Abstract
We report the use genetic algorithms (GAs) as a search mechanism for the discovery of linear causal models when using two Bayesian metrics for linear causal models, a Minimum Message Length (MML) metric [10] and a full posterior analysis (BGe) [3]. We also consider two structure priors over causal models, one giving all variable orderings for models with the same arc density equal prior probability (P1) and one assigning all causal structures with the same arc density equal priors (P2). Evaluated with Kullback-Leibler distance prior P2 tended to produce models closer to the true model than P1 for both metrics, with MML performing slightly better than BGe. By contrast, when using an evaluation metric that better reflects the nature of the causal discovery task, namely a metric that compares the results of predictive performance on the effect nodes in the discovered model P1 outperformed P2 in general, with MML and BGe discovering models of similar predictive performance at various sample sizes. This supports our conjecture that the P1 prior is more appropriate for causal discovery.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
G.F. Cooper and E. Herskovits. A Bayesian method for the induction of probabilistic networks from data. Machine Learning, 9:309–347, 1992.
D. Heckerman. A Bayesian approach to learning causal networks. Proc of the 11th Conf on Uncertainty in AI, pp. 285–295, 1995. Morgan Kaufmann.
D. Heckerman and D. Geiger. Learning Bayesian networks: A unification for discrete and Gaussian domains. Proc of the 11th Conf on Uncertainty in AI, pp. 274–284, San Francisco, California, 1995. Morgan Kaufmann.
D. Heckerman, D. Geiger, and D.M. Chickering. Learning Bayesian networks. Machine Learning, 20:197–243, 1995.
J.R. Neil and K.B. Korb. The MML evolution of causal models. Technical Report 98/17, School of Computer Science, Monash University, 1998.
J.R. Neil and K.B. Korb. The evolution of causal models: A comparison of bayesian metrics and structure priors. Technical Report 1999/27, School of Computer Science, Monash University, 1999.
J. Pearl. Probabilistic Reasoning in Intelligent Systems. Morgan Kaufmann, 1988.
C.S. Wallace and P.R. Freeman. Estimation and inference by compact coding. Jrn Royal Stat Soc (Series B), 49:240–252, 1987.
C.S. Wallace and K.B. Korb. Learning linear causal models by MML sampling. Thompson (ed.) Causal Models and Intelligent Data Management, Springer, 1999.
C.S. Wallace, K.B. Korb, and H. Dai. Causal discovery via MML. L. Saitta, editor, Proc of the 13th Int Conf on Machine Learning, pp. 516–524, 1996.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1999 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Neiland, 1.R., Korb, K.B. (1999). The Evolution of Causal Models: A Comparison of Bayesian Metrics and Structure Priors. In: Zhong, N., Zhou, L. (eds) Methodologies for Knowledge Discovery and Data Mining. PAKDD 1999. Lecture Notes in Computer Science(), vol 1574. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-48912-6_57
Download citation
DOI: https://doi.org/10.1007/3-540-48912-6_57
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-65866-5
Online ISBN: 978-3-540-48912-2
eBook Packages: Springer Book Archive