D. Software
Refine
Year of publication
- 2015 (2) (remove)
Document Type
- Master's Thesis (1)
- ZIB-Report (1)
Language
- English (2)
Has Fulltext
- yes (2) (remove)
Is part of the Bibliography
- no (2)
Keywords
- MIPLIB (1)
- Mixed Integer Programming (1)
- Node merging (1)
- Parallel processing (1)
- Racing ParaSCIP (1)
- Steiner Tree Problem (1)
- Steiner Tree Problem Variants (1)
- Ubiquity Generator Framework (1)
Institute
Spawned by practical applications, numerous variations of the classical Steiner tree problem in graphs have been studied during the last decades. Despite the strong relationship between the different variants, solution approaches employed so far have been prevalently problem-specific.
In contrast, we pursue a general-purpose strategy resulting in a solver able to solve both the classical Steiner tree problem and ten of its variants without modification. These variants include well-known problems such as the prize-collecting Steiner tree problem, the maximum-weight connected subgraph problem or the rectilinear minimum Steiner tree problem. Bolstered by a variety of new methods, most notably reduction techniques, our solver is not only of unprecedented versatility, but furthermore competitive or even superior to specialized state-of-the-art programs for several Steiner problem variants.
This paper describes how we solved 12 previously unsolved mixed-integer program-
ming (MIP) instances from the MIPLIB benchmark sets. To achieve these results we
used an enhanced version of ParaSCIP, setting a new record for the largest scale MIP
computation: up to 80,000 cores in parallel on the Titan supercomputer. In this paper
we describe the basic parallelization mechanism of ParaSCIP, improvements of the
dynamic load balancing and novel techniques to exploit the power of parallelization
for MIP solving. We give a detailed overview of computing times and statistics for
solving open MIPLIB instances.