基于锥模型的一种新非单调信赖域算法  

A New Nonmonotone Trust Region Algorithms Based On Conic Model

作　　者：邢治业[1] XING Zhi-ye(Department of Basic Course,Shanxi Engineering Vocational College, Taiyuan 030009, China)

机构地区：[1]山西工程职业技术学院基础部,山西太原030009

出　　处：《信阳农林学院学报》2018年第4期114-117,共4页Journal of Xinyang Agriculture and Forestry University

摘　　要：本文结合目前流行的非单调技术,将其应用于锥模型信赖域算法中,提出一种新的非单调无约束优化算法,与传统非单调信赖域算法相比,该算法不仅不需要重解子问题,而且充分利用了迭代函数的信息,并且每一步迭代保证Hesse的正定性,一定条件下,证明了该算法的收敛性。In this paper,combining the conic trust region algorithm with the popular non-monotone techniques,we present a new nonmontonic unconstrained optimization algorithms.Compared with the traditional nonmonotone trust region algorithm,the new algorithm not only does not need to resolve the sub-problems,but also makes full use of the information of the iterated function,,and maintains the positive definiteness of Hesse matrix at each interation.Under certain conditions,the global convergence of this new method is proved.

关 键 词：无约束优化 信赖域算法 锥模型 非单调技术 全局收敛性 

分 类 号：O224[理学—运筹学与控制论]