Sparse probabilistic classifiers - Archive ouverte HAL Access content directly
Conference Papers Year : 2007

Sparse probabilistic classifiers

Abstract

The scores returned by support vector machines are often used as a confidence measures in the classification of new examples. However, there is no theoretical argument sustaining this practice. Thus, when classification uncertainty has to be assessed, it is safer to resort to classifiers estimating conditional probabilities of class labels. Here, we focus on the ambiguity in the vicinity of the boundary decision. We propose an adaptation of maximum likelihood estimation, instantiated on logistic regression. The model outputs proper conditional probabilities into a user-defined interval and is less precise elsewhere. The model is also sparse, in the sense that few examples contribute to the solution. The computational effciency is thus improved compared to logistic regression. Furthermore, preliminary experiments show improvements over standard logistic regression and performances similar to support vector machines.
Fichier principal
Vignette du fichier
icml.pdf (216.64 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-00442746 , version 1 (22-12-2009)

Identifiers

Cite

Romain Hérault, Yves Grandvalet. Sparse probabilistic classifiers. ICML '07: Proceedings of the 24th international conference on Machine learning, Jun 2007, New York, NY, USA, United States. pp.337--344, ⟨10.1145/1273496.1273539⟩. ⟨hal-00442746⟩
51 View
72 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More