Laboratoire de Mathématiques Raphaël Salem

UMR 6085 CNRS - Université de Rouen Normandie

In multivariate kernel density estimation, the bandwidth selection remains a challenge in terms of algorithmic performance and quality of the resulting estimation. A recently developped method, the Penalized Comparison to Overfitting (PCO), is compared to other usual bandwidth selection methods for multivariate and univariate kernel density estimation. In particular, the cross-validation and plug-in estimators are numerically investigated and compared to PCO.

We consider two models of Markov processes observed (continuous time) in white Gaussian noise. One is telegraph signal with two states (ergodic case) and the second is partially observed linear system with small noise in observation equation. For both models we construct the MLE and One-step MLEs of the unknown parameters and show the consistency, asymptotic normality, convergence of moments and the asymptotic minimax efficiency of these estimators.

This work deals with multi state systems that we model by means of semi-Markov processes. The sojourn times are seen to be independent not identically distributed random variables and assumed to belong to a general class of distributions that is closed under maxima. We obtain maximum likelihood estimators of the parameters of interest and investigate their asymptotic properties. Plug-in type estimators are furnished for various reliability quantities related to the system under study.

I will talk about recent progress related to nonsingular Bernouli shifts. Properties of conservativeness, ergodicity, Krieger's type, K-property will be under discussion. "General shifts", i.e. those which are not the natural extensions of nonsingular endomorphisms, will be also considered

We consider the problem of estimating function in a periodic regression in continuous time with the Levy noise by discrete time observations. We use the model selection approach and develop a new adaptive procedure, which involves special modifications of the well-known James-Stein estimates, to improve the accuracy of the basic weighted least square estimates.

Nous considérons un modèle stochastique décrivant une population structurée par la taille. L’approche étudiée est motivée par la modélisation des divisions cellulaires et par la détection du vieillissement cellulaire en biologie. La population est représentée par une mesure ponctuelle évoluant suivant un processus aléatoire déterministe par morceaux. Nous étudions ici l’estimation non-paramétrique du noyau régissant les divisions, dans le cas où l’arbre de division n’est pas complètement observé.

Sahar Albosaily est la doctorante de Serge Pergamenchtchikov.

Cet exposé rentre dans le cadre du séminaire de vulgarisation de la Fédération Normandie-Mathématiques.

Pour en savoir plus, consulter la page du séminaire