In this paper, we study the problem of information preservation when decomposing a single Bayesian network into a set of smaller Bayesian networks. We present a method that losslessly decomposes a Bayesian network so that no conditional independency information is lost and no extraneous conditional independency information is introduced during the decomposition.
Subjects: 3.4 Probabilistic Reasoning; 9. Foundational Issues
Submitted: Feb 11, 2007