Saltar a la navegació principal Saltar a la cerca Vés al contingut principal

A novel weakly supervised problem: Learning from positive-unlabeled proportions

  • University of the Basque Country

Producció científica: Capítol del llibre/Acta del congrésActa de congrésAvaluat per experts

4 Cites (Scopus)

Resum

Standard supervised classification learns a classifier from a set of labeled examples. Alternatively, in the field of weakly supervised classification different frameworks have been presented where the training data cannot be certainly labeled. In this paper, the novel problem of learning from positive-unlabeled proportions is presented. The provided examples are unlabeled and the only class information available consists of the proportions of positive and unlabeled examples in different subsets of the training dataset. An expectation-maximization method that learns Bayesian network classifiers from this kind of data is proposed. A set of experiments has been designed with the objective of shedding light on the capability of learning from this kind of data throughout different scenarios of increasing complexity.

Idioma originalAnglès
Títol de la publicacióAdvances in Artificial Intelligence - 16th Conference of the Spanish Association for Artificial Intelligence, CAEPIA 2015, Proceedings
EditorsJosé M. Puerta, José A. Gámez, Bernabé Dorronsoro, Bruno Baruque, Alicia Troncoso, Edurne Barrenechea, Mikel Galar
EditorSpringer Verlag
Pàgines3-13
Nombre de pàgines11
ISBN (imprès)9783319245973
DOIs
Estat de la publicacióData de publicació - 2015
Publicat externament
Esdeveniment16th Conference of the Spanish Association for Artificial Intelligence, CAEPIA 2015 - Albacete, Espanya
Durada: 9 de nov. 201512 de nov. 2015

Sèrie de publicacions

NomLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volum9422
ISSN (imprès)0302-9743
ISSN (electrònic)1611-3349

Congrés

Congrés16th Conference of the Spanish Association for Artificial Intelligence, CAEPIA 2015
País/TerritoriEspanya
CiutatAlbacete
Període9/11/1512/11/15

Fingerprint

Navegar pels temes de recerca de 'A novel weakly supervised problem: Learning from positive-unlabeled proportions'. Junts formen un fingerprint únic.

Com citar-ho