An Importance Sampling Algorithm Based on Evidence Pre-propagation

Precision achieved by stochastic sampling al­ gorithms for Bayesian networks typically de­ teriorates in face of extremely unlikely ev­ idence. To address this problem, we pro­ pose the Evidence Pre-propagation Impor­ tance Sampling algorithm (EPIS-BN), an importance sampling algorithm that com­ putes an approximate importance function using two techniques: loopy belief propaga­ tion [19, 25] and E-cutoff heuristic [2]. We tested the performance of EPIS-BN on three large real Bayesian networks: ANDES [3], CPCS [21], and PATHFINDER [11]. We observed that on each of these networks the EPIS-BN algorithm outperforms AIS­ BN [2], the current state of the art algorithm, while avoiding its costly learning stage.