LazyPPL: laziness and types in non-parametric probabilistic programs

Hugo Paquet, Sam Staton

Abstract: We introduce LazyPPL, a prototype probabilistic programming library for Haskell. The library emphasises the clarifying power of types, and the connection between non-parametric, stochastic processes and lazy (call by need) evaluation. We illustrate the power of the language with natural specifications of infinite structures including Poisson point processes, Gaussian processes, and Dirichlet Process clustering.