Sensitivity Analysis in Markov Networks

This paper explores the topic of sensitivity analysis in Markov networks, by tackling questions similar to those arising in the context of Bayesian networks: the tuning of parameters to satisfy query constraints, and the bounding of query changes when perturbing network parameters. Even though the distribution induced by a Markov network corresponds to ratios of multi-linear functions, whereas the distribution induced by a Bayesian network corresponds to multi-linear functions, the results we obtain for Markov networks are as effective computationally as those obtained for Bayesian networks. This similarity is due to the fact that conditional probabilities have the same functional form in both Bayesian and Markov networks, which turns out to be the more influential factor. The major difference we found, however, is in how changes in parameter values should be quantified, as such parameters are interpreted differently in Bayesian networks and Markov networks.