University of Bamberg
open search form
You are here

Information Systems and Applied Computer Sciences

Communication Networks

  1.  University of bamberg
  2. Faculties
  3. Information Systems and Applied Computer Sciences
  4. Subject groups
  5. Computer Science
  6. Communication Networks
  7. Teaching
  8. PhD Courses

Rare Event Simulation Methods for Stochastic Models of Computer and Communication Systems

PhD Course, October 2-6, 2006

Duration: 20 hours

Instructor: Dr. rer. nat. Werner Sandmann


General Information

The course is part of the joint PhD program of the European Network of Excellence EuroNGI - Design and Engineering of the Next Generation Internet. Applicants from the EuroNGI community please check the official procedure on the EuroNGI web pages. Interested PhD students who are not members of the EuroNGI please send me an email.

As a closely related event, directly after the PhD Course, the 6th International Workshop on Rare Event Simulation (RESIM2006) takes place in Bamberg from October 8-10.  PhD students are cordially invited and particularly encouraged to participate.

Location

The course will be held at the University of Bamberg, Building Feldkirchenstr. 21; see travel information. Participants are responsible for their accommodation.

Lecture Room: F381

Course Times

The lectures will be given on four days, five hours each day at the times from 10-13 and 14-16. That is

  • Monday, October 2, 10-13 and 14-16
  • Wednesday, October 4, 10-13 and 14-16
  • Thursday, October 5, 10-13 and 14-16
  • Friday, October 6, 10-13 and 14-16

Note that Tuesday, October 3 is a public holiday in Germany.

Prerequisites

  • Basics of probability, statistics, and Markov chains
  • Basics of queueing theory
  • Basics of simulation output data analysis

Course Contents

1 Rare Events
   1.1 Introduction and Motivation
   1.2 Characterizing Rare Events: A Taste of Large Deviations
   1.3 The Problem of Rare Event Simulation
   1.4 Approaches to Simulation Speed-up

2 Importance Sampling
   2.1 General Basics
   2.2 Efficiency Criteria
   2.3 Classical Change of Measure
   2.4 Applications and Examples

3 Importance Sampling for Markovian Models
   3.1 Review of Markov Chains
   3.2 Formal Basis of Importance Sampling for Markov Chains
   3.3 Optimal Importance Sampling for Markovian Models
   3.4 Application to Higher Level Model Descriptions
   3.5 Heuristics for Tandem Jackson Networks
   3.6 Heuristics for Markovian Reliability Models
   3.7 Attractor - Rare Set Framework and Cyclic Approach

4 Adaptive Importance Sampling
   4.1 Unified Parametrization
   4.2 Direct Variance Minimization
   4.3 The Cross Entropy Method

5 Introduction to RESTART