Premium
A sub‐feasible correction step for non‐linear programming methods with equality constraints
Author(s) -
Nikolayzik Tim,
Büskens Christof
Publication year - 2008
Publication title -
pamm
Language(s) - English
Resource type - Journals
ISSN - 1617-7061
DOI - 10.1002/pamm.200810779
Subject(s) - hessian matrix , solver , mathematical optimization , nonlinear programming , nonlinear system , computer science , parametric statistics , linear programming , mathematics , optimization problem , algorithm , statistics , physics , quantum mechanics
Due to the sparsity of large–scale optimazation problems it can be useful to use the exact hessian matrix instead of update techniques for solving nonlinear optimization problems. Herewith it is possible to use techniques of the parametric sensitivity analysis to improve the feasibility of the NLP–problem in each iteration step of a solver. The main idea is to treat the measurable error between the general nonlinear equality constraints and the linear approximations inside the QP subproblems as a linear perturbation in the constraints. By special real–time optimization approximations this error can be reduced nearly without any extra computational costs. First results will be presented. (© 2008 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)