Skip to Main content Skip to Navigation
Conference papers

Choix de modèles quand la vraisemblance est incalculable

Christine Keribin 1, 2, *
* Corresponding author
2 SELECT - Model selection in statistical learning
LMO - Laboratoire de Mathématiques d'Orsay, Inria Saclay - Ile de France
Abstract : Penalised likelihood criteria such as BIC are popular methods for model selection and require to compute the maximised likelihood. Unfortunately, this maximised likelihood can be untractable, as it is the case for the latent block model (LBM). LBM is a mixture model for co-clustering, allowing to perform the simultaneous clustering of rows and columns of large data matrices. Due to the complex dependence between the row and column class membership variables conditionally to the observations, approximations are needed to perform the estimation step of the EM algorithm, leading to a lower bound of the maximised likelihood. For the same reason, the usual asymptotic approximation used to derive BIC is itself questionable. On the other hand, the integrated completed likelihood criterion (ICL) is exactly computed for LBM, but requires to investigate the influence of hyperparameters. Links between both criteria are analyzed and comparison with Bayesian inference is discussed.
Document type :
Conference papers
Complete list of metadata

https://hal.inria.fr/hal-01260761
Contributor : Christine Keribin <>
Submitted on : Friday, January 22, 2016 - 3:37:14 PM
Last modification on : Tuesday, May 4, 2021 - 3:37:10 AM

Identifiers

  • HAL Id : hal-01260761, version 1

Citation

Christine Keribin. Choix de modèles quand la vraisemblance est incalculable. 47èmes Journées de Statistique de la SFdS, Jun 2015, Lille, France. ⟨hal-01260761⟩

Share

Metrics

Record views

376

Files downloads

60