avviso di seminario Universita' di Pavia

["Guido Consonni" <guido.consonni@unipv.it>]

UNIVERSITA' DEGLI STUDI DI PAVIA
DIPARTIMENTO DI ECONOMIA POLITICA E METODI QUANTITATIVI
DIPARTIMENTO DI MATEMATICA "F.CASORATI"
DOTTORATO DI RICERCA IN MATEMATICA E STATISTICA
***********************************************************************


                         AVVISO DI SEMINARIO


Giovedi' 11 luglio, alle ore 14.00, nell'aula del Consiglio della Facolta'
di Economia, via San Felice 7, Pavia


                            JEAN MICHEL MARIN
                  Université Paris IX Dauphine
Centre de Recherche en Mathématique de la Décision



terra' un seminario dal titolo:

Population Monte Carlo for Ion Channel Restoration.



Abstract

Ion channels are protein molecules that are intimately involved in the
transmission of information through the nervous system.
Their behaviour cannot be observed directly. The deficiencies of the
observation process make the study of ion channel data difficult.
Several methods have been proposed for making inference about
the parameters of the underlying process.

In the bayesian paradigm, the ion channel model is a special case
of discrete time hidden semi-Makov model for wich there exists no
explicit polynomial time formula for the posterior distribution of the
hidden
process, as opposed to the hidden Markov model with the forward-
backward formula of Baum and Petrie. From a computationnal (MCMC)
point of view, there is therefore no way of eliminating this hidden
process to simulate directely the parameter conditional on the observables,
as was done in the hidden Markov model.

We study through the ion channel model the behaviour of an iterated
importance sampling scheme, called population Monte Carlo.

We compare this approach with the equivalent MCMC algorithm and we
show the superior robustness of population Monte Carlo to a possibly
poor choice of the proposal distribution. Moreover, we study through this
model the degeneracy phenomenon associated with repeated calls to the
iterated importance sampling scheme.


************************************************************
Guido Consonni
Dipartimento di Economia Politica e Metodi Quantitativi
Via S. Felice 7
27100 Pavia
ITALY

Tel: +39 0382 506 225
Fax: +39 0382 304226

email: guido.consonni@unipv.it

http://economia.unipv.it/gconsonni