Skip to content
Permalink
master
Switch branches/tags
Go to file
 
 
Cannot retrieve contributors at this time

ForneyLab.jl

A Julia toolbox for automatic generation of (Bayesian) inference algorithms.

Given a probabilistic model, ForneyLab generates efficient Julia code for message-passing based inference. It uses the model structure to generate an algorithm that consists of a sequence of local computations on a Forney-style factor graph (FFG) representation of the model.

Package Features

  • User friendly syntax for specification of probabilistic models.
  • Automatic generation of message passing algorithms including
  • Support for hybrid models combining discrete and continuous latent variables.
  • Evaluation of free energy as a model performance measure.
  • Combination of distinct inference algorithms under a unified paradigm.
  • Features composite nodes that allow for flexible hierarchical design in terms of model structure and algorithms.

Resources

How to get started?

Head to the Getting started section to get up and running with ForneyLab.