2009.08.25 |
| Date | Fri Aug 28 |
| Time | 10:15 — 11:15 |
| Location | Turing 014 |
This talk is postponed to a later date and so will not take place Fridy August 28.
Title: Importance sampling for failure probabilities in computing and data transmission
Friday August 28, 2009, at 10:15-11:15 in Turing-014 at Department of Computer Science.
Søren Asmussen, University of Aarhus, Department of Mathematical Sciences
Abstract: Importance sampling is a technique for improving the performance of crude Monte Carlo simulation. One of the main areas of application is simulation of small probabilities (rare event simulation) where there are many examples where the improvement one can obtain roughly corresponds from going from exponential time algorithms to polynomial time ones.
I give a short introduction to the method and a short survey of classical examples. The last part of the talk is a more detailed case study. A task like the execution of a computer program or the transmission of a file may fail and then needs to restarted. How does one do importance sampling for estimating the rare event probability ofexcessive delays due to failures?
See www.daimi.au.dk/~bjarke/compmath/ for details.