

Home > Research > Bayesian Networks

The Bayesian Network (BN) formalism is one of the dominant representations for modeling uncertainty in intelligent systems. A BN is a probabilistic graphical model of a joint probability distribution over a set of statistical variables. Bayesian inference on a BN answers probabilistic queries about the variables and their influence relationships. We are investigating improvements of stochastic simulation algorithms for BN inference.

Bayesian Network Inference

H. Yu and R. van Engelen, Refractor Importance Sampling, in the proceedings of the Conference on Uncertainty in Artificial Intelligence (UAI), 2008.
This paper introduces Refractor Importance Sampling (RIS), an improvement to reduce error variance in Bayesian network importance sampling propagation under evidential reasoning. RIS approaches the optimal importance function by applying localized arc changes to minimize the divergence between the evidenceadjusted importance function and the optimal importance function.
R. van Engelen, Approximating Bayesian Belief Networks by Arc Removal, in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 19, no. 8, August 1997, pages 916920.
This paper proposes a general framework for approximating Bayesian belief networks based on model simplification by arc removal. The approximation method aims at reducing the computational complexity of probabilistic inference on a network at the cost of introducing a bounded error in the prior and posterior probabilities inferred.
