Preliminary program:

Tuesday, June 28, 2016

8:30-9:00	Registration
9:00	Opening
9:00-10:15	Opening address László Gerencsér: An interaction between queueing and change detection Session 1: Queueing applications (chair: Giuliano Casale) Bruce Jones, Sally McClean and David Stanford: Modelling Mortality and Discharge of Hospitalized Stroke Patients Caglar Tunc and Nail Akar: Performance Modeling of Delay-based Dynamic Speed Scaling
10:15-10:50	Break
10:50-12:30	Session 2: Numerical methods (chair: Nail Akar) Dario Bini, Guy Latouche and Beatrice Meini: Shift techniques for Quasi-Birth-and-Death processes: Canonical factorizations and matrix equations Gábor Horváth and Miklós Telek: Matrix-analytic solution of second order Markov fluid models by using matrix-quadratic equations Nigel Bean, Giang Nguyen, Malgorzata O'Reilly and Vikram Sunkara: A numerical framework for computing the limiting distribution of a stochastic fluid-fluid process Dario Andrea Bini, Stefano Massei and Leonardo Robol: Efficient cyclic reduction for QBDs with rank structured blocks: algorithm and analysis
12:30-14:00	Lunch
14:00-15:40	Session 3: Fluid queues (chair: Giang Nguyen) Aviva Samuelson, Malgorzata O'Reilly and Nigel Bean: Classification of Generators for Stochastic Fluid Models: New Riccati equation for Psi Eleonora Deiana, Guy Latouche and Marie-Ange Remiche: Fluid flows with jumps at the boundary Gábor Horváth: A novel approach for the efficient waiting time and queue length analysis of Markov-modulated fluid priority queues Sarah Dendievel and Guy Latouche: Perturbation analysis of Markov modulated fluid models
15:40-16:15	Break
16:15-17:30	Session 4: Asymptotics and stability (chair: Peter Taylor) Yuanyuan Liu, Pengfei Wang and Yiqiang Zhao: Exact Stationary Tail Asymptotics for a Markov Modulated Two-Demand Model --- In Terms of a Kernel Method Masakiyo Miyazawa: Martingale decomposition for large queue asymptotics Alexander Rumyantsev and Evsey Morozov: Stability criterion of a MAP/PH-multiserver model with simultaneous service
18:00-20:00	Boat trip on the Danube river

Wednesday, June 29, 2016

9:00-10:15	Invited talk Kousha Etessami: The computational complexity of analyzing infinite-state structured Markov chains and structured Markov decision processes Session 5: Branching processes (chair: Beatrice Meini) Peter Braunsteins and Sophie Hautphenne: An algorithmic approach to the extinction of branching processes with countably infinitely many types
10:15-10:50	Break
10:50-12:30	Session 6: Discrete queues (chair: Malgorzata O'Reilly) Li Xia, Qi-Ming He and Attahiru Alfa: Optimal Control of Service Rates in a MAP/M/1 Queue Rostislav Razumchik: Stationary distribution in MAP/PH/1 queue with bi-level hysteretic control of arrivals Nail Akar: Waiting Times in BMAP/BMSP/1 Queues Qi-Ming He: An MAP/PH/K Queue with Constant Impatient Time
12:30-14:00	Lunch
14:00-15:40	Session 7: Systems with multiple queues (chair: David Stanford) Malgorzata O'Reilly and Werner Scheinhardt: Analysis of tandem fluid queues Eleni Vatamidou, Ivo Adan, Maria Vlasiou and Bert Zwart: Asymptotic error bounds for truncated buffer approximations of a 2-node tandem queue Giuliano Casale, Gabor Horvath and Juan F. Perez: A Matrix-Analytic Approximation for Closed Queueing Networks with General FCFS Nodes Kevin Granville and Steve Drekic: On a 2-Class Polling Model with Class-dependent Reneging, Switchover Times, and Phase-type Service
15:40-16:15	Break
16:15-17:30	Session 8: Markov modulated Brownian motion (chair: Masakiyo Miyazawa) Nigel Bean, Giang Nguyen and Federico Poloni: Algorithms for Stationary Distributions of Fluid Queues: Interpretations and Re-interpretations Guy Latouche and Giang Nguyen: Markov-modulated Brownian motion and the flip-flop fluid queue: A symbiotic relationship Giang T. Nguyen and Federico Poloni: Componentwise accurate numerical methods for Markov-modulated Brownian motion
19:00-	Conference banquet

Thursday, June 30, 2016

9:00-10:15	Session 9: Risk, insurance and survival systems (chair: Srinivas Chakravarthy) Florin Avram, Andrei Badescu, Martijn Pistorius and Landy Rabehasaina: On a class of dependent Sparre Andersen risk models with application Mogens Bladt and Oscar Peralta: Parisian ruin for Markov additive risk processes with phase-type claims Salim Khamis Serhan, Bo Friis Nielsen and Mogens Bladt: Estimation of discretely observed Markov Jump Processes with applications in survival analysis
10:15-10:50	Break
10:50-12:30	Session 10: Distributions, processes and fitting (chair: Guy Latouche) Rosa Lillo, Joanna Rodriguez and Pepa Ramírez-Cobo: Dependence patterns related to the BMAP Bo Friis Nielsen and Azucena Campillo Navarro: On Order Statistics of Matrix--Exponential distributions Luz Judith Rodriguez Esparza and Fernando Baltazar-Larios: A Stochastic EM algorithm for construction of Mortality Tables Sophie Hautphenne and Katharine Turner: Fitting Markovian binary trees using global and individual population data
12:30-14:00	Lunch
14:00-15:15	Session 11: Markov models (chair: Dario Bini) Jevgenijs Ivanovs, Guy Latouche and Peter Taylor: Lattice and Non-Lattice Markov Additive Models Haibo Yu: Monotonicity, convexity and comparability of some functions associated with block-monotone Markov chains and their applications to queueing systems Stella Kapodistria and Zbigniew Palmowski: Matrix geometric approach for random walks in the quadrant
15:15-15:50	Break
15:50-16:15	Session 12: Population models (chair: Sophie Hautphenne) Matthieu Simon and Claude Lefèvre: SIR epidemics with stages of infection
16:15-16:40	Panel discussion

Opening address László Gerencsér: An interaction between queueing and change detection A basic feature of a single server queue is waiting time recorded at the time of arrival of a new customer, following a simple non-linear dynamics. By a remarkable coincidence the same dynamics is obtained for what is called the the Page-Hinkley detector designed for real-time change detection of stochastic processes. According to this an alarm is given if the detector (an equivalent to waiting time) exceeds a prefixed threshold. A result on the empirical tail-probabilities of the detector under very general technical conditions will be presented, assuming no change, thus providing an upper bound for the false alarm rate. These results translate to results on empirical large deviations for waiting times. The above mentioned technical condition can be verified in a variety of interesting special cases to be briefly presented. References: Gerencsér, L. and Prosdocimi, C. (2011): Input-output properties of the Page-Hinkley detector. Systems and Control Letters, 60 (7). pp. 486-491. Gerencsér, L., Prosdocimi, C. and Vágó, Zs. (2013): Change detection for finite dimensional Gaussian linear systems - a bound for the almost sure false alarm rate. In: Proc. of the 2013 European Control Conference (ECC), 2013-07-17 - 2013-07-19, Zürich, Switzerland. Short bio: László Gerencsér received his M.Sc. and Doctorate degree in mathematics at  Eötvös Loránd University, Budapest, Hungary in 1969 and 1970, respectively. Since 1970 he has been with MTA SZTAKI (Computer and Automation Research Institute, Hungarian Academy of Sciences) with intermissions abroad. His first research papers are written in graph theory, including a widely cited paper with A.Gyárfás on a generalized Ramsey problem. He started his academic carrier in operations research, from which he moved to control and systems theory, in particular to system identification. In 1986 he held a one year visiting position at the Department of Mathematics, Chalmers University of Technology, Göteborg, Sweden. A major milestone in his academic life was a visit to the Department of Electrical Engineering, McGill University, Montreal, Quebec, Canada, from 1988 to 1991, as a visiting professor. He was awarded a Széchenyi Professorship with  Eötvös Loránd University for the period of 1997 to 2001. His international activities include membership of the Conference Editorial Board of the IEEE Control Systems Society, 1998-2002, and his serving as an associate editor for the SIAM Journal of Control and Optimization, 1998-2004, and for the IEEE Transactions on Automatic Control, 2001-2004. He was the IPC Chair of the 19th International symposium on Mathematical Theory of Networks and Systems (MTNS 2010) Budapest, 2010. He is also the founder of the multidisciplinary Conferences on Applied Mathematics of the János Bolyai Mathematical Society.

Invited talk Kousha Etessami: The computational complexity of analyzing infinite-state structured Markov chains and structured Markov decision processes I will survey a series of results over recent years on the computational complexity of key analysis problems for several families of infinite-state structured Markov chains (MCs) and structured Markov decision processes (MDPs). A key aspect of our results is algorithmic bounds for computing the least fixed point (the least non-negative solution) for monotone systems of nonlinear (min/max)-polynomial equations which arise for these stochastic models and MDPs. In particular, I will describe algorithms based on Newton's method, combined with P-time preprocessing steps, which yield polynomial time algorithms (in the standard Turing model of computation) for computing, to arbitrary desired accuracy, the G matrix of quasi-birth death processes (QBDs), and the vector of extinction probabilities of multi-type branching processes (or Markovian trees). We then consider extensions of these purely stochastic models to MDPs. We describe a Generalized Newton's Method (GNM), which employs linear programming in each iteration, and we use GNM to obtain a P-time algorithm for computing, to desired accuracy, the vector of optimal (maximum or minimum) extinction probabilities for a given Branching MDP. We also study one-counter MDPs, which generalize discrete-time QBDs with control, and we give algorithms and complexity bounds for both qualitative and quantitative analysis of optimal termination probabilities for one-counter MDPs. Finally, we consider a model more general than the above stochastic models and MDPs, called recursive Markov chains (RMCs) and Recursive MDPs (RMDPs). RMCs are closely related to probabilistic pushdown systems and to tree-like-QBDs. We show that, in the worst-case, any non-trivial approximation of the termination probabilities of a RMC (or tree-like-QBD) is PosSLP-hard, and thus approximation in NP would already yield a breakthrough in exact numerical computation. For RMDPs (or tree-like-QBD-MDPs), we show that computing any non-trivial approximation of their optimal termination probability is not even computable at all (with any complexity). (This talk describes a series of results together with Alistair Stewart and Mihalis Yannakakis, as well as some older results with other co-authors.) Short bio: Kousha Etessami is an associate professor (reader) in the Laboratory for Foundations of Computer Science in the School of Informatics at the University of Edinburgh. In the past he was a member of the research staff at Bell Laboratories. His research interests are in theoretical computer science, including: algorithmic game theory, equilibrium computation, algorithmic analysis and verification of probabilistic systems, Markov decision processes, stochastic games, analysis of infinite-state systems, model checking, automata, logic, and computational complexity theory.