From Dirichlet Process mixture models to spectral clustering
Stefano Tonellato ()
Additional contact information
Stefano Tonellato: Department of Economics, University Of Venice CÃ Foscari
No 2017:33, Working Papers from Department of Economics, University of Venice "Ca' Foscari"
Abstract:
This paper proposes a clustering method based on the sequential estimation of the random partition induced by the Dirichlet process. Our approach relies on the Sequential Importance Resampling (SIR) algorithm and on the estimation of the posterior probabilities that each pair of observations are generated by the same mixture component. Such estimates do not require the identification of mixture components, and therefore are not affected by label switching. Then, a similarity matrix can be easily built, allowing for the construction of a weighted undirected graph, where nodes represent individuals and edge weights quantify the similarity between pairs of individuals. The paper shows how, in such a context, spectral clustering techniques can be applied in order to identify homogeneous groups.
Keywords: Dirichlet process priors; sampling importance resampling; weighted graph; laplacian; spectral clustering (search for similar items in EconPapers)
JEL-codes: C11 C38 (search for similar items in EconPapers)
Pages: 18 pages
Date: 2017
References: View references in EconPapers View complete reference list from CitEc
Citations:
Downloads: (external link)
https://www.unive.it/web/fileadmin/user_upload/dip ... _tonellato_33_17.pdf First version, anno (application/pdf)
Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.
Export reference: BibTeX
RIS (EndNote, ProCite, RefMan)
HTML/Text
Persistent link: https://EconPapers.repec.org/RePEc:ven:wpaper:2017:33
Access Statistics for this paper
More papers in Working Papers from Department of Economics, University of Venice "Ca' Foscari" Contact information at EDIRC.
Bibliographic data for series maintained by Sassano Sonia ().