A New Branch-and-Bound Method for Global Optimization

K. Madsen and S. Zertchaninov

For a copy of this paper, either

Abstract

The Interval Branch-and-Bound Method for global optimization over a compact right parallelepiped, parallel to the coordinate axes, forms the basis of a new stochastic method which can be applied even when function values are only known as black boxes.

The stochastic method is described and convergence to the set of global minimizers is proved under the assumption that the number of stationary points is finite. The method has been implemented in C++, and numerical experiments with some well known test problems in up to 10 variables are described. Comparisons are made with the interval method.

Technical report 05/98

Last modified March 31, 1998
IMM HomePage