Derivative Free Inexact-restoration Algorithm for Constrained Optimization

MARÍA B. AROUXÉT; NÉLIDA ECHEBEST; ELVIO A. PILOTTA

Buenos Aires

Congreso; ALIO-INFORMS Joint International Meeting; 2010

Sociedad Argentina de Informática / Association of Latin-Iberoamerican Operational Research Societies / Institute for Operations Research and the Management Sciences

Derivative free optimization (DFO) methods are designed for solving optimization problems where the derivatives of the objective function and the constraints are not available because they come from industrial measure or are the results of a computer simulation. We propose a new algorithm for DFO based on the Inexact Restoration methods (Martínez-Pilotta, 2000). In our approach, the subproblems are performed appropriately in order not to use derivatives. We also present numerical experiments.