摘要
针对适应度计算复杂问题,提出了一种基于改进信赖域方法的代理模型的适应度计算方法。该方法采用差分计算过程来代替信赖域方法中搜索时的导数计算过程,从而避免了在计算中遇到的函数不可导使算法无法正常进行的问题。并采用此改进的信赖域方法与微粒群算法(PSO)相结合更新采样空间,提高采样效率。再结合径向基神经网络(RBF)代理模型,进一步提高算法的优化性能。在几类典型的测试函数中进行仿真实验,结果表明本文算法具有较好的寻优能力。
Aimed at the fitness calculation complex problems, a method based on improved trust region of fitness calculation method of the Surrogate model is put forward. The method adopts the difference calculation process in- stead of a trust region method in search of derivative calculation process avoiding that the calculation of normal function cannot be take derivative. The sampling efficiency is improved using the improved trust region method combined with particle swarm optimization (PSO) to update the sampling space. Combining RBF neural network (RBF) surrogate model, the optimization performance of the algorithm is further improved. In several kinds of typ- ical test function simulation experiments, the results show that the algorithm has better search ability
出处
《太原科技大学学报》
2017年第2期93-97,共5页
Journal of Taiyuan University of Science and Technology
基金
国家自然科学基金资助项目(61403272)
校博士启动基金(20122026)
东北大学流程工业综合自动化国家重点实验室开放课题基金资助项目
关键词
信赖域
径向基函数
微粒群算法
适应度函数
差分法
radial basis function, trust region, particle swarm optimization, fitness function, difference method