Derivative-free nonlinear programming

Author(s)

Lucas Garcia Pedroso

• Document type: Doctoral Thesis
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Instituto de Matemática, Estatística e Computação Científica
• Defense date: 2009-08-28

Examining board members

José Mario Martínez Pérez; Roberto Andreani; Orizon Pereira Ferreira; Elizabeth Wegner Karas; Yuan Jin Yun

• Advisor: José Mario Martínez Pérez; Maria Aparecida Diniz Ehrhardt

Abstract

We propose in this work a derivative-free Augmented Lagrangian algorithm for the general problem of optimization. We consider the method due to Andreani, Birgin, Martínez and Schuverdt, eliminating the derivative computations in the algorithm by making suitable modifications on the stopping criterion. The good theoretical results of the method were mantained, as convergence under the CPLD constraint qualification and the limitation of the penalty parameter. Numerical experiments are presented, and the most relevant of them is an example of derivative-free problem based on the simulation of areas of figures on the plane. (AU)

• FAPESP's process: 04/15635-2 - Derivative-free nonlinear programming
• Grantee: Lucas Garcia Pedroso
• Support Opportunities: Scholarships in Brazil - Doctorate