On June 23rd, 2003 - we have released new versions of the HUGIN Graphical User Interface (v6.3) and HUGIN Decision Engine (v6.1).
The main new features of this release are:
Thread-safe HUGIN Decision Engine
User specified data preprocessing plugins
Parameter learning wizard
Various improvements of HUGIN Graphical User Interface
Data conflict analysis is the activity of detecting, tracing, and resolving possible conflicts among observations of variable values. The Hugin Graphical User Interface and HUGIN Decision Engine now support identification of potential conflicts in observations and hypothesis driven conflict resolution.
The HUGIN Graphical User Interface and HUGIN Decision Engine now support the use of case files. Case files make it possible for the user to save evidence entered in a domain in a case file for easy reload.
The HUGIN Decision Engine has been extended to support safe usage in a multi-threaded application. It is now possible to safely use the HUGIN Decision Engine in multi-threaded applications where threads process their own local domains or applications where threads access a common domain. In the last scenario, it is necessary to explicitly ensure mutual exclusion.
Support for loading user specified preprocessing plugins have been included in the learning wizards of the HUGIN Graphical User Interface. This enables the user to use her own data preprocessing functionality directly in the HUGIN Graphical User Interface.
The functionality of the HUGIN Graphical User Interface has been extended with an additional Learning Wizard - the EM Learning Wizard. The EM Learning Wizard increases the user support for learning the parameters of Bayesian networks from data. The EM Learning Wizard includes support for data acquisition (database connectivity via ODBC, interface to Oracle databases, and plain text files), data preprocessing, specification of domain expert knowledge on parameters, generation of randomized conditional probability distributions, etc.
The HUGIN Graphical User Interface has been improved with various new features. This includes:
Import of network information (groups, colors, node sizes, etc).
This is very useful in documentation of networks, but also for structural learning of Bayesian networks, where loading and saving structural constraints is now supported.
Curved and piecewise linear arrows between nodes. This makes it possible to avoid having arrows and nodes overlap.
View the junction tree structure (useful for conflict analysis).
Individual node sizes
Improved print functionalities (e.g multi-page networks and selection of printer).
Finally, the heuristic measure used by the HUGIN Decision Engine as part of the total-weight triangulation has been improved to give better results on large networks.
User feedback form¶
HUGIN Expert A/S has a continuous interest in the opinion of the users of HUGIN software and we wish to improve our software in line with our customers needs. For this reason the HUGIN Graphical User Interface has been extended with a questionnaire, which we kindly ask our users to complete. The user feedback will help us to improve our software and to guide the development of the software and keep it exciting to use. We sincerely hope that many of you will spend a few minutes in helping us improving ourselves.