Indirect Elicitation of NIN-AND Trees in Causal Model Acquisition
To specify a Bayes net, a conditional probability table, often of an effect conditioned on its n causes, needs to be assessed for each node. Its complexity is generally exponential in n and hence how to scale up is important to knowledge engineering. The non-impeding noisy-AND (NIN-AND) tree causal model reduces the complexity to linear while explicitly expressing both reinforcing and undermining interactions among causes. The key challenge to acquisition of such a model from an expert is the elicitation of the NIN-AND tree topology. In this work, we propose and empirically evaluate two methods that indirectly acquire the tree topology through a small subset of elicited multi-causal probabilities. We demonstrate the effectiveness of the methods in both human-based experiments and simulation-based studies.
KeywordsTrue Model Tree Topology Causal Model Causal Interaction Conditional Probability Table
Unable to display preview. Download preview PDF.
- [Galan and Diez(2000)]Galan, S., Diez, F.: Modeling dynamic causal interaction with Bayesian networks: temporal noisy gates. In: Proc. 2nd Inter. Workshop on Causal Networks, pp. 1–5 (2000)Google Scholar
- [Heckerman and Breese(1996)]
- [Kahneman et al(1982)Kahneman, Slovic, and Tversky]Kahneman, D., Slovic, P., Tversky, A. (eds.): Judgment under uncertainty: heuristics and biases. Cambridge University Press, Cambridge (1982)Google Scholar
- [Lemmer and Gossink(2004)]
- [Maaskant and Druzdzel(2008)]Maaskant, P., Druzdzel, M.: An independence of causal interactions model for opposing influences. In: Jaeger, M., Nielsen, T. (eds.) Proc. 4th European Workshop on Probabilistic Graphical Models, Hirtshals, Denmark, pp. 185–192 (2008)Google Scholar
- [Xiang(2010a)]Xiang, Y.: Acquisition and computation issues with NIN-AND tree models. In: Myllymaki, P., Roos, T., Jaakkola, T. (eds.) Proc. 5th European Workshop on Probabilistic Graphical Models, Finland, pp. 281–289 (2010a)Google Scholar
- [Xiang(2010b)]Xiang, Y.: Generalized non-impeding noisy-AND trees. In: Proc. 23th Inter. Florida Artificial Intelligence Research Society Conf., pp. 555–560 (2010b)Google Scholar
- [Xiang and Jia(2007)]
- [Xiang et al(2009a)]
- [Xiang et al(2009b)]
- [Zagorecki and Druzdzel(2004)]Zagorecki, A., Druzdzel, M.: An empirical study of probability elicitation under Noisy-OR assumption. In: Proc. 17th Inter. Florida Artificial Intelligence Research Society Conf., pp. 880–885 (2004)Google Scholar