PREVAL : Probability, modelling and evaluation of information processing systems

About the team : Activity report Research reports Theses Theme : Bio Biological systems Paris - Rocquencourt research center

PREVAL has been dissolved on 01/01/2007

Team Presentation

PREVAL is the continuation of the MEVAL team (1982-2001). Its core activity is to achieve a good understanding of the behavior of random systems stemming not only from computer science and telecommunications, but also from physics and biology. The approach proposed by the team is essentially the development and resolution of mathematical models, and in addition, simulation. The competence acquired by the team more than covers the spectrum of stochastic modeling. In this respect, it is worth recalling how beneficial the recourse to probability theory is in numerous branches of modern science (physics, economics, biology).
The themes include at the same time methodological aspects and special models. The scientific approach always consist in proposing general methods, based on concrete problems, that can be used to obtain quantitative results concerning the efficiency, stability and control of such and such a structure.

Research themes

Research is eclectic and touches on varied application fields: computer networks, transportation networks, statistical physics, graphs and random structures. The macroscopic temporal and spatial analysis of these various objects inevitably leads to the study of stochastic processes that are often physically relevant (stay duration in a system, client or data distribution, stationary regime).

Original theories have been developed within the team during the last twenty years in the following areas of expertise.

• Networks and random walks: resolution of functional equations in several complex variables, classification of Markov chains in polyhedra with boundaries using equivalent dynamical systems, methods to construct semi-martingales to determine the stability conditions.
• Large systems: when the size or volume of a system grows, chaos propagation or phase transition phenomena occur, as in classical physics. This is what is called the thermodynamic limit. Applications are numerous (telecommunications, transportation, biology).
• Grammars, graphs and random complexes: we build new theoretical bridges between physics and computer science (random chains, enumeration problems, quantum gravity).
• Large deviations: the point is to estimate the probability of rare events, that can however be crucial (hurricanes, diverging nuclear reactor, congested Internet).
• A recent theme concerns bioinformatics: chemical networks, metabolism modeling, time evolution of large molecules.

International and industrial relations

PREVAL maintains continuing relations with the universities of Berkeley, Braunschweig, Cambridge, Columbia, Moscow, Novosibirsk, Monterey, San Diego, as well as with CWI and Eurandom.
Several contractual initiatives were carried out with the DRET (large systems) and France Télécom (broadband network study). PREVAL is currently a partner of the national IMARA (La Route Automatisée) program.

Scientific leader

Guy FAYOLLE
+33 1 39 63 53 96
guy.fayolle@inria.fr