Today we are releasing a new version of the HUGIN software (v. 8.5)
The main new features of this release are support for adaptation of class parameters in Object-Oriented Bayesian Networks including Dynamic Bayesian Networks as a special case. HUGIN now supports parameter adaptation in network classes using either fractional update or Online Expectation-Maximization (EM). Previously, the parameter adaptation in case of Object-Oriented Bayesian Networks was performed in the tables of the run-time instance. Now the adaptation is performed in the network class tables. Furthermore, a new HUGIN Decision Engine Application Programming Interface for the Python programming language is introduced in this release.
HUGIN Graphical User Interface v. 8.5
The HUGIN Graphical User Interface has been improved with various new features. These include:
– Parameter adaptation in network classes using either fractional update or Online Expectation-Maximization (EM) including supporting functionality in the Data Frame window.
– In the Learning Wizard and the Structure Learning Wizard, it is now possible to specify an initial structure of the graph prior to running the greedy search-and-score structure learning algorithm.
– A new Distance Analysis dialog is introduced. This is a tool to compute the distance between the (conditional) probability tables defined in two different models over the same set of nodes. The tool supports Hellinger distance and weighted Hellinger distance as measures of the distance between the probability distributions of two models.
– The feature selection functionality of the Learning Wizard and the Structure Learning Wizard has been improved.
– It is now possible to copy state labels from a numeric node to all numeric nodes without a state label.
– Other minor improvements.
Finally, work has been done to improve the performance of the HUGIN Graphical User Interface.
HUGIN Decision Engine v. 8.5
The HUGIN Decision Engine has been extended with the following features:
– The HUGIN Decision Engine now supports two new methods for parameter adaptation in Object-Oriented Bayesian Networks including Dynamic Bayesian Networks as a special case.
– A new HUGIN Decision Engine Application Programming Interface for the Python programming language.
– Support for Visual Studio 2017