A generic framework for a compilation-based inference in probabilistic and possibilistic networks

Probabilistic and possibilistic networks are important tools proposed for an efficient representation and analysis of uncertain information. The inference process has been studied in depth in these graphical models. We cite in particular compilation-based inference which has recently triggered the attention of several researchers. In this paper, we are interested in comparing this inference mechanism in the probabilistic and possibilistic frameworks in order to unveil common points and differences between these two settings. In fact, we will propose a generic framework supporting both Bayesian networks and possibilistic networks (product-based and min-based ones). The proposed comparative study points out that the inference process depends on the specificity of each framework, namely in the interpretation of the handled uncertainty degrees (probability⧹possibility) and appropriate operators (∗⧹min and +⧹max).