HUGS: Combining Exact Inference and Gibbs Sampling in Junction Trees

Dawid, Kjrerulff & Lauritzen (1994) provided a preliminary description of a hybrid between Monte-Carlo sampling methods and exact lo­ cal computations in junction trees. Utiliz­ ing the strengths of both methods, such hy­ brid inference methods has the potential of expanding the class of problems which can be solved under bounded resources as well as solving problems which otherwise resist ex­ act solutions. The paper provides a detailed description of a particular instance of such a hybrid scheme; namely, combination of ex­ act inference and Gibbs sampling in discrete Bayesian networks. We argue that this com­ bination calls for an extension of the usual message passing scheme of ordinary junction trees.