ULBIBM International Chair

in Computer Science

Modeling And Optimisation





Febuary-June 1999
Department of Computer Science

Inaugural Lesson: Febuary 9



organised with the support of


FNRS

IBM









Index

Lessons

Numerical methods for markov chains

Queueing Networks and Mobile Communications

Partial Evaluation and Automatic Program Optimisation

Models for telecommunications

Location

Leçon inaugurale: Telecommunication Modeling

Registration





Lessons

Numerical methods for Markov chains

16-23 Febuary, 2 March 1999

Solvay Auditorium, 10-12 a.m., 2-4 p.m.

Dr. Beatrice Meini

Department of Mathematics, University of Pisa



One of the main problems related to the numerical solution of Markov chains is the computation of the probability invariant vector. The aim of these lectures is to introduce both the fundamental numerical methods and the more advanced techniques for finite as well as infinite Markov chains with M/G/1 structure, which are of particular interest for telecommunication modeling. In the finite case we recall the known direct methods based on matrix factorizations and the iterative methods based on regular splittings. In the infinite case we have more specific computational issues, like the solution of a nonlinear matrix equation. We analyze the classical functional iteration methods and present new strategies for improving the convergence rate, based on Toeplitz matrix computations, on doubling techniques and on cyclic reduction.

Beatrice Meini received her degree cum laude in Mathematics in 1993, and her PhD degree in 1998 at the University of Pisa, Italy. Then, she obtained a senior fellowship from INDAM (the Institute of High Mathematics), and a post doctoral fellowship from CNR (the National Research Council). Now she has a permanent position, as researcher, in the Department of Mathematics of the University of Pisa. Her main research activity is in the field of the design and analysis of numerical algorithms for the solution of problems of large dimension and with a strong structure; particular interest is devoted to computational issues related to the solution of structured Markov chains.

Queueing Networks and Mobile Communications

9-16-23-30 March 1999

Solvay Auditorium, 10-12 a.m., 2-4 p.m.

Prof. Dr. Rudolf Mathar

Aachen University of Technology



Markov chains in discrete and continuous time are a main modeling tool. In particular, solving the balance equations for transition probabilities and intensities allows to describe the corresponding systems in equilibrium; for reversible chains the balance equations can be easily solved, and in Markovian tandem networks all intermediate arrival and departure processes are recognized as Poissonian. Open and closed Jackson Networks are another general model. By applying reversibility and truncation, the product form of the steady state distribution is preserved even in very general cases. The lectures will conclude with some interesting applications of queueing networks in mobile communication modeling, optimal register location strategies and optimal channel allocation in cellular radio networks.



Rudolf Mathar received his Dipl.-Math. and his Dr. Rer. Nat. degree in mathematics from the Aachen University of Technology, Germany, in 1978 and 1981, respectively. From 1986 to 1988, he worked as a lecturer in computer science at the European Business School and in an optimization research group at the University of Augsburg. He is currently a Professor of Stochastics at the Aachen University of Technology. He is a cofounder of the Center of Network Optimization (CNO), and a cofounder and manager of Telecommunications Network Consulting ltd. His research interests include applications to mobile communication systems through performance analysis and optimization of stochastic networks.

Partial Evaluation and Automatic Program Optimisation

20-27 April, 4-11 May 1999

Solvay Auditorium, 10-12 a.m., 2-4 p.m.

Dr. Michael Leuschel

Department of Electronics and Computer Science,

University of Southampton




Because of their simplicity and purity, declarative (e.g., functional and logic) programming languages are particularly amenable to reasoning and automatic transformation techniques. The idea often is to automatically transform an elegant (but inefficient) program into an efficient (but inelegant) program. Partial evaluation is an automatic transformation technique that has received much attention. In this course we shall also study a number of more aggressive automatic transformation techniques, yielding both greater speedups and more difficulty in controlling the transformation process. These latter difficulties pose many interesting problems which we will study in the course.

Michael Leuschel received a degree in Computer Science from the Université Libre de Bruxelles in 1990 and a Master of Artificial Intelligence from the Katholieke Universiteit Leuven in 1993, where he also received his Ph.D in 1997. Then he obtained a post-doctoral fellowship of the Belgian Fund for Scientific Research. He is currently a lecturer at the Department of Electronics and Computer Science of the University of Southampton. His research focuses on program analysis, transformation, and specialisation of declarative languages (mainly logic and functional programming). He is also interested in metaprogramming, termination, deductive databases, and formal methods.



Models for telecommunications

8-25 May, 1 June 1999

Solvay Auditorium, 10-12 a.m., 2-4 p.m.

Dr. Peter Taylor

Teletraffic Research Centre, University of Adelaide



A loss network consists of a set of resources accessed by customers, in such a way that if the set of needed resources is not available, the customer is blocked and lost from the system. For instance, in a packet-switched broadband network such as an ATM network, one may regard each connection as requiring an amount of bandwidth from the links that it traverses. If the required bandwidth is available on all links of a route between a connection's origin and destination then that connection is accepted onto the network, if not the connection is lost. Loss networks can also be used to model cellular phone systems, digital packet switched integrated services networks and database access problems. In these talks we shall discuss some properties of loss networks, giving particular emphasis to performance analysis, management of these networks and paradoxical behaviour that can occur.



Peter Taylor received a B.Sc. in 1980 and a Ph.D. In applied mathematics in 1987 at the University of Adelaide, where he is now an associate professor in the Department of Applied Mathematics. He also worked three years for the Australian Public Service in Canberra. His research interests are in stochastic processes and applied probability, with particular emphasis on teletraffic and biological applications. He is currently a Co-Director of the Teletraffic Research Centre at the University of Adelaide and has recently been appointed an associate editor of Queueing Systems.



Location