A new nonmonotone trust-region method of conic model for solving unconstrained optimization (Q2654183): Difference between revisions

Unconstraint minimization of a twice continuously differentiable objective function \(f: \mathbb{R}^n\to\mathbb{R}\) is considered. The authors propose a new trust-region method for solving this problem. The method is based on the conic model. The global convergence of the method is proved under the assumptions that the level set \(\Omega= \{x\in\mathbb{R}^n; f(x)\leq f(x^0)\}\) (\(x^0\) is the initial iterate) is bounded, the gradient of \(f\) is Lipschitz continuous, and sequence of matrices \(\{B_k\}\) generated by the proposed method is uniformly bounded. In the further part of the paper, conditions for local convergence are presented and the local convergence is proved. Numerical experiments with the proposed method indicate its efficiency in comparison with other related methods known from the literature.