Metric Semantics and Full Abstractness for Action Refinement and Probabilistic Choice

Abstract

This paper provides a case-study in the field of metric semantics for probabilistic programming. Both an operational and a denotational semantics are presented for an abstract process language pr, which features action refinement and probabilistic choice. The two models are constructed in the setting of complete ultrametric spaces, here based on probability measures of compact support over sequences of actions. It is shown that the standard toolkit for metric semantics works well in the probabilistic context of pr, e.g. in establishing the correctness of the denotational semantics with respect to the operational one. In addition, it is shown how the method of proving full abstraction —as proposed recently by the authors for a nondeterministic language with action refinement—can be adapted to deal with the probabilistic language pr as well.