Borch: A Deep Universal Probabilistic Programming Language

Borch is a scalable deep universal probabilistic programming language built on top of PyTorch. It makes it easy to write anything from hierarchical Bayesian models to probabilistic deep neural networks with billions of parameters, and supports one-line conversion of any PyTorch neural network into a Bayesian model.

The package is open source and available on PyPI and GitLab.

See also the announcement blog post for context on why we built Borch and what it enables.

@misc{belcher2022borch,
  title={Borch: A Deep Universal Probabilistic Programming Language},
  author={Belcher, Lewis and Gudmundsson, Johan and Green, Michael},
  year={2022},
  eprint={2209.06168},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2209.06168}
}