基于拟牛顿法的同时扰动随机逼近算法  

Simultaneous perturbation stochastic approximation algorithm based on quasi-Newton method

作　　者：张华军[1] 赵金[1] 罗慧[1] 

机构地区：[1]华中科技大学自动化学院,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2014年第9期1-4,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61273174;61034006;60874047)

摘　　要：基于拟牛顿法原理,结合同时扰动随机逼近算法特性提出了一种搜索方向dk的计算方法,从而提高了同时扰动随机逼近算法的收敛速度和逼近精度.针对典型优化问题分别比较了改进后的同时扰动随机逼近算法、标准同时扰动随机逼近算法及二阶同时扰动随机逼近算法的优化性能,数值分析结果表明:改进后的算法在逼近精度上均优于其他两种算法,收敛速度介于其他两种算法之间.In order to improve convergence speed and approximation precision of simultaneous perturbation stochastic approximation（SPSA）,lessons were drawn from Broyden-Fletcher-Goldfarb-Shanno（BFGS）quasi-Newton method,and a computing method of search direction dkbased on SPSA was provided.A typical optimization problem as numerical analysis case was used,and characteristics of the improved SPSA were compared with standard SPSA and the second order SPSA.The results of numerical analysis indicate that the improved SPSA is better than the other two methods at approximation precision aspect,and the performance of convergence speed is between basic SPSA and the second order SPSA.

关 键 词：拟牛顿法 同时扰动随机逼近 搜索方向 收敛速度 逼近精度 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]