Probabilistic Constraint Handling Rules

Classical Constraint Handling Rules (CHR) provide a powerful tool for specifying and implementing constraint solvers and programs. The rules of CHR rewrite constraints (non-deterministically) into simpler ones until they are solved. In this paper we introduce an extension of Constraint Handling Rules (CHR), namely Probabilistic CHRs (PCHR). These allow the probabilistic “weighting” of rules, specifying the probability of their application. In this way we are able to formalise various randomised algorithms such as for example Simulated Annealing. The implementation is based on source-to-source transformation (STS). Using a recently developed prototype for STS for CHR, we could implement probabilistic CHR in a concise way with a few lines of code in less than one hour.