Today – 6 November 2007 – we have released new versions of the HUGIN Graphical User Interface (v6.9) and HUGIN Decision Engine (v6.7).
The main new features of this release are:
-Parameter Sensitivity Analysis in HUGIN Graphical User Interface and APIs. This includes a Parameter Sensitivity Wizard in the HUGIN Graphical User Interface
-Internationalization of HUGIN Graphical User Interface (distributions include support for English and Japanese languages)
-An Adaptation Wizard in the HUGIN Graphical User Interface
Hugin Graphical User Interface v6.9:
The HUGIN Graphical User Interface has been improved with various new features. This includes:
-A Parameter Sensitivity Analysis Wizard. The Parameter Sensitivity Analysis Wizard aids the user in the process of identifying the most influential (conditional probability) parameters of a model and analyzing their effects on the “output” probabilities of the model.
-A color chart is displayed in conjunction with a node table for discrete nodes. The purpose of the color chart is to show where the large values are concentrated.
-The Learning Wizard now supports the specification of constraints between a node and a subset of nodes.
-The performance of the EM learning algorithm has been improved significantly by saving the initial clique potentials to memory, if possible. This improves the time efficiency of learning CPTs significantly, but increases the memory usage.
-A “Save to memory” option has been added to Network Properties. The save-to-memory operation stores a copy of the initial clique potentials in memory. This implies a faster initialization process which may improve efficiency of performing multiple propagations (e.g., as part of the Analysis Wizard).
-An adaptation wizard has been added. The adaptation wizard allows for performing batch adaptation using a number of HUGIN case and data files. Experience counts and posterior marginal distributions can be recorded as they develop during adaptation, and can be plotted as a graph.
-Japanese language support.
-Functionality has been added for computing the joint probability distribution over a set of discrete chance nodes.
-The HTML help pages have been extended with search facilities. In addition, the HTML help pages have been updated with a case and data files tutorial. Also, revisions have been made to the tutorial on adaptation and EM learning.
-Node and Domain descriptions support the use of HTML tags.
-It is possible to resize the dialog for inserting likelihood evidence.
-The conflict resolution button is enabled when evidence has been propagated. This allows the user to investigate subsets of the evidence when there is no conflict in the entire set of evidence.
-A JDBC interface has been added to the Wizards supporting database connectivity.
-A Cases/Beliefs panel has been added to the Analysis Wizard. This allows the user to compute and display the posterior probability of a hypothesis for a large number of cases.
-GUI response when working with large networks has been speeded up.
-Belief bars can be forced to render in scientific notation.
-The status bar reports the mean and variance for discrete numerical nodes when selected in run-mode.
-The status bar reports the mean and variance for continuous nodes when selected in run-mode.
-The selection of a continuous chance node reports the size of the node table to the status bar.
-The title of a network window now specifies whether a network is a “net” (i.e., non class model) or a “class” model.
Finally, efforts have been put into improving the stability of the HUGIN Graphical User Interface.
Hugin Decision Engine v6.7:
The HUGIN Decision Engine has been extended with the following features:
-Sensitivity analysis: Functions to aid in the process of identifying the most influential (conditional probability) parameters of a model and analyzing their effects on the “output” probabilities of the model are now provided.
-New statistical distributions: LogNormal, Triangular, and PERT.
-An optional location parameter has been added to the Gamma, Exponential, and Weibull distributions.
-Truncated continuous distributions can be expressed using a new truncation operator: It is possible to specify double as well as single truncation (single truncation is either left or right truncation).
-It is now possible to combine expressions of different types in conditional expressions. This permits specification of relationships that are sometimes probabilistic and sometimes deterministic.
-The performance of the operation for computing marginals on sets of nodes has been improved.
-The performance of inference has been improved for cases in which a memory backup is not available.