Today we are releasing a new version of the HUGIN software (v7.2). The main new features of this release are:
– A new Application Programming Interface (API) for PDAs. The HUGIN Decision Engine now has an API that makes it possible to run applications using HUGIN functionality on a PDA.
– HUGIN software has been extended with a new Monte Carlo algorithm to find the most probable configurations of a set of nodes in a network.
– Parameter estimation from data for object-oriented networks using the Expectation Maximization (EM) algorithm is now supported by the HUGIN Graphical User Interface (GUI). The parameter estimation is performed using the EM algorithm supporting missing values in data.
The HUGIN Graphical User Interface has been improved with various new features. This includes:
– Functionality to identify non-requisite parents of a decision node in a limited-memory incluence diagrams has been included.
– Value of Information Dialog: now supports value of information analysis on continuous nodes through a simple approximation.
– Evidence Sensitivity Analysis Dialog: the what-if tab now supports evidence on continuous nodes.
– Analysis Wizard: the Accuracy pane is enabled for networks with both discrete and continuous nodes.
– A new triangulation method has been introduced. This triangulation method is now the default method used by the tool.
– Improved support for automatic update when new builds are released.
– The status bar now displays the size of the policy for a selected decision node.
– The status bar now displays the size of the utility function associated with a selected utility node.
– Functionality to flip the network upside down has been included.
– Other minor improvements.
Finally, efforts have been put into improving the performance of the HUGIN Graphical User Interface.
The HUGIN Decision Engine has been extended with the following features:
– An algorithm for finding the “requisite” parents of a decision node in a LIMID has been implemented.
– A Monte Carlo algorithm for finding the most probable configurations of a set of nodes has been implemented.
– A new triangulation method has been implemented: This method triangulates each prime component using all of the elimination based triangulation heuristics and uses the best result.
This triangulation method is now used by the compilation operation when compiling untriangulated domains.
– As the elimination based triangulation heuristics may produce non-minimal triangulations, an extra pass that removes “redundant fill-in edges” has been added to these heuristics.