Stan is freedom-respecting, open-source software. -- mc-stan.org
Stan is amazing. It’s our tool of choice for building generative models. The project is open-source and we are committed to supporting the open-source community.
Stan is a probabilistic programming language, specifically designed for statistical models. Generable uses Stan to express generative models, a subset of the possible types of models that can be written in the Stan language. These are statistical models can be used to simulate data. The hope is to encode a generative model that approximates the actual data generation process well enough to make decisions based on the inferences from the statistical model.
I started working on Stan in 2011 while working with Andrew Gelman at Columbia University. This was before there were any users. Literally 0 users. At this time, the Stan language didn’t exist and the “model” had to be written directly in C++. Even at this stage, the development team wanted Stan to be a permissively licensed, open-source project. That decision allowed Stan to be adopted freely by a larger audience. Stan is now a much bigger project than the initial team had imagined.
At Generable, we’re putting generative models into the hands of more people by turning them into products. Our customers are downstream from the Stan programs we’re writing. We provide APIs that are abstracted to the appropriate level for our customers allowing them to focus on interpreting inferences and predictions in the context of their decision making process. (Which also includes removing the burden of maintaining Stan infrastructure.)
Commitment to Open-Source
Generable is committed to the open-source Stan project. As we work on our products, we will continue to contribute features and improvements to Stan. Here are some of the ways we are currently involved in the community:
As users. We present and publish models, give talks, and are active participants on the Stan forums.
As developers. We contribute code, review code, and work on new features.