Local and global probabilistic-logic inference in algebraic bayesian networks: a matrix-vector description and issues of sensitivity

Algorithms and equations for the probabilistic-logic inference of algebraic Bayesian networks are presented in the paper. All types of global consistency are considered and a matrix-vector formalization of consistency conditions is proposed. The paper summarizes results in local posterior inference for different kinds of knowledge patterns. Moreover in this paper we conduct a sensitivity analysis of first problem of a posterior inference for the knowledge pattern built over the ideal of disjuncts and formulate a linear programming problem to find the described estimates.