COLLOQUIUM
DEPARTMENT OF MATHEMATICS AND STATISTICS
OAKLAND UNIVERSITY
ROCHESTER, MICHIGAN 48309
Ambros Gleixner
Zuse Institute Berlin
Bound tightening techniques in mixed-integer nonlinear programming
State-of-the-art
solvers for generic mixed-integer nonlinear programming utilize a
linear relaxation within a branch-and-cut search. We will give a short
introduction to computational MINLP and present recent advances in
propagation techniques. The focus will be on improvements of
optimization-based bound tightening, an expensive method that minimizes
and maximizes each variable over a linear relaxation. We show how dual
information from the LP solutions can be exploited to speed up the
branch-and-bound search. This is joint work with Timo Berthold, Stefan
Weltge, and Stefan Vigerske.
Tuesday, May 28, 2013
3:00– 4:00 P.M.
372 Science and Engineering Building
(Refreshments at 2:30-3:00 PM in the kitchen area adjacent to 368 SEB)