A nonmonotone trust-region method of conic model for unconstrained optimization (Q939515)

The authors study a non-monotone trust-region method for solving the general unconstrained optimization problem. In the first section they outline the background for this approach, namely the trust-region methods of quadratic model and monotone trust-region methods. In the second section, the main algorithm is presented, which consists of a non-monotone trust-region algorithm based on the conic model. The algorithm is further studied in the third section where the convergence properties are examined and several relevant theorems are proven. The article concludes with a section containing the results of numerical experimentation and a list of relevant references.