Active-set Newton methods for mathematical programs with vanishing constraints
- 206 Downloads
Mathematical programs with vanishing constraints constitute a new class of difficult optimization problems with important applications in optimal topology design of mechanical structures. Vanishing constraints usually violate standard constraint qualifications, which gives rise to serious difficulties in theoretical and numerical treatment of these problems. In this work, we suggest several globalization strategies for the active-set Newton-type methods developed earlier by the authors for this problem class, preserving superlinear convergence rate of these methods under weak assumptions. Preliminary numerical results demonstrate that our approach is rather promising and competitive with respect to the existing alternatives.
KeywordsMathematical program with vanishing constraints Constraint qualification Optimality condition Active-set method Sequential quadratic programming Semismooth Newton method Global convergence
The authors thank the anonymous referees for useful comments. In particular, Remark 4.1 is in response to the question raised by one of the referees.
- 2.Achtziger, W., Hoheisel, T., Kanzow, C.: A smoothing-regularization approach to mathematical programs with vanishing constraints. Preprint 284, Institute of Mathematics, University of Würzburg (2008) Google Scholar
- 8.Facchinei, F., Pang, J.-S.: Finite-Dimensional Variational Inequalities and Complementarity Problems. Springer, New York (2003) Google Scholar
- 17.Hoheisel, T., Kanzow, C., Schwartz, A.: Convergence of a local regularization approach for mathematical programs with complementarity or vanishing constraints. Optim. Methods Softw. doi: 10.1080/10556788.2010.535170
- 19.Izmailov, A.F., Pogosyan, A.L.: On active-set methods for mathematical programs with vanishing constraints. In: Bereznyov, V.A. (ed.) Theoretical and Applied Problems of Nonlinear Analysis, pp. 18–49. CCAS, Moscow (2009) (In Russian) Google Scholar