FactorGraph

FactorGraph is a Julia package for constructing factor graphs and running belief propagation algorithms.

The package supports Gaussian factor graphs with linear Gaussian factors and discrete factor graphs with finite-state variables and factor tables. It provides sum-product and min-sum inference, Gaussian belief propagation in moment and canonical forms, several message schedules, forward-backward inference for tree-structured factor graphs, and dynamic graph updates.

Documentation

The documentation is organized around the two model families supported by the package. The Gaussian Models section covers linear Gaussian factor graphs, iterative Gaussian belief propagation, min-sum inference, and forward-backward inference on trees. The Discrete Models section covers finite-state factor graphs, sum-product inference, min-sum inference, and tree inference.

The Examples section shows complete workflows built from these APIs, including how to construct, update, visualize, and run inference on factor graphs. The API section collects the public types, graph construction functions, inference objects, validation utilities, visualization utilities, and printing helpers.

Installation

FactorGraph is available through the Julia package manager. Install it with:

pkg> add FactorGraph

Then load it with:

using FactorGraph

Gaussian Quick Start

using FactorGraph

variables = [
    GaussianVariable(:x1, 1),
    GaussianVariable(:x2, 1)
]

factors = [
    GaussianFactor(:x1, 1.0, 1.0, 0.1),
    GaussianFactor(:x2, 2.0, 1.0, 0.1),
    GaussianFactor(:x1, :x2, -1.0, [1.0 -1.0], 0.2)
]

graph = factorGraph(variables, factors)
inference = moment(graph)

gbp!(graph, inference; iterations = 30)

Discrete Quick Start

using FactorGraph

variables = [
    DiscreteVariable(:x1, 2; states = [:off, :on]),
    DiscreteVariable(:x2, 2; states = [:low, :high])
]

factors = [
    DiscreteFactor(:x1, [0.8, 0.2]),
    DiscreteFactor(:x1, :x2, [1.0 0.2; 0.1 0.9])
]

graph = factorGraph(variables, factors)
inference = sumproduct(graph)

gbp!(graph, inference; iterations = 10)