BNLearner.h

Go to the documentation of this file.

/****************************************************************************
 *   This file is part of the aGrUM/pyAgrum library.                        *
 *                                                                          *
 *   Copyright (c) 2005-2026 by                                             *
 *       - Pierre-Henri WUILLEMIN(_at_LIP6)                                 *
 *       - Christophe GONZALES(_at_AMU)                                     *
 *                                                                          *
 *   The aGrUM/pyAgrum library is free software; you can redistribute it    *
 *   and/or modify it under the terms of either :                           *
 *                                                                          *
 *    - the GNU Lesser General Public License as published by               *
 *      the Free Software Foundation, either version 3 of the License,      *
 *      or (at your option) any later version,                              *
 *    - the MIT license (MIT),                                              *
 *    - or both in dual license, as here.                                   *
 *                                                                          *
 *   (see https://agrum.gitlab.io/articles/dual-licenses-lgplv3mit.html)    *
 *                                                                          *
 *   This aGrUM/pyAgrum library is distributed in the hope that it will be  *
 *   useful, but WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,          *
 *   INCLUDING BUT NOT LIMITED TO THE WARRANTIES MERCHANTABILITY or FITNESS *
 *   FOR A PARTICULAR PURPOSE  AND NONINFRINGEMENT. IN NO EVENT SHALL THE   *
 *   AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER *
 *   LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,        *
 *   ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR  *
 *   OTHER DEALINGS IN THE SOFTWARE.                                        *
 *                                                                          *
 *   See LICENCES for more details.                                         *
 *                                                                          *
 *   SPDX-FileCopyrightText: Copyright 2005-2026                            *
 *       - Pierre-Henri WUILLEMIN(_at_LIP6)                                 *
 *       - Christophe GONZALES(_at_AMU)                                     *
 *   SPDX-License-Identifier: LGPL-3.0-or-later OR MIT                      *
 *                                                                          *
 *   Contact  : info_at_agrum_dot_org                                       *
 *   homepage : http://agrum.gitlab.io                                      *
 *   gitlab   : https://gitlab.com/agrumery/agrum                           *
 *                                                                          *
 ****************************************************************************/
 

#ifndef GUM_LEARNING_BN_LEARNER_H
#define GUM_LEARNING_BN_LEARNER_H
 
#include <algorithm>
#include <sstream>
#include <vector>
 
#include <agrum/agrum.h>
 
#include <agrum/BN/learning/BNLearnUtils/IBNLearner.h>
#include <agrum/BN/learning/priors/DirichletPriorFromBN.h>
 
namespace gum {
  namespace learning {
    class BNLearnerListener;

    template < GUM_Numeric GUM_SCALAR >
    class BNLearner final: public IBNLearner {
      public:
      // ##########################################################################
      // ##########################################################################


      BNLearner(std::string_view                  filename,
                const std::vector< std::string >& missingSymbols = {"?"},
                const bool                        induceTypes    = true);


      BNLearner(const DatabaseTable& db);

      BNLearner(std::string_view                   filename,
                const gum::BayesNet< GUM_SCALAR >& src,
                const std::vector< std::string >&  missing_symbols = {"?"});

      BNLearner(const BNLearner&);

      BNLearner(BNLearner&&);

      ~BNLearner() override;

 
      // ##########################################################################
      // ##########################################################################

      BNLearner& operator=(const BNLearner&);

      BNLearner& operator=(BNLearner&&) noexcept;


      BayesNet< GUM_SCALAR > learnBN();

      BayesNet< GUM_SCALAR > learnParameters(const DAG& dag, bool takeIntoAccountScore = true);

      BayesNet< GUM_SCALAR > learnParameters(const BayesNet< GUM_SCALAR >& bn,
                                             bool takeIntoAccountScore = true);

      BayesNet< GUM_SCALAR > learnParameters(bool take_into_account_score = true);

      std::string toString() const;

      std::vector< std::tuple< std::string, std::string, std::string > > state() const;

      void copyState(const BNLearner< GUM_SCALAR >& learner);
 
      //=== === add return to certain methods in order to chain command
      BNLearner< GUM_SCALAR >& setInitialDAG(const DAG& dag);

      BNLearner< GUM_SCALAR >& useEM(const double epsilon, const double noise = default_EM_noise);

      BNLearner< GUM_SCALAR >& useEMWithRateCriterion(const double epsilon,
                                                      const double noise = default_EM_noise);

      BNLearner< GUM_SCALAR >& useEMWithDiffCriterion(const double epsilon,
                                                      const double noise = default_EM_noise);

      BNLearner< GUM_SCALAR >& forbidEM();

      BNLearner< GUM_SCALAR >& EMsetEpsilon(const double eps);

      BNLearner< GUM_SCALAR >& EMdisableEpsilon();

      BNLearner< GUM_SCALAR >& EMenableEpsilon();

      BNLearner< GUM_SCALAR >& EMsetMinEpsilonRate(const double rate);

      BNLearner< GUM_SCALAR >& EMdisableMinEpsilonRate();

      BNLearner< GUM_SCALAR >& EMenableMinEpsilonRate();

      BNLearner< GUM_SCALAR >& EMsetMaxIter(const Size max);

      BNLearner< GUM_SCALAR >& EMdisableMaxIter();

      BNLearner< GUM_SCALAR >& EMenableMaxIter();

      BNLearner< GUM_SCALAR >& EMsetMaxTime(const double timeout);

      BNLearner< GUM_SCALAR >& EMdisableMaxTime();

      BNLearner< GUM_SCALAR >& EMenableMaxTime();

      BNLearner< GUM_SCALAR >& EMsetPeriodSize(const Size p);
 
      // for pyagrum support
      using IBNLearner::EMPeriodSize;

      BNLearner< GUM_SCALAR >& EMsetVerbosity(const bool v);
 
      BNLearner< GUM_SCALAR >& useScoreAIC();
 
      BNLearner< GUM_SCALAR >& useScoreBD();
 
      BNLearner< GUM_SCALAR >& useScoreBDeu();
 
      BNLearner< GUM_SCALAR >& useScoreBIC();
 
      BNLearner< GUM_SCALAR >& useScoreK2();
 
      BNLearner< GUM_SCALAR >& useScoreLog2Likelihood();
 
      BNLearner< GUM_SCALAR >& useNoPrior();
 
      BNLearner< GUM_SCALAR >& useBDeuPrior(double weight = 1.0);
 
      BNLearner< GUM_SCALAR >& useSmoothingPrior(double weight = 1);
 
      BNLearner< GUM_SCALAR >& useDirichletPrior(std::string_view filename, double weight = 1);
 
      BNLearner< GUM_SCALAR >& useDirichletPrior(const gum::BayesNet< GUM_SCALAR >& bn,
                                                 double                             weight = 1);
 
      BNLearner< GUM_SCALAR >& useGreedyHillClimbing();
 
      BNLearner< GUM_SCALAR >& useExtendedGreedyHillClimbing();
 
      BNLearner< GUM_SCALAR >& useGreedyThickThinning();
 
      BNLearner< GUM_SCALAR >& setGreedyThickThinningReversals(bool allow);
 
      bool greedyThickThinningReversals() const;
 
      BNLearner< GUM_SCALAR >& useLocalSearchWithTabuList(Size tabu_size   = 100,
                                                          Size nb_decrease = 2);
 
      BNLearner< GUM_SCALAR >& useK2(const Sequence< NodeId >& order);
 
      BNLearner< GUM_SCALAR >& useK2(const std::vector< NodeId >& order);
 
      BNLearner< GUM_SCALAR >& useMIIC();
 
      BNLearner< GUM_SCALAR >& usePC();
 
      BNLearner< GUM_SCALAR >& useFCI();
 
      BNLearner< GUM_SCALAR >& useFCIChi2Test();
 
      BNLearner< GUM_SCALAR >& useFCIG2Test();
 
      BNLearner< GUM_SCALAR >& setFCIAlpha(double alpha);
 
      BNLearner< GUM_SCALAR >& setFCIMaxPathLength(Size max_len);
 
      BNLearner< GUM_SCALAR >& setFCIExhaustiveSepSet(bool exhaustive);
 
      bool fciExhaustiveSepSet() const;
 
      BNLearner< GUM_SCALAR >& useChi2Test();
 
      BNLearner< GUM_SCALAR >& useG2Test();
 
      BNLearner< GUM_SCALAR >& setPCAlpha(double alpha);
 
      BNLearner< GUM_SCALAR >& setPCStable(bool stable);
 
      BNLearner< GUM_SCALAR >& setPCMaxCondSetSize(Size max_k);
 
      BNLearner< GUM_SCALAR >& setPCUnshieldedColliderSorted(bool sorted);
 
      BNLearner< GUM_SCALAR >& useNMLCorrection();
 
      BNLearner< GUM_SCALAR >& useMDLCorrection();
 
      BNLearner< GUM_SCALAR >& useNoCorrection();
 
      BNLearner< GUM_SCALAR >& setMaxIndegree(Size max_indegree);
 
      BNLearner< GUM_SCALAR >& setSliceOrder(const NodeProperty< NodeId >& slice_order);
 
      BNLearner< GUM_SCALAR >&
          setSliceOrder(const std::vector< std::vector< std::string > >& slices);
 
      BNLearner< GUM_SCALAR >& setTotalOrder(const std::vector< std::string >& order);
 
      BNLearner< GUM_SCALAR >& setTotalOrder(const Sequence< NodeId >& order);
 
      BNLearner< GUM_SCALAR >& setForbiddenArcs(const ArcSet& set);
 
      BNLearner< GUM_SCALAR >& addForbiddenArc(const Arc& arc);
 
      BNLearner< GUM_SCALAR >& addForbiddenArc(NodeId tail, NodeId head);
 
      BNLearner< GUM_SCALAR >& addForbiddenArc(std::string_view tail, std::string_view head);
 
      BNLearner< GUM_SCALAR >& eraseForbiddenArc(const Arc& arc);
 
      BNLearner< GUM_SCALAR >& eraseForbiddenArc(NodeId tail, NodeId head);
 
      BNLearner< GUM_SCALAR >& eraseForbiddenArc(std::string_view tail, std::string_view head);
 
      BNLearner< GUM_SCALAR >& addMandatoryArc(const Arc& arc);
 
      BNLearner< GUM_SCALAR >& addMandatoryArc(NodeId tail, NodeId head);
 
      BNLearner< GUM_SCALAR >& addMandatoryArc(std::string_view tail, std::string_view head);
 
      BNLearner< GUM_SCALAR >& eraseMandatoryArc(const Arc& arc);
 
      BNLearner< GUM_SCALAR >& eraseMandatoryArc(NodeId tail, NodeId head);
 
      BNLearner< GUM_SCALAR >& eraseMandatoryArc(std::string_view tail, std::string_view head);
 
      BNLearner< GUM_SCALAR >& addPossibleEdge(const Edge& edge);
 
      BNLearner< GUM_SCALAR >& addPossibleEdge(NodeId tail, NodeId head);
 
      BNLearner< GUM_SCALAR >& addPossibleEdge(std::string_view tail, std::string_view head);
 
      BNLearner< GUM_SCALAR >& erasePossibleEdge(const Edge& edge);
 
      BNLearner< GUM_SCALAR >& erasePossibleEdge(NodeId tail, NodeId head);
 
      BNLearner< GUM_SCALAR >& erasePossibleEdge(std::string_view tail, std::string_view head);
 
      BNLearner< GUM_SCALAR >& setMandatoryArcs(const ArcSet& set);
 
      BNLearner< GUM_SCALAR >& setPossibleEdges(const EdgeSet& set);
 
      BNLearner< GUM_SCALAR >& setPossibleSkeleton(const UndiGraph& skeleton);
 
      BNLearner< GUM_SCALAR >& addNoParentNode(NodeId node);
 
      BNLearner< GUM_SCALAR >& addNoParentNode(std::string_view name);
 
      BNLearner< GUM_SCALAR >& eraseNoParentNode(NodeId node);
 
      BNLearner< GUM_SCALAR >& eraseNoParentNode(std::string_view name);
 
      BNLearner< GUM_SCALAR >& addNoChildrenNode(NodeId node);
 
      BNLearner< GUM_SCALAR >& addNoChildrenNode(std::string_view name);
 
      BNLearner< GUM_SCALAR >& eraseNoChildrenNode(NodeId node);
 
      BNLearner< GUM_SCALAR >& eraseNoChildrenNode(std::string_view name);
 
      BNLearner< GUM_SCALAR >& allowArcAdditions(bool allow);
 
      BNLearner< GUM_SCALAR >& allowArcDeletions(bool allow);
 
      BNLearner< GUM_SCALAR >& allowArcReversals(bool allow);
 
      BNLearner< GUM_SCALAR >& allowArcTriangleDeletions(bool allow);
 
      bool isConstraintBased() const;
 
      bool isScoreBased() const;
 
      protected:
      void createPrior_() override;
 
      private:
      BayesNet< GUM_SCALAR > _prior_bn_;

      NodeProperty< Sequence< std::string > > _labelsFromBN_(std::string_view              filename,
                                                             const BayesNet< GUM_SCALAR >& src);

      void _checkDAGCompatibility_(const DAG& dag);

      BayesNet< GUM_SCALAR > _learnParameters_(const DAG& dag, bool takeIntoAccountScore);

      std::pair< std::shared_ptr< ParamEstimator >, std::shared_ptr< ParamEstimator > >
          _initializeEMParameterLearning_(const DAG& dag, bool takeIntoAccountScore);

      BayesNet< GUM_SCALAR > _learnParametersWithEM_(const DAG& dag, bool takeIntoAccountScore);

      BayesNet< GUM_SCALAR > _learnParametersWithEM_(const BayesNet< GUM_SCALAR >& bn,
                                                     bool takeIntoAccountScore);
    };

    template < GUM_Numeric GUM_SCALAR >
    std::ostream& operator<<(std::ostream& output, const BNLearner< GUM_SCALAR >& learner);
  } /* namespace learning */
} /* namespace gum */

#include <agrum/BN/learning/BNLearner_tpl.h>
 
#endif /* GUM_LEARNING_BN_LEARNER_H */