无约束最优化锥模型拟牛顿信赖域方法的收敛性(英)

本文研究无约束最优化雄模型拟牛顿信赖域方法的全局收敛性．文章给出了确保这类方法全局收敛的条件．文章还证明了，当用拆线法来求这类算法中锥模型信赖域子问题的近似解时，确保全局收敛的条件得到满足

Convergence of Conic Quasi-Newton Trust Region Methods for Unconstrained Minimization

This paper considers the global convergence property of a model method which is based on conic model and quasi-Newton update and uses trust region strategy. Conditiions arc given under which the model method is globally convergent. It is also proved that when the trust region subproblems in the model method arc solved using a dogleg path method, the convergence conditions are satisfied.