对无约束优化问题提出了一个带线搜索的非单调自适应信赖域算法,每次迭代都充分利用当前迭代点包含的二次信息自动产生一个信赖域半径.在试探步不被接受时,算法并不重解信赖域子问题,而采用非单调线搜索,从而减少了计算量.在适当的条件下,证明了算法的全局收敛性.
A line search nonmonotonic self - adaptive trust region algorithm was presented for unconstrained optimization problems in this paper. Each iteration took full advantage of the current iteration points ineluding secondary information to automatically generate a trust region radius. When the trial step is not accepted, the non monotone line search is used in the algorithm instead of re - solution of the trust region subproblem, thereby reducing the amount of computation. Under suitable conditions, the global convergence of the algorithm was proved.