无约束最优化的一个信赖域梯度路径法

A Trust Region Gradient Path Method for Unconstrained Optimization

本文给出一种易于实现的解无约束最优化问题的信赖域梯度路径法．方法对海色矩阵无正定的限制，保留了信赖域方法的特色．并证明了方法的全局收敛性和在某些条件下的二次收敛性．

In this paper we propose a eaisier gredient path search methods to solve trust region problems arising from unconstrained optimization problems. The Hessian matrix don't need to be positive and the merits of trust region methods are remained. The global Convergence is proved. Under certain conditions the solutions obtained by the method is the quadratic convergence.