BLOCKING GIBBS SAMPLING IN VERY LARGE PROBABILISTIC EXPERT-SYSTEMS

We introduce a methodology for performing approximate computations in very complex probabilistic systems (e.g. huge pedigrees). Our approach , called blocking Gibbs, combines exact local computations with Gibbs sampling in a way that complements the strengths of both. The methodol ogy is illustrated on a real-world problem involving a heavily inbred pedigreee containing 20000 individuals. We present results showing tha t blocking-Gibbs sampling converges much faster than plain Gibbs sampl ing for very complex problems.