A nonmonotone trust region method for unconstrained optimization (Q814729)

The authors propose a combination of a nonmonotone technique, fixed step length and the trust region method, in order to minimize a twice continuously differentiable unconstrained function, in the aim to improve the algorithm of \textit{J. Nocedal} and \textit{Y. Yuan} [Appl. Optim. 14, 153--175 (1998; Zbl 0909.90243)] and make it more effective in practical implementation. The main difference between the proposed method and the original method is that in the former one a step length is computed by a line search when the trial step is not successful, whereas in the present method a step length is defined by a formula. It is proved under mild conditions that the algorithm is global and superlinear convergent. Some numerical test on known test problems with the original and the new proposed modification are provided.