RESIM 2012 - Program

An archive of all talks you find here.

Monday June 25

11:00 – 12:00Registration and coffee
12:00 – 13:30Lunch
13:30 – 13:45


Geir E. Øien
Dean of the Faculty of Information Technology, Mathematics and Electrical Engineering
13:45 – 14:45

Invited presentation 1

Chair: Poul E. Heegaard, NTNU
Speaker: Ad Ridder
Affiliation: Department of Econometrics and Operations Research, Faculty of Economics and Business Administration, Vrije Universiteit, Amsterdam, The Netherlands
Title: Minimum cross-entropy methods for rare-event simulation
Abstract: In this work we consider an importance sampling simulation procedure for estimating rare-event probabilities. Typical examples are buffer overflows in queueing systems or machine unavailability in reliability systems. The main idea of our approach is to associate with the original problem an auxiliary single-constrained convex optimization program of a special type, namely, the minimization of the Kullback-Leibler divergence, subject to an integral equation. We propose to do the importance sampling from the solution of this program. For a special case we prove that the optimal pdf is a zero variance pdf. Next we consider more general problems and obtain state- and time-dependent importance sampling estimators, which are logarithmic efficient. Finally, we present results of simulation experiments in a range of applications including comparisons between our algorithms and others from the literature. The talk is based on joint work with Reuven Rubinstein and Thomas Taimre.
14:45 – 15:15Break
15:15 – 16:45

Session 1: RESTART - Splitting

Chair: Michele Pagano, University of Pisa

Tuesday June 26

09:00 – 10:00

Invited presentation 2

Chair: Bjarne E. Helvik, NTNU
Speaker: Henrik Hult
Affiliation: Department of Mathematics, Royal Institute of Technology (KTH), Stockholm, Sweden
Title: Efficient Monte Carlo algorithms for computing high quantiles
Abstract: Extreme risks in a financial or actuarial context is often quantified by a quantile or other tail-based risk measure. Efficient calculation of such risk measures is important as they influence management and allocation decisions. In many ways the computation of a high quantile is similar to computing the probability of the event that the portfolio value falls below a low threshold. The main difference is that there may be a need to compute many rare-event probabilities for an accurate evaluation of a risk measure. In this talk we present several adapted importance sampling algorithms where the adaption takes advantage of the shape of the asymptotic tail to improve accuracy. The algorithms are illustrated in the context of computing high quantiles for a heavy-tailed random walk. The talk is partly based on joint work with Jens Svensson and Cecile Mercadier.
10:00 – 10:30Break
10:30 – 12:00

Session 2: Reliability and risk

Chair: Gerardo Rubino, INRIA
12:00 – 13:30Lunch
13:30 – 15:00

Session 3: Sampling of rare events

Chair: Pieter-Tjerk de Boer, University of Twente
15:00 – 15:30Break
15:30 – 16:30

Session 4: Applied techniques

Chair: José Villén-Altamirano, Technical University of Madrid

Social event

Wednesday June 27

09:00 – 10:00

Session 5: Performance of importance sampling techniques

Chair: Paul Dupuis, Brown University
10:00 – 10:30Break
10:30 – 12:00

Session 6: Random walks

Chair: Werner Sandmann, Clausthal University of Technology
12:00 – 13:30Lunch