pygms 0.3.0

Python Graphical Models Toolbox

pyGMs : A Python toolbox for Graphical Models

This code provides a simple Python-based interface for defining probabilistic graphical models (Bayesian networks, factor graphs, etc.) over discrete random variables, along with a number of routines for approximate inference. It is being developed for use in teaching, as well as prototyping for research.

The code currently uses NumPy for representing and operating on the table-based representation of discrete factors, and SortedContainers for some internal representations. Smaller portions use networkx and scipy as well.

Installation

Simply download or clone the repository to a directory pyGMs, and add its parent directory to your Python path, either:

$ export PYTHONPATH=${PYTHONPATH}:/directory/containing/

or in Python

import sys
sys.path.append('/directory/containing/')