aGrUM/pyAgrum 0.22.9 released

Posted on mar. 29 mars 2022 in News

ANNOUNCE: aGrUM 0.22.9

aGrUM/pyAgrum 0.22.9 is out.

Changelog for 0.22.9

This tag is a pre-relase for 1.0.0 (!).

  • aGrUM

    • Added a new Multithreaded facility which supports easily both openMP and STL.
    • Enabled exceptions raised by threads to be catched.
    • Made gum::CredalNetworks and gum::learning::BNLearner use the new multithreading facility.
    • Made a fully new architecture for scheduling inferences.
    • Added a sequential and a parallel schedulers for inferences.
    • Enabled gum::LazyPropagation and gum::ShaferShenoy to use schedulers for their inferences.
    • gum::DiscretizedVariable can now be declared as "empirical". Meaning that the lower and upper ticks are not always hard limits.
    • improve a bit API for gum::IntegerVariable.
  • pyAgrum

    • add a way to export BN samples as a pandas.DataFrame instead of csv files in pyAgrum.BNDatabaseGenerator and in the function pyAgrum.generateSamples(....).
    • gum.BNLearner can now take a pandas.DataFrame as data source in its constructor.
    • Add support for default number of thread in gum.config.
    • Added methods to get/set the number of threads used by pyAgrum.BNLearner, pyAgrum.LazyPropagation and pyAgrum.ShaferShenoy.
    • small change in pyAgrum.skbn.Discretizer.audit : show the domain size for discrete variable.
    • better graphical diff between BN, even if a node is missing using pyAgrum.bn_vs_bn functionalities.
    • empirical gum.DiscretizedVariable used in pyAgrum.skbn.Discretizer
    • new configuration for (LaTeX) fractions in gum.lib.notebook.showCPT (see tutorial)
    • update a bit pyAgrum.IntegerVariable wrapper and documentation.