Learning, Adaptation, and Tuning

In Chapter 2, we mainly considered modeling situations where a domain expert establishes the structure of the network. With respect to estimating the potentials for the model (the conditional probabilities), it can be based on combinations of theoretical considerations, a database of cases, and pure subjective estimates. You may be so fortunate that you have a large database of cases or you expect to collect cases in the future. In that case, you would like to exploit the information for model building or for future change.