Prof. Emo Welzl and Prof. Bernd Gärtner
|Mittagsseminar Talk Information|
Date and Time: Thursday, June 05, 2008, 12:15 pm
Duration: This information is not available in the database
Location: OAT S15/S16/S17
Speaker: Piotr Sankowski (ETH Zurich)
The seminar will be dedicated to the study of online optimization problems, where the demands are given by a probability distribution, e.g., for online Steiner tree if at each time instant, the demand vertex is a uniformly random vertex from the graph, which needs to be connected before the next (random) vertex arrives. During the seminar I will introduce the notion of competitiveness that can be used in this stochastic setting. Next, I will show that one can beat the tight logarithmic bound known for online Steiner tree, i.e., prove that if each demand vertex is an independent draw from some probability distribution, a variant of the natural greedy algorithm achieves an average cost that is at most a constant away from the average optimal cost. Finally, I will talk about stochastic version of the online set cover problem, and show that here again one can improve over standard online algorithms. Astonishingly, the presented algorithms can be used to construct universal/oblivious solutions for the studied problems as well.
Automatic MiSe System Software Version 1.4803M | admin login