Building Probabilistic Models
We have seen in Chapter 3 that the knowledge base of a probabilistic expert system consists of a set of variables and a probabilistic model describing the relationships among them. We have also seen that all the information about the relationships among a set of variables is contained in the joint probability distribution (JPD) of the variables. Thus, the performance of a probabilistic expert system hinges on the correct specification of the JPD. Therefore, an important task for expert systems developers is to specify the JPD as accurately as possible. Human experts often collaborate to achieve this objective.
KeywordsDirected Acyclic Graph Undirected Graph Conditional Independence Human Expert Joint Probability Distribution
Unable to display preview. Download preview PDF.