Expert systems and probabilistic network models.

*(English)*Zbl 0867.68099
Texts and Monographs in Computer Science. Berlin: Springer. xiv, 605 p. (1996).

The book is an extensive treatment of Bayesian probabilistic methods in the construction of expert systems. In addition to a synthesis of a very extended modern literature (the bibliography extends over 16 pages), the reader will find techniques that allow the programmer to control the consistency of the rules introduced from expert advice. This is a point that is usually glossed over in implementation of rule-based systems but it is easily followed in probabilistic systems. Several chapters are devoted to integrating graph based and probabilistic methods and to the propagation of evidence (the latter being what expert systems are all about). At every step, the reader will find pseudo-code programs to implement the techniques treated. Measures of quality of Bayesian networks are also treated.

Reviewer: H.Guggenheimer (West Hempstead)