Propagation describes the act of computing posterior probability distributions and expected utilities for unobserved discrete chance nodes and decision nodes, and distribution functions for continuous chance nodes given the values for the observed nodes (i.e., the evidence).
Four different propagation algorithms are available: Sum normal, max normal, sum fast retraction, and max fast retraction. Each one of these can be activated by selecting the appropriate Propagate item of the Network Menu. In addition, two of the most common propagation algorithms (sum normal and max normal) can be accessed by pressing one of the propagation buttons of the Run Mode Tool Bar (see Figure 1). If auto propagation has been specified, propagation will take place each time new evidence is entered.