Global optimization of quadratically constrained problems

Conference · Sat Dec 31 04:00:00 EST 1994

OSTI ID:36118

Haugland, D

The problem of minimizing a quadratic function subject to quadratic constraints is studied. No assumptions concerning definiteness of any of the quadratic forms are made. In order to locate a global solution, a deterministic tree search algorithm is given. The suggested method is based on convex annexations of the original problem leading to linear subproblems. Unlike most familiar methods in deterministic global optimization, the proposed algorithm does not possess the property of exhaustive partition rules. This opens for a shorter search path to optimum, and thereby savings in computational requirements. The paper presents the theory of the algorithm, including a proof of convergence to a global optimum. Numerical experiments on standard test problems show good prospectives of the method. Finally we discuss computational aspects and some approaches to further speed-up of the algorithm.

• OSTI ID: 36118
• Report Number(s): CONF-9408161--
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
ALGORITHMS
MATRICES
NONLINEAR PROBLEMS
NONLINEAR PROGRAMMING
NUMERICAL SOLUTION