Abstract
We consider dynamically generating linear constraints (cutting planes) to tighten relaxations for polynomial optimization problems. Many optimization problems have feasible set of the form \(S \cap P\), where S is a closed set and P is a polyhedron. Integer programs are in this class and one can construct intersection cuts using convex “forbidden” regions, or S-free sets. Here, we observe that polynomial optimization problems can also be represented as a problem with linear objective function over such a feasible set, where S is the set of real, symmetric matrices representable as outer-products of the form \(xx^T\). Accordingly, we study outer-product-free sets and develop a thorough characterization of several (inclusion-wise) maximal intersection cut families. In addition, we present a cutting plane approach that guarantees polynomial-time separation of an extreme point in \(P\setminus S\) using our outer-product-free sets. Computational experiments demonstrate the promise of our approach from the point of view of strength and speed.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
This BoxQP relaxation only adds the “diagonal” McCormick estimates \(X_{ii} \le x_i\).
References
Andersen, K., Louveaux, Q., Weismantel, R.: An analysis of mixed integer linear sets based on lattice point free convex sets. Math. Oper. Res. 35(1), 233–256 (2010)
Andersen, K., Louveaux, Q., Weismantel, R., Wolsey, L.A.: Inequalities from two rows of a simplex tableau. In: Fischetti, M., Williamson, D.P. (eds.) IPCO 2007. LNCS, vol. 4513, pp. 1–15. Springer, Heidelberg (2007). https://doi.org/10.1007/978-3-540-72792-7_1
Anstreicher, K.M.: Semidefinite programming versus the reformulation linearization technique for nonconvex quadratically constrained quadratic programming. J. Glob. Optim. 43, 471–484 (2009)
Atamtürk, A., Narayanan, V.: Conic mixed-integer rounding cuts. Math. Program. 122, 1–20 (2010)
Averkov, G.: On finite generation and infinite convergence of generalized closures from the theory of cutting planes. arXiv preprint arXiv:1106.1526 (2011)
Balas, E.: Intersection cuts—a new type of cutting planes for integer programming. Oper. Res. 19(1), 19–39 (1971)
Bao, X., Sahinidis, N.V., Tawarmalani, M.: Multiterm polyhedral relaxations for nonconvex, quadratically constrained quadratic programs. Optim. Methods Softw. 24(4–5), 485–504 (2009)
Basu, A., Conforti, M., Cornuéjols, G., Zambelli, G.: Maximal lattice-free convex sets in linear subspaces. Math. Oper. Res. 35(3), 704–720 (2010)
Basu, A., Conforti, M., Cornuéjols, G., Zambelli, G.: Minimal inequalities for an infinite relaxation of integer programs. SIAM J. Discrete Math. 24(1), 158–168 (2010)
Belotti, P., Góez, J.C., Pólik, I., Ralphs, T.K., Terlaky, T.: On families of quadratic surfaces having fixed intersections with two hyperplanes. Discrete Appl. Math. 161(16–17), 2778–2793 (2013)
Bienstock, D., Michalka, A.: Cutting-planes for optimization of convex functions over nonconvex sets. SIAM J. Optim. 24, 643–677 (2014)
Bienstock, D., Chen, C., Muñoz, G.: Outer-product-free sets for polynomial optimization and oracle-based cuts. arXiv preprint arXiv:1610.04604 (2016)
Bonami, P., Günlük, O., Linderoth, J.: Globally solving nonconvex quadratic programming problems with box constraints via integer programming methods. Math. Program. Comput. 10(3), 333–382 (2018)
Borozan, V., Cornuéjols, G.: Minimal valid inequalities for integer constraints. Math. Oper. Res. 34(3), 538–546 (2009)
Burer, S.: Optimizing a polyhedral-semidefinite relaxation of completely positive programs. Math. Program. Comput. 2(1), 1–19 (2010)
Chen, C., Atamtürk, A., Oren, S.S.: A spatial branch-and-cut method for nonconvex QCQP with bounded complex variables. Math. Program. 165(2), 549–577 (2017)
Conforti, M., Cornuéjols, G., Daniilidis, A., Lemaréchal, C., Malick, J.: Cut-generating functions and S-free sets. Math. Oper. Res. 40(2), 276–391 (2014)
Cornuéjols, G., Wolsey, L., Yıldız, S.: Sufficiency of cut-generating functions. Math. Program. 152, 1–9 (2013)
Dax, A.: Low-rank positive approximants of symmetric matrices. Adv. Linear Algebra Matrix Theory 4(3), 172–185 (2014)
Dey, S.S., Wolsey, L.A.: Lifting integer variables in minimal inequalities corresponding to lattice-free triangles. In: Lodi, A., Panconesi, A., Rinaldi, G. (eds.) IPCO 2008. LNCS, vol. 5035, pp. 463–475. Springer, Heidelberg (2008). https://doi.org/10.1007/978-3-540-68891-4_32
Dey, S.S., Wolsey, L.A.: Constrained infinite group relaxations of MIPs. SIAM J. Optim. 20(6), 2890–2912 (2010)
Eckart, C., Young, G.: The approximation of one matrix by another of lower rank. Psychometrika 1(3), 211–218 (1936)
Fischetti, M., Ljubić, I., Monaci, M., Sinnl, M.: A new general-purpose algorithm for mixed-integer bilevel linear programs. Oper. Res. 65(6), 1615–1637 (2017)
Floudas, C.A., et al.: Handbook of Test Problems in Local and Global Optimization, vol. 33. Springer, Boston (2013). https://doi.org/10.1007/978-1-4757-3040-1
Freund, R.M., Orlin, J.B.: On the complexity of four polyhedral set containment problems. Math. Program. 33(2), 139–145 (1985)
Ghaddar, B., Vera, J.C., Anjos, M.F.: A dynamic inequality generation scheme for polynomial programming. Math. Program. 156(1–2), 21–57 (2016)
Gomory, R.E., Johnson, E.L.: Some continuous functions related to corner polyhedra. Math. Program. 3(1), 23–85 (1972)
Guennebaud, G., Jacob, B., et al.: Eigen v3 (2010). http://eigen.tuxfamily.org
Hillestad, R.J., Jacobsen, S.E.: Reverse convex programming. Appl. Math. Optim. 6(1), 63–78 (1980)
Kılınç-Karzan, F.: On minimal valid inequalities for mixed integer conic programs. Math. Oper. Res. 41(2), 477–510 (2015)
Kocuk, B., Dey, S.S., Sun, X.A.: Matrix minor reformulation and SOCP-based spatial branch-and-cut method for the AC optimal power flow problem. Math. Program. Comput. 10(4), 557–596 (2018)
Krishnan, K., Mitchell, J.E.: A unifying framework for several cutting plane methods for semidefinite programming. Optim. Methods Softw. 21, 57–74 (2006)
Lasserre, J.B.: Global optimization with polynomials and the problem of moments. SIAM J. Optim. 11, 796–817 (2001)
Laurent, M.: Sums of squares, moment matrices and optimization over polynomials. In: Putinar, M., Sullivant, S. (eds.) Emerging Applications of Algebraic Geometry, pp. 157–270. Springer, New York (2009). https://doi.org/10.1007/978-0-387-09686-5_7
Locatelli, M., Schoen, F.: On convex envelopes for bivariate functions over polytopes. Math. Program. 144, 1–27 (2013)
Lovász, L.: Geometry of numbers and integer programming. In: Mathematical Programming: Recent Developments and Applications, pp. 177–210 (1989)
Luedtke, J., Namazifar, M., Linderoth, J.: Some results on the strength of relaxations of multilinear functions. Math. Program. 136(2), 325–351 (2012)
McCormick, G.P.: Computability of global solutions to factorable nonconvex programs: part I - convex underestimating problems. Math. Program. 10(1), 147–175 (1976)
Meeraus, A.: GLOBALLib. http://www.gamsworld.org/global/globallib.htm
Mirsky, L.: Symmetric gauge functions and unitarily invariant norms. Q. J. Math. 11(1), 50–59 (1960)
Misener, R., Floudas, C.A.: Global optimization of mixed-integer quadratically-constrained quadratic programs (MIQCQP) through piecewise-linear and edge-concave relaxations. Math. Program. 136(1), 155–182 (2012)
Modaresi, S., Kılınç, M.R., Vielma, J.P.: Intersection cuts for nonlinear integer programming: convexification techniques for structured sets. Math. Program. 155, 1–37 (2015)
MOSEK ApS: The MOSEK Fusion API for C++ 8.1.0.63 (2018). https://docs.mosek.com/8.1/cxxfusion/index.html
Qualizza, A., Belotti, P., Margot, F.: Linear programming relaxations of quadratically constrained quadratic programs. In: Lee, J., Leyffer, S. (eds.) Mixed Integer Nonlinear Programming, pp. 407–426. Springer, New York (2012). https://doi.org/10.1007/978-1-4614-1927-3_14
Rikun, A.D.: A convex envelope formula for multilinear functions. J. Glob. Optim. 10(4), 425–437 (1997)
Saxena, A., Bonami, P., Lee, J.: Convex relaxations of non-convex mixed integer quadratically constrained programs: extended formulations. Math. Program. 124, 383–411 (2010)
Saxena, A., Bonami, P., Lee, J.: Convex relaxations of non-convex mixed integer quadratically constrained programs: projected formulations. Math. Program. 130, 359–413 (2011)
Sen, S., Sherali, H.D.: Nondifferentiable reverse convex programs and facetial convexity cuts via a disjunctive characterization. Math. Program. 37(2), 169–183 (1987)
Sherali, H.D., Fraticelli, B.M.P.: Enhancing RLT relaxations via a new class of semidefinite cuts. J. Glob. Optim. 22, 233–261 (2002)
Shor, N.Z.: Quadratic optimization problems. Sov. J. Circ. Syst. Sci. 25, 6 (1987)
Tardella, F.: Existence and sum decomposition of vertex polyhedral convex envelopes. Optim. Lett. 2, 363–375 (2008)
Tawarmalani, M., Richard, J.P.P., Xiong, C.: Explicit convex and concave envelopes through polyhedral subdivisions. Math. Program. 138, 1–47 (2013)
Tawarmalani, M., Sahinidis, N.V.: Convex extensions and envelopes of lower semi-continuous functions. Math. Program. 93, 247–263 (2002)
Tuy, H.: Concave programming under linear constraints. Sov. Math. 5, 1437–1440 (1964)
Vandenbussche, D., Nemhauser, G.: A branch-and-cut algorithm for nonconvex quadratic programs with box constraints. Math. Program. 102(3), 559–575 (2005)
Acknowledgements
We would like to thank the anonymous reviewers for their valuable comments. This research was partly supported by award ONR N00014-16-1-2889, Conicyt Becas Chile 72130388 and The Institute for Data Valorisation (IVADO).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix
Appendix
Proof
(Theorem 4). The OPF property is given by Lemma 2, so maximality remains. Let C be a set described by (5a) or (5b). It suffices to construct, for every symmetric matrix \(\bar{X} \not \in C\), such that \(Z-\bar{X}\in \text {int}(C)\). This implies \(Z \in \text {int} ( \text {conv}(C\cup \bar{X}))\). Denote the submatrices of \(\bar{X}, Z\):
Furthermore, for convenience let us define the following:
Construction for (5a): Suppose \(\bar{X}\) violates (5a). We propose the following:
where the last equality follows from \(\lambda _1^2 + \lambda _2^2 = 1\). This implies
and since \( \Vert ((b_Z-\bar{b})+(c_Z-\bar{c}))/2,((a_Z-\bar{a})-(d_Z-\bar{d}))/2\Vert _2 = 0 \), we conclude \(Z-\bar{X} \in \text {int}(C)\).
Construction for (5b): If \(\bar{X}\) violates (5b), we use the following construction:
We conclude \(Z-\bar{X} \in \text {int}(C)\) as before, since \( \Vert ((a_Z-\bar{a})+(d_Z-\bar{d}))/2,((b_Z-\bar{b})-(c_Z-\bar{c}))/2\Vert _2 = 0.\) It remains to set the other entries of Z and to show it is an outer product.
Claim
For each condition (i)–(viii), \(a_Zd_Z=b_Zc_Z\) and all diagonal elements among \(a_Z,b_Z,c_Z,d_Z\) are nonnegative.
Proof: First consider conditions (i)–(iv). By construction of (6):
The second equality is derived from the following identity:
Nonnegativity of diagonal elements follows from \(\Vert \bar{q}, \bar{r}\Vert _2 \ge \max \{|\bar{q}|,|\bar{r}|\}\). In case (i) only \(a_Z\) or \(d_Z\) can be diagonal elements, and they are both nonnegative. The other cases can be directly verified. Similarly, for conditions (v)–(viii):
The second equality is derived from the following identity:
Nonnegativity of diagonal elements follows from the same argument as before, by using the fact that \(\Vert \bar{p}, \bar{s}\Vert _2 \ge \max \{|\bar{p}|,|\bar{s}|\}\). \(\blacksquare \)
To maintain symmetry we set \(Z_{i_1,j_1}=Z_{j_1,i_1}\), \(Z_{i_1,j_2}=Z_{j_2,i_1}, Z_{i_2,j_1}=Z_{j_1,i_2}\), \(Z_{i_2,j_2}=Z_{j_2,i_2}\). Now denote \(\ell = [i_1,i_2,j_1,j_2]\). If \(a_Z=b_Z=c_Z=d_Z=0\), then we simply set all other entries of Z equal to zero and so Z is the outer-product of the vector of zeroes. Otherwise, consider the following cases.
Case 1: \(\ell \) has 4 unique entries. Suppose w.l.o.g we have an upper-triangular entry \((i_1<i_2<j_1<j_2)\) and furthermore suppose that \(b_Z\) is nonzero. Then set
and set all remaining entries of Z to zero. Other orderings of indices or the use of a different nonzero entry is handled by relabeling/rearranging column/row order.
Case 2: \(\ell \) has three unique entries. Then, exactly one of \(a_Z,b_Z,c_Z,d_Z\) is a diagonal entry, and so cases (i)–(iii), (v)–(vii) apply. If in any of these cases \(a_Z\) or \(d_Z\) is on the diagonal, by construction \(|b_Z|=|c_Z|\). As \(a_Zd_Z=b_Zc_Z\), we have \(b_Z=c_Z=0\) iff exactly one of \(a_Z\) or \(d_Z\) is zero. Likewise, if \(b_Z\) or \(c_Z\) is a diagonal element, then \(|a_Z|=|d_Z|\) and so \(a_Z=d_Z=0\) iff exactly one of \(b_Z\) or \(c_Z\) are zero.
Suppose \(a_Z\) is a nonzero diagonal entry. We propose:
where \(\ell '\) are the unique entries of \(\ell \). If \(a_Z=0\) and on the diagonal, then we replace \(b_Z^2/a_Z\) and \(c_Z^2/a_Z\) with \(|d_Z|\). If \(b_Z,c_Z\) or \(d_Z\) is on the diagonal, we use the same construction but with relabeling/rearranging column/row order.
Case 3: \(\ell \) has two unique entries. All remaining entries of Z are set to zero.
For all cases, our construction ensures that all diagonal entries of Z are nonnegative, and all \(2\times 2\) minors are zero; by Proposition 1, Z is an outer-product. \(\square \)
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Bienstock, D., Chen, C., Muñoz, G. (2019). Intersection Cuts for Polynomial Optimization. In: Lodi, A., Nagarajan, V. (eds) Integer Programming and Combinatorial Optimization. IPCO 2019. Lecture Notes in Computer Science(), vol 11480. Springer, Cham. https://doi.org/10.1007/978-3-030-17953-3_6
Download citation
DOI: https://doi.org/10.1007/978-3-030-17953-3_6
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-17952-6
Online ISBN: 978-3-030-17953-3
eBook Packages: Computer ScienceComputer Science (R0)