A Comparison of Stochastic Logic Programs and Bayesian Logic Programs

First-order probabilistic models are recognized as efficient frameworks to represent several realworld problems: they combine the expressive power of first-order logic, which serves as a knowledge representation language, and the capability to model uncertainty with probabilities. Among existing models, it is usual to distinguish the domain-frequency approach from the possible-worlds approach. Bayesian logic programs (BLPs, which conveniently encode possible-worlds semantics) and stochastic logic programs (SLPs, often referred to as a domain-frequency approach) are promising probabilistic logic models in their categories. This paper is aimed at comparing the respective expressive power of these frameworks. We demonstrate relations between SLPs’ and BLPs’ semantics, and argue that SLPs can encode the same knowledge as a subclass of BLPs. We introduce extended SLPs which lift the latter result to any BLP. Converse properties are reviewed, and we show how BLPs can define the same semantics as complete, range-restricted, non-recursive SLPs. Algorithms that translate BLPs into SLPs (and vice versa) are provided, as well as worked examples of the intertranslations of SLPs and BLPs.