What is GraphPPL.jl?
GraphPPL.jl is a next-gen PPL that allows a general, high-level, all purpose Domain Specific Language (DSL) for probabilistic programming. It is designed to be a backend-agnostic and user-friendly PPL that can be used to specify a wide range of probabilistic models. The engine transforms a series of mathematical statements such as x ~ Normal(0, 1) into a factor graph containing the necessary information to perform inference. Next to this engine, GraphPPL.jl contains an implementation of a nested model specification, allowing users to specify models in a hierarchical manner. This allows for a more modular and reusable way of specifying models, and is especially useful for specifying models with a hierarchical structure.
RxInfer.jl 3.0
RxInfer.jl is a Julia package containing an inference engine for factor graphs. With the release of GraphPPL.jl, we have decided to integrate the nested model specification of GraphPPL.jl into RxInfer.jl. This renews the user-interface of RxInfer.jl and allows for a more modular and reusable way of specifying models. The renewed RxInfer.jl is now powered by GraphPPL.jl and mdoel specification is therefore more concise and powerful.