Incorporating Evidence in Bayesian Networks with the Select Operator

In this paper, we propose that the select operator in relational databases be adopted for incorporating evidence in Bayesian networks. This approach does not involve the construction of new evidence potentials, nor the associated computational costs of multiplying the evidence potentials into the knowledge base. The select operator also provides unified treatment of hard and soft evidence in Bayesian networks. Finally, some query optimization rules, involving the select operator implemented in relational databases, can be directly incorporated into probabilistic expert systems.