The main new feature of this release is functionality for learning the structure of a Bayesian network from data in parallel using the PC algorithm. The parallel implementation is based on the use of threads and requires all data in main memory.
This release also includes a number of improvements made to the Data Frame window for data processing and model performance evaluation.
HUGIN Graphical User Interface v8.2¶
The HUGIN Graphical User Interface has been improved with various new features. These include:
The Data Frame window has been improved with a number of new features to support model evaluation. In particular, the Data Frame window has been extended with functionality to calculate the confusion matrix for classification models where the class variable has more than two states, leave-one-out cross validation, the option to specify the value of a real-valued function node in the data set as well as functionality to show multiple curves in the same plot and to plot values against the case index.
The PC algorithm has been extended with functionality to learn the structure of a Bayesian network from data in parallel using threads. This means the time to learn the structure of a Bayesian network from data can be reduced by taking advantage of multiple cores on the computer.
The DBN functionality has been extended with a rewind button making it easy to reset a DBN without performing a new compilation of the model.
The Node Properties dialog has been improved.
Selected edges are now highlighted in Edit-mode.
New functionality to illustrate the strengths of dependence relations in a Bayesian network has been included. This is based on computing the pairwise mutual information between connected nodes in the graph.
Other minor improvements.
Finally, work has been done to improve the performance of the HUGIN Graphical User Interface.
HUGIN Decision Engine v8.2¶
The HUGIN Decision Engine has been extended with the following features:
The Table Generator has a new operator to specify the state index of a parent node in an expression. This is particularly useful when a node has parent nodes of subtype Interval.
The HUGIN Web Service API now supports the use of JSON objects for more efficient communication between a client and the server. This significantly reduces the communication overhead when the server has to perform a large number of HUGIN function calls.
The HUGIN Decision Engine supports parallel PC structure learning using threads.