Probabilistic reasoning in intelligent systems: networks of plausible inference
Probabilistic reasoning in intelligent systems: networks of plausible inference
Hi-index | 0.00 |
In this paper, we introduce two new algorithms for approximate inference of Bayesian networks that use edge deletion techniques. The first reduces a network to its maximal weight spanning tree using the Kullback-Leibler information divergence as edge weights, and then runs Pearl's algorithm on the resulting tree for linear-time inference. The second algorithm deletes edges from the triangulated graph until the biggest clique in the triangulated graph is below a desired bound, thus placing a polynomial time bound on inference. When tested for efficiency, these two algorithms perform up to 10,000 times faster than exact techniques. See www.cis.ksu.edu/~jas3466/research.html for more information.