Inference in hybrid causal belief networks

摘要

Causal belief networks are compact representations of uncertain causal knowledge under the belief function framework. On these networks, we can handle and compute the effect of observations and interventions using the propagation process. This paper explores propagation algorithm for hybrid causal belief networks (i.e., some conditional distributions are defined per edge and some others are specified for more than one cause) allowing the experts to express their beliefs in a flexible way.