12.07.2015 Views

ILOG CPLEX 11.0 User's Manual

ILOG CPLEX 11.0 User's Manual

ILOG CPLEX 11.0 User's Manual

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

convex QPs, Q must be positive semi-definite; that is, x T Qx ≥ 0 for every vector x, whetheror not x is feasible. For concave maximization problems, the requirement is that Q must benegative semi-definite; that is, x T Qx ≤ 0 for every vector x. It is conventional to use the sameterm, positive semi-definite, abbreviated PSD, for both cases, on the assumption that amaximization problem with a negative semi-definite Q can be transformed into an equivalentPSD.For a separable function, it is sufficient to check whether the individual diagonal elements ofthe matrix Q are of the correct sign. For a nonseparable case, it may be less easy to decide inadvance the convexity of Q. However, <strong>ILOG</strong> <strong>CPLEX</strong> detects this property during the earlystages of optimization and terminates if the quadratic objective term in a QP is found to benot PSD.For a more complete explanation of quadratic programming generally, a text, such as one ofthose listed in Further Reading on page 39 of the preface of this manual, will be helpful.Entering QPs<strong>ILOG</strong> <strong>CPLEX</strong> supports two views of quadratic objective functions: a matrix view and analgebraic view.◆ Matrix View on page 229◆ Algebraic View on page 230◆ Examples for Entering QPs on page 230◆ Reformulating QPs to Save Memory on page 231Matrix ViewIn the matrix view, commonly found in textbook presentations of QP, the objective functionis defined as 1/2 x T Qx+c T x, where Q must be symmetric and positive semi-definite for aminimization problem, or negative semi-definite for a maximization problem. This view issupported by the MPS file format and the Callable Library routines, where the quadraticobjective function information is specified by providing the matrix Q. Thus, by definition,the factor of 1/2 must be considered when entering a model using the matrix view, as it willbe implicitly assumed by the optimization routines.Similarly, symmetry of the Q matrix data is required; the MPS reader will return an errorstatus code if the file contains unequal off-diagonal components, such as a nonzero value forone and zero (or omitted) for the other.This symmetry restriction applies to quadratic programming input formats rather than thequadratic programming problem itself. For models with an asymmetric Q matrix, eitherexpress the quadratic terms algebraically, as described in Algebraic View on page 230, or<strong>ILOG</strong> <strong>CPLEX</strong> <strong>11.0</strong> — USER’ S MANUAL 229

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!