Release 8.7

December 6, 2018

The main new features of this release are support for temporal clones of continuous variables in dynamic Bayesian networks and a Decision Sensitivity to Evidence dialog. Furthermore, a number of new features of the HUGIN Decision Engine are introduced in this release.

HUGIN Graphical User Interface v8.7

The HUGIN Graphical User Interface has been improved with various new features. These include:

  • It is now possible to create a temporal clone of a continuous chance node in dynamic Bayesian networks (DBNs). This means that a DBN may include both discrete and continuous chance nodes as well as have temporal clones of each category of node. In the case of mixed discrete and continuous temporal clones, there are restrictions on the compilation of the model.

  • The HUGIN Graphical User Interface has been extended with a Decision Sensitivity to Evidence dialog. This dialog enables the user to investigate the impact of changing the value of an observed chance node on the maximum expected utility of a decision node. This is useful for identifying the most and least important observations made prior to making a decision.

  • It is now possible in the Preference Pane of the HUGIN Graphical User Interface to deselect the use of fixed font size. This is useful when running HUGIN Graphical User Interface on a high resolution small screen.

  • The support for using the mouse wheel to zoom has been improved.

  • In the Conflict Analysis dialog it is now possible to specify a maximum subset size on the evidence when performing partial conflict analysis to reduce the computational cost.

  • In the Conflict Analysis dialog it is now possible to investigate the contribution of each individual finding to a conflict.

Finally, work has been done to improve the performance of the HUGIN Graphical User Interface.

HUGIN Decision Engine v8.7

The HUGIN Decision Engine has been extended with the following features:

  • The HUGIN Decision Engine now supports three new methods for learning the structure of restricted Bayesian networks from data. The algorithms make it possible to learn the structure of a Rebane-Pearl polytree, a Tree-Augmented Naive Bayes Model and a Chow-Liu tree from data.

  • The HUGIN Decision Engine Application Programming Interface for the Python programming language has been updated to use double precision for floating-point computations.

  • The run-time performance of the algorithm for optimal triangulation has been improved.