aGrUM/pyAgrum 3.3.2 released

  • pyAgrum

    • parents, children, family, descendants, ancestors (BN, ID) and neighbours (MRF) are the C++ methods:
      parents/children also accept a set of ids or a list of names, and a wrong argument (e.g. a float or a
      negative id) raises TypeError instead of pyagrum.InvalidArgument (or of being silently accepted).
    • PAG.addNodes returns a set of ids, as for the other graphs (it returned a list).
    • generateSample (overloaded on name_out), mutilateBN, MEU and meanVar are annotated.
    • Bug fixes: memory corruption in BNDatabaseGenerator.varOrder() for node ids above 256, reference counting in
      PRMexplorer, memory leaks in Tensor.argmin/argmax, MEU/meanVar and when parsing Python sequences.
      BNLearner.useK2/setSliceOrder with a wrong argument raise TypeError (instead of SystemError).
  • aGrUM

    • Leaf modules (CM, CN, ID, PRM): explicit template instantiations tagged GUM_PUBLIC_<MODULE> for shared builds.
    • DAGmodel (BN, ID, CausalModel): parents, children, family, ancestors and descendants of a missing
      node id raise NotFound (they returned an empty set, or {id} for family).
    • Removed gum::learning::SimpleMiic (legacy, unused since Miic replaced it).