Stochastic Sampling Algorithms: Likelihood Sampling

From DSL

This algorithm is described in (Fung & Chang 1990) and in (Shachter & Peot 1990). Our implementation is based on (Fung & Chang 1990).

The likelihood sampling algorithm makes an attempt at improving the efficiency of the probabilistic logic sampling algorithm by instantiating only non-evidence nodes. Each sample is weighted by the likelihood of evidence given the partial sample generated. It is a simple algorithm with little overhead that generally performs well and certainly better than probabilistic logic sampling in cases with observed evidence.