Hybrid Propagation in Junction Trees

We introduce a methodology for performing approximate computations in complex probabilistic expert systems, when some components can be handled exactly and others require approximation or simulation. This is illustrated by means of a modified version of the familiar `chest-clinic' problem. Keywords: probabilistic expert system; graphical model; local computation; Monte Carlo. 1 Introduction Markov Chain Monte-Carlo (MCMC) methods have over the last decade become increasingly popular as a computational tool in complex stochastic systems of various type (Gelfand & Smith 1990, Thomas, Spiegelhalter & Gilks 1992, Gelman & Rubin 1992, Geyer 1992, Smith & Roberts 1993). The methods are flexible, easy to implement, and computing time tends to scale manageably with the size of the problem under consideration. Their main disadvantages are associated with their reliance on pseudorandom number generators (Ripley 1987), some difficulty in deciding whether the necessary equilibrium has been reach..