In the class of mixtures with random covariates, the generalized linear Gaussian cluster-weighted model (GLGCWM) has been recently proposed; in each mixture component, it models the response variable within the exponential family of distributions and the vector of real-valued covariates according to the multivariate Gaussian distribution. Due to the number of free parameters of each covariance matrix of the component Gaussian distributions, a family of fourteen parsimonious GLGCWMs is here introduced by applying some constraints on the eigen decomposition of these matrices. This novel family of models is also applied to a real data set where it gives good classification performance, especially when compared with more established mixture-based approaches.
Parsimony in Mixtures with Random Covariates
INGRASSIA, Salvatore;PUNZO, ANTONIO
2013-01-01
Abstract
In the class of mixtures with random covariates, the generalized linear Gaussian cluster-weighted model (GLGCWM) has been recently proposed; in each mixture component, it models the response variable within the exponential family of distributions and the vector of real-valued covariates according to the multivariate Gaussian distribution. Due to the number of free parameters of each covariance matrix of the component Gaussian distributions, a family of fourteen parsimonious GLGCWMs is here introduced by applying some constraints on the eigen decomposition of these matrices. This novel family of models is also applied to a real data set where it gives good classification performance, especially when compared with more established mixture-based approaches.File | Dimensione | Formato | |
---|---|---|---|
Ingrassia & Punzo - CLADAG 2013.pdf
solo gestori archivio
Licenza:
Non specificato
Dimensione
1.5 MB
Formato
Adobe PDF
|
1.5 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.