Prof. Emo Welzl and Prof. Bernd Gärtner
|Mittagsseminar Talk Information|
Date and Time: Tuesday, September 11, 2012, 12:15 pm
Duration: 30 minutes
Location: NO C44
Speaker: Uwe Schöning (Universität Ulm)
The development of local search algorithms for the satisfiability problem has resulted in increasingly sophisticated concepts during the last years. As it turns out, a very simple and straightforward approach which is just based on probability distributions controlling the selection of the next flip variable results in an algorithm being able to beat all previous (at least up to 2011) local search algorithms, with respect to the benchmark sets used at the SAT competition.
For calculating probability distributions, the make- and the break-values of the flip variables under consideration are important, where make=number of additionally satisfied clauses, and break=number of additionally falsified clauses, after flipping the respective variable assignment. It seems reasonable that (make - break) is a natural score parameter which should be used to calculate flip probabilities.
Automatic MiSe System Software Version 1.4803M | admin login