A Characterization of Bayesian Network Structures and its Application to Leaming
Proceedings of the Sixth International Workshop on Artificial Intelligence and Statistics, PMLR R1:203-210, 1997.
We present an analysis of the minimal I-map relation between Bayesian network structures and dependency models. This includes a partial order characterisation of the structures, and the connection between the relation and the arc reversal operation. Two applications of this analysis are presented. The first is a simple condition for identifying equivalence between Bayesian network structures, and the second is an exact learning algorithm based on the partial order characterisation.