A Direct Propagation Method in Singly Connected Causal Belief Networks with Conditional Distributions for all Causes

摘要

Existing algorithms of propagation in belief networks deal with inference of observations when conditional distributions are initially defined per edge. The aim of this paper is to propose a direct method of causal inference of both observations and interventions on the causal belief networks quantified with the belief function theory where conditional beliefs are defined for all parents without having to transform the network into a junction tree. We explain how it is still possible to use the disjunctive rule of combination DRC and the generalized Bayesian theorem GBT to perform this propagation.