This repository contains two examples that compare different Probabilistic Programming Languages (PPLs) and their approaches. The first example "Hello Coin" is an introductiory example where we want to infer the bias of a coin based on observed coinflips. We will compare all 5 selected PPLs with each other and showcase different main features of this languages. In the second example "Deep Dive - Linear Regression with Outlier Detection" we will look at different approaches to inference. We will compare black-box inference with block inference and custom proposers. All three types are implemented in Gen. For comparisson Turing.jl showcases the limitations of black-box inference only languages (Compositional Inference using Gibbs Sampling is left out of this comparison, but might show some improvements for turing).
- Stan
- Beanmachine
- Pyro
- Turing.jl
- Gen
Unfinished. Please come back to this repository in a couple of days for system setup instructions and a docker image.