Description

This artifact provides an implementation of a probabilistic programming language that (i) features a coroutine-based paradigm for implementing generative models and custom inference guides (e.g., proposals for importance sampling), and (ii) uses a novel guide-type system to ensure that the distributions specified by a model and its guide have the same support; as a consequence, the model-guide pair is provably sound for probabilistic inference.

Assets

Provenance

The source code is part of the authors’ research project. The virtual machine is packaged using Vagrant (https://www.vagrantup.com/).

License