynamic Learning Ability |
| |
| XU Chao,SHAN Zhi-yong,XU Hao-hao.A Hierarchical Evolutionary Particle Swarm Optimization Algorithm with Dynamic Learning Ability[J].Introduction of Educational Technology,2021,20(1):128-131. |
| |
| Authors: | XU Chao SHAN Zhi-yong XU Hao-hao |
| |
| Institution: | 1. School of Information Science and Technology， Donghua University； 2. Ministry of Education， Digital Textile Research Center， Shanghai 201620， China |
| |
| Abstract: | In order to optimize the high-dimensional multi-extreme problem of particle swarm optimization （PSO） which is easy to fall into the problem of local extremum， combining hierarchical evolution and dynamic learning strategies，a hierarchical evolutionary particle swarm optimization algorithm with dynamic learning ability （DHEPSO） is proposed. The algorithm first divides the particles into different levels according to the particle fitness value， and then the particles at different levels adopt different evolution strategies to avoid the rapid loss of population diversity in the later iterations， and dynamically adjust the particle learning according to the different levels of the particles’ ability to improve the convergence speed of the algorithm while ensuring the convergence accuracy of the algorithm. Finally， the algorithm is applied to four typical test functions. The results show that DHEPSO achieves the global optimum quickly except for the morbid function， and the improvement of the problem dimension has a smaller impact on the global convergence ability of the algorithm. which proves that the algorithm has good stability. |
| |
| Keywords: | particle swarm optimization high-dimensional multi-extreme value learning ability hierarchical evolution |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
