@inproceedings{fd8b713bfa9740b7a878b8584262db8a,
title = "A novel weakly supervised problem: Learning from positive-unlabeled proportions",
abstract = "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.",
keywords = "Bayesian network models, Label proportions, Positive-unlabeled learning, Structural EM method, Weakly supervised classification",
author = "Jer{\'o}nimo Hern{\'a}ndez-Gonz{\'a}lez and I{\~n}aki Inza and Lozano, \{Jose A.\}",
note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2015.; 16th Conference of the Spanish Association for Artificial Intelligence, CAEPIA 2015 ; Conference date: 09-11-2015 Through 12-11-2015",
year = "2015",
doi = "10.1007/978-3-319-24598-0\_1",
language = "English",
isbn = "9783319245973",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "3--13",
editor = "Puerta, \{Jos{\'e} M.\} and G{\'a}mez, \{Jos{\'e} A.\} and Bernab{\'e} Dorronsoro and Bruno Baruque and Alicia Troncoso and Edurne Barrenechea and Mikel Galar",
booktitle = "Advances in Artificial Intelligence - 16th Conference of the Spanish Association for Artificial Intelligence, CAEPIA 2015, Proceedings",
address = "Germany",
}