[Forum SIS] Avviso di seminario: Bartolucci a DISMEQ (Dipartimento di Statististica e Metodi Quantitativi, Milano-Bicocca)
Fulvia Pennoni
fulvia.pennoni a unimib.it
Lun 7 Nov 2016 12:09:54 CET
-----------------------------------------------------------------------------
Dipartimento di Statistica e Metodi Quantitativi
Via Bicocca degli Arcimboldi, 8 - 20126 Milano
-----------------------------------------------------------------------------
“STOCHASTIC BLOCK MODELS: INFERENTIAL DEVELOPMENTS IN THE CONTEXT OF
STATIC AND DYNAMIC SOCIAL NETWORKS”
Francesco Bartolucci
University of Perugia (INSA)
Lunedì 14 Novembre ore 11.00 Ed. U7, 2° piano, aula 2062 (sezione di
demografia)
__________________________________________________________
Abstract:
Stochastic block models (SBMs) are commonly used in social network
analysis to discover communities and clusters of individuals having a
similar behavior. Despite the simplicity of the assumptions of these
models, inference remains problematic because the model likelihood
cannot be computed and this limits their applicability. The talk is
focused on two known methods of inference for SBMs and compares these
methods with a new proposal that shows a great potential. The known
methods rely on composite likelihood and variational inference. The
proposed method is based on a classification likelihood, but has
features in common with the previous two. The extension of SBMs for the
analysis of longitudinal social network data is also illustrated,
including some features regarding reciprocity. An application based on
the Enron email dataset is used as an illustration, together with a
series of simulations.
__________________________________________________________
Si allega la locandina.
Cordiali saluti
Fulvia Pennoni
--
Fulvia Pennoni
Department of Statistics and Quantitative Methods
University of Milano-Bicocca
http://www.statistica.unimib.it/utenti/pennoni/
-------------- parte successiva --------------
Un allegato non testuale è stato rimosso....
Nome: 2016-11-14 Locandina Bartolucci.pdf
Tipo: application/pdf
Dimensione: 90724 bytes
Descrizione: non disponibile
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20161107/52c6c01e/attachment-0001.pdf>
Maggiori informazioni sulla lista
Sis