<< Click to Display Table of Contents >> Navigation:  Building blocks of QGeNIe > Inference algorithms > EPIS Sampling

The Estimated Posterior Importance Sampling (EPIS) algorithm is described in (Yuan & Druzdzel 2003). This is quite likely the best stochastic sampling algorithm for discrete Bayesian networks available. It produces results that are even more precise than those produced by the AIS-BN algorithm and in case of some networks produces results that are an order of magnitude more precise. The EPIS-BN algorithm uses loopy belief propagation to compute an estimate of the posterior probability over all nodes of the network and then uses importance sampling to refine this estimate. In addition to being more precise, it is also faster than the AIS-BN algorithm, as it avoids the costly learning stage of the latter.