***************************************************************************
UNIVERSITA' DEGLI STUDI DI ROMA "LA SAPIENZA"
DIPARTIMENTO DI STUDI GEOECONOMICI, LINGUISTICI, STATISTICI,
STORICI PER L'ANALISI REGIONALE
***************************************************************************
AVVISO DI SEMINARIO
Il giorno 28 ottobre p.v., alle ore 16.30, nell'auletta del Dipartimento
(Ala amministrazione - quarto piano della Facolta' di Economia,
via del Castro Laurenziano 9
il Prof. CHRISTIAN P. ROBERT
della Universite Paris IX "Dauphine"
terrą un seminario dal titolo
"POPULATION MONTE CARLO".
Tutti gli interessati sono invitati a partecipare.
Brunero Liseo
======================================================================
Abstract
While MCMC methods have come to dominate the Bayesian computational area,
there are still difficulties in devising adaptive methods that are
satisfactory both from a theoretical and a methodological point of view.
We reanalyze in this talk the Population Monte Carlo method that
encompasses much more adaptive and local importance sampling schemes than
thought previously. By introducing a temporal dimension to the selection
of the importance function, an adaptive perspective can be achieved at
little cost, with the same theoretical justification as the standard
importance sampling technology. We have shown that the PMC scheme is a
viable alternative to MCMC schemes in missing data settings, among others
for the stochastic volatility model. We will present in addition recent
results on the convergence properties of the most standard PMC schemes.
(This is joint work with O. Cappe, R. Douc, A. Guillin, J.M. Marin, and E.
Moulines.)
Link: http://www.ceremade.dauphine.fr/CMD/preprints02/0215ihp2.ps.gz