Expert Systems: Uncertainty and Learning

Booknews

A general introduction to expert systems for readers with a previous knowledge of probability, statistics, and programming in Pascal or Lisp. Particularly addresses issues of uncertainty and learning. Points out the deficiencies of uncertainty propagation methods, describing the most common methods and noting where errors are most likely to be made. Discusses the lack of learning ability in most expert systems, and suggests solutions, drawing from log-linear, regression, discriminant, or causal network models. Annotation c. Book News, Inc., Portland, OR (booknews.com)