Bayesian Networks and Influence Diagrams
Finn Verner Jensen · 1997
Bayesian networks are graphical probabilistic models consisting of variables and cause-e ect relations between them. First Bayesian networks are de ned and a couple of examples are given. It is illustrated how Bayesian networks are used for various task like calculating updated probabilities for variables given evidens, calculating probabilities for speci c con gurations of variables, calculating con gurations of maximal probability and analysis for con icting evidens. In uence diagrams are Bayesian networks argumented with special variables for actions. They are used for calculating optimal strategies for sequences of actions. It is nally shown how the problem of deciding between various information sources can be solved in situations with only one set of action options. These and other issues are treated more deeply in Jensen (1996). 1