Constraints for probabilistic logic programming

In knowledge representation, one commonly distinguishes definitions of predicates from constraints. This distinction is also useful for probabilistic programming and statistical relational learning as it explains the key differences between probabilistic programming languages such as ICL, ProbLog and Prism (which are based on definitions) and statistical relational learning languages such as Markov Logic (based on constraints). This motivates us to extend ProbLog with constraints; the resulting cProbLog in a sense unifies ProbLog and Markov Logic and is strictly more expressive than either of them.

Publication year:2012

Accessibility:Open