ProBT 1.9: New features and functionalities

Probayes released ProBT© 1.9 at the end of June 2008. This version includes new algorithms such as Expectation-Maximization, Hidden Markov Models, Junction Tree, Maximum Likelihood estimate and structure learning. All these algorithms improve the flexibility of the ProBT© toolbox. For instance, the Junction Tree algorithm will allow improving the computation time when asking multiple questions from the same Bayesian model. Similarly, the structure-learning module is an invaluable tool. It allows finding dependencies according to provided data when little knowledge about the variables dependencies is available. The serialization module of the toolbox was also improved. Inferred conditional and unconditional questions can now be serialized for its transmission or storage. This allows model reuse and opens the possibility of parallel Bayesian computation as well as the use of external software to communicate with the toolbox. Introspection utilities were added in ProBT© allowing to find the parameters that were used to construct the ProBT© objects. ProBT© is used in more than 15 countries around the world.