Semidefinite optimization for solving boolean quadratic problems

Authors

  • A. I. Kosolap State Higher Education Establishment “Ukrainian State University of Chemical Technology”, 40, Naberezhna Peremogy str., Dnipro, Ukraine
  • A. S. Peretiatko State Higher Education Establishment “Ukrainian State University of Chemical Technology”, 40, Naberezhna Peremogy str., Dnipro, Ukraine

Keywords:

quadratic optimization, boolean optimization, boolean quadratic problems, semidefinite optimization, semidefinite relaxation, lower and upper bounds.

Abstract

Abstract.  Purpose.  We consider boolean quadratic optimization problems. Such problems arise in market economy, planning,  project management, artificial intelligence, optimal design and are NP-hard. In this paper we propose a procedure for finding the  lower and upper bounds of the objective functions of boolean quadratic problems.  Methodology. In this paper we propose to convert  the boolean quadratic problem to the general quadratic optimization problem, and then apply semi definite relaxation to it. Findings. The using of semidefinite relaxation has allowed us to find lower bounds of the objective function or global minimum point with a given accuracy in boolean quadratic problems. Moreover, unlike other methods, testing the  optimality of the found solution doesn`t require exponential time and is quite simple. The upper bound of the objective function, which is found by using the lower es timate  of semidefinite relaxation as a starting point for solving the original problem by local search methods, in 91% of cases coincided with the global minimum of the initial problem.  Originality. A new procedure for finding the upper and lower bounds of objective function was proposed.  Practical value.  The considered method of solution can  be used for finding the solution in applications that can be  formulated as boolean quadratic problems. Comparative numerical experiments confirm the efficiency of this approach to solvin g such problems.

Author Biographies

A. I. Kosolap, State Higher Education Establishment “Ukrainian State University of Chemical Technology”, 40, Naberezhna Peremogy str., Dnipro

Dr. Sc. (Phys.-Math.), Prof.

A. S. Peretiatko, State Higher Education Establishment “Ukrainian State University of Chemical Technology”, 40, Naberezhna Peremogy str., Dnipro

Ph.D. (Phys.-Math.)

References

Kosolap A.I. and Peretiatko A.S. Chislennaya effektivnost metodov poluopredelennoy optimizatsii [Numerical efficiency of semidefinite optimization methods]. Problemy upravleniya i informatiki [Problems of control and informatics], 2014, no. 2, pp. 56–64. (in Russian).

Peretiatko A.S. Napivvyznachena optymizatsiia dlia zadachi klasterizatsii danykh k-clustering [Semidefinite optimization for k-klustering problem]. Materialy Vseukrainskoi naukovo-metodychnoi konferentsii «Problemy matematychnogo modeliuvannia (25.05–27.05.2016)» [Proc. of the All Ukrainian Scientific and Methodical Conf. «Problems of mathematical modelling»]. Dniprodzerzinsk, 2016, pp. 57–60. (in Ukrainian).

Blekherman G., Parrilo P.A.and Rekha R.T. Semidefinite Optimization and Convex Algebraic Geometry. SIAM, 2013, 476 p. Available at: http://www.math.washington.edu/~thomas/frg/frgbook/SIAMBookFinalv Nov12-2012.pdf.

Ding Y., Ge D., Wolkowicz H. On Equivalence of Semidefinite Relaxations for Quadratic Matrix Programming. 2010, 23 p. Available at: www.optimization-online.org/DB_FILE/2010/04/2606.pdf.

Horst R., Tuy H. Global Optimization: Deterministic Approaches. 3rd ed.,Berlin, Springer–Verlag, 1996, 727 p.

Karger D., Motwani R., Sudan M. Approximate Graph Coloring by Semidefinite programming. 18 p. Available at: http://dl.acm.org/citation.cfm?id=274791.

Karoussi E. Data Mining K-Clustering Problem. University of Agder, 2012, 80 p. Available at:

https://brage.bibsys.no/xmlui/bitstream/handle/11250/137565/masteroppgave.pdf?sequence=1.

Kenneth V.P., Storn R.M., Lampinen J.A. Differential Evolution. A Practical Approach to Global Optimization. BerlinHeidelberg: Springer-Verlag, 2005, 542 p.

Nesterov Y.E., Nemirovskii A. S. Interior point polynomial algorithms in convex programming. SIAMStudies in Applied Mathematics,SIAM:Philadelphia, 1994, Vol. 13, 405 p.

Nocedal J., Wright S.J. Numerical optimization. Springer, 2006, 685 p.

O’Donnell R. Advanced Approximation Algorithms. Lecture 14: Semidefinite Programming and Max-Cut. 2008, 6 p. Available at: www.cs.cmu.edu/ ~anupamg/adv-approx/lecture14.pdf.

Roumili H., Keraghel A., Yassine A. Infeasible Interior Point Method for Semidefinite Programs. Applied Mathematical Sciences, 2007, Vol. 1, No. 21, pp. 1009–1018. Available at: http://www.m-hikari.com/ams/ams-password-2007/ams-password21-24-2007/roumiliAMS21-24-2007.pdf.

Wen Z., Goldfarb D., Yin W. Alternating Direction Augmented Lagrangian Methods for Semidefinite Programming. Math. Prog, 2010, №2, pp. 203–230. Available at: http://mpc.zib.de/index.php/MPC/article/viewFile/40/20.

Published

2016-09-27

Issue

Section

Computer systems and information technologies in education, science and management