In this paper, we present a trust region method for unconstrained optimization problems with locally Lipschitz functions. For this idea, at first, a smoothing conic model sub-problem is introduced for the objective function, by the approximation of steepest descent method. Next, for solving the conic sub-problem, we presented the modified convenient curvilinear search method and equipped it with Armijo condition such that in each repetition, the objective function is sufficiently reduced. Then, the convergence property of this method is proved. Finally, the presented method is implemented in MATLAB environment, and numerical results are compared with the non-smooth trust region method
Rights and permissions | |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |