An active trust-region algorithm for solving constrained optimization problem

Abstract

The thesis is planned to work in the area of numerical analysis. More precisely in the further development of numerical methods for solving general nonlinear programming problems. The quadratic penalty function is also used as a merit function to decide whether the trial step should be accepted. The convergence theory of this algorithms is based on the fraction of cauchy conditions. In this algorithm the dogleg method will be used to compute the trial step.