zbMATH — the first resource for mathematics

Handling interventions with uncertain consequences in belief causal networks. (English) Zbl 1252.68298
Greco, Salvatore (ed.) et al., Advances in computational intelligence. 14th international conference on information processing and management of uncertainty in knowledge-based systems, IPMU 2012, Catania, Italy, July 9–13, 2012. Proceedings, Part III. Berlin: Springer (ISBN 978-3-642-31717-0/pbk; 978-3-642-31718-7/ebook). Communications in Computer and Information Science 299, 585-595 (2012).
Summary: Interventions are tools used to distinguish between mere correlations and causal relationships. These standard interventions are assumed to have certain consequences, i.e. they succeed to put their target into one specific state. In this paper, we propose to handle interventions with uncertain consequences. The uncertainty is formalized with the belief function theory which is known to be a general framework allowing the representation of several kinds of imperfect data. Graphically, we investigate the use of belief function causal networks to model the results of passively observed events and also the results of interventions with uncertain consequences. To compute the effect of these interventions, altered structures namely, belief mutilated graphs and belief augmented graphs with uncertain effects are used.
For the entire collection see [Zbl 1251.68017].

68T37 Reasoning under uncertainty in the context of artificial intelligence
PDF BibTeX Cite
Full Text: DOI