Patrick THIRAN   Résumé Teaching Research Collaborators Publications

Address Office Location: BC 208 Electronic Mail: Patrick.Thiran {at} epfl.ch Phone Numbers: +41-21-693 5601 +41-21-693 2623 (secr.) +41-21-693 6610 (fax) Postal Address:  LCA3-I&C  (BC) Station 14  EPFL  CH-1015 Lausanne  Switzerland

Résumé

I am an associate professor at the School of Computer and Communication Sciences of EPFL, Switzerland.

I received an electrical engineering degree from the Université Catholique de Louvain, Louvain-la-Neuve, Belgium, an M.S. degree in electrical engineering from the University of California at Berkeley, USA, and the PhD degree from the Swiss Federal Institute of Technology at Lausanne (EPFL), in 1996.

I did my PhD in the Circuits & Systems group (now Laboratory of Nonlinear Systems) of EPFL. My PhD advisor was Prof. Martin Hasler. In 1996, I joined the Communication Networks Laboratory of the same institution. This lab later became part of the Laboratory for Computer Communications and their Applications (LCA), where I became an adjunct professor (professeur titulaire) in 1998. I was with Sprint Advanced Technology Labs in Burlingame, California, in the IP group, in 2000-01. I became an assistant professor at EPFL in 2002 and an associate professor in 2006.

I served as an associate editor for the IEEE Transactions on Circuits and Systems from July 1997 until July 1999, and I currently serve as an associate editor for the IEEE/ACM Transactions on Networking since 2006. I also serve(d) in the Technical Programm Committee of various conferences in Networking including ACM Sigcomm, Sigmetrics, IMC, CoNext, Mobihoc and IEEE Infocom. I am a Fellow of the Belgian American Educational Foundation. I received the 1996 EPFL Doctoral Prize and the 2008 Crédit Suisse Teaching Award.

Teaching

2010-11:

• Information Sciences (First year Communication Systems and Computer Science undergraduate program).
• Stochastic models for communication systems (Third year Communication Systems undergraduate program)
• "Networks out of Control" : Models and Methods for Large Scale Random Networks (Communication Systems graduate program, with Prof. Matthias Grossglauser).
• TCP/IP Networking(Communication Systems graduate program, with Prof. Jean-Yves Le Boudec).

Research

Self-organized wireless networks

• How do the properties of multi-hop wireless ad-hoc and sensor networks scale with their size ? Our current effort is devoted to the one of the most basic properties of such a network: connectivity. Most research results address the full connectivity of a network over a finite area, with a number of nodes tending to infinity. In this work instead, we let the area tend to infinity, but we keep the node density finite. We also included the effect of interferences, using percolation theory - more information on NCCR MICS IP1 . This research is carried out in the framework of the National Center of Competence in Mobile Information and Communication Systems (NCCR MICS).
• One of the most challenging and important problems in multi-hop ad hoc networks is to know how many bits per second and meter they can transport, i.e. their capacity. Since the pioneering work of Gupta and Kumar, some upper and lower bounds are available, under various strong assumptions. Among these is a centralized scheduling scheme, which is optimized to yield the highest possible capacity. In practice, this scheduling needs to be distributed among the nodes. We are investigating algorithms that implement such a distributed Medium Access Control, and yet achieve a high transport capacity. More information on NCCR MICS IP1 . This research is carried out in the framework of the National Center of Competence in Mobile Information and Communication Systems (NCCR MICS).

Network vulnerability and monitoring

• Fault location in optical networks is hampered by a multitude of factors: the redundancy and the lack of coordination (internetworking) of the managements at the different layers (WDM, SDH/SONET, ATM, IP); the large number of alarms a single failure can trigger; the difficulty in detecting some failures and the resulting need to cope with missing or false alarms. We have implemented centralized algorithms for locating multiple failures at the physical layer of a WDM network, within the COBNET project. In the project DICS-1830 of the Hasler Foundation, we extend these algorithms for more general networks. In particular, we use elements of Boolean algebra to optimally configure probing schemes in network tomography.
  
• We have recently devised a novel survivable mapping algorithm of logical (IP) topologies on top of physical (WDM) topologies. This algorithm is, to date, the fastest one, but ignores physical link capacity constraints. This work is carried out in the project DICS-1830 of the Hasler Foundation. We have also shown, in collaboration with Sprintlabs, that a differentiated protection in IP/WDM networks, which offers applications the choice between receiving a very high availability insurance and receiving more overall throughput, and  combines a simple scheduling mechanism in IP routers, with a reduced amount of fiber redundancy at the physical layer, can yield a substantial increase in bandwidth usage.
• Traffic statistical modelling is driven by real traffic traces collected in the Sprint backbone network, and ranges from the local queuing features of a core router to the analysis of end-to-end flows transiting in the backbone.
  

• Network calculus is a collection of results based on Min-Plus algebra, which applies to deterministic queuing systems found in communication networks. It can be used for example to understand the computations for delays used in the IETF guaranteed service, why re-shaping delays can be ignored in shapers, how to optimally smooth multimedia streams, and much more. The Network Calculus Book is available as Springer Verlag LNCS 2050, or online on our Network Calculus web server thanks to a generous arrangement from Springer.

• ABE (Alternative Best Effort) is a service for IP networks, which offers applications the choice between receiving a lower end-to-end delay and receiving more overall throughput within a best-effort service class (unlike differentiated services proposals). Every best effort packet is marked as either green or blue. Green packets receive a low, bounded queueing delay. To ensure blue packets do not suffer as a result, green flows receive less throughput during bouts of congestion. 

• Feedback is also used in low-power transmission schemes in digital VLSI. As systems-on-chips evolve towards networks interconnecting basic building blocks (CPU, memory,...) taken from a library of predesigned components, communication protocols for data exchanges between these building blocks need to be devised, taking constraints from VLSI in account. We are investigating the use of feedback to design adaptive, self-calibrating low-power transmission schemes for a bus interconnecting VLSI subsystems in a joint project with LAP-EPFL (Processor Architectures Laboratory).
  

Locally Coupled Neural Networks

• Locally Coupled Neural Networks are arrays of nonlinear dynamical systems that are only connected to their nearest neighbors. Their high dimensionality makes their analysis challenging, their regularity makes it more tractable than in other more complex systems. Two models of particular interest are the Kohonen Self-Organizing Map, where I studied effects due to hardware imperfections in collaboration with the Mantra team, and the Cellular Neural Network, whose interest resides in an analog implementation for applications requiring a fast preprocessing of data. More information in the book Dynamics and Self-Organization of Locally Coupled Neural Networks.

Collaborators

Currently at EPFL

Florence Benezit, PhD student, graduated from Ecole Polytechnique in France. She is co-supervised by Prof. Martin Vetterli and myself

Mathilde Durvy, PhD student, graduated in Communication Systems from EPFL. Thesis topic:  Spatial reuse and MAC protocols in multi-hop wireless networks.

Maciej Kurant, PhD student, graduated in Telecommunication from Gdansk University of Technology in Poland, and in Communication Systems from EPFL pre-doc school.

Hung Nguyen, PhD student, graduated in Electrical Engineering from University of Adelaide in Australia, and in Communication Systems from EPFL pre-doc school. Thesis topic: failure location and network tomography.

Frédéric Worm, PhD student, graduated in Communication Systems from EPFL, and joined LAP-EPFL (Processor Architectures Laboratory). He is co-supervised by Prof. Paolo Ienne and myself. Thesis topic: Low-power transmission schemes for on-chip networks.

Graduated PhD from EPFL

Dr Olivier Dousse, graduated in Physics from EPFL. He obtained his PhD degree in 2005, for which he was awarded the 2005 Honorable Mention for the ACM Doctoral Dissertation Award. Thesis topic: Asymptotic properties of wireless multi-hop networks. He is now a senior resercher at Deutsche Telekom Laboratories, in Berlin.

Dr Carmen Mas was my first PhD student. She graduated in Telecommunication Engineering from Polytechnic University of Catalonia, Spain. She obtained her PhD degree in 2000. Topic: Fault location algorithms for Optical Networks. She then joined the Development Programmes Department of Intracom, and she is now an assistant professor with the Athens Information Technology Center of Excellence for Graduate Research and Education AIT, Greece.

Previous postdocs at EPFL

Dr Chadi Barakat came as a Postdoc in 2001-02. He holds his PhD degree in Networking and Distributed Systems from the University of Nice Sophia-Antipolis. He is now a permanent researcher in the PLANETE group at INRIA Sophia Antipolis, France.

Dr Can Emre Koksal, postdoc, holds his PhD degree in EECS from MIT, USA. He came as a post-doc in 2004-06, and is currently an assistant professor at Ohio State University.

Dr Petteri Mannersalo came as a Postdoc, under an ERCIM/SARIT Fellowship, in 2003-04. He holds his PhD degree in EECS from the Helsinki University of Technology, Espoo. He is a Research Scientist at VTT Technical Research Centre of Finland Information Technology Telecommunications.

Previous visitors at EPFL

Prof Massimo Franceschetti, was a visiting professor during summer 2005. He is an assistant professor at UC San Diego, USA.

Dr Claudio Serpico was a visiting PhD student in the Circuits & Systems group in 1995, working with me on Cellular Neural Networks. He got his PhD from University Federico II of Napoli in 1998, and is now a permanent researcher (Ricercatore) at the same university in Italy.

Prof Gianluca Setti was a visiting PhD student in the Circuits & Systems group in 1994-95, working with me on Cellular Neural Networks. He got his PhD from University of Bologna in 1997, and is now an associate professor at University of Ferrara, Italy.

Christina Tavoularis, was a visiting PhD student in 2004-05, from Cornell University, USA.

Publications

Recent Publications on line from 1998

Journal papers until 1998

[J2] T.Roska, T. Boros, A.Radvanyi, P. Thiran and L.O. Chua, "Detecting moving and standing objects using cellular neural networks", International Journal of Circuit Theory and Applications, vol. 20, pp.613-628, September 1992.

[J3] P. Thiran, "Influence of the Boundary Conditions on Behaviour of Cellular Neural Networks", IEEE Transactions on Circuits and Systems, Part I, Vol. Cas-40 Part I, pp.207-212, March 1993.

[J4] P. Thiran, V. Peiris, P. Heim and B. Hochet, "Quantization Effects in Digitally Behaving Circuit Implementations of Kohonen Networks", IEEE Transactions on Neural Networks, vol. NN-5, pp. 450-458, May 1994.

[J5]P. Thiran, M. Hasler, "Self-Organization of a One-Dimensional Kohonen Network With Quantized Weights and Inputs", Neural Networks, vol. 7, 1994, pp. 1427-1439.

[J6] P. Thiran and M. Hasler, "Study of Kohonen network with a discrete state space", Mathematics and Computers in Simulation (MATCOM), vol. 38, pp. 189-197, 1995.

[J7] P. Thiran and M. Hasler, "Information storage using stable and unstable oscillations: an overview", International Journal of Circuits Theory and Applications, Vol. 24, 57-67, 1996.

[J8] P. Thiran and G. Setti, "Information storage and retrieval in an associative memory based on one-dimensional maps" , International Journal of Electronics, vol. 79, pp. 815-821, 1995.

[J9] P. Thiran, K. R. Crounse, L. O. Chua and M. Hasler, "Pattern formation Properties of Autonomous Cellular Neural Networks", IEEE Transactions on Circuits and Systems, Part I, vol. CAS-42 (I), vol CAS 42, pp. 757-774, October 1995.

[J10] T. Cornu, P. Ienne, D. Niebur, P. Thiran and M. Viredaz, "Design, Implementation, and Test of a Multi-Model Systolic Neural Network Accelerator", Scientific Programming, Vol. 5, pp. 47-61, 1996.

[J11] K. R. Crounse, L. O. Chua, P. Thiran and G. Setti, "Characterization and Dynamics in Cellular Neural Networks", International Journal of Bifurcation and Chaos, vol 6, pp. 1703-1724, September 1996.

[J12] P. Ienne, P. Thiran and N. Vassilas, "Modified Self-Organizing feature Map algorithms for efficient hardware implementation", IEEE Transactions on Neural Networks, vol. NN-8, pp. 315-330, March 1997.

[J13] P. Thiran, G. Setti and M. Hasler, "An Approach to Information Propagation in Cellular Neural Networks, Part I: Local Diffusion", IEEE Transactions on Circuits and Systems, Part I, vol CAS-45(8), pp. 777--789, August 1998.

[J14] G. Setti, P. Thiran and C. Serpico "An Approach to Information Propagation in Cellular Neural Networks, Part II: Global Propagation", IEEE Transactions on Circuits and Systems, Part I, vol CAS-45(8), pp. 790--811, August 1998.

Last modified: October 16, 2006.