A Bayesian network is a network of evidence and deriveed properties/facts linked by conditional probabilities that can be used to work out the probabilities of various facts given evidence. The same network can typically be used to make inferences in different dircetions, for example working out the probability of seeing someone with an umbrella if it is rainy, or the probability that it is rainy if you spot someone with an umbrella.
Used on page 49
Example Bayesian network