Generalizations of the noisy-or model

0447357 - ÚTIA 2016 RIV CZ eng J - Journal Article
Vomlel, Jiří
Generalizations of the noisy-or model.
Kybernetika. Roč. 51, č. 3 (2015), s. 508-524. ISSN 0023-5954
R&D Projects: GA ČR GA13-20012S
Institutional support: RVO:67985556
Keywords : Bayesian networks * noisy-or model * classification * generalized linear models
Subject RIV: JD - Computer Applications, Robotics
Impact factor: 0.628, year: 2015
http://library.utia.cas.cz/separaty/2015/MTR/vomlel-0447357.pdf

We generalize the noisy-or model. The generalizations are three-fold. First, we allow parents to be multivalued ordinal variables. Second, parents can have both positive and negative influences on their common child. Third, we describe how the suggested generalization can be extended to multivalued child variables. The major advantage of our generalizations is that they require only one parameter per parent. We suggest a model learning method and report results of experiments on the Reuters text classification data. The generalized noisy-or models achieve equal or better performance than the standard noisy-or. An important property of the noisy-or model and of its generalizations suggested in this paper is that it allows more efficient exact inference than logistic regression models do.
Permanent Link: http://hdl.handle.net/11104/0249419