Abstract
In this paper, we propose an algorithm for the solution of nonlinear constrained programming. This algorithm does not use any penalty function or a filter. Instead, it uses the idea of maximal constraint violation to guarantee global convergence. The infeasibility of each iterate is controlled by a progressively decreasing limit. At each iteration, a normal step which is obtained by solving a tightened normal subproblem and a tangential step which is a solution of a tangential subproblem are computed successively. The algorithm reduces the value of objective function or improves feasibility according to the relation of the tangential predicted reduction and the predicted reduction of constraint violation achieved by the trial step. The framework of the algorithm ensures the global convergence without assuming regularity or boundedness of iterate sequence. Moreover, the algorithm does not need any restoration phase. We also include some preliminary yet promising numerical results on some standard test problems in constrained optimization.
Similar content being viewed by others
References
Bielschowsky, R.H., Gomes, F.A.M.: Dynamic control of infeasibility in equality constrained optimization. SIAM J. Optim. 19(3), 1299–1325 (2008)
Bongartz, I., Conn, A., Gould, N., Saunders, M., Toint, Ph.: A numerical comparison between the LANCELOT and MINOS packages for large-scale constrained optimization: The complete numerical results. Report 97/14 (1997)
Bongartz, I., Conn, A., Gould, N., Toint, Ph.: CUTE: Constrained and unconstrained testing environment. ACM Trans. Math. Softw. 21(1), 123–160 (1995)
Chen, Z.: A penalty-free-type nonmonotone trust-region method for nonlinear constrained optimization. Appl. Math. Comput. 173(2), 1014–1046 (2006)
Chen, Z., Han, J., Han, Q.: A globally convergent trust region algorithm for optimization with general constraints and simple bounds. Acta Math. Appl. Sin. 15(4), 425–432 (1999)
Conn, A., Gould, N., Toint, Ph.: LANCELOT: A Fortran Package for Large-Scale Nonlinear Optimization (Release A). Springer, Berlin (1992)
Conn, A., Gould, N., Toint, Ph.: Trust-Region Methods. SIAM, Philadelphia (2000)
COPS: Constrained optimization problem set. http://www.mcs.anl.gov/~more/cops
Dolan, E., Moré, J.: Benchmarking optimization software with performance profiles. Math. Program. 91(2), 201–213 (2002)
Fletcher, R., Gould, N., Leyffer, S., Toint, Ph., Wächter, A.: Global convergence of a trust-region SQP-filter algorithm for general nonlinear programming. SIAM J. Optim. 13(3), 635–659 (2002)
Fletcher, R., Leyffer, S.: Nonlinear programming without a penalty function. Math. Program. 91(2), 239–269 (2002)
Gomes, F., Maciel, M., Martínez, J.: Nonlinear programming algorithms using trust regions and augmented Lagrangians with nonmonotone penalty parameters. Math. Program. 84(1), 161–200 (1999)
Gould, N., Toint, Ph.: Nonlinear programming without a penalty function or a filter. Math. Program. 122(1), 155–196 (2010)
Liu, X., Yuan, Y.: A sequential quadratic programming method without a penalty function or a filter for nonlinear equality constrained optimization. SIAM J. Optim. 21(2), 545–571 (2011)
Murtagh, B., Saunders, M.: Minos 5.4 user’s guide (revised). Technical Report SOL 83-20R, Department of Operations Research, Stanford University, Stanford, CA 94305, USA, 1993. Revised (1995)
Qiu, S., Chen, Z.: Global and local convergence of a class of penalty-free-type methods for nonlinear programming. Appl. Math. Model. 36(7), 3201–3216 (2012)
Qiu, S., Chen, Z.: A new penalty-free-type algorithm based on trust region techniques. Appl. Math. Comput. 218(22), 11,089–11,099 (2012)
Solodov, M.: Global convergence of an SQP method without boundedness assumptions on any of the iterative sequences. Math. Program. 118(1), 1–12 (2009)
Su, K.: A modified penalty-free-type nonmonotone SQP method. In: Computational Sciences and Optimization (CSO), 2011 Fourth International Joint Conference on, pp. 462–466. IEEE, New York (2011)
Ulbrich, M., Ulbrich, S.: Non-monotone trust region methods for nonlinear equality constrained optimization without a penalty function. Math. Program. 95(1), 103–135 (2003)
Xu, D., Han, J., Chen, Z.: Nonmonotone trust-region method for nonlinear programming with general constraints and simple bounds. J. Optim. Theory Appl. 122(1), 185–206 (2004)
Xue, W., Shen, C., Pu, D.: A penalty-function-free line search SQP method for nonlinear programming. J. Comput. Appl. Math. 228(1), 313–325 (2009)
Yamashita, H.: A globally convergent quasi-newton method for equality without a penalty function for nonlinear constrained optimization. Technical Report, Mathematical Systems Inc., Tokyo, Japan (1979). Revised September 1982
Yamashita, H., Yabe, H.: A globally convergent trust-region SQP method without function. Technical Report, Mathematical Systems Inc., Tokyo, Japan (2003). Revised July 2007
Zoppke-Donaldson, C.: A tolerance-tube approach to sequential quadratic programming with applications. Ph.D. Thesis, Department of Mathematics and Computer Science, University of Dundee, Dundee, Scotland, UK (1995)
Acknowledgements
The authors would like to thank the referees whose comments have helped us greatly to improve the presentation.
Author information
Authors and Affiliations
Corresponding author
Additional information
This work was supported by the Fundamental Research Funds for the Central Universities 2014QNA62 and Chinese NSF grant 11371273 and 11171247.
Appendix
Appendix
See Table 1.
Rights and permissions
About this article
Cite this article
Qiu, S., Chen, Z. A Globally Convergent Penalty-Free Method for Optimization with Equality Constraints and Simple Bounds. Acta Appl Math 142, 39–60 (2016). https://doi.org/10.1007/s10440-015-0013-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10440-015-0013-6
Keywords
- Nonlinear programming
- Constrained optimization
- Penalty-free algorithm
- Trust region method
- Global convergence