Robustness Analysis of Bayesian Networks with Global Neighborhoods

This paper presents algorithms for robustness analysis of Bayesian networks with global neighborhoods. Robust Bayesian inference is the calculation of bounds on posterior values given perturbations in a probabilistic model. We present algorithms for robust inference (including expected utility, expected value and variance bounds) with global perturbations that can be modeled by -contaminated, constant density ratio, constant density bounded and total variation classes of distributions. c 1996 Carnegie Mellon University This research is supported in part by NASA under Grant NAGW-1175. Fabio Cozman was supported under a scholarship from CNPq, Brazil.