Les chaînes de Markov à mémoire de longueur variable constituent une classe de sources probabilistes. Il sera question dans cet exposé d'existence et unicité de mesure invariante pour une collection d'exemples de chaînes. Nous nous intéresserons également au comportement asymptotique d'une marche aléatoire dont les longueurs de sauts ne sont pas forcément intégrables. Les lois de sauts dépendent partiellement du passé de la trajectoire. Plus précisément, la probabilité de monter ou de descendre dépend du temps passé dans la direction dans laquelle le marcheur est en train d'avancer. Un critère de récurrence/transience s'exprimant en fonction des paramètres du modèle sera énoncé. Suivront plusieurs exemples illustrant le caractère instable du type de la marche lorsqu'on perturbe légèrement les paramètres.
Les travaux décrits dans cet exposé ont été faits en collaboration avec B. Chauvin, F. Paccaut et N. Pouyanne ou B. de Loynes, A. Le Ny et Y. Offret.[-]

We consider a simple stochastic model for the spread of a disease caused by two virus strains in a closed homogeneously mixing population of size N. In our model, the spread of each strain is described by the stochastic logistic SIS epidemic process in the absence of the other strain, and we assume that there is perfect cross-immunity between the two virus strains, that is, individuals infected by one strain are temporarily immune to re-infections and infections by the other strain. For the case where one strain has a strictly larger basic reproductive ratio than the other, and the stronger strain on its own is supercritical (that is, its basic reproductive ratio is larger than 1), we derive precise asymptotic results for the distribution of the time when the weaker strain disappears from the population, that is, its extinction time. We further consider what happens when the difference between the two reproductive ratios may tend to 0.
This is joint work with Fabio Lopes.[-]

The continuous time Bayesian networks (CTBNs) represent a class of stochastic processes, which can be used to model complex phenomena, for instance, they can describe interactions occurring in living processes, in social science models or in medicine. The literature on this topic is usually focused on the case, when the dependence structure of a system is known and we are to determine conditional transition intensities (parameters of the network). In the paper, we study the structure learning problem, which is a more challenging task and the existing research on this topic is limited. The approach, which we propose, is based on a penalized likelihood method. We prove that our algorithm, under mild regularity conditions, recognizes the dependence structure of the graph with high probability. We also investigate the properties of the procedure in numerical studies to demonstrate its effectiveness .[-]

In this presentation, we shall discuss the reconstruction of demographic parameters based on the genetic variability observed within a sample of individual DNA. In the family of models that we consider, the statistics describing this genetic diversity (number of mutations, distribution of the mutations amongst individuals in the sample) depend on a more or less coarse ‘resolution of (i.e., level of information on) the hidden genealogical tree that relates the sampled individuals. Considering the optimal resolution thus allows to greatly improve the exploration of the space of possible genealogies when computing the likelihood of demographic parameters, compared to classical methods based on full labelled trees such as Kingmans coalescent. We shall focus on two examples, based on works with Raazesh Sainudiin (Uppsala Univ.) and with Julia Palacios (Stanford Univ.), Sohini Ramachandran (Brown Univ.) and John Wakeley (Harvard Univ.).[-]

We consider stochastic models of scalable biological reaction networks in the form of continuous time pure jump Markov processes. The study of the mean field behavior of such Markov processes is a classical topic, with fundamental results going back to Kurtz, Athreya, Ney, Pemantle, etc. However, there are still questions that are not completely settled even in the case of linear reaction rates. We study two such questions. First is to characterize all possible rescaled limits for linear reaction networks. We show that there are three possibilities: a deterministic limit point, a random limit point and a random limit torus. Second is to study the mean field behavior upon the depletion of one of the materials. This is a joint work with Lai-Sang Young.[-]