A Nonmontone Self-adaptive Trust Region Algorithm Based on a Simple Quadratic Model

Based on the simple quadratic model,nonmonotone technique and adaptive technique are combined to propose a nonmonotone self-adaptive trust region algorithm for unconstrained optimization problems.The global and superlinear convergence results of the nonmonotone self-adaptive trust region method are proved under mild conditions.Theoretical analysis shows that the algorithm is efficient and attractive for large-scaled optimization problems.