Nonlinear Programming Without a Penalty Function or a Filter, M.Sc. Thesis Sharif University of Technology ; Mahdavi Amiri, Nezamedin (Supervisor)
Abstract
A new method, recently introduced in the literature, is discussed for solving equality constrained nonlinear optimization problems. This method does not use a penalty or a barrier function, or a filter, and yet its global convergence to first-order stationary points can be proved. The method uses different trust regions to cope with the nonlinearities of the objective function and the constraints, and admits inexact SQP steps not lying exactly in the nullspace of the local Jacobian. We implement the method in MATLAb 7.7 software environment and test the resulting program on a collection of CUTEr problems. The numerical results are promising and confirm the global convergence of the method
Cataloging briefNonlinear Programming Without a Penalty Function or a Filter, M.Sc. Thesis Sharif University of Technology ; Mahdavi Amiri, Nezamedin (Supervisor)
Abstract
A new method, recently introduced in the literature, is discussed for solving equality constrained nonlinear optimization problems. This method does not use a penalty or a barrier function, or a filter, and yet its global convergence to first-order stationary points can be proved. The method uses different trust regions to cope with the nonlinearities of the objective function and the constraints, and admits inexact SQP steps not lying exactly in the nullspace of the local Jacobian. We implement the method in MATLAb 7.7 software environment and test the resulting program on a collection of CUTEr problems. The numerical results are promising and confirm the global convergence of the method
Find in contentBookmark
|
|