NumFOCUS is pleased to announceStan as our newest fiscally sponsored project. Stan pairs a user-focused probabilistic programming language with a state-of-the-art computational library, enabling automatic inference for a large class of statistical models. Stan is used extensively in both industry and academia; work citing Stan spans the social and natural sciences and includes the analysis of the recent LIGO gravitational wave observation.
Stan is open source, BSD-licensed, and features among other things:
- Full Bayesian inference using the No-U-Turn sampler, a variant of Hamiltonian Markov chain Monte Carlo
- Approximate Bayesian inference using automatic differentiation variational inference
- Penalized maximum likelihood estimation using L-BFGS optimization
- User interfaces for the command line, Julia, MATLAB, Python, R, and Stata
- A stand-alone, extensible C++ library for reverse-mode automatic differentiation
- A supporting fully-templated matrix, linear algebra, and probability special function library with derivatives
Stan development began in 2011 at Columbia University and the development team has since grown to include core contributors from other institutions across the United States as well as Canada, Europe and South America.
With the addition of Stan, NumFOCUS now fiscally sponsors twelve different open source data science projects. To make a donation to support Stan or any of NumFOCUS’ other projects, click here.