LISP implementation of a Bayesian DAG Classifier
Constructs a DAG whose nodes are interrelated belief probabilities. As new information is acquired, algorithm works by calling 5 recursively interrelated formulas and 3 update rules to propogate messages that update parent and child probabilities.
For a full description of the theory and algorithm, see Neopolitan, Probabilistic reasoning in expert systems: theory and algorithms, pp 160-162, 217-231.