Eigenvalue-based algorithm and analysis for nonconvex QCQP with one constraint

A nonconvex quadratically constrained quadratic programming (QCQP) with one constraint is usually solved via a dual SDP problem, or Moré’s algorithm based on iteratively solving linear systems. In this work we introduce an algorithm for QCQP that requires finding just one eigenpair of a generalized eigenvalue problem, and involves no outer iterations other than the (usually black-box) iterations for computing the eigenpair. Numerical experiments illustrate the efficiency and accuracy of our algorithm. We also analyze the QCQP solution extensively, including difficult cases, and show that the canonical form of a matrix pair gives a complete classification of the QCQP in terms of boundedness and attainability, and explain how to obtain a global solution whenever it exists.

[1]  Jos F. Sturm,et al.  A Matlab toolbox for optimization over symmetric cones , 1999 .

[2]  Y. Moon,et al.  Finding a positive definite linear combination of two Hermitian matrices , 1983 .

[3]  Henry Wolkowicz,et al.  The generalized trust region subproblem , 2014, Comput. Optim. Appl..

[4]  Nicholas I. M. Gould,et al.  Solving the Trust-Region Subproblem using the Lanczos Method , 1999, SIAM J. Optim..

[5]  Arkadi Nemirovski,et al.  Lectures on modern convex optimization - analysis, algorithms, and engineering applications , 2001, MPS-SIAM series on optimization.

[6]  Aaas News,et al.  Book Reviews , 1893, Buffalo Medical and Surgical Journal.

[7]  Henry Wolkowicz,et al.  The trust region subproblem and semidefinite programming , 2004, Optim. Methods Softw..

[8]  Kim-Chuan Toh,et al.  SDPT3 -- A Matlab Software Package for Semidefinite Programming , 1996 .

[9]  Chao Yang,et al.  ARPACK users' guide - solution of large-scale eigenvalue problems with implicitly restarted Arnoldi methods , 1998, Software, environments, tools.

[10]  G. Fehmers,et al.  An algorithme for quadratic optimization with one quadratic constraint and bounds on the variables , 1998 .

[11]  J. J. Moré.  Generalizations of the trust region problem , 1993 .

[12]  Stephen J. Wright,et al.  Numerical Optimization , 2018, Fundamental Statistical Inference.

[13]  Nicholas I. M. Gould,et al.  Trust Region Methods , 2000, MOS-SIAM Series on Optimization.

[14]  Akiko Takeda,et al.  Solving the Trust-Region Subproblem By a Generalized Eigenvalue Problem , 2017, SIAM J. Optim..

[15]  Leiba Rodman,et al.  Canonical Forms for Hermitian Matrix Pairs under Strict Equivalence and Congruence , 2005, SIAM Rev..

[16]  Akiko Takeda,et al.  Solving Generalized CDT Problems via Two-Parameter Eigenvalues , 2016, SIAM J. Optim..

[17]  R. C. Thompson,et al.  Pencils of complex and real symmetric and skew matrices , 1991 .

[18]  R. C. Thompson,et al.  The characteristic polynomial of a principal subpencil of a Hermitian matrix pencil , 1976 .

[19]  Danny C. Sorensen,et al.  Minimization of a Large-Scale Quadratic FunctionSubject to a Spherical Constraint , 1997, SIAM J. Optim..

[20]  Akiko Takeda,et al.  Global Optimization Methods for Extended Fisher Discriminant Analysis , 2014, AISTATS.

[21]  Tamás Terlaky,et al.  A Survey of the S-Lemma , 2007, SIAM Rev..

[22]  Franz Rendl,et al.  A semidefinite framework for trust region subproblems with applications to large scale minimization , 1997, Math. Program..

[23]  Nicholas J. Higham,et al.  An Improved Arc Algorithm for Detecting Definite Hermitian Pairs , 2009, SIAM J. Matrix Anal. Appl..

[24]  Henry Wolkowicz,et al.  Local nonglobal minima for solving large-scale extended trust-region subproblems , 2015, Comput. Optim. Appl..

[25]  Duan Li,et al.  SOCP reformulation for the generalized trust region subproblem via a canonical form of two symmetric matrices , 2016, Math. Program..

[26]  Stephen A. Vavasis,et al.  Quadratic Programming is in NP , 1990, Inf. Process. Lett..

[27]  Yong Xia,et al.  Duality and solutions for quadratic programming over single non-homogeneous quadratic constraint , 2012, J. Glob. Optim..

[28]  Danny C. Sorensen,et al.  A New Matrix-Free Algorithm for the Large-Scale Trust-Region Subproblem , 2000, SIAM J. Optim..

[29]  Stephen P. Boyd,et al.  Convex Optimization , 2004, Algorithms and Theory of Computation Handbook.

[30]  W. Gander,et al.  A constrained eigenvalue problem , 1988 .

[31]  D. Sorensen,et al.  4. The Implicitly Restarted Arnoldi Method , 1998 .

[32]  Ruey-Lin Sheu,et al.  A Revisit to Quadratic Programming with One Inequality Quadratic Constraint via Matrix Pencil , 2013, 1312.1400.

[33]  Jack Dongarra,et al.  Templates for the Solution of Algebraic Eigenvalue Problems , 2000, Software, environments, tools.

[34]  Sartaj Sahni,et al.  Computationally Related Problems , 1974, SIAM J. Comput..

[35]  Danny C. Sorensen,et al.  Algorithm 873: LSTRS: MATLAB software for large-scale trust-region subproblems and regularization , 2008, TOMS.

[36]  Jorge J. Moré,et al.  Computing a Trust Region Step , 1983 .

[37]  V. Mehrmann,et al.  Möbius transformations of matrix polynomials , 2015 .

[38]  James Demmel,et al.  The generalized Schur decomposition of an arbitrary pencil A–λB—robust software with error bounds and applications. Part I: theory and algorithms , 1993, TOMS.

[39]  Gene H. Golub,et al.  Matrix computations , 1983 .