On May 22, 2015, Yanting Chen defended her PhD thesis "Random Walks in the Quarter-Plane: Invariant Measures and Performance Bounds". Prior to this ceremony, a seminar was held.
Date&time: May 22, 10:30 – 17.30
Location of Mini Symposium: Carre-2H
Location of the Defense: Berkhoff Auditorium
10:30 - 10:45 Welcome coffee and tea
10:45 - 11:15 Ivo Adan (Eindhoven University of Technology)
11:30 - 12:00 Joost-Pieter Katoen (University of Twente/ RWTH Aachen University)
12:10 - 13:20 Lunch at Ravelijn Atrium
13:30 - 14:00 Nico van Dijk (University of Amsterdam/ University of Twente)
14.30 - 14.40 Introduction to the defense
14.45 – 15.30 PhD defense of Yanting Chen
15.45 – 16.15 Ceremony
16.30 – 17.30 Reception
10:45 - 11:15 Ivo Adan (Eindhoven University of Technology):
Stretching the limits of the compensation approach
Abstract: This talk reviews some results on the use of compensation arguments to solve the steady-state equations for two-dimensional Markov processes. It also briefly discusses some more recent results pointing to possible extensions of the applicability of this approach.
11:30 - 12:00 Joost-Pieter Katoen (University of Twente/ RWTH Aachen University):
Stochastic Petri Nets Revisited
Abstract: Analysis algorithms for Generalized stochastic Petri nets (GSPNs) are restricted to confusion-free nets. We discuss how this restriction can be overcome. In addition, we show counterexamples to some classical results on ergodicity in unbounded SPNs and present new results that `repair' these flaws.
13:30 - 14:00 Nico van Dijk (University of Amsterdam/ University of Twente):
How large can an error be?
Abstract: Markov chains are known to be analytically hard and computationally expensive if not prohibitive. As elegantly shown by Yanting’s thesis, special multi-dimensional birth-death structures might therefore be required. The effect of a parameter perturbation (e.g. as by statistical inaccuracy), or of a simplifying system modification (e.g. as for analytic solvability), or of a state space truncation (e.g. for computational reduction) is thus of interest. It will be argued that their quantifaction all comes down to one and the same: its bounding of so-called bias terms. Two practical illustrations will be provided.