CONSTRUCTING EFFICIENT BELIEF NETWORK STRUCTURES WITH EXPERT PROVIDEDINFORMATION

We present a technique to construct efficient belief network structure s for application areas where large amounts of data are available and information on the ordering of the variables can be obtained from doma in experts. We identify classes of networks that are efficient for pro pagating beliefs. We formulate the problem as one of determining the b elief network representation from a given class that best represents t he data. We use the I-Divergence measure which is known to have certai n desirable properties for evaluating different approximations. We pre sent some theoretical findings that characterize the nature of solutio ns that are obtained. These theoretical results lead to an efficient s olution procedure for finding the best network representation. We also discuss other information that may be reasonably obtained from expert s, and show how such information leads to improving the efficiency of the technique to find the best network structure.