DirichletPriorFromBN_tpl.h

Go to the documentation of this file.

/****************************************************************************
 *   This file is part of the aGrUM/pyAgrum library.                        *
 *                                                                          *
 *   Copyright (c) 2005-2025 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 2005-2025                            *
 *       - 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                           *
 *                                                                          *
 ****************************************************************************/
#pragma once
 
 
#include <agrum/BN/learning/priors/DirichletPriorFromBN.h>

#ifndef DOXYGEN_SHOULD_SKIP_THIS
 
namespace gum::learning {

  template < typename GUM_SCALAR >
  DirichletPriorFromBN< GUM_SCALAR >::DirichletPriorFromBN(const DatabaseTable& learning_db,
                                                           const BayesNet< GUM_SCALAR >* priorbn) :
      Prior(learning_db), _prior_bn_(priorbn) {
    GUM_CONSTRUCTOR(DirichletPriorFromBN)
  }

  template < typename GUM_SCALAR >
  DirichletPriorFromBN< GUM_SCALAR >::DirichletPriorFromBN(const DirichletPriorFromBN& from) :
      Prior(from), _prior_bn_(from._prior_bn_) {
    GUM_CONS_CPY(DirichletPriorFromBN)
  }

  template < typename GUM_SCALAR >
  DirichletPriorFromBN< GUM_SCALAR >::DirichletPriorFromBN(DirichletPriorFromBN&& from) noexcept :
      Prior(std::move(from)), _prior_bn_(std::move(from._prior_bn_)) {
    GUM_CONS_MOV(DirichletPriorFromBN)
  }

  template < typename GUM_SCALAR >
  DirichletPriorFromBN< GUM_SCALAR >* DirichletPriorFromBN< GUM_SCALAR >::clone() const {
    return new DirichletPriorFromBN(*this);
  }

 
  template < typename GUM_SCALAR >
  DirichletPriorFromBN< GUM_SCALAR >::~DirichletPriorFromBN() {
    GUM_DESTRUCTOR(DirichletPriorFromBN)
  }

  template < typename GUM_SCALAR >
  DirichletPriorFromBN< GUM_SCALAR >&
      DirichletPriorFromBN< GUM_SCALAR >::operator=(const DirichletPriorFromBN& from) {
    if (this != &from) {
      Prior::operator=(from);
      _prior_bn_ = from._prior_bn_;
    }
    return *this;
  }

  template < typename GUM_SCALAR >
  DirichletPriorFromBN< GUM_SCALAR >&
      DirichletPriorFromBN< GUM_SCALAR >::operator=(DirichletPriorFromBN&& from) {
    if (this != &from) {
      Prior::operator=(std::move(from));
      _prior_bn_ = from._prior_bn_;
    }
    return *this;
  }

 
  template < typename GUM_SCALAR >
  INLINE PriorType DirichletPriorFromBN< GUM_SCALAR >::getType() const {
    return PriorType::DirichletPriorType;
  }

 
  template < typename GUM_SCALAR >
  INLINE bool DirichletPriorFromBN< GUM_SCALAR >::isInformative() const {
    return (this->weight_ != 0.0);
  }

  template < typename GUM_SCALAR >
  INLINE void DirichletPriorFromBN< GUM_SCALAR >::setWeight(const double weight) {
    Prior::setWeight(weight);
  }

  template < typename GUM_SCALAR >
  INLINE void
      DirichletPriorFromBN< GUM_SCALAR >::addJointPseudoCount(const IdCondSet&       idset,
                                                              std::vector< double >& counts) {
    if (this->weight_ == 0.0) return;
    const auto [X, Y] = idset.toNodeSets();
 
    gum::Instantiation Ijoint;
    for (auto i = std::size_t(0); i < idset.size(); i++) {
      Ijoint.add(_prior_bn_->variable(idset.ids()[i]));
    }
 
    _addCountsForJoint_(Ijoint, X + Y, counts);
  }

  template < typename GUM_SCALAR >
  INLINE void DirichletPriorFromBN< GUM_SCALAR >::addConditioningPseudoCount(
      const IdCondSet&       idset,
      std::vector< double >& counts) {
    if (this->weight_ == 0.0) return;
    const auto [X, Y] = idset.toNodeSets();
    gum::Instantiation Ijoint;
    for (auto i = idset.nbLHSIds(); i < idset.size(); i++)
      Ijoint.add(_prior_bn_->variable(idset.ids()[i]));
    _addCountsForJoint_(Ijoint, Y, counts);
  }
 
  template < typename GUM_SCALAR >
  void DirichletPriorFromBN< GUM_SCALAR >::_addCountsForJoint_(Instantiation&         Ijoint,
                                                               const NodeSet&         joint,
                                                               std::vector< double >& counts) {
    const auto size = counts.size();
    if (size != Ijoint.domainSize())
      GUM_ERROR(ArgumentError,
                "The size of counts ("
                    << size << ") does not match with the size of the asked prior (" << joint
                    << " : " << Ijoint.domainSize() << ")")
 
    if (joint.empty()) {
      counts[0] += weight_;
      return;
    }
 
    LazyPropagation lazy(_prior_bn_);
 
    Tensor< GUM_SCALAR > p;
    if (joint.size() == 1) {
      const auto& target = *joint.begin();
      lazy.addTarget(target);
      lazy.makeInference();
      p = lazy.posterior(target);
    } else {
      lazy.addJointTarget(joint);
      lazy.makeInference();
      p = lazy.jointPosterior(joint);
    }
    if (weight_ != 1) p.scale(weight_);
 
    Ijoint.setFirst();
    for (auto i = std::size_t(0); i < size; ++i) {
      counts[i] += p[Ijoint];
      Ijoint.inc();
    }
  }
}   // namespace gum::learning
 
#endif /* DOXYGEN_SHOULD_SKIP_THIS */