本文给出了一种新的信赖域算法。该算法以变化的速率来调整信赖域半径的大小。在由信赖域子问题产生的试探步不被接受的情况下,新算法采用线搜索的方法得到下一个迭代点。同时算法采用非单调的技术来加速算法的收敛效果。文中给出了新算法的全局收敛性分析和数值试验的结果。
We propose a new trust region algorithm which the trust region radius is updated at a variable rate. Moreover, the new algorithm performs a backtracking line search from the failed point instead of resolving the trust region subproblem. A nonmonotonic criterion is also used to speed up the convergence. We establish the global convergence of the new algorithm. Numerical results are also presented.