- Message from the Dean
- History
- Education and Research
- Staff Introduction
- Seminars & Events
- Distinctive Programs
- Access
- Job Openings
- Publications
- Related Links
- Contacts
Seminars
Filtered performance diagrams for analyzing MINLP solver performance
- Hold Date
- 2014-06-17 10:30〜2014-06-17 12:00
- Place
- #317, Faculty of Mathematics building, Ito Campus
- Object person
- Speaker
- Ambros Gleixner (ZIB)
Abstract:
General-purpose solvers that address large and heterogeneous problem classes like mixed-integer nonlinear programming (MINLP) necessarily combine a variety of algorithmic techniques in their solution process. In this talk, we present recent advances in the constraint integer programming-based MINLP solver SCIP with a special focus on analyzing the computational impact of individual solver components such as branching strategies, separation routines, bound tightening techniques, and primal heuristics. We propose the use of so-called filtered performance diagrams in order to exhibit average performance impact not only on the overall collection of a set of benchmark instances, but also on subsets of instances of increasing hardness.