[Statlist] Research Seminar in Statistics | *FRIDAY 23 MAY 2025* | GSEM, University of Geneva

gsem-support-instituts gsem-support-instituts at unige.ch
Mon May 19 14:18:49 CEST 2025


Dear all,

We are pleased to invite you to our next Research Seminar, organized by Professor Davide La Vecchia on behalf of the Research Institute for Statistics and Information Science
< https://www.unige.ch/gsem/en/research/institutes/risis/team/ >.

FRIDAY 23 MAY 2025, at 11:15 am, Uni Mail M 4220.

On the Asymptotics of Importance Weighted Variational Inference
Dr. Badr-Eddine CHÉRIEF-ABDELLATIF, CNRS, Paris, France
https://badreddinecheriefabdellatif.github.io/


ABSTRACT:
For complex latent variable models, the likelihood function is not available in closed form. In this context, a popular method to perform parameter estimation is Importance Weighted Variational Inference. It essentially maximizes the expectation of the logarithm of an importance sampling estimate of the likelihood with respect to both the latent variable model parameters and the importance distribution parameters, the expectation being itself with respect to the importance samples. Despite its great empirical success in machine learning, a theoretical analysis of the limit properties of the resulting estimates is still lacking. In this talk, we will study the consistency of IWVI when both the Monte Carlo and the observed data sample sizes go to infinity simultaneously. We also discuss normality and efficiency under additional conditions relating the rate of growth between the Monte Carlo and the observed data samples sizes. We distinguish several regimes related to the smoothness of the importance ratio. 

> View the Research Seminar agenda: < https://www.unige.ch/gsem/en/research/seminars/risis/ >

Regards,

Birgit

Birgit Müller-Marreros 
Assistant to the Research Institutes
file:///\\nasac-faculty.isis.unige.ch\GSEM_HOME_PAPAT\mullermb\11_PRIVE\2_TO-DO-liste_Bilaterales\00_Passage_autre_poste\Mes%20exemples%20dès%20le%2022.4.25\RISIS\Mail.unige.ch







More information about the Statlist mailing list