A Randomized Approximation Algorithm of Logic Sampling

In recent years, researchers in decision analysis and artifi­ cial intelligence (AI) have used Bayesian belief networks to build models of expert opinion. Using standard methods drawn from the theory of computational complexity, work­ ers in the field have shown that the problem of exact prob­ abilistic inference on belief networks almost certainly requires exponential computation in the worst case [3]. We have previously described a randomized approximation scheme, called BN-RAS, for computation on belief net­ works [1, 2, 4]. We gave precise analytic bounds on the convergence of BN-RAS and showed how to trade running time for accuracy in the evaluation of posterior marginal probabilities. We now extend our previous results and dem­ onstrate the generality of our framework by applying sim­ ilar mathematical techniques to the analysis of convergence for logic sampling [7], an alternative simulation algorithm for probabilistic inference.