Using Four Cost Measures to Determine Arc Reversal Orderings
Four cost measures s1, s2, s3, s4 were recently studied for sorting the operations in Lazy propagation with arc reversal (LPAR), a join tree propagation approach to Bayesian network inference. It has been suggested to use s1 with LPAR, since there is an effectiveness ranking, say s1, s2, s3, s4, when applied in isolation. In this paper, we also suggest to use s1 with LPAR, but to use s2 to break s1 ties, s3 to break s2 ties, and s4 to break s3 ties. Experimental results show that sometimes there is a noticeable gain to be made.
KeywordsBayesian networks arc reversal join tree propagation
Unable to display preview. Download preview PDF.
- 1.Butz, C.J., Chen, J., Konkel, K., Lingras, P.: A Formal Comparison of Variable Elimination and Arc Reversal in Bayesian Network Inference. Intell. Dec. Analysis 3(3), 173–180 (2009)Google Scholar
- 2.Butz, C.J., Konkel, K., Lingras, P.: Join Tree Propagation Utilizing both Arc Reversal and Variable Elimination. Intl. J. Approx. Rea. (2010) (in press)Google Scholar
- 7.Hansen, P.F., Pedersen, P.T.: Risk Analysis of Conventional and Solo Watch Keeping. Research Report, Department of Naval Architecture and Offshore Engineering. Technical University of Denmark (1998)Google Scholar
- 12.Madsen, A.L.: An Empirical Evaluation of Possible Variations of Lazy Propagation. In: 20th Conference in Uncertainty in Artificial Intelligence, pp. 366–373. AUAI Press, Arlington (2004)Google Scholar
- 17.Olmsted, S.: On Representing and Solving Decision Problems. Ph.D. Thesis, Department of Engineering Economic Systems. Stanford University, Stanford, California (1983)Google Scholar