[Forum SIS] Seminario Roberto Rocci, Università G. d'Annunzio di Chieti-Pescara
dibattis a unich.it
dibattis a unich.it
Mar 12 Giu 2018 11:51:31 CEST
********************************************************************************
AVVISO DI SEMINARIO
********************************************************************************
Giovedì 15 giugno alle ore 15:00, presso l'aula dottorandi
dell'Università G. d'Annunzio, Viale Pindaro, 42, Pescara
si terrà il seminario:
FINITE MIXTURES FOR SIMULTANEOUS CLUSTERING AND REDUCTION OF MATRIX
VALUED OBSERVATIONS
Prof. Roberto ROCCI
Dipartimento di Economia e Finanza
Università degli Studi di Roma Tor Vergata
Abstract:
Finite mixture models are often used to classify two-way (units by
variables) data.
However, two issues arise: model complexity and recovering the true
clustering structure.
Indeed, a huge number of variables and/or occasions implies a large
number of model parameters, while the existence of noise dimensions
could mask the true cluster structure.
The approach adopted in the present work is to reduce the number of
model parameters by identifying a sub-space containing the information
needed to classify the observations
nesting a PCA-like reparameterization into the classification model.
This also helps in identifying and discarding noise dimensions that
could coincide with some observed variables.
The aforementioned problems become more serious in the case of
three-way (units by variables by occasions) data where the
observations are matrices rather than vectors.
Our approach is extended to this case by nesting a three-way
PCA-like reparameterization (named Tucker2) into the classification
model.
This allows us to reduce the number of parameters identifying noise
dimensions for the variables and/or the occasions.
The effectiveness of the proposals is assessed through a simulation
study and an application to real data.
Il seminario è organizzato dal Dipartimento di Scienze Filosofiche,
Pedagogiche ed Economico-Quantitative, Università G. d'Annunzio di
Chieti-Pescara.
I colleghi interessati sono cordialmente invitati a partecipare.
Tonio Di Battista
Maggiori informazioni sulla lista
Sis