Summary: The minisymposium will deal with numerical methods for optimization which deliver guaranteed bounds for the optimal value of the objective function and/or the minimizers. The methods can be applied to a large class of (potentially non-differentiable or discontinuous) optimization problems.
As opposed to 'classical' numerical optimization methods, these methods use information on the whole domain rather than on a finite number of isolated points. For certain classes of problems these algorithms are competitive (with respect to speed) to well-established non-validating algorithms.
Day: July 7, 1995 Time: 09.30
S. Berner, Bergische Universität-GH Wuppertal, Germany:
Parallel validated global optimization
T. Csendes, Jozsef Attila University, Hungary;
D. Ratz, University of Karlsruhe, Germany:
Subdivision direction selection in interval methods for global
optimization
C. Jansson, Technische Universität Hamburg-Harburg, Germany:
An expansion principle in validated global optimization
D. Ratz, Universität (TH) Karlsruhe, Germany:
Improved techniques for gap-treating and box-splitting in interval
Newton Gauss-Seidel steps
K. Madsen, The Technical University of Denmark, Denmark:
Real versus interval methods for global optimization