Mini-Symposium on Performance Analysis of Markov Processes
Date: September 20, 2018
Location: Ravelijn 2503
Lunch: Ravelijn Atrium
Program
10:30 - 10:45 Welcome coffee and tea
10:45 - 11:15 Stella Kapodistria (Eindhoven University of Technology)
11:30 - 12:00 Rudesindo Núñez-Queija (University of Amsterdam)
12:10 - 13:15 Lunch at Ravelijn Atrium
13:15 - 13:45 Konstantin Avrachenkov (INRIA Sophia Antipolis, France)
14.30 - 16.00 PhD defense of Xinwei Bai
Performance Bounds for Random Walks in the Positive Orthant
[Location: Building Waaier, Room 4]
16:00 – 17:30 Reception
Speakers & Abstracts
10:45 - 11:15 Stella Kapodistria (Eindhoven University of Technology):
Performance analysis for random time-limited polling models
Abstract: We consider a polling model with a service discipline based on the randomly timed gated service: the server resides at a queue for a random amount time and does not switch even if the queue becomes empty but only when the random residing clock expires. Such a service discipline has two advantages, as it enables to: i) keep the frequency of switching at a predetermined level (thus controlling the total cost, if there is a switching cost); ii) balance the time that the server spends in each queue (since, contrary to exhaustive or gated service disciplines, this discipline does not depend on the number of customers present in the various queues). This polling model violates the branching property, thus a direct analytic derivation of the joint queue length distribution turns out to be very difficult. To overcome this issue, we explore on the one hand the use of (parametric) perturbation and on the other hand the reduction of the original problem to a boundary value problem. Furthermore, we compute the marginal workload distribution and prove a decomposition property that permits us to carry out the heavy traffic and/or heavy (light) tail asymptotic analysis. (joint work with Mayank Saxena, Onno Boxma, and Rudesindo Núñez-Queija)
Biography: Stella Kapodistria is assistant professor in the section Stochastics of the Department of Mathematics and Computer Science in Eindhoven University of Technology. Since 2014, she is a member of the Networks project and is involved in the research program Smart Maintenance and Manufacturing of the Data Science Center of TU/e. Moreover, since 2013, she has been involved in the development of prognostic and diagnostic algorithms for the DAISY and the DAISY4OFFSHORE projects on wind turbines, and for the Impulse Data Science program with Philips. She serves on the editorial board of the Probability in the Engineering and Informational Sciences international journal and is a guest editor of the Annals of Operation Research international journal.
Her main research interests are in applied probability and stochastic operations research, with a particular interest on the modelling of processes and on the decision making under uncertainty of practices arising in energy and maintenance. The research topics that motivate her research tend to be real-life problems arising from practical situations, which can be analyzed using traditional and non-traditional tools from stochastic processes and optimization.
11:30 - 12:00 Rudesindo Núñez-Queija (University of Amsterdam):
Some queueing problems motivated from road traffic modelling
Abstract: In this talk we will discuss the analysis of a few queueing models that are motivated from three problems arising in road traffic and logistics. The first problem we discuss is the control of two sequentially connected service systems, of which the second has limited buffer capacity and the means of control lies in the regulation of the operating speed of the first system. The second problem concerns the control of heavily loaded traffic light systems in an urban network. If time allows, we will discuss the capacity analysis of a queueing model for highway access and unsignalized road intersections.
This work has been done in collaboration with several people including Daphne van Leeuwen, Peter Kovacs, Mayank Saxena, and Abhishek.
Biography: Rudesindo (Sindo) Núñez Queija did his PhD research at CWI in Amsterdam and obtained his PhD in Applied Mathematics from Eindhoven University of Technology in 2000. He was subsequently affiliated with INRIA Sophia Antipolis, Eindhoven University of Technology, TNO Information and Communication Technology, University of Amsterdam and was a visiting professor at the University of Toulouse (2015-2016). He is currently part-time affiliated with CWI and with the Korteweg - De Vries Mathematical Institute (Faculty of Science of the University of Amsterdam).
His main research interests are in queueing theory and its application to stochastic networks, particularly information and communication networks and - more recently - road traffic networks. He has particularly worked on the analysis of processor sharing systems and scheduling in multi-class service systems. Currently much of his work concentrates on the use of asymptotic analysis techniques (heavy traffic, fluid scaling, perturbation techniques) to study complex stochastic systems.
13:15 - 13:45 Konstantin Avrachenkov (INRIA Sophia Antipolis, France):
Singularly perturbed linear programs and their application to MDP
Abstract: Firstly I review different types of singularly perturbed linear
programs and their regularizations. Singular perturbation, or discontinuity
of the optimal solution with respect to the perturbation parameter, can
arise from either the discontinuity in the rank of the constraints or from
the violation of the Slater's condition. Various regularizations can be
constructed to avoid dealing with ill-conditioned problems. Then, secondly,
I discuss the application of the singularly perturbed linear programs and
their regularizations to Markov decision processes.
Biography: Konstantin Avrachenkov received Master degree in Control Theory from
St. Petersburg State Polytechnic University (1996), Ph.D. degree in Mathematics
from University of South Australia (2000) and Habilitation from University
of Nice Sophia Antipolis (2010). Currently, he is a Director of Research at
Inria Sophia Antipolis, France. He is an associate editor of International
Journal of Performance Evaluation and ACM TOMPECS. His main research interests
are Markov processes, singular perturbation theory, optimization, game theory,
and analysis and control of complex network systems.