Prof. Emo Welzl and Prof. Bernd Gärtner
|Mittagsseminar Talk Information|
Date and Time: Friday, May 16, 2014, 12:15 pm
Duration: 30 minutes
Location: CAB G61
Speaker: Carola Doerr (Laboratoire d'Informatique de Paris 6)
The star discrepancy of a point set is a measure for its irregularity. Point sets of low star discrepancy value allow Quasi-Monte Carlo integration with small approximation error. Obtaining such point sets is thus one of the classic and well-studied tasks in numerical integration theory. While most research has focused on computing point sets with nice asymptotic behavior, not many constructions are known for optimizing the star discrepancy in fixed dimensions and for a fixed number of points. For all practical purposes, it is thus questionable if the existing constructions are satisfactory. In this talk we will see that indeed there is much room for improvement. We use a simple randomized search heuristic to find point sets that have a star discrepancy value that is much smaller than what theoretical results predict.
Automatic MiSe System Software Version 1.4803M | admin login