A truncated SQP algorithm for solving nonconvex equality constrained optimization problems

Laurent Chauvier (Laurent.Chauvierartelys.fr)
Antonio Fuduli (Antonio.Fuduliunile.it)
Jean Charles Gilbert (Jean-Charles.Gilbertinria.fr)

Abstract: An algorithm for solving equality constrained optimization problems is proposed. It can deal with nonconvex functions and uses a truncated conjugate algorithm for detecting nonconvexity. The algorithm ensures convergence from remote starting point by using line-search. Numerical experiments are reported, comparing the approach with the one implemented in the trust region codes ETR and Knitro.

Keywords: equality constraint, exact penalty function, global convergence, line-search, Newton's method, nonconvex optimization, sequential quadratic programming, truncated conjugate gradient algorithm

Category 1: Nonlinear Optimization (Constrained Nonlinear Optimization )

Citation: Research Report RR-4346, INRIA, BP 105, 78153 Le Chesnay, France

Download: [Compressed Postscript]

Entry Submitted: 01/03/2002
Entry Accepted: 01/03/2002
Entry Last Modified: 01/03/2002

Modify/Update this entry