Introducing the HUGIN System

What Is It?

The HUGIN System is a tool enabling you to construct model based decision support systems in domains characterized by inherent uncertainty. The models supported are Bayesian belief networks and their extension influence diagrams. The HUGIN System allows you to define both discrete nodes and to some extent continuous nodes in your models.

You have the opportunity to use the HUGIN System through HUGIN Runtime - an easy-to-use graphical environment. You can also use the HUGIN API (Application Program Interface) which comes as a library for C (or C++). More information about the versions of the HUGIN System that are for sale is found in the versions section.

The HUGIN System can be used to construct models as components in an application (mostly) in the area of decision support and expert systems. The application can communicate with the constructed component models either through DDE or by using the HUGIN API.

Basic Concepts

Before you can use the HUGIN System, you should at least understand the concept of Bayesian belief networks which is described in the basic concepts section. This guides you through the construction of a small Bayesian network.

In extension to Bayesian belief networks lies the concept of influence diagrams allowing you to create decision support directly through certain action nodes. If you are not familiar with influence diagrams you can also read about them in the basic concepts section.

In the basic concepts section you can also find a specification of the Bayesian belief network technology supported by the HUGIN System.

Examples provide one of the best ways to learn about a new concept. You can get an idea of the possibilities of Bayesian belief networks and influence diagrams in the examples section.

A more complete mathematical description of Bbns and Influence diagrams can be found in the book "Introduction to Bayesian Networks" by Finn Verner Jensen.

Learning More about HUGIN

When you feel that you know enough about the area of Bayesian belief networks and influence diagrams, the easiest way to get to know the HUGIN System is by experimenting with HUGIN Runtime (the easy-to-use graphical environment). The HUGIN Runtime tutorials section contains tutorials to help you construct your first Bayesian belief networks and influence diagrams. The HUGIN Runtime manual section contains more specific details about how to use HUGIN Runtime.

In the HUGIN API section, you can read about the opportunity to include HUGIN Bayesian belief networks and influence diagrams in C (or C++) programs.

You can get a detailed view of the components of the HUGIN System by reading the components desciption.

The Origin of HUGIN

During an ESPRIT project on diagnosing neuromuscular diseases, the Bayesian belief network MUNIN was constructed. A research group at Aalborg University worked on developing correct and efficient calculation methods for the diagnosis problem. Some results had at that time been obtained by American researchers, but a very obstinate problem still remained, which prevented Bayesian belief networks from being used in the construction of expert systems. The problem was know as the rumour problem: you may hear the same story through several different channels; but still the story may originate from the same source. Without knowing whether or not your channels are independent, you cannot combine the statements correctly.

In Bayesian belief networks the rumour problem appears when a cause can influence the same event through different paths in the network.

The problem was solved and general methods were made available to be used in any domain which can be modelled by a Bayesian belief network.

The methods have been programmed into a general development system, which is easy to use for anyone who wishes to construct an expert system based on Bayesian belief networks. The system is called HUGIN. Over the years the system has been extended with the facility of influence diagrams. Also, there has been research in trying to allow continuous nodes in the models. This has been added to the system in the case of continuous nodes with a Gaussian (normal) distribution.


Back