New Release – HUGIN 7.6

Today we are releasing a new version of the HUGIN software (v7.6). The main new features of this release are support for learning the structure of a Bayesian network with both continuous and discrete variables from a database of cases, new functionality in the HUGIN Graphical User Interface (GUI) for reading data and preprocessing data, and a new input panel of the HUGIN GUI.

HUGIN Graphical User Interface 7.6

The HUGIN Graphical User Interface has been improved with various new features. This includes:

– The constraint-based structure learning algorithms, i.e., the PC and NPC algorithms for learning structure, have been extended to support learning the structure of Bayesian networks with both continuous (CG) and discrete variables under the assumption that continuous nodes are conditional linear Gaussian and discrete nodes cannot have continuous parents.

– The functionality of the HUGIN GUI for reading and representing data has been improved significantly. The major changes include the use of a new flexible internal data structure for representing data as well as a completely reengineered data preprocessor for manipulating raw data.

– A new Input Panel has been added to HUGIN GUI Run-Time mode. The Input Panel includes a list of predefined nodes that the user can enter evidence on and the possibility to change the value of function nodes. This, for instance, makes it easy to enter evidence on a set of nodes as well as to change constants of expressions for tables defined as function nodes.

– The new input panel supports the computation of the p-percentile for interval nodes.

– The Value of Information Dialog now supports the calculation of mutual information between all subsets of nodes.

– The Learning Wizard now reports the table size of a selected node. This is helpful for determining if the data set is sufficient to justify the complexity of a network.

– Function and utility nodes can be specified as interface nodes.

– Function nodes can now be specified as parents of non-function nodes.  This can be used to provide input to the table generation facility.

Finally, efforts have been put into improving the performance of the HUGIN Graphical User Interface.

HUGIN Decision Engine v7.6

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

– The structure learning algorithm now supports learning of networks with continuous (CG) nodes.

– Function nodes can now be specified as parents of non-function nodes.  This can be used to provide input to the table generation facility.