Finite Gaussian mixture models provide a powerful and widely employed probabilistic approach for clustering multivariate continuous data. However, the practical usefulness of these models is jeopardized in high-dimensional spaces, where they tend to be over-parameterized. As a consequence, different solutions have been proposed, often relying on matrix decompositions or variable selection strategies. Recently, a methodological link between Gaussian graphical models and finite mixtures has been established, paving the way for penalized model-based clustering in the presence of large precision matrices. Notwithstanding, current methodologies implicitly assume similar levels of sparsity across the classes, not accounting for different degrees of association between the variables across groups. We overcome this limitation by deriving group-wise penalty factors, which automatically enforce under or over-connectivity in the estimated graphs. The approach is entirely data-driven and does not require additional hyper-parameter specification. Analyses on synthetic and real data showcase the validity of our proposal.

Group-Wise Shrinkage Estimation in Penalized Model-Based Clustering / A. Casa, A. Cappozzo, M. Fop. - In: JOURNAL OF CLASSIFICATION. - ISSN 0176-4268. - 39:3(2022), pp. 648-674. [10.1007/s00357-022-09421-z]

Group-Wise Shrinkage Estimation in Penalized Model-Based Clustering

A. Cappozzo
Secondo
;
2022

Abstract

Finite Gaussian mixture models provide a powerful and widely employed probabilistic approach for clustering multivariate continuous data. However, the practical usefulness of these models is jeopardized in high-dimensional spaces, where they tend to be over-parameterized. As a consequence, different solutions have been proposed, often relying on matrix decompositions or variable selection strategies. Recently, a methodological link between Gaussian graphical models and finite mixtures has been established, paving the way for penalized model-based clustering in the presence of large precision matrices. Notwithstanding, current methodologies implicitly assume similar levels of sparsity across the classes, not accounting for different degrees of association between the variables across groups. We overcome this limitation by deriving group-wise penalty factors, which automatically enforce under or over-connectivity in the estimated graphs. The approach is entirely data-driven and does not require additional hyper-parameter specification. Analyses on synthetic and real data showcase the validity of our proposal.
Model-based clustering; Penalized likelihood; Sparse precision matrices; Gaussian graphical models; Graphical lasso; EM algorithm
Settore SECS-S/01 - Statistica
2022
https://link.springer.com/article/10.1007/s00357-022-09421-z
Article (author)
File in questo prodotto:
File Dimensione Formato  
2105.07935.pdf

accesso aperto

Descrizione: Article
Tipologia: Pre-print (manoscritto inviato all'editore)
Dimensione 884.25 kB
Formato Adobe PDF
884.25 kB Adobe PDF Visualizza/Apri
s00357-022-09421-z.pdf

accesso aperto

Descrizione: Article
Tipologia: Publisher's version/PDF
Dimensione 2.05 MB
Formato Adobe PDF
2.05 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2434/1030199
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 2
social impact