On Minimizing Objective and KKT Error in a Filter Line Search Strategy for an Interior Point Method
This paper carries out a numerical study of filter line search strategies that aim at minimizing the objective function and the Karush-Kuhn-Tucker (KKT) vector error in order to encourage global convergence of interior point methods. These filter strategies are implemented in an infeasible primal-dual interior point framework for nonlinear programming. First, we propose a filter that has four components measuring primal feasibility, complementarity, dual feasibility and optimality. The different measures arise from the KKT conditions of the problem. Then, we combine the KKT equations defining a different two-dimensional filter technique. The versions have in common the objective function as the optimality measure. The primary assessment of these techniques has been done with a well-known collection of small- and medium-scale problems.
KeywordsNonlinear programming Interior point method Filter method Line search
Unable to display preview. Download preview PDF.
- 2.Costa, M.F.P., Fernandes, E.M.G.P.: Comparison of interior point filter line search strategies for constrained optimization by performance profiles. International Journal of Mathematics Models and Methods in Applied Sciences 1, 111–116 (2007)Google Scholar
- 6.Fletcher, R., Leyffer, S., Toint, P.: A brief history of filter methods, Report ANL/MCS-P1372-0906, Argonne National Laboratory (2006)Google Scholar
- 11.Vanderbei, R.J.: LOQO: An interior-code for quadratic programming. Technical report SOR-94-15, Princeton University, Statistics and Operations Research (1998)Google Scholar