Abstract

In this talk I will present an initial version of a Haskell library designed to ease the construction and use of MCMC samplers. I will first introduce the MCMC method and motivate its application on a concrete example - the Gaussian Mixture Model (GMM). I will then describe an MCMC sampler for a GMM, and show how this sampler can be made more modular by using combinators from the library. The goal is to convince the audience of the benefits of applying paradigms such as higher order functions and lazy evaluation to the domain of MCMC sampling. No prior experience with MCMC samplers is required, and questions are strongly encouraged.