Skip to content

Latest commit

 

History

History
1491 lines (1193 loc) · 71 KB

CHANGELOG.md

File metadata and controls

1491 lines (1193 loc) · 71 KB
Raw

aGrUM Changelog

Changelog for 1.17.2

  • pyAgrum
    • Fixed some bugs in pyAgrum.lib.shapley (thanks Stefano Mariani !).
    • Changed last MN->MRF (pyAgrum.{loadMRF|saveMRF|availableMRFExts}).
    • Improved pickling of pyAgrum.skbn.BNClassifier.

Changelog for 1.17.1

  • aGrUM

    • Fixed extra semicolon.

  • pyAgrum

    • Fixed some pyAgrum types in generated files.
    • Added check for pandas in tests.
    • Renamed files in module pyAgrum.clg.
    • Fixed METADATA for pypi.

Changelog for 1.17.0

  • pyAgrum
    • Added support for python 3.13.
    • Added generator pyAgrum.Instantiation.loopIn() to iterate on an pyAgrum.Instantiation.
    • Significantly accelerated pyAgrum.Potential.fillFromDistribution().
    • Major acceleration of pyAgrum.clg.ForwardSampling.make_sample().

Changelog for 1.16.0

  • aGrUM

    • Added 3 behaviors for generating sampling from gum::DiscretizedVariable in gum::BNDatabaseGenerator : INTERVAL/MEDIAN/RANDOM. The default was INTERVAL and is now RANDOM.
    • Added timeout for gum::BNDatabaseGenerator to prevent slowish (maybe infinite) rejection sampling.
    • Added two new constraints for learning :gum::learning::NoParentNode,gum::learning::NoChildNode.
    • Added these new constraints in gum::BNLearner : addNoParentNode(nodeid) and addNoChildrenNode(nodeid).
    • (internal) Improved act's scripts.

  • pyAgrum

    • Dropped support of python 3.9 (following NEP29).
    • Added 3 behaviors for generating sampling from pyAgrum.DiscretizedVariable in pyAgrum.BNDatabaseGenerator : INTERVAL/MEDIAN/RANDOM. The default was INTERVAL and is now RANDOM.
    • Added timeout for pyAgrum.BNDatabaseGenerator to prevent slowish (maybe infinite) rejection sampling.
    • Added two new constraints pyAgrum.BNLearner.addNoParentNode() and pyAgrum.BNLearner.addNoChildrenNode().
    • (internal) Standardized licence in python files
    • Added new pyAgrum.causal.CausalEffectsEstimation (thanks to Thierry Rioual).
    • pyAgrum.skbn.BNClassfiercan now be pickled (thanks to Valentin Vassilev).
    • Added module pyAgrum.lib.shapley (included in pyAgrum.lib.explain) to allow the use of ShapValues not only from notebooks (thanks to Stefano Mariani).
    • Added a parameter filename to pyAgrum.lib.shapley.ShapValues's methods for saving the figure instead of showing it (thanks to Stefano Mariani).

Changelog for 1.15.1

  • aGrUM

    • Added gum::Potential<GUM_SCALAR>::memoryFootprint and gum::IBayesNet::memoryFootprint (thanks to Benjamin Datko).
    • Fixed a bug when subscripting gum::DiscretizedVariable with the (last) interval (thanks to Gaspard Ducamp).
    • Removed gcc warnings (psabi, false positive uninitialized).

  • pyAgrum

    • Added pyAgrum.Potential.memoryFootprint and pyAgrum.BayesNet.memoryFootprint (thanks to Benjamin Datko).
    • Fixed a bug when subscripting gum::DiscretizedVariable with the (last) interval (thanks to Gaspard Ducamp).
    • Fixed a bug in pyAgrum.skbn.BNDiscretizer for NoDiscretization method with given paramDiscretization.
    • Moved function pyAgrum.lib.export.toFast to method pyAgrum.{BayesNet|InfluenceDiagram|MRF}.toFast().
    • Moved pyAgrum.skbn.BNDiscretizer to pyAgrum.lib.discretizer.
    • Worked on documentation for CTBN
    • Fixed fontcolor configuration for causal model.

Changelog for 1.15.0

We continue to add (full python) experimental models. In 1.15, we introduce the CTBN model (Continuous Time Bayesian Network) featuring, as usual, modelisation and representation, inference (exact and sampling) and a learning algorithm.

  • aGrUM

    • Added gum::NodeId gum::EssentialGraph::idFromName(const std::string& name) and const std::string& gum::EssentialGraph::nameFromId(gum::NodeId node).

  • pyAgrum

    • Added pyAgrum.EssentialGraph.idFromName(str)->int and pyAgrum.EssentialGraph.nameFromId(int)->str
    • Improved documentation of pyAgrum.lib.explain
    • Better pyAgrum.clg.CLG.toDot() and pyAgrum.clg.CLG._repr_html().
    • New model Continuous Time Bayesian Network pyAgrum.ctbn.
    • Formatted and adjustments in pyAgrum.ctbn.
    • Updated documentations for python experimental models notebooks.
    • Updated thumbnails for python experimental models notebooks.
    • Added serialization (pickle) for CLG and CTBN (consistent with other models in pyAgrum).
    • Improved pyAgrum.lib.utils.{apply_}dot_layout
    • Added pyAgrum.lib.utils.async_html2image for exported HTML as png or pdf (notably for pyAgrum.lib.notebook.getSideBySide and pyAgrum.lib.notebook.getPotential).

Changelog for 1.14.1

  • aGrUM

    • Fixed compilation errors for clang18.

  • pyAgrum

    • Fixed python version for readthedocs.

Changelog for 1.14.0

  • aGrUM

    • gum::Potential::marg{Xyz}{In|Out} are renamed gum::Potental::{xyz}{In|Out} to cut down on some verbosity. (for instance : p.margSumOut becomes p.sumOut)
    • Better error messages for gum::BNLearner<GUM_DATA>

  • pyAgrum

    • Dropped support of python 3.8 (following NEP29).
    • Fixed some memory leaks (in special cases) found by Christophe Gonzales.
    • new pyAgrum.lib.explain.generalizedMarkovBlanket{Names} and documentation.
    • pyAgrum.Potential.marg{Xyz}{In|Out} are renamed pyAgrum.Potental.{xyz}{In|Out} to cut down on some verbosity. (for instance : p.margSumOut becomes p.sumOut)
    • Better error messages for pyAgrum.BNLearner
    • Added new python's experimental model : Conditional Linear Gaussian (pyAgrum.clg.*) : model, learning and inference.

Changelog for 1.13.2

  • aGrUM

    • Better (compact) format for numerical labels.
    • Better uniformly distributed random discrete distribution.

  • pyAgrum

    • Better (compact) format for numerical labels.
    • Better uniformly distributed random discrete distribution.
    • Continue to implement some automatic pyAgrum.Potential filled tools with pyAgrum.Potential.fillFromExpression.
    • Worked on better plots for probability+utility
    • Speeded up gum.BayesNet.addArcs() with jit memory allocation
    • Slight adjustments to histogram display.
    • Fixed documentation of gum.BNLearner.mutualInformation.
    • Fixed (deprecated) arguments in pyAgrum.skbn.BNClassifier's methods.

Changelog for 1.13.1

This release is mainly about documentation.

  • pyAgrum
    • Fixed last references to LIP6's notebooks instead of rtd's notebooks (thanks to Benjamin Datko).
    • Added 'Book of Why's notebooks to readthedoc's documentation.
    • Updated Notebooks and thumbnails.
    • Improved documentation coverage for pyAgrum class methods (from 86.19% to 87.05%)
    • gum.BNLearner.learnEssentialGraph for all learning algorithm.
    • Better wrapped of pyAgrum.BNLearner.setPossibleEdges().

Changelog for 1.13

  • Internal

    • New servers for (linux and windows) CI !

  • aGrUM

    • Fast syntax for empirical gum::DiscretizedVariable: name suffixed by '+'.
    • Reorganized a bit gum::DiscreteVariable's hierarchy.
    • Worked on quasi-continuous (numerical discrete variable with large domain size) :
      • Logical operators (&,|, ~) to combine evidence (monodimensionnal gum::Potential not forced to sum to 1).
      • New gum::Potential and gum::BayesNet methods to generate numerical evidence : evEq, evIn, evLt, evGt.
      • Fast syntax now accepts any type as default (and not only the size of the domain).

  • pyAgrum

    • Fixed a bug in `pyAgrum.lib.bn2scores.
    • Fast syntax for empirical pyAgrum.DiscretizedVariable: name suffixed by '+'.
    • Reorganized pyAgrum.lib._colors.py and pyAgrum.lib.utils.py.
    • Changed a default for pyAgrum.Instantiation.todict.
    • Improved (a lot) histograms for pyAgrum.DiscretizedVariable.
    • Worked on quasi-continuous (numerical discrete variable with large domain size) :
      • Logical operators (&,|, ~) to combine evidence (monodimensionnal pyAgrum.Potential not forced to sum to 1).
      • New pyAgrum.Potential and pyAgrum.BayesNet methods to generate numerical evidence : evEq, evIn, evLt, evGt.
      • Fast syntax now accepts any type as default (and not only the size of the domain).
      • pyAgrum.Potential accepted as evidence for all inference.
      • Long-awaited pyAgrum.Potential.fillFrom{Function|Distribution}.
    • Added pyAgrum.config.{push|pop}.
    • Removed obsolete code for python2.7.
    • Removed some oldest depreactionWarning.
    • Added 'iter' for graphs, graphical models, variable and potentials.

Changelog for 1.12.1

  • aGrUM
    • Added a new job dured CI for sphinx documentation.
    • Fixed a typo in gum::MeeRules::propagate(s).
  • pyAgrum
    • Fixed a typo in pyAgrum.MeekRules.propagate(s).
    • Added new rules for 'python typification' of generated files.
    • Fixed and updated documentations and notebooks.

Changelog for 1.12.0

  • aGrUM

    • Added a new class gum::MeekRules to complete a learning PDAG to CPADG or BN.
    • Prevented 'infinities' in the definition of numerical discrete variable.
    • Added a new constructor for gum::PDAG fromgum::MixedGraph.
    • Better (stricter) operator== between gum::DiscreteVariable.
    • Added a new static constructor gum::{Undi|Di}Graph::completeGraph(int n).
    • Added closestIndex() and closestLabel() methods for all numerical gum::DiscreteVariable.
    • Fixed an exception raised in inferences due to the removal of uninformative potentials contained only barren variables
    • new syntax for fastVariable : [a:b:n] = gum.Discretiezd with n intervalls from a to b.

  • pyAgrum

    • Pickle for graphs and graphicals models !
    • Prevented 'infinities' in the definition of numerical discrete variable.
    • Added a new class pyAgrum.MeekRules to complete a learning PDAG to CPADG or BN.
    • Better (stricter) __eq__ and __ne__ between pyAgrum.DiscreteVariable.
    • Added a new constructor for pyAgrum.PDAG frompyAgrum.MixedGraph.
    • Added a new expert discretization method to pyAgrum.skbn.BNDiscretizer (thanks to Ahmed Mabrouk).
    • pyAgrum.skbn.BNDiscretizer can now take csv filename for input.
    • Added pyAgrum.fastGraph to create pyAgrum.{Di|Undi|Mixed}Graph.
    • Dataframes can now come from pandas or polars (thanks to Mads Linksou).
    • Added a compact mode for histogram for discretized variable controlled by histogram_mode histogram_epsilon in gum.config (thanks to Gaspard Ducamp).
    • Added a new static constructor pyAgrum.{Undi|Di}Graph::completeGraph(int n).
    • Added subscripted to graphical models : g[name] gives now a direct access to the pyAgrum.DiscreteVariable by its name.
    • New syntax for pyAgrum.fast{Var|BN|ID|MRF} : [a:b:n] = pyAgrum.DiscretizedVariable with n intervalls from a to b.
    • Added closestIndex() and closestLabel() methods for all numerical pyAgrum.DiscreteVariable.
    • Added g.adjacencyMatrix() for all graphs and graphical models (thanks to Kenneth Lee).
    • Added new bgcolor parameter (None by default) for showROC/PR in pyAgrum.lib.bn2roc and pyAgrum.skbn (thanks to Clara Charon).
    • Added optional dialect parameter to pyAgrum.lib.bn2roc.computeScores.
    • new syntax for fastVariable : [a:b:n] = pyAgrum.Discretiezd with n intervalls from a to b.

Changelog for 1.11.0

  • Internal

    • Removed some warnings in CMakeLists.txt
    • Refactored CI, cleaned wheelhouse

  • aGrUM

    • Added gum::LazyPropagation::mpe{Log2Posterior} to compute the Most Probable Explanation with some evidence.
    • Improved MIIC algorithm and application of Meek Rules in ordre to to better solve (partially) oriented cycle problems.
    • Fixed a bug in gum::BNLearner : EM parameter estimation bug due to a cache used in record counters.
    • Fixed a bug in gum::GibbsSampling : some evidence may not be taken into account (thanks to Micromix@discord).
    • gum::PDAG must accept partially directed cycle (thanks to Kenneth Lee).
    • 'infinities' is now forbidden in the definition of numerical gum::DiscreteVariable.

  • pyAgrum

    • Updated and added 3.12 in metadatas.
    • Added pyAgrum.LazyPropagation.mpe{Log2Posterior} to compute the Most Probable Explanation with some evidence.
    • Removed reference to sys.log and to setuptools.
    • Removed legacy setuptools related files.
    • Fixed a bug in pyAgrum.BNLearner : EM parameter estimation bug due to a cache used in record counters.
    • Fixed a bug in pyAgrum.GibbsSampling : some evidence may not be taken into account (thanks to Micromix@discord).
    • Fixed a bug in frontdoor detection in a causal model with multiple connected components.
    • Fixed some regexp expressions.
    • pyAgrum.PDAG must accept partially directed cycle (thanks to Kenneth Lee).
    • 'infinities' is now forbidden in the definition of numerical pyAgrum.DiscreteVariable.

Changelog for 1.10.0

  • aGrUM

    • added contraints to MIIC BN learning algorithm.
    • MIIC is now the default for gum::BNLearner.
    • Fixed some minor bugs in gum::BNLearner.
    • Removed old 3off2 BN learning algorithm (use MIIC instead).
    • added gum::MIIC::learnPDAG and give access to this method via gum::BNLearner.

  • pyAgrum

    • package for python 3.12 !
    • New configuration for specification of graphviz engine and rankdir .
    • Constraint for MIIC algorithm in pyAgrum.BNLearner.
    • MIIC is now the default for pyAgrum.BNLearner.
    • Fixed minor bugs in pyAgrum.BNLearner.
    • new method pyAgrum.BNLearner.learnPDAG.
    • Back to classical representation of do() operator in pyAgrum.causal.
    • Examples from the Book Of Why (Pearl) are now in the sphinx documentation.
    • The name of a loaded BN should not include the complete path but only the name of the file (thanks to Jules Cassan).

Changelog for 1.9.0

  • aGrUM

    • learning algorithm gum::learning::MIIC can use the weighted databases.
    • Internal improvements for act tool, cmake and compilers (clang).

  • pyAgrum

    • New visualisation for gum::DiscretizedVariable + new config to select this visualisation.
    • pyAgrum.BNLearner can use now the weighted databases for all learning algorithms.
    • Documentation improvements.
    • pyAgrum.lib.bn2roc
      • added new functions get{ROC|PR}points().
      • accepted pandas.DataFrame as data source (datasrc).
      • added Fbeta (beta!=1) scores to bn2roc.
      • added F-Beta threshold on ROC and PR curves.
      • bn2roc functions now force many parameters to be keyword-arguments in order to prevent the risk of mixed arguments.
      • added new functions anim{ROC|PR}.
    • pyAgrum.skbn.Discretizer can propose a set of labels (that includes the labels from the database) when "NoDiscretization" is selected. (see tutorial 52-Classifier_Discretizer).

Changelog for 1.8.3

  • aGrUM

    • Removed GCC warnings about overloaded virtual methods.
    • Fixed a bug in jointree inference in rare cases (when an irrelevant soft evidence is added).

  • pyAgrum

    • Fixed a bug in jointree inference in rare cases (when an irrelevant soft evidence is added).
    • Removed deprecated methods from pyAgrum<1.1.0.
    • Added new syntax for set (conditional) marginal with dict : p[{"A":0,"C":1}]={"no":0.3,"yes":0.7}.
    • small updates for pyAgrum's documentation.

Changelog for 1.8.2

  • aGrUM

    • Fixed a bug (still) in gum::PDAG::cSeparation (thanks to Keneth Lee).
    • Fixed a bug in gum::BIFXMLIdWriter::save.
    • typos in gum::PDAG::toDot().

  • pyAgrum

    • Fixed a bug in pyAgrum/{load|save}ID.
    • Wrapped gum::PDAG::moralizedAncestralGraph().
    • Added a warned when importing pyAgrum.lib.notebook if graphviz is not installed.

Changelog for 1.8.1

  • aGrUM

    • Fixed bugs in gum::PDAG::cSeparation (thanks to Keneth Lee).
    • Used 'table' instead of 'default' for marginal cpt in gum::BIFWriter.
    • Added std::cout<<(gum::Timer& t)

  • pyAgrum

    • Updated build image for rtd.
    • Improved documentations.
    • Documentation for pyAgrum.InformationTheory.
    • Uniformized cmap to cmapNode in pyAgrum.lib.notebook.
    • Added beeswarmplot for shapley values figures in pyAgrum.lib.explain.

Changelog for 1.8.0

  • aGrUM

    • Added new class gum::InfomationTheory that regroups every information theory concepts (entropies, conditional entropies and mutual information, ...) with only one inference, which allows also conditioned information theory concepts.
    • Added gum::Potential::expectedValue()
    • Added the alternative fast syntax for NumericalDiscreteVariable : {1.5:3.5:3} meaned from 1.5 to 3.5 in 3 steps.
    • Updated GraphicalModel::exists(const std::string&) : return false instead of throwed an exception.
    • Added gum::graphicalModels::variables() and a new alias for gum::VariableSet.
    • Fixed a bug in gum::ShaferShenoyInference for joint posterior when some nodes in the target received hard evidence.
    • Improved error message w.r.t gum::*::jointPosterior.
    • Added mingw as a target for CI.
    • Fixed dangled pointers in testsuite, only detected by MSVC.
    • Optimized gum::PDAG::cSeparation
    • Renamed gum::Set::isProper{Sub|Super}Set to gum::Set::isStrict{Sub|Super}Set.

  • pyAgrum

    • swig>4.1 is now needed.
    • Enabled installation with poetry (thanks to Lorenzo Conti)
    • Added new class pyAgrum.InfomationTheory that regroups every information theory concepts (entropies, conditional entropies and mutual information, ...) with only one inference, which allows also conditioned information theory concepts.
    • Added pyAgrum.Potential.expectedValue().
    • Restructured a bit the sphinx documentation.
    • Down-exported the read-only API for every types of discrete Variable to pyAgrum.DiscreteVariable. (see VariablesTestSuite.testExportDerivedReadOnlyAPIforDiscreteVariable)
    • Updated notebooks to the new down-exported read-only API of pyAgrum.DiscreteVariable.
    • Added the alternative fast syntax for NumericalDiscreteVariable : {1.5:3.5:3} meaned from 1.5 to 3.5 in 3 steps.
    • Updated docs for new fast syntax.
    • Updated GraphicalModel::exists(const std::string&) : return false instead of throw an exception.
    • Initiated a new notebook about the interaction with ipywidget
    • Optimized and correctly wrapped gum::PDAG::cSeparation (thanks to Kenneth Lee).
    • New functions pyAgrum.lib.explain.nestedMarkovBlankets and pyAgrum.lib.explain.nestedMarkovBlanketsNames.

Changelog for 1.7.1

  • aGrUM

    • Added gum::{AllDiscreteVariable}.toFast() to write the variable in fast syntax.
    • Fixed a bug in BN's random generators : domain_size incremented by 1 (thanks to Kenneth Lee).
    • Fixed a bug in fast syntax : 'A[n]' with n<2.

  • pyAgrum

    • Added pyAgrum.{AllDiscreteVariable}.toFast() to write the variable in fast syntax.
    • Fixed a bug in BN's random generators (e.g. pyAgrum.randomBN) : domain_size incremented by 1 (thanks to Kenneth Lee).
    • Fixed a bug in fast syntax : 'A[n]' with n<2.
    • Modifyed (lightly) the structure for documentation.

Changelog for 1.7.0

  • aGrUM
    • Add gum::Potential::sign and gum::Potential::new_sign
    • Bug fix in causalImpact found by Musfiqur Rahman
    • Generalization and rationalization of access of aGrUM's generator of pseudo-random values
    • Added gum::Timer::toString().
    • Improved BN random generator.
    • get rid of sprintf as much as possible.
    • Fixed random seed.
    • Fixed gum::Potential::random() : not a distribution.
    • Updated Fast syntax : enumerated consecutive integers (e.g {1|2|3|4}) creates a gum::RangeVariable[1,4] rather than an gum::IntegerVariable{1|2|3|4}.
  • pyAgrum
    • Add pyAgrum.Potential.sign and pyAgrum.Potential.new_sign
    • Bug fix in causalImpact found by Musfiqur Rahman
    • Added pyAgrum.Timer.__str()__.
    • Improved BN random generator.
    • Fixed random seed.
    • Fixed pyAgrum.Potential.random() : not a distribution.
    • Added access to gum.fastVariable().
    • Updated Fast syntax : enumerated consecutive integers (e.g {1|2|3|4}) creates a pyAgrum.RangeVariable[1,4] rather than an pyAgrum.IntegerVariable{1|2|3|4}.
    • add a new module to export in different formats (for now just GML).

Changelog for 1.6.1

Quick fix for the configuration of pyAgrum documentation generation by readthedoc.

Changelog for 1.6.0

The main points of this tag is the renamed of Markov network (MarkovNet, MN) for the better known Markov Random Field (MarkovRandomField, MRF) and a new reader/writer of XDSL format (Genie/Smile) for Bayesian networks. Other improvements and corrections have naturally also been made.

  • aGrUM

    • Renamed gum::MarkovNet to gum::MarkovRandomField. Renamed gum::*MN* to gum::*MRF* when necessary.
    • Fixed glitches and bugs induced or revealed by gum::MarkovNet->gum::MarkovRandomField.
    • new XDSL Reader/writer for Bayesian network.
    • Renamed gum::Learning::BNLearner::learnMixedStructure to gum::Learning::BNLearner::learnPDAG
    • worked on documentation : better rendered for doxygen pages.
    • Renamed gum::dSeparation to gum::dSeparationAlgorithm.

  • pyAgrum

    • Renamed pyAgrum.MarkovNet to pyAgrum.MarkovRandomField. Renamed pyAgrum.*MN* to pyAgrum.*MRF* when necessary.
    • new XDSL Reader/writer for Bayesian network.
    • Renamed pyAgrum.BNLearner.learnMixedStructure() to pyAgrum.BNLearner.learnPDAG().
    • For figure contained nodes drawn by matplotlib (e.g. inference), use the same font for all nodes (default from matplotlib) (thanks to Jonathon Blackford).
    • Worked on documentation : better rendered for readthedocs pages, improved structuration, new thumbnail image for some notebooks/tutorials.
    • Significant improvement of the documentation coverage.
    • Improved gum.DiscreteVariable.to[typeOfVariable], renamed as gum.DiscreteVariable.as[typeOfVariable] and added documentations.

Changelog for 1.5.2

Mainly documentation and new analytics for the different sites.

  • pyAgrum
    • Added a new and very easy to use pyAgrum.randomBN that creates a BN from a list of names.
    • Removed obsolete pyAgrum.generateCSV replaced by pyAgrum.generateSample. Fixed tests.
    • Fixed and improved pyAgrum documentation.

Changelog for 1.5.1

  • aGrUM

    • Improved gum::BNLearner::state().
    • (internals) new macro GUM_TEST and GUM_INACTIVE_TEST for CI

  • pyAgrum

    • Improved gum.BNLearner.state().
    • Improved style for readthedoc (pygments and no StickySideBar)
    • Improved documentation coverage (for methods) from 84.9% to 90.7% : more than 100 newly covered methods.

Changelog for 1.5.0

  • aGrUM

    • (internals) Updated act --stats.
    • (internals) Reorganized build folder : build/{aGrUM|pyAgrum}/{debug|release}.
    • (internals) Improved cmake & CIs.
    • Threadsafe graphs and Graphical Models.
    • Added new graph gum::PDAG (Partially Directed Acyclic Graph).
    • Renamed gum::MixedGraph::adjacents with correct graph notion : gum::MixedGraph::boundary.
    • Initialized the majority of the end/rend iterators at compile time.
    • Added AVL binary search trees: gum::AVLTree.
    • Added priority queues that can be iterated in order: gum::SortedPriorityQueue.
    • Better messages for gum::BayesNet::check().

  • pyAgrum

    • (internals) New docker images with linux gcc 11 for wheels.
    • (internals) CMake: Use FindPython module.
    • (internals) Removed some (false positive) warned notifications from swig.
    • Fixed NaN bugs for new versions of Graphviz.
    • Improved gum.lib.notebooks.flow and light/dark theme compatibility.
    • Renamed pyAgrum.MixedGraph.adjacents with correct graph notion : pyAgrum.MixedGraph.boundary.
    • New graph class gum.PDAG (Partially Directed Acyclic Graph).
    • Fixed a small typo when displayed function as Potential.
    • better messages for pyAgrum.BayesNet.check().
    • More robust gum.explain.showInformation() w.r.t. NaN.
    • Fixed typos in documentation.
    • Improved the organization of ReadTheDoc documentation.

Changelog for 1.4.1

  • pyAgrum

    • bugfixes in pyAgrum.lib.[causal.]notebook.

Changelog for 1.4.0

  • aGrUM

    • better toString for BN : show the memory used by the parameters.

  • pyAgrum

    • added conda and pip packages for python 3.11
    • better __str__ for BN : show the memory used by the parameters.
    • bug fix in obsolete pyAgrum.BNLearner.useNoAPriori().
    • bug fix when displayed a pyAgrum.causal.CausalFormula generated by do-Calculus : retrieved the original doing and knowing sets.

Changelog for 1.3.2

  • aGrUM

    • gum::BNDataGenerator : Add a way to generate sampls from Bayesian Network with evidence
    • More informative error message for gum::DatabaseTable::insertRow().
    • Code structure for checked syntax before saved a Bayesian Network. Instantiated for BFI and DSL format.
    • Improved code for gum::BinSearchTree.
    • Better error messages for gum::DSLReader.

  • pyAgrum

    • pyAgrum.BNDataGenerator.drawSamples() : Add a way to generate sampls from Bayesian Network with evidence.
    • More informative error message for gum::DatabaseTable::insertRow().
    • pyAgrum.lib.dBN : better dot representation of unrolled BNs.
    • pyAgrum.lib.notebook.show() : bugs fixed for some types of parameters.
    • pyAgrum.config : new types accessors : config.as{Int|Float|Bool}.
    • Updated documentations for pyAgrum.lib.notebook.show(), pyAgrum.BNDatabaseGenerator.drawSamples(), pyAgrum.skbn.
    • New argument when drawed graphs : label on arcs (thanks to Mahdi Hadl Ali).

Changelog for 1.3.1

  • aGrUM

    • Updated readme

  • pyAgrum

    • Fixed warnings in gum.Potential.topandas().
    • Allowed label or index when addressed variables in dictionary for gum.Instantiation (such as in gum.Potential.extract).
    • Allowed dictionnary for mapped variables in pyAgrum.Potential.fillWith.
    • Fixed typos in documentation of pyAgrum.Potential.fillWith.
    • Fixed css for sphinx notebooks gallery.

Changelog for 1.3.0

  • aGrUM

    • Fixed errors in doxygen configuration.
    • Added a new class of discrete variable gum::NumericalDiscreteVariable whose support is a (finite) set of double.
    • Fixed new clang warnings.
    • Fixed a bug in gum::Potential::max() when all values are negative (thanks to Zakarie Aloui).
    • Added a new method gum::learning::BNLearner::score() which computes the currently selected score for a node and its parents.
    • Added a test for a variance beed not negative in gum::ShaferSheoyLIMIDInference (thanks to Benjamin Datko).
    • Added new methods gum::learning::BNLeaner::mutualInformation() and gum::learning::BNLearner::correctedMutualInformation().

  • pyAgrum

    • Added a new class of discrete variable gum.NumericalDiscreteVariable whose support is a (finite) set of double.
    • Updated notebooks w.r.t. this new variable (16-Examples_quasiContinuous and 18-Examples-BayesianBetaCoin).
    • Added a new method gum.BNLearner.score() which computes the currently selected score for a node and its parents.
    • Added new methods gum.BNLeaner.mutualInformation() and gum.BNLearner.correctedMutualInformation().

Changelog for 1.2.0

  • aGrUM

    • Moved named convention from apriori to prior.
    • gum::CliqueGraph::addNode{WithId} with the same syntax as in other graphs.
    • [act] Improved the compliance for various microsoft compilers.
    • [act] option -windows is now -compiler and accepts gcc or clang when platform is linux or mac.
    • [act] options --clangpath add --gccpath to specify the path of the compilers (if those compilers are not in PATH).
    • Changed internal representation of gum::Exception (now derived from std::exception).
    • Added a new prior : gum::learning::DirichelPriorFromBN.
    • Renamed genericBNLearner to IBNLearner.
    • Added gum::learning::BNLearner::useDirichletPriorFromBN.

  • pyAgrum

    • Moved named convention from apriori to prior.
    • Added pyAgrum.BNLearner.useDirichletPriorFromBN.
    • Added a new function gum::mutilateBN.
    • pyAgrum.CliqueGraph.addNode{WithId} with the same syntax as in other graphs.

Changelog for 1.1.1

  • aGrUM

    • Fix a regression that was slowed down the inference in credal networks.
    • Better error messages when IOError caused by writed in 'bad' files.
    • Typos in several error messages.
    • Fix warnings from gcc 12.1 and integrate linter proposals.

  • pyAgrum

    • PEP 3149 implemented.
    • Small typos in pyAgrum.causal.
    • Added acces to the database built for learning in pyAgrum.skbn.BNClassifier.preparedData().
    • Small bug when represented Potential as fractions.
    • Small bug in type deduction from database for variable in pyAgrum.skbn.
    • notebooks now added in the sphinx documentation (and then in readthedocs site).

Changelog for 1.1.0

  • aGrUM

    • Parallelized inference for Markov Net : gum::ShaferShenoyMNInference.
    • Added a new gum::fastVar(std::string) factory function to easily build variable such as in gum::*::fastPrototype(std::string). See Documentation. This syntax is now used in every graphical model with gum::{BayesNet|MarkovNet|etc}::add(std::string) to easily add a new variable in the model.
    • Added a new graphical representation of cliqueGraph : gum::CliqueGraph::mapToDot().

  • pyAgrum

    • gum.ShaferShenoyMNInference is now parallelized too.
    • Wrapped the new fast syntax for created variable in functions gum.{BayesNet|etc}.add(str). See Documentation.
    • Introduced a better type induction for discrete variables when created by gum.skbn.BNDiscretizer. gum.IntegerVariable and gum.RangeVariable are now induced from the data.
    • Added gum.JunctionTree.map() (in notebook) to show a condensed map of a the junction tree.
    • Still worked on documentation (One must imagine Sisyphus happy).

Changelog for 1.0.0

The aGrUM's team is very proud to announce the release of aGrUM/pyAgrum 1.0.0!

This long journey sometimes strewn with pitfalls, but which often brought great satisfactions, began with this first commit (subversion):

add8dbff5 | 13 years ago | phw | test?

aGrUM/pyAgrum is now definitely no longer a test :-) Parallelization of inference in Bayesian networks is the important feature that allows us to symbolically assert a level of quality of aGrUM. The 1.0.0 release shows that we also believe the API will remain stable in a medium term.

pyAgrum continues its path as a wrapper of aGrUM but with more and more specific features. Some of them are intended to be implemented in C++ and eventually integrated in aGrUM (causality, etc.), others will certainly remain specificities of pyAgrum.

In the near future, for aGrUM and pyAgrum, we expect many more new features and releases ! Stay tuned!

  • aGrUM

    • VariableElimination, ShaferShenoy and LazyPropagation are now parallelized.
    • Better use of d-separation in ShaferShenoy and LazyPropagation.
    • Better initialization/registrations used Meyers singleton.
    • Better 0-dimensional Potential.
    • new gum::IBayesNet::check() to test if the BN is completely and well-defined.

  • pyAgrum

    • Better documentation and argument order for gum.BNLearner.__init__.
    • Better numpystyle for docstrings.
    • Better tests and notebooks.
    • Better signature for gum.Potential.arg{max|min}.
    • New gum.IBayesNet.check() to test if the BN is completely and well-defined.
    • More consistent API : with_labels default is True everywhere.

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 multithreaded facility.
    • Made a fully new architecture for scheduled 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". Meaned 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 missed used 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.

Changelog for 0.22.8

  • aGrUM

    • Remove clang 13 warnings (thanks to Julien Schueller).
    • Add the ability to chain many methods in gum::learning::BNLearner.

  • pyAgrum

    • Reorganization of notebooks.

    • Add pyAgrum.lib.image.export[Inference] with no filename in order to export a numpy.array ready for matplotlib.pyplot.imshow() (thanks to Mads Lindskou).

    • Improve pyAgrum.bn_vs_bn graphical diff between BNs with many style configurations in pyAgrum.config.

    • Cosmetic improvements for pyAgrum.lib.bn2scores.

    • Add an access to pyAgrum.lib.explain.getInformationGraph in order to export the image (thanks to Dennis Mac R).

    • pyAgrum.BNLearner

      • Add the ability to chain many methods in pyAgrum.BNLearner.
      • Add an easy-to-use pyAgrum.BNLearner.fitParameters to learn in-place the parameters of the argument.
      • Add pyAgrum.BNLearner.learnEssentialGraph.

    • pyAgrum.lib.notebook

      • Add _repr_html_ to pydot.Dot.
      • Add an easier-to-use pyAgrum.lib.notebook.flow.add(obj,caption).
      • Add the ability to chain many methods in pyAgrum.lib.notebook.flow.
      • Move from title to caption as argument for pyAgrum.lib.notebook.flow's methods.
      • Improve graphical representation for pyAgrum.lib.notebook.showInformation().
      • Add a new argument ncols to pyAgrum.lib.notebook.sideBySide in order to use it as a grid (and not only as a line).

Changelog for 0.22.7

Mostly several quick fixes.

  • aGrUM

    • Small bugfix on Markov random field with only one factor.
    • Better behavior for ordered nodes when gum::MarkovNetwork::addFactor.
    • Better behavior for gum::ShaferShenoyMNInference::jointPosterior when a node of the joint is hard-observed.

  • pyAgrum

    • Update and fix documentation (mostly for readthedocs and typehints).
    • Fix and improve ticks when plotted distribution with large domainSize.
    • Better behavior for ordered nodes when pyAgrum.MarkovNetwork.addFactor.
    • Better behavior for pyAgrum.ShaferShenoyMNInference.jointPosterior when a node of the joint is hard-observed.

Changelog for 0.22.6

  • aGrUM

    • aGrum is now compiled used c++20 (updated cmake).
    • Fix 'distribute' phase in some cases for gum::ShaferShenoyLIMIDInference (Thanks to Michael Kontoulis).
    • Remove a lot of allocators as template parameters.
    • act: Improve architecture.
    • act: Use of argparse instead of the obsolete optparse.
    • Worked on CI : Worked on CI : removed some obsolete platform (MSVC 15), added some new (MSVC 22)
    • use std::mt19937 instead of rand() (thanks to Julien Schueller).

  • pyAgrum

    • Drop support for python 3.7 (following NEP29).
    • Drop support for 32bits architectures (following numpy).
    • Use of pydot instead of pydotplus.
    • Several typos in code and documentation of gum.lib (thanks to Benjamin Datko).
    • Improve notebooks (thanks to Clara Charon).

Changelog for 0.22.5

This release mainly deals with provided a workaround to a pydotplus/pyparsed bug affected several graph visualizations in pyAgrum.

  • pyAgrum

    • workaround to the pydotplus/pyparsed bug : "added an empty node".
    • remove the use of deprecated distutils in act.
    • (approximate) fix labels in x-axis for lined-histogram.

Changelog for 0.22.4

This release aims to provide a number of new configurations for aGrUM and pyAgrum:

  • aGrUM

    • builder for windows used msvc2022 (new option for act)
    • aGrUM is now compiled used c++17

  • pyAgrum

    • package for MacOS arm64 (M1)
    • package for python 3.10 (limited to 64 bits)
    • customization title of plots in pyAgrum.lib.dynamicBN

Changelog for 0.22.3

  • pyAgrum

    • new option in gum.config for percent or just number in histograms.
    • better wrappers for gum.Instantiation.variablesSequence() and pyAgrum.MixedGraph.adjacents(id).
    • better wrappers for pyAgrum.MixedGraph.{mixedOrientedPath|mixedUnorientedPath}.
    • refreshed all requirements.txt.
    • better visualisation for pyagrum.lib.DynamicBN.
    • improvements in documentations.
    • aarch64 version for linux package

Changelog for 0.22.2

  • aGrUM

    • Minor change in CSVParser (BNLearner etc.): if a token is surrounded by quote characters, those are not part of the very token anymore.

  • pyAgrum

Changelog for 0.22.1

  • aGrUM

    • fix issue #69 (no more final destructor),
    • update Coco/R parsers (notations and explicit casts),
    • better gum::SyntaxError (access to filename),
    • remove many redundant ';' in testsuites,
    • better option --stats for act,
    • gum.[model].fastPrototype now accepts multiline specifications.

  • pyAgrum

    • better pyAgrum.SyntaxError treated as python's SyntaxError,
    • much better annotations for types declaration in python codes,
    • improved documentation,
    • update pyAgrum.causal,
    • gum.fast[Model] now accepts multiline specifications,
    • gum.DiscreteVariable are now hashable.

Changelog for 0.22.0

As planned, 0.22.0 is the first version of pyAgrum that does not support python>3.6 (included 2.7).

  • aGrUM

    • fix issue #27
    • (act) remove (hopefully) all the codes to support both python. In particular, there is no more options for act to choose the targeted version of python.
    • (ci/deploy) removed 2.7 tests and deploy (thanks to @Aspard)
    • better and customized type induction when learning Bayesian networks from CSV.
    • new constructor for gum::learning::BNLearner to activate/deactivate the type induction when readed a csv file. (thanks to @gonzalesc)

  • pyAgrum

    • remove a large part of the codes dedicated to python2 in the wrapper (wrapper/python/generated-files2) and in pyAgrum.lib. To be cont'd.
    • many improvements due to linter (pylint especially) in pyAgrum.lib.
    • graphical improvement in pyAgrum.lib.bn2roc thanks to Clara Charon.
    • new constructor for gum.BNLearner to activate/deactivate the type induction when readed a csv file.

Changelog for 0.21.0

Contrary to what was said in the 0.20.0 changelog, we decided to remove support for python 2.7 before the 1.0 release.

This tag (0.21.0) is the last version that supports python 2.7. We are already worked and will deliver a 0.22.0 tag as soon as possible, which will be dedicated to this move and will then be the first tag without python 2.7 support.

The next tag (0.22.0) will be the (new) last minor version before the release of agrum/pyAgrum 1.0.0 (:fist: :smirk: ).

  • aGrUM

    • New type for discrete variable (gum::IntegerVariable) which represents a set of non-consecutive integers.
    • New syntax for gum::IntegerVariable in gum::*::fastPrototype : a{-3|0|3}.
    • Change in syntax for gum::MarkovNetwork::fastPrototype : the link are represented by -- instead of -.
    • New gum::BNLearner::state() which gives a view of the activated options in the learner (scores, priors, algorithms, constraints, etc.).
    • New gum::BNLearner::toString() which gives a stred representation of gum::BNLearner::state().
    • Add a new CI for last gcc (g++11 for now).
    • Code optimizations for hash function for small-sized values.
    • Better hierarchy for exceptions.
    • MLEstimator should lead to an error when divided by 0.

  • pyAgrum

    • New type for discrete variable (pyAgrum.IntegerVariable) which represents a set of non-consecutive integers.
    • New syntax for pyAgrum.IntegerVariable in pyAgrum.fast* : a{-3|0|3}.
    • Change in syntax for pyAgrum.fastMN : the links are represented by -- instead of -.
    • New pyAgrum.BNLearner.state() which gives a view of the activated options in the learner (scores, priors, algorithms, constraints, etc.).
    • New pyAgrum.BNLearner.__str__() which gives a stred representation of gum::BNLearner::state().
    • Documentations and notebooks updated w.r.t. this new features.
    • Added ShapValues for BN in pyAgrum.lib.explain (see notebook).
    • Added pyAgrum.lib.explain.independenceListForPairs().
    • Other improvements in pyAgrum.lib.explain and the corresponded notebook and documentations.
    • Updated notebooks for classifiers.
    • Better hierarchy for exceptions.
    • Removed unnecessary and obsolete codes by deleted pyAgrum.lib._utils.
    • 'Terminology clash' between 'Laplace's adjustment' and 'Smoothing' : use more generic 'Smoothing' everywhere now.
    • MLEstimator should lead to an error when divided by 0.

Changelog for 0.20.3

  • aGrUM

    • Refactoring/fixed MIIC and better heuristic for orientations for constraint-based learning algorithms.
    • Updated guidelines and new convention for private methods/attributes.
    • Changed behaviour of gum::MixedGraph::mixed{Oriented|Unoriented}Path : no misuse of exception when no path is found.

  • pyAgrum

    • Refactored MIIC and better heuristic for orientations for constraint-based learning algorithms.
    • Changed behaviour of pyAgrum.MixedGraph.mixed{Oriented|Unoriented}Path : no misuse of exception when no path is found.
    • Updated new pyAgrum.Potential's methods and documentation.
    • New tool for layout in notebooks : pyAgrum.notebook.flow.
    • New gum.config options for background colors in CPT : potential_color_0 and potential_color_1.
    • New module pyAgrum.lib.explain.

Changelog for 0.20.2

  • aGrUM

    • Add a check on parameters when builded a gum::credal::CredalNet from BNmin and BNmax: 'are Pmin<=Pmax' ?".
    • Fix a bug, and a visualisation of results on decision nodes with deterministic optimal strategy in gum::InfluenceDiagram.

  • pyAgrum

    • Add a check on parameters when builded a gum.CredalNet from BNmin and BNmax: 'are Pmin<=Pmax' ?".
    • Fix a bug and add a better visualisation of results on decision nodes with deterministic optimal strategy in gum. InfluenceDiagram.
    • Add some options for notebook and influence diagrams in gum.config. Notably, add a gum.notebook. show_inference_time
    • Fixes and typos in notebooks
    • Finally, add a worked version of gum.lib.notebook.exportInference to create pdf from an inference. With gum. lib.notebook.export, it is now possible to export all kind of pyAgrum's graphs into pdf from a notebook.
    • new methods: pyAgrum.Potential.topandas(),pyAgrum.Potential.tolatex()

Changelog for 0.20.1

  • aGrUM

    • Fix an infamous bug: monocycle in DAG (thanks to Guy, GabF and Joanne). This bug did not propagate to graphical models (especially BNs).

  • pyAgrum

    • new site for tutorials.
    • renamed and reorganized many tutorials
    • sync'ed documentation (readthedocs) with the new URLs for notebooks

Changelog for 0.20.0

0.20.0 is the last minor release before 1.0.0.

  • aGrUM

    • Workaround for OMP with MVSC
    • Refreshed doxygen configuration file
    • Graph methods for childrenand parents of sets of nodes.
    • Renamed core/math/math.h to core/math/math_utils.h to avoid clash names and false warnings from linters
    • work on CIs
    • Fix and typos from F.Keidel
    • Improved API of gum::BayesNetFragment (for instance, non-implemented gum::BayesNetFragment::VariableNodeMap).
    • Major changes (and typos and bug fixes) in gum::CN::CredalNetwork's API
    • Fixed minor bugs in inference for gum::CN::CredalNet
    • (internal) re-organized files and folders for Credal Networks
    • (internal) fixed bug in organization of inline/tpl/source files for gum::credal::lp::LpInterface

  • pyAgrum

    • (internal) Better logic and automatic generation for the multiple "requirements.txt".
    • Graph methods for childrenand parents of sets of nodes.
    • pyAgrum.notebook.export and pyAgrum.notebook.exportInference to export as png, pdf(, etc.) PGM and inference in PGM
    • Fix several tests
    • Fix and typos from F.Keidel
    • Added gum.Instantiation.addVarsFromModel and allowed chained gum.Instantiation.add().
    • Fixed some broken links in documentation.
    • Updated gum.skbn for non-binary classifier (see notebooks).
    • Improved notebooks for classifiers.
    • Major changes in gum.CredalNetwork's API
    • Specific visualisation for credal networks
    • Graphical visualisation of inference with credal networks (Visual Credal networks)
    • Added some example for credal networks in notebooks

Changelog for 0.19.3

  • pyAgrum
    • missed graphical (not correctly wrapped) methods in gum::InfluenceDiagram
    • fix falsely raised exception leaded to incomplete generation of documentation and wheels.
    • pyAgrum.lib.ipython improved.
    • pyAgrum's documentation refreshed a bit.
    • pyAgrum.skbn improved.
    • several typos in notebooks and testsuites.

Changelog for 0.19.2

  • aGrUM
    • bugfix for EssentialGraph (thanks to M.Lasserre).

Changelog for 0.19.1

  • aGrUM
    • bugfix for InfluenceDiagram with all-negative utilities (thanks to B.Enderle).
    • [internal] typos and reorganization for act's modules.

Changelog for 0.19.0

Mainly : important changes for Influence Diagram (aGrUM and pyAgrum) and for BayesNet classifiers compliant to scikit-learn's API (pyAgrum).

  • aGrUM

    • new and better inference for Influence Diagrams and LIMIDs (gum::ShafeShenoyLIMIDInference).
    • new builder for Influence Diagram gum::InfluenceDiagram::fastPrototype.
    • bugfixes.

  • pyAgrum

    • wrapper and notebook functions for new inference and new methods for influence diagram.
    • new module skbn for BayesNet classifier compatible with sklearn (classification and discretization) with optimized predict method and specific structural learning for fit (Naïve Bayes, TAN, Chow-Liu tree, and others learning aGrUM's algorithms). Several discretization methods are implemented.
    • minor graphical improvements.
    • remove old deprecated class/method (since pyAgrum 0.12.0).
    • Improved documentation (readthebook).
    • bugfixes.

Changelog for 0.18.2

Mainly bugfixes and internal improvements.

  • aGrUM

    • bugs fixed for gum::MarkovNet and gum::ShaferShenoyMNInference.
    • typo in the name of odbc library for mac.

  • pyAgrum

    • packages for python 3.9 (except win32).
    • better error message for DuplicateElement in operations between gum.Potential.
    • [internal] improvements for builded wheels.
    • deprecated PyEval_CallObject.
    • [internal] improvements for pyAgrum's tests.

Changelog for 0.18.1

  • aGrUM

    • Direct access to gum::<graphicalmodel>::isIndependent(X,Y,Z).
    • Direct access to direct access to ancestors and descendants().
    • Update API with node names for putFirst/reorganize/VI/I.

  • pyAgrum

    • bug fixed on wrapped {Edge|Arc}Part (thanks to Arthur Esquerre-Pourtère).
    • bug fixed for some UTF8 names.
    • Direct access to gum::<graphicalmodel>::isIndependent(X,Y,Z).
    • Direct access to ancestors() and descendants().
    • Update API with node names for putFirst/reorganize/VI/I.

Changelog for 0.18.0

  • aGrUM

    • MarkovNet's model, UAI file format and inference (incremental ShaferShenoy).
    • Bug fix in MIIC learning algorithm.
    • Bug fix in gum::GammaLog2 approximations for very small values.
    • Updated and enhanced gum::GraphicalBNComparator.
    • Enhanced API for gum::MixedGraph (build a MixedGraph from other graphs).
    • API changes for gum::MultiDimAggregator (consistant behavior without parent).
    • new gum::MultidimmAggegator : Sum.
    • Minor API changes for gum::Potential (normalizeAsCPT,minNonZero,maxNonOne).
    • Minor API changes for graphical models (gum::DAGModel and gum::UGModels).
    • [internal] adopted more classical convention for named pr{otected|ivate} methods and attributes.
    • [internal] Updated sources for MVSC 2019.

  • pyAgrum

    • MarkovNet's model, UAI file format and inference (incremental ShaferShenoy).
    • Bug fix in MIIC learning algorithm.
    • Updated and enhanced pyAgrum.GraphicalBNComparator.
    • Enhanced API for pyAgrum.MixedGraph (build a MixedGraph from other graphs).
    • API changes for pyAgrum::MultiDimAggregator (consistent behavior without parent).
    • new pyAgrum::MultidimmAggegator : Sum.
    • Minor API changes for pyAgrum::Potential (normalizeAsCPT,minNonZero,maxNonOne).
    • In gum.lib.bn2roc : bugfix for ROC, access to significant_digit for predict, add Precision-Recall graph.

Changelog for 0.17.3

  • aGrUM

    • improved version of MIIC's learning algorithm.
    • add access to pseudo count with gum::BNLearner::pseudoCount.
    • fix a bug in inference with gum::InfluenceDiagram.
    • improved API for Influence Diagram : accessor with variable names (instead of only NodeId).
    • VS2019's compiler is now supported by act/CMakeLists.txt.
    • reorganized resources for testsuite.

  • pyAgrum

    • improved version of MIIC's learning algorithm.
    • add access to pseudo count with gum.BNLearner.pseudoCount.
    • fix a bug in inference with gum.InfluenceDiagram.
    • improved API for Influence Diagram : accessor with variable names (instead of only NodeId).

Changelog for 0.17.2

  • aGrUM

    • fix a bug in graphChangeGeneratorOnSubDiGraph (thanks @yurivict for the issue).
    • fix a bug in LazyPropagation due to a (rare) improper optimization.
    • improve projection and combination codes for MultiDim hierarchy.
    • reorganization of source codes and internal structure.
    • refreshed a bit the c++ examples.
    • preliminary works on undirected graphical models.

  • pyAgrum

    • forgotten description for pip packages.
    • typo for special char (':' for instance) with pydotplus.

Changelog for 0.17.1

(really) minor patch

  • aGrUM

    • O3PRMBNReader can now read a BN from an o3prm file with a unique class even it the name of the class is not the name of the file.

  • pyAgrum

    • internal changes for wheel generations.
    • updated description for packages.
    • optimized loops and inference for ROC and classifiers.
    • minor improvements for pyAgrum's tests (logged and restrictions for python2).

Changelog for 0.17.0

  • pyAgrum

    • Update requirements
    • Improve classifier.py
    • Documentation improvements
    • Remove '?' from names in some resources files concerned Asia
    • Updated API for pyAgrum.Instantiation (accessor used name of variables)
    • Important internal updates for the relation between numpy.array and pyAgrum.Potential (with a significant speed-up)
    • Add operators between pyAgrum.Potential and numbers
    • Fix a bug when used pydotplus with size=None
    • Fix minor bugs in pyAgrum.fastBN and in pyAgrum.O3prmBNWriter/Reader
    • Add pyAgrum.Potential.log2() method and pyAgrum.log2(pyAgrum.Potential) function
    • Add pyAgrum.BayesNet.clear() method

  • aGrUM

    • Updated API for gum::Instantiation (accessor used name of variables)
    • Add operators between gum::Potential<GUM_SCALAR> and GUM_SCALAR
    • Fix minor bugs in gum::fastPrototype and in gum::O3prmBNWriter/Reader
    • Add gum::multiDimDecorator::erase(std::string& name) (mainly used as gum::Potential::erase(std::string& name))
    • Add gum::BayesNet<GUM_SCALAR>::clear() method

Changelog for 0.16.4

  • pyAgrum

    • first version of pyAgrum.lib.classifier providing a class pyAgrum.BNClassifier wrapping a BN as a classifier with a scikitlearn-like API.
    • Fix bug in pyAgrum.lib.notebook.showPotential with explicit digit param
    • Add a pyAgrum.Potential.loopIn() to iterate inside a Potential
    • Enhanced API for pyAgrum.InfluenceDiagram
    • Documentation improvements
    • remove package for python 3.4 and 3.5 (following NEP29). But 2.7 is still maintained.
    • add package for python 3.8

  • aGrUM

    • still worked on CI
    • Enhanced API for gum::InfluenceDiagram

Changelog for 0.16.3

  • pyAgrum
    • wrapper for the class gum::BayesNetFragment
    • typos in dot methods for Influence Diagrams and Causal Models

Changelog for 0.16.2

  • pyAgrum
    • remove the use of 'f-strings' in pyAgrum.lib.notebook.py

Changelog for 0.16.1

  • aGrUM

    • improve the syntax for BN specification used gum::fastPrototype
    • improve several CMakeFiles.txt and doxygen documentation
    • add CI for python 2.7
    • refresh gum::BayesNet::toString()
    • API change : gum::MarkovBlanket()::{mb()->gum::MarkovBlanket::dag()}

  • pyAgrum

    • pyAgrum.config object for customization (see notebook 08-configForPyAgrum)
    • improved the syntax for BN specification used pyAgrum.fastBN
    • improved pyAgrum's documentation
    • add pyAgrum.causal.counterfactual and pyAgrum.causal.counterfactualModel (see notebook 55-Causality_Counterfactual)
    • improve plots for histograms (particularly : now svg image by default)
    • add pyAgrum.Potential.fullWithFunction()
    • add pyAgrum.{any graph-like}.connectedComponents()
    • add pyAgrum.BayesNet.parents() and pyAgrum.BayesNet.children() with the name of node.
    • pyAgrum.MarkovBlanket(bn,var,level) build now a level-th order Markov Blanket (by default level=1)
    • add access to constants pyAgrum.VarType_{Discretized|Labelized|Range} when tested pyAgrmu.{any discrete variable}.varType()
    • API change : pyAgrum.MarkovBlanket().{mb()->dag()}
    • add pyAgrum.lib.bn_vs_bn.GraphicalBNComparator.hamming()

Changelog for 0.16.0

  • aGrUM

    • bug fixed in gum::BNLearner::learnParameters() in some cases with ML estimation (without priors). Better bootstrap for EM.
    • bug fixed for variables order in the gum::Potential generated by gum::BayesNet::reverseArc()
    • new gum::DiGraph::hasDirectedCycle
    • new methods in gum::Potential for random generations and pertubations

  • pyAgrum

    • improvements in pyAgrum.lib.causality (direct access to backdoor and frontdoor, typos in doCalculus, etc.)
    • pyAgrum.lib.notebook's visualisations of graph can be adapted for dark or light themes (in jupyterlab for instance) with forDarkTheme()and forLightTheme()
    • pyAgrum.lib.notebook.{show|get}Inference() can now have colored arcs (see 06-colouringBNs.ipynb in the notebooks)
    • improvements on the documentation framework (w.r.t. readthdocs)
    • bug fixed for variables order in the pyAgrum.Potential generated by pyAgrum.BayesNet.reverseArc()
    • new pyAgrum.DiGraph.hasDirectedCycle
    • new methods in pyAgrum.Potential for random generations and pertubations
    • better error messages when added arcs in a Bayesian Network
    • API change for joint targets in exact inference : only as set of node ids or names of variable

Changelog for 0.15.2

  • aGrUM
    • fix a (rare) bug in counters for contingency tables for chi2 and G2 (bug found by Bastien Chassagnol)
  • pyAgrum
    • fix fscore in gumlib/bn_vs_bn.py
    • API : Wherever a list of strings is the type for an argument, a single stred "x" can be used instead of ["x"]
    • workaround for weird bug when displayed matplotlib-generated svg in notebooks (for gnb.showInference)
    • fix AUC computation in gumlib/bn2roc.py

Changelog for 0.15.1

  • aGrUM
    • add forgotten addPossibleEdge constraint for LocalSearchWithTabuList learning algorithm in BNLearner.
    • bug fix in exact inference leaded to an erroneous exception.
    • Better iterated random test for sampled inference
  • pyAgrum
    • update path for new notebooks location for CI and documentation
    • nightly build with pip : use now pip install pyAgrum-nightly
    • API change in pyAgrum.lib.dynamicBN : plotFollowUnrolled(lovars, dbn, T, evs)
    • add forgotten addPossibleEdge constraint for LocalSearchWithTabuList learning algorithm in BNLearner.
    • bug fix in exact inference leaded to an erroneous exception.
  • internal
    • new values for act -j option (number of jobs for compilation) : all, except1, half, halfexcept1.
    • several bugfixes in CI

Changelog for 0.15.0

  • LGPLV3 new LGPL3 licence for aGrUM/pyAgrum
  • aGrUM
    • bug fix with openMP in BNLearner::setDatabaseWeight
    • new BNLearner::recordWeight() and BNLearner::databaseWeight()
    • new BNLearner::setRecordWeight()
  • pyAgrum
    • nightly builds with pip : pip install pyAgrum-nightly
    • bug fix with openMP in BNLearner::setDatabaseWeight
    • new BNLearner.recordWeight() and BNLearner.databaseWeight()
    • new BNLearner.setRecordWeight()
    • bug fix and minor API changes in pyAgrum.causality

Changelog for 0.14.3

  • aGrUM
    • new constraint for structural learning : possibleEdge forces the tested edges to be taken from the sets of possibleEdges
    • new methods BNLearner::addPossibleEdge(const gum::Edge&) and BNLearner::setPossibleSkeleton(const gum::UndiGraph&)
    • Fix a bug in gum::IndepTestG2
    • Access to scores from BNLearner : BNLearner::G2()
  • pyAgrum
    • new methods BNLearner.addPossibleEdge(x,y) and BNLearner.setPossibleSkeleton(undigraph)
    • fix a bug in causality's identification algorithm
    • Access to scores from BNLearner : BNLearner.G2()
    • add tests and fix typos in notebooks

Changelog for 0.14.2

  • aGrUM
    • bug fixes in learning (3off2/miic and learnMixedGraph)
    • removed redundant gum::BNLearner::setAprioriWeight
  • pyAgrum
    • pyAgrum.lib.notebook.showInference can now use svg format
    • use of the svg format by default for graphs and drawings in pyAgrum.lib.notebook
    • refreshed notebooks
    • removed redundant pyAgrum.BNLearner.setAprioriWeight
    • added forgotten wrapper for pyAgrum.BNLearner.useBDeuPrior
    • changed the representation of causal model (special node for latent variable)
    • extended documentation

Changelog for 0.14.1

  • aGrUM
    • chaintool for compilation with microsfot visual C++ 17 (act --msvc17 and act --msvc17_32)
  • pyAgrum
    • fixed a missed importation of the pyAgrum.causal module in pypi packages
    • updated sphynx version for pyAgrum's ReadTheDoc

Changelog for 0.14.0

  • aGrUM
    • support for mingw64 + bugfix for mingw (act --mingw64)
    • Access to scores from BNLearner : BNLearner::Chi2 and BNLearner::logLikelihood
    • bug fix in KL[...]::bhattacharya
    • add KL[...]::jsd (Jensen-Shannon divergence)
    • renamed gum::[...]]KL classes into gum::[...]distance because they provide access to KL but also to Hellinger, Bhattacharya distances and Jensen-Shanon divergence.
  • pyAgrum
    • pyAgrum.causality (do-calculus and causal identification !)
    • JunctionTreeGenerator (formerly JTGenerator) can now expose the eliminationOrder and can drive the triangulation with a partial order of the nodes.
    • Access to scores from BNLearner : BNLearner::Chi2 and BNLearner::logLikelihood
    • bug fix in pyAgrum.lib.notebook
    • bug fix in KL[...]::bhattacharya
    • add KL[...]::jsd (Jensen-Shannon divergence)
    • renamed pyAgrum.[...]]KL classes into pyAgrum::[...]distance because they provide acces to KL but also to Hellinger, Bhattacharya distances and Jensen-Shanon divergence.
    • fix some scratches in pyAgrum documentation

Changelog for 0.13.6

  • aGrUM
    • Compilation issue for clang4 fixed
    • remove all pre-compiled float instanciations of aGrUM's templates (and significantly reduce the size of all libraries)
    • add the configuration files needed for interactive notebooks on mybinder.org
  • pyAgrum
    • minor changes in notebooks

Changelog for 0.13.5

Changelog for 0.13.4

  • pyAgrum
  • aGrUM
    • parametric EM !!
    • several internal improvements
    • learning: major update of the scores, independence tests and record counters: They can now be used on subsets of databases (e.g., for cross validation), the ids of the nodes need not correspond to indices of columns in the database. The interfaces of these classes have been simplified.
    • learning: all the scores have been speeded-up
    • learning: new score fNML has been introduced
    • learning: Dirichlet prior has been improved: the variables in its database need not be in the same order as those of the learning database
    • learning: all the score-related testunits have been improved
    • learning: the documentations of the scores have been improved
    • learning: the corrected mutual information of 3off2 has been improved
    • BNLearner: now supports cross validation
    • New method : JointTargetedInference::jointMutualInformation for any set of variables in the BN

Changelog for 0.13.3

  • pyAgrum
    • pip : wheels for mac/windows/linux for python 2.7,3.{4-7}
    • anaconda : compilation for maxOS/anaconda64/python3 should be fixed
    • updated tests
    • updated pyAgrum.lib
    • updated posterior histograms for notebooks (added mean/stdev for RangeVariable and DiscretizedVariable)
    • new functions for coloured and graphically compared BNs
    • improved documentation
  • aGrUM
    • fixed bugs for DiscreteVariable with domainSize()<=1 (particularly when added in Potential)
    • improved CMakeFiles.txt
    • improved documentation
    • fixed UAI format for read and write
    • BNLearner.setSliceOrder with list of list of names (and not only with ids)
    • improved error messages
    • fixed learnParameters
    • multi-thread support for learning

Changelog for 0.13.2

  • aGrUM/pyAgrum
    • fixed bugs in Potential::fillWith
    • removed unsafe and ambiguous Potential::fastKL and kept safe Potential::KL

Changelog for 0.13.1

  • aGrUM
    • variable: new methods to set bounds with doubles in ContinuousVariable
    • Changed the code of Instantiation's hash functions to make it compliant with windows mingw implementation
  • TestUnits
    • fixed bug in RawDatabaseTable test unit

Changelog for 0.13.0

  • aGrUM
    • inference: Loopy Belief Propagation (LBP)
    • inference: new approximated inference : Monte-Carlo/Importance/Weighted Sampling + the same used LBP as a Dirichlet prior (Loopy...).
    • learning: new algorithm 3off2 and miic
    • learning: new database handled framework (allows for coped with missed values and with different types of variables)
    • learning: possibility to load data from nanodbc databases (e.g., postgres, sqlite)
    • learning: add a progress Listener/Signaler in BNDatabaseGenerator
    • potential: API extension (findAll,argmax,argmin,fillWith(pot,map))
    • variable: new constructor for LabelizedVariable with labels as vector of stred + posLabel(std::string)
    • variable: new constructor with vector of ticks for gum::DiscretizedVariable
    • graph: API extension (addNodes(n))
    • graph: API change (addNode(id)->addNodeWithId(id))
    • Changes and bug fixe in in BIF and NET writer/reader
  • pyAgrum
    • wheels for python 3.3 and 3.4
    • access to the new learning framework used BNLearner
    • access to the new inference algorithms
    • new methods Instantiation.fromdict and Instantiation.todict
    • DiscreteVariable.toDiscretized/toLabelized/toRange copy the variable instead of gived a (not readonly) reference
  • O3PRM
    • new syntax for types
    • read and write Bayesian Network with O3PRM syntax
  • Documentations
  • act
    • new command guideline for a few easy checks
  • many bug fixes

Changelog for 0.12.0

  • API
    • new class EssentialGraph
    • new class MarkovBlanket
    • improved targets in MarginalTargettedInference
  • pyAgrum
    • update notebooks
    • new swig-based documentation framework
    • transparent background for dot graphs
    • more windows-compliant agrum.lib.bn2csv
  • aGrUM
    • PRM bug fixes
    • improved CI in gitlab
    • improved exception messages in BN learning and O3PRM
    • improved act

Changelog for 0.11.2

  • aGrUM
    • a lot of internal changes for CI in gitlab (especially for future automatic generation of wheels)
    • learning: correct identification of stred labels beginned with digits
    • learning: labels from CSV are now alphabetically sorted
    • fix an issue with sql.h
  • pyAgrum
    • notebooks as tests (now in wrappers/pyAgrum/notebooks)
    • updated requirements
    • some improvements in doc
    • pyagrum.lib.ipython: emulation of 'pyagrum.lib.notebook' for ipython graphical console (within spyder for instance)
    • pyagrum.lib.bn2csv: csv file with labels of variables instead of index (parameter with_labels:boolean)
    • pyagrum.lib.bn2roc: use a csv with labels by default (parameter with_labels:boolean)

Changelog for 0.11.1

  • 2 typos found in pyAgrum.lib.notebook

Changelog for 0.11.0

  • internal
    • worked on continuous integration with gitlab
    • aGrUM/pyAgrum to be compilable with g++-4.8
    • aGrUM/pyAgrum to be compilable with win32
    • pyAgrum wheels generation used act for 'pip' tool
  • aGrUM
    • removed some unused data structure (AVLTree)
    • fixed bug in localSearchWithTabuList learning class
    • Remove wrong parallel estimations for learning (now correct but sequential)
    • worked on docs
    • API change : add BayesNet::minimalCondSet(NodeSet&,NodeSet&) (migration from pyAgrum to aGrUM)
    • API change : add JointTargettedInference::evidenceJointImpact()
  • pyAgrum
    • API changes : pyAgrum.lib.bn2graph (BN2dot, BNinference2dot, proba2histo)
    • API changes : pyAgrum.lib.pretty_print (bn2txt, cpt2txt)
    • API changes : pyAgrum.lib.notebook : uniformed parameters evs (first) and targets (second) order.
    • API changes : pyAgrum.lib.notebook : showEntropy->showInformation
    • updated sphinx help generation
    • fix CNMonteCarloSampling not recognized as ApproximationScheme
    • enhanced showInformation with Mutual Information on arcs
    • API change : added BayesNet.minimalCondSet(set_of_targets,set_of_evs) (as wrapper)
    • API change : added LazyInference.evidenceJointImpact(set_of_targets,set_of_evs)

Changelog for 0.10.4

  • Add new approximated inference : LBP (aGrUM and pyAgrum)
  • Fix bugs in LazyPropagation and ShaferShenoy inference
  • Refresh some codes in learning module
  • Update (and simplify) CMakeLists.txt for new swig 3.0.11
  • Add some project files (included this CHANGELOG.md)
  • Refresh pyAgrum notebooks with matplotlib2

Changelog for 0.10.3

  • Only bug fixes in tests

Changelog for 0.10.2

  • New method for BayesNet : minimalCondSet
  • New method for all inference : evidenceImpact
  • Potential has a (single) value even if no dimension.
  • Bug fix for LazyPropagation
  • Typos for Visual C++ compiler
  • Many internal changes

Changelog for 0.10.1

  • aGrUM
    • Fix GCC compilation
    • ParamEstimator::setMaxThread new method
  • pyAgrum
    • VariableElimination and ShaferShenoy inference
    • new addJointTarget and jointPosterior methods for exact inference
    • pyAgrum.getPosterior now uses VariableElimination
    • Fix pyAgrum.lib.notebook error for python2
    • pyAgrum now linked with static library aGrUM
    • pyAgrum.so (linux) size significantly reduced

Changelog for 0.10.0

  • aGrUM
    • Improvements in inference : New target/evidence-driven incremental inference scheme used relevant reasoned used by Lazy/Shafer-Shenoy/Variable Elimination algorithms. Relevant reasoned leads to a major improvement of the inference (see RelevanceReasoning.html).
  • pyAgrum
    • LazyPropagation API follow the new inference scheme (add/removeTarget, add/remove/chgEvidence)
  • Installers used pip or anaconda.

Changelog for 0.9.3

Tag 0.9.3 has not been properly announced. Still, many changes in this release :

  • Many bug fixes and API glitch/improvement
    • Many internal reorganisations (compilation, test, jenkins, etc.)
    • Many change in the C++ code in order to be more c++11/14
    • Bug fix in learning
    • Many Doxygen improvements
    • Many refactors and bug fix in PRM
  • Improvements
    • dynamic BN in pyAgrum
    • nanodbc support for pyAgrUM
    • O3PRMBNReader in pyAgrum (read a prm to a BN)
    • PRMExplorer in pyAgrum
    • UAI reader/writer for BayesNet
    • Algebra of potentials (operators on Potential)
    • pyAgrum.lib.notebook refactored and simplified
    • updated lrs version for credal networks
  • Windows
    • aGrUM/pyAgrum compilation on windows used Visual Studio 2015

Changelog for 0.9.2

  • aGrUM
    • Improvements in Inference
      • old LazyPropagation renamed JunctionTreeInference,
      • Improved LazyPropagation ~30% faster,
      • Bug fix and other improvements for relevance reasoned features.
    • Improvements for Probabilistic Relational Models
      • model refinements : e.g. parameterized classes, specification of CPTs used formula, etc.
      • bug fixes and other improvements in dedicated inference algorithms,
      • improved and fixed documentations
      • new file format for Bayesian network : o3prmBNReader (readed a BN by grounded a system)
    • learning API still improved
      • BNLearner templatized
      • new feature for BNLearner : used a BN to specicfy variables and their modalities,
      • bug fixes and improvement for parameter learning.
    • other bug fixes and improvements in aGrUM architecture
      • aGrUM g++5.1-ready
      • etc.
  • pyAgrum
    • small bugs fixed and reorganisation

Changelog for 0.9.1

  • aGrUM
    • Improvement in learning algorithms
    • learning from databases with fewer rows than there are processors
    • method to BNLearner to learn parameters from a BN's DAG
    • static lib compilation for aGrUM
    • bug fixes and other improvements
  • pyAgrum
    • Compiled for Python 3 or Python 2 (default is python3, python2 if no python3.). New option for act to choose which python : --python={2|3}.
    • gumLib has moved and changed its name (in the pyAgrum package) : pyAgrum.lib
    • Improved API for learning (changeLabel/parameter learning/ etc.)
    • Improved graphs manipulation
    • bug fixes and other improvements

Changelog for 0.9.0

Aside from many bug fixes and general improvements such as performance optimizations in various areas, some changes are especially noteworthy:

  • Functionality : Structural and parameter learning for Bayesian networks
  • Model : Credal Networks, FMDP used Multi-Valued Decision Diagrams
  • Language : migration to modern C++(11/14)
  • Core : Improvements and optimization of basic data structures in aGrUM/core