GitHub

GraphPPL.jl Documentation

Welcome to the documentation of GraphPPL.jl, a Probabilistic Programming Language for Julia for specifying probabilistic models in a form of a factor graph. GraphPPL is a high-level backend-agnostic PPL that supports nested graphical models, allowing hierarchical modeling and model specification. GraphPPL materializes your probabilistic models as a factor graph. Additionally, it support a plugin system, that allows specification of inference specific information for different methods, e.g. variational inference. GraphPPL is designed to be a flexible and extensible PPL, and supports user-defined nodes and transformations.

It is important to note that GraphPPL.jl is not an inference package and does not run inference in the specified models. For inference, you may need a GraphPPL.jl compatible package, for example RxInfer.jl.

Table of Contents

Index