Intrinsic Dimension Estimation by Maximum Likelihood in Isotropic Probabilistic PCA - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Journal Articles Pattern Recognition Letters Year : 2011

Intrinsic Dimension Estimation by Maximum Likelihood in Isotropic Probabilistic PCA

Abstract

A central issue in dimension reduction is choosing a sensible number of dimensions to be retained. This work demonstrates the surprising result of the asymptotic consistency of the maximum likelihood criterion for determining the intrinsic dimension of a dataset in an isotropic version of Probabilistic Principal Component Analysis (PPCA). Numerical experiments on simulated and real datasets show that the maximum likelihood criterion can actually be used in practice and outperforms existing intrinsic dimension selection criteria in various situations. This paper exhibits and outlines the limits of the maximum likelihood criterion. It leads to recommend the use of the AIC criterion in specific situations. A useful application of this work would be the automatic selection of intrinsic dimensions in mixtures of isotropic PPCA for classification.
Fichier principal
Vignette du fichier
revision_PPCAds.pdf (191.51 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-00440372 , version 1 (10-12-2009)
hal-00440372 , version 2 (25-01-2010)
hal-00440372 , version 3 (11-07-2011)

Identifiers

Cite

Charles Bouveyron, Gilles Celeux, Stéphane Girard. Intrinsic Dimension Estimation by Maximum Likelihood in Isotropic Probabilistic PCA. Pattern Recognition Letters, 2011, 32 (14), pp.1706-1713. ⟨10.1016/j.patrec.2011.07.017⟩. ⟨hal-00440372v3⟩
2972 View
1093 Download

Altmetric

Share

Gmail Facebook X LinkedIn More