Belief Updating in Bayesian Networks

Abstract

In this chapter, we present algorithms for probability updating. An efficient updating algorithm is fundamental to the applicability of Bayesian networks. As shown in Chapter 2, access to P($\mathcal{U}$, e) is sufficient for the calculations. However, because the joint probability table increases exponentially with the number of variables, we look for more efficient methods. Unfortunately, no method guarantees a tractable calculation task. However, the method presented here represents a substantial improvement, and it is among the most efficient methods known.