With the explosion of interest in applications of machine learning, there has also been renewed interest in probabilistic programming languages (PPL). PPLs tend to be viewed as (nontrivial) extensions of existing programming languages. Nontrivial not only because most extensions do not simply add some new primitives, though some do, but rather because they entirely change the semantics: rather than denoting values, programs now denote distributions. And, as with programming languages in general, these follow some underlying paradigm, be it imperative, functional, or logical.
展开▼
