Skip to main content
Log in

Narrowing the difficulty gap for the Celis–Dennis–Tapia problem

  • Full Length Paper
  • Series B
  • Published:
Mathematical Programming Submit manuscript

Abstract

We study the Celis–Dennis–Tapia (CDT) problem: minimize a non-convex quadratic function over the intersection of two ellipsoids. In contrast to the well-studied trust region problem where the feasible set is just one ellipsoid, the CDT problem is not yet fully understood. Our main objective in this paper is to narrow the difficulty gap that occurs when the Hessian of the Lagrangian is indefinite at all Karush–Kuhn–Tucker points. We prove new sufficient and necessary conditions both for local and global optimality, based on copositivity, giving a complete characterization in the degenerate case.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1

Similar content being viewed by others

Notes

  1. [20, Cor.5.4] offers simple sufficient conditions on the problem data \(({\mathsf Q},{\mathsf q},{\mathsf A},{\mathsf a})\) that rule out this phenomenon, namely that no off-diagonal entry of \({\mathsf Q}\) or \({\mathsf A}{\mathsf A}^\top \) is strictly positive and \(\left\{ {\mathsf A}{\mathsf a}, -{\mathsf q}\right\} \subseteq {\mathbb R}^n_+\,\). Note that this particularly implies \({\mathsf H}(\bar{u},\bar{v})\bar{{\mathsf x}}\in {\mathbb R}^n_+\) by (23).

  2. We are indebted to W. Schachinger for this observation (personal communication). Note that when \(n=2\), this means that \({\mathsf H}(\bar{u}(t),\bar{v}(t))\) is \(\Gamma _{\!\!\mathrm red}(\bar{{\mathsf x}})\)-copositive for either \(t=0\) or \(t=1\), but in higher dimensions the relations between the curvatures vary with \({\mathsf d}\), so that whether \(t\) is 0 or 1 in the condition \({\mathsf d}^\top {\mathsf H}(\bar{u}(t),\bar{v}(t)) {\mathsf d}\ge 0\) depends on \({\mathsf d}\).

  3. We used BFGS to minimize the exact penalty function \(p({\mathsf x})=f({\mathsf x})+\rho \max (r({\mathsf x}),0)+\rho \max (s({\mathsf x}),0)\), for some \(\rho >0\) that is increased as needed to ensure feasibility. As discussed in [21], BFGS is a surprisingly effective method to find local minimizers of nonsmooth, nonconvex functions such as \(p\). Since one can expect only to find local minimizers in general, we did this repeatedly from 10 randomly generated starting points for each problem instance, selecting the result \(\bar{{\mathsf x}}\) with the lowest value of \(f\) as our candidate for the global minimizer. The results summarized in Table 1 and discussed below show that in by far the majority of cases, global optimality was confirmed, and in all except one of 70,000 tests at least local optimality was confirmed. (At the suggestion of M. Fampa, we also conducted a similar experiment using up to \(2^n\) starting points instead of 10, stopping as soon as global optimality was confirmed; the outcome was similar.)

  4. The case \(n=2\) receives particular attention in the recent theoretical study [29].

References

  1. Ai, W., Zhang, S.: Strong duality for the CDT subproblem: a necessary and sufficient condition. SIAM J. Optim. 19(4), 1735–1756 (2009)

    Article  MATH  MathSciNet  Google Scholar 

  2. Beck, A., Eldar, Y.: Strong duality in nonconvex quadratic optimization with two quadratic constraints. SIAM J. Optim. 17(3), 844–860 (2006)

    Article  MATH  MathSciNet  Google Scholar 

  3. Bienstock, D., Michalka, A.: Polynomial solvability of variants of the trust-region subproblem. In: Proceedings of the Twenty-Fifth Annual ACM-SIAM Symposium on Discrete Algorithms, pp. 380–390, Portland, Oregon, USA (2014)

  4. Bomze, I.M.: Copositivity for second-order optimality conditions in general smooth optimization problems. Preprint NI13033-POP, Isaac Newton Institute, Cambridge UK (2013)

  5. Bomze, I.M.: Copositivity conditions for global optimality in indefinite quadratic programming problems. Czechoslov. J. Oper. Res. 1, 7–19 (1992)

    MATH  Google Scholar 

  6. Borwein, J.M.: Necessary and sufficient conditions for quadratic minimality. Numer. Funct. Anal. Optim. 5, 127–140 (1982)

    Article  MATH  MathSciNet  Google Scholar 

  7. Burer, S., Anstreicher, K.: Second-order-cone constraints for extended trust-region subproblems. SIAM J. Optim. 23(1), 432–451 (2013)

    Article  MATH  MathSciNet  Google Scholar 

  8. Celis, M.R., Dennis, Jr., J.E., Tapia, R.A.: A trust region strategy for nonlinear equality constrained optimization. In: Numerical Optimization, 1984 (Boulder, Colo., 1984), pp. 71–82. SIAM, Philadelphia, PA (1985)

  9. Chen, X.-D., Yuan, Y.-X.: On local solutions of the Celis–Dennis–Tapia subproblem. SIAM J. Optim. 10(2), 359–383 (1999)

    Article  MATH  MathSciNet  Google Scholar 

  10. Chen, X.-D., Yuan, Y.-X.: Strong duality for the CDT subproblem: a necessary and sufficient condition. J. Comput. Math. 19(2), 113–124 (2009)

    MathSciNet  Google Scholar 

  11. Conn, A.R., Gould, N.I.M., Toint, P.L.: Trust-region methods. In: MPS/SIAM Series on Optimization. Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA (2000)

  12. Contesse, L.B.: Une caractérisation complète des minima locaux. Numer. Math. 34, 315–332 (1980)

    Article  MATH  MathSciNet  Google Scholar 

  13. Dickinson, P.J.C., Gijben, L.: On the computational complexity of membership problems for the completely positive cone and its dual. Comput. Optim. Appl. 57(2), 403–415 (2014)

    Article  MATH  MathSciNet  Google Scholar 

  14. Ding, B.: A Parametric Solution for Local and Global Optimization. PhD thesis, University of Waterloo, Ontario, Canada (1996)

  15. Grigoriev, D., Pasechnik, D.V.: Polynomial time computing over quadratic maps I: sampling in real algebraic sets. Comput. Complex. 14, 20–52 (2005)

    Article  MATH  MathSciNet  Google Scholar 

  16. Higham, N.J.: Accuracy and stability of numerical algorithms. In: Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA, Second edition (2002)

  17. Hiriart-Urruty, J.-B.: Global optimality conditions in maximizing a convex quadratic function under convex quadratic constraints. J. Global Optim. 21(4), 445–455 (2001)

    Article  MathSciNet  Google Scholar 

  18. Jargalsaikhan, B.: Indefinite copositive matrices with exactly one positive eigenvalue or exactly one negative eigenvalue. Electron. J. Linear Algebra 26, 754–761 (2013)

    MATH  MathSciNet  Google Scholar 

  19. Jeyakumar, V., Li, G.: Trust-region problems with linear inequality constraints: exact SDP relaxation, global optimality and robust optimization. Math. Program., to appear. 2013, doi:10.1007/s10107-013-0716-2

  20. Jeyakumar, V., Lee, G.M., Li, G.: Alternative theorems for quadratic inequality systems and global quadratic optimization. SIAM J. Optim. 20(2), 983–1001 (2009)

    Article  MATH  MathSciNet  Google Scholar 

  21. Lewis, A.S., Overton, M.L.: Nonsmooth optimization via quasi-Newton methods. Math. Program. 141(1-2, Ser. B), 135–163 (2013)

  22. Majthay, A.: Optimality conditions for quadratic programming. Math. Program. 1, 359–365 (1971)

    Article  MATH  MathSciNet  Google Scholar 

  23. Martínez, J.M.: Local minimizers of quadratic functions on Euclidean balls and spheres. SIAM J. Optim. 4(1), 159–176 (1994)

    Article  MATH  MathSciNet  Google Scholar 

  24. Murty, K.G., Kabadi, S.N.: Some NP-complete problems in quadratic and nonlinear programming. Math. Program. 39(2), 117–129 (1987)

    Article  MATH  MathSciNet  Google Scholar 

  25. Nesterov, Y., Wolkowicz, H., Ye, Y.: Nonconvex quadratic optimization. In: Wolkowicz, H., Saigal, R., Vandenberghe, L. (eds.) Handbook of Semidefinite Programming, pp. 361–420. Kluwer, New York (2000)

    Chapter  Google Scholar 

  26. Nocedal, J., Wright, S.J.: Numerical Optimization, 2nd edn. Springer Series in Operations Research and Financial Engineering. Springer, New York (2006)

  27. Peng, J., Yuan, Y.-X.: Optimality conditions for the minimization of a quadratic with two quadratic constraints. SIAM J. Optim. 7(3), 579–594 (1997)

    Article  MATH  MathSciNet  Google Scholar 

  28. Pólik, I., Terlaky, T.: A survey of the S-lemma. SIAM Rev. 49(3), 371–418 (2007)

    Article  MATH  MathSciNet  Google Scholar 

  29. Yang, B., Burer, S.: A two-variable analysis of the two-trust-region subproblem. Preprint, submitted to SIAM J. Optim., http://www.optimization-online.org/DB_HTML/2013/07/4126.html (2013)

  30. Ye, Y., Zhang, S.: New results on quadratic minimization. SIAM J. Optim. 14(1), 245–267 (2003)

    Article  MATH  MathSciNet  Google Scholar 

  31. Yuan, Y.-X.: On a subproblem of trust region algorithms for constrained optimization. Math. Program. 47(1, (Ser. B)), 53–63 (1990)

Download references

Acknowledgments

The first author is grateful to Cambridge University and to New York University for the hospitality he enjoyed during his research stays as a visiting fellow. The second author was supported in part by the U.S. National Science Foundation under grant DMS-1317205. Both authors are indebted to an anonymous referee and the Managing Guest Editor for their diligence and helpful constructive remarks.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Immanuel M. Bomze.

Additional information

Dedicated to John E. Dennis Jr. and Richard A. Tapia on the occasion of their 75th birthdays.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Bomze, I.M., Overton, M.L. Narrowing the difficulty gap for the Celis–Dennis–Tapia problem. Math. Program. 151, 459–476 (2015). https://doi.org/10.1007/s10107-014-0836-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10107-014-0836-3

Keywords

Mathematics Subject Classification

Navigation