GitHub

BipartiteFactorGraphs.jl

A performant implementation of bipartite factor graphs in Julia

Overview

BipartiteFactorGraphs.jl provides a type-stable implementation of bipartite factor graphs built on top of Graphs.jl. The package is designed for high performance while maintaining a clean, intuitive API.

Bipartite factor graphs are a specialized type of graph where nodes are divided into two distinct sets:

  • Variable nodes: Represent variables in a probabilistic model
  • Factor nodes: Represent relationships or constraints between variables

This structure is particularly useful for probabilistic graphical models, Bayesian inference, message passing algorithms, factor graph algorithms, and more.

Key Features

  • Type-stable implementation with parametric types for variables, factors, and edges
  • Efficient data storage and retrieval for node and edge properties
  • Specialized query functions for bipartite factor graphs
  • Compatible with the Graphs.jl ecosystem

Installation

using Pkg
Pkg.add("BipartiteFactorGraphs")

or via the Julia REPL:

] add BipartiteFactorGraphs

Quick Start

Here's a simple example demonstrating how to create and work with a bipartite factor graph:

using BipartiteFactorGraphs

# Create a graph with Float64 variable data, String factor data, and Int edge data
g = BipartiteFactorGraph(Float64, String, Int)

# Add variables
v1 = add_variable!(g, 1.0)
v2 = add_variable!(g, 2.0)
v3 = add_variable!(g, 3.0)

# Add factors
f1 = add_factor!(g, "sum")
f2 = add_factor!(g, "product")

# Connect variables and factors with edges
add_edge!(g, v1, f1, 10)
add_edge!(g, v2, f1, 20)
add_edge!(g, v2, f2, 30)
add_edge!(g, v3, f2, 40)

# Query the graph
println("Number of variables: ", num_variables(g))
println("Number of factors: ", num_factors(g))
println("Neighbors of factor f1: ", variable_neighbors(g, f1))
println("Edge data between v2 and f1: ", get_edge_data(g, v2, f1))
Number of variables: 3
Number of factors: 2
Neighbors of factor f1: [1, 2]
Edge data between v2 and f1: 20

Where to go next?

Index