[Forum SIS] DEI - Università di Catania: AVVISO DI SEMINARI

[antonio punzo]

********************************************************************************

AVVISO DI SEMINARI
********************************************************************************

Il Dipartimento di Economia e Impresa dell’Università di Catania ha
organizzato i seminari:



*Parsimonious multiple scaled mixtures*

*Brian Franczak*

*Department of Mathematics & Statistics, *

*McMaster University Hamilton, Ontario, Canada*







Lunedì 13 Luglio dalle ore 11:00

In Aula 5



Università di Catania

Palazzo delle Scienze, Corso Italia, 55, Catania





*Abstract*:



The Mixtures of multiple scaled distributions have garnered increased
attention in the last few years. Generalizations of the multivariate-t and
shifted asymmetric Laplace distributions have been introduced, and both are
shown to exhibit untraditional physical features that can be beneficial
when performing cluster analysis. One issue with the aforementioned
mixtures is that their covariance structures become highly parameterized as
the dimension of the data increases. Since the multiple scaled
distributions are formulated using an eigen-decomposed covariance matrix, a
natural way to introduce parsimony is by constraining the constituent
elements of this decomposition. This leads to two families of mixture
models, that we call parsimonious multiple scaled mixtures (PMSMs).

Interestingly, the PMSMs can be derived using two different stochastic
relationships. As such, in addition to introducing the PMSMs, we compare
these two stochastic relationships and discuss their advantages when
deriving the parameter estimates for our families of mixture models. We
demonstrate the PMSMs abilities using well-known real data sets.





I colleghi interessati sono cordialmente invitati a partecipare.











*Partially Supervised Biclustering of Gene Expression Data with
Applications in Nutrigenomics Biomarker Discovery*

*Monica Wong*

*Department of Mathematics & Statistics, *

*McMaster University Hamilton, Ontario, Canada*







Lunedì 13 Luglio dalle ore 12:00

In Aula 5



Università di Catania

Palazzo delle Scienze, Corso Italia, 55, Catania





*Abstract*:



The estimation of a family of parsimonious Gaussian mixture models for the
biclustering of high-dimensional gene expression data is introduced.
Previously, a family of parsimonious Gaussian mixture models was introduced
which is based on the mixtures of factor analyzers model. Our family
extends these models into the biclustering framework by including a binary
and row-stochastic factor loadings matrix. This particular form of factor
loadings matrix results in a block-diagonal covariance matrix, which is a
useful property in specific gene expression applications in nutrigenomics
biomarker discovery. Knowledge of the factor loadings matrix is useful in
this application and is reflected in the partially supervised nature of the
algorithm. Parameter estimates are obtained through a variant of the
expectation-maximization algorithm and the best fitting model is selected
using the Bayesian information criterion. We demonstrate our family of
models using both simulated and real data.





I colleghi interessati sono cordialmente invitati a partecipare.





Antonio Punzo




-- 
Dr. Antonio Punzo
Research Fellow in Statistics - University of Catania.
Corso Italia 55, 95129 Catania - Italy.
-------------- parte successiva --------------
Un allegato HTML è stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20150706/8925d1de/attachment-0001.html>