【摘要】 对于信赖域子问题产生的搜索方向 ,利用一个给定的步长到达下一迭代点 ,证明了算法的全局收敛性 ,从而既避免了信赖域方法中在一次迭代多次求解信赖域子问题的困难 ,又避免了线搜索方法中为达到下一迭代点而多次调用函数值的计算 .更多还原

【Abstract】 A trust region trial step is utilized as a search direction and the next iteration is obtained via a given steplength. It not only avoids the difficulty of repeated solving trust region subproblems in an iterate in trust region algorithm but also avoids repeated computing function value in linear search algorithm.更多还原