Conférence de Vincent Danos organisée par le département d'informatique.
We present a method to compute approximate descriptions of a class of stochastic systems. For the method to apply, the system must be presented as a Markov chain on a state space consisting in graphs or graph-like objects, and jumps must be described by transformations which follow a finite set of local rules. The method is a form of static analysis and uses a technique which is reminiscent of theories of critical pairs in term rewriting systems. Its output is a system of coupled ordinary differential equations (ODE) which tracks the mean evolution of the number of (typically small) subgraphs. In some cases, these ODEs form an exact and finite description of these mean numbers. But even when the ODE description is only an approximation, it can often reveal interesting properties of the original system.
Vincent Danos est directeur de recherches CNRS dans l'équipe preuves, programmes, systèmes à l' Université Paris-Diderot.Cliquer ICI pour fermer
Dernière mise à jour : 16/12/2014