关键词: 无约束最优化, 信赖域算法, 大步长Armijo线搜索, 收敛, 数值实验

Abstract:

The authors propose a new trust region algorithm with simple quadratic models and larger Armijo line search rule. The global convergence property of the  new method is proved under the condition that the gradient of function is uniformly continuous. Numerical results show
that the new algorithm is efficient, and attractive for large-scale optimization problems.

Key words: Unconstrained optimization, Trust region method\ Larger Armijo line search rule, Convergence, Numerical experiment