Symbolic Probabilitistic Inference in Large BN2O Networks

A BN20 network is a two level belief net in which parent interactions are modeled using the noisy-or interaction model. In this paper we discuss application of the SPI local expression language [1] to effi­ cient inference in large BN20 networks. In particular, we show that there is sig­ nificant structure which can be exploited to improve over the Quickscore result. We further describe how symbolic tech­ niques can provide information which can significantly reduce the computation required for computing all cause poste­ rior marginals. Finally, we present a novel approximation technique with pre­ liminary experimental results.