保守修正BFGS算法及全局收敛性  

A Cautious BFGS-type Method and Its Global Convergence

作　　者：张继伟[1] 

机构地区：[1]集美大学理学院,福建厦门361021

出　　处：《集美大学学报（自然科学版）》2010年第3期228-233,共6页Journal of Jimei University：Natural Science

基　　金：国家自然科学基金资助项目(10471036);集美大学科研基金资助项目(C60657)

摘　　要：基于新拟牛顿方程,提出一类保守修正BFGS算法.该算法的特点是:即使当目标函数是非凸函数时,该算法仍然是全局收敛的.在适当的条件下,该算法具有局部超线性收敛性.初步的数值实验表明,该算法是有效的.A cautious BFGS-type algorithm based on a new secant equation was this version of BFGS algorithm was that this algorithm possessed global convergence even when the objective function was nonconvex. In the proper conditions, the cautious BFGS-type algorithm was superlinearly convergent. The numerical experiments showed that the cautious BFGS-type algorithm was promising.

关 键 词：BFGS算法 全局收敛性 超线性收敛性 

分 类 号：O221.2[理学—运筹学与控制论] O224[理学—数学]