粒子群算法对高维问题的优化研究  

Study on the High Dimensional Problem Optimization by Particle Swarm Algorithm

作　　者：郝武伟[1] 李俊吉[2] HAO Wu-wei;LI Jun-ji(Economic Management Department, Shanxi Traffic Vocational and Technical College, Taiyuan 030031, China;Personnel and Education Division, Taiyuan University of Science and Technology, Taiyuan 030024, China)

机构地区：[1]山西交通职业技术学院经济管理系,太原030031 [2]太原科技大学人事教育处,太原030024

出　　处：《控制工程》2018年第5期870-877,共8页Control Engineering of China

基　　金：国家自然科学基金(61472269)

摘　　要：基于Q-learning机器学习技术的粒子群优化算法（PSO）可以提高PSO对高维问题的优化效果。首先,缩小粒子群的种群大小,通过Q-learning机器学习技术管理PSO粒子的行为;然后,Q-learning机器学习技术根据粒子的性能自适应地切换粒子的操作,性能好的操作受到奖赏,性能差的操作受到惩罚;最终,通过Q-learning学习技术的全局寻优能力来弥补PSO局部优化能力的不足。通过多组仿真实验的结果表明,该算法提高了PSO算法对高维问题的优化性能与收敛速度。Particle swarm optimization algorithm（PSO） based on Q-learning machine learning technique can improve the optimization effect of PSO on high dimensional problems. Firstly, the size of particle swarm is reduced, and the behavior of particles are managed by Q-learning machine learning technique; then, the operations of particles are switched adaptively by Q-learning machine learning technique according to the particles, the operations is rewarded if their performances are improved, otherwise, they are punished; lastly, the disadvantage of local optimization of PSO is fixed by Q-learning technique. Several simulation experimental results show that the proposed algorithm improves the optimization effect and convergence speed of PSO algorithm when applied to high dimensional problems.

关 键 词：粒子群优化算法 机器学习 收敛速度 组合问题 局部优化 全局优化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]