Mixture models - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Book Sections Year : 2017

Mixture models

Abstract

Finite mixture models are one of the probabilistic frameworks which reach an especially diverse community of people, including statisticians and practitioners (scientific or not). Initial reasons for being confronted with mixtures may be different for impacted communities but lead finally to close interconnections between them. Indeed, applied statisticians and practitioners usually discover finite mixture models from the numerous application fields where they meet numerous successes. It typically gathers {none,un,semi-} supervised classification and density estimation. The keys of these successes are both their high meaningfulness and flexibility. However, flexibility is in return a matter of algorithmic and mathematical questionings for methodological and theoretical statisticians. In particular, it addresses estimation and model selection issues, on both computational and mathematical aspects. But, solutions to be provided to these issues highly beneficiate to depend on initial related application fields.
Fichier principal
Vignette du fichier
JES2014-chap1.pdf (3.22 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-01252671 , version 1 (18-01-2016)

Identifiers

  • HAL Id : hal-01252671 , version 1

Cite

Christophe Biernacki. Mixture models. J-J. Droesbeke; G. Saporta; C. Thomas-Agnan. Choix de modèles et agrégation, Technip, 2017, 9782710811770. ⟨hal-01252671⟩
223 View
184 Download

Share

Gmail Facebook Twitter LinkedIn More