Application of a genetic algorithm for plausible justification of observations

摘要

The idea of graphical expert systems was initiated by Pearl [5]) and worked-out by Lauritzen and Spiegelhalter [2]. Its generalization to different uncertainty formalisms was proposed by Shafer, Shenoy and Mellouli [6] and refined to the form of so-called Valuation-Based Systems by Shenoy [7]. Under probabilistic context the knowledge about interrelationships among variables from a fixed set X* is represented on the qualitative level by a directed graph G=(X*,E*) and on the quantitative level by a set of conditional probabilities p(x|Pa(x))) where Pa(x) stands for the set of parents of x in the graph G. Now, the joint probability distribution can be represented as the product of all the probabilities p(x|Pa(x)).