#include <agrum/FMDP/learning/fmdpLearner.h>

Inheritance diagram for gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >:

Collaboration diagram for gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >:

Public Member Functions
Constructor & destructor.
	FMDPLearner (double learningThreshold, bool actionReward, double similarityThreshold=0.05)
	Default constructor.
	~FMDPLearner ()
	Default destructor.
Initialization
void	initialize (FMDP< double > *fmdp)
	Initializes the learner.
MultiDimFunctionGraph< double > *	_instantiateFunctionGraph_ ()
	Initializes the learner.
MultiDimFunctionGraph< double > *	_instantiateFunctionGraph_ (Int2Type< IMDDILEARNER >)
	Initializes the learner.
MultiDimFunctionGraph< double > *	_instantiateFunctionGraph_ (Int2Type< ITILEARNER >)
	Initializes the learner.
VariableLearnerType *	_instantiateVarLearner_ (MultiDimFunctionGraph< double > target, gum::VariableSet &mainVariables, const DiscreteVariable learnedVar)
	Initializes the learner.
VariableLearnerType *	_instantiateVarLearner_ (MultiDimFunctionGraph< double > target, gum::VariableSet &mainVariables, const DiscreteVariable learnedVar, Int2Type< IMDDILEARNER >)
	Initializes the learner.
VariableLearnerType *	_instantiateVarLearner_ (MultiDimFunctionGraph< double > target, gum::VariableSet &mainVariables, const DiscreteVariable learnedVar, Int2Type< ITILEARNER >)
	Initializes the learner.
RewardLearnerType *	_instantiateRewardLearner_ (MultiDimFunctionGraph< double > *target, gum::VariableSet &mainVariables)
	Initializes the learner.
RewardLearnerType *	_instantiateRewardLearner_ (MultiDimFunctionGraph< double > *target, gum::VariableSet &mainVariables, Int2Type< IMDDILEARNER >)
	Initializes the learner.
RewardLearnerType *	_instantiateRewardLearner_ (MultiDimFunctionGraph< double > *target, gum::VariableSet &mainVariables, Int2Type< ITILEARNER >)
	Initializes the learner.
Incremental methods
bool	addObservation (Idx actionId, const Observation *obs)
	Gives to the learner a new transition.
void	updateFMDP ()
	Starts an update of datastructure in the associated FMDP.

Private Types
using	VariableLearnerType
using	RewardLearnerType
using	VarLearnerTable = HashTable< const DiscreteVariable, VariableLearnerType >

Private Attributes
FMDP< double > *	_fmdp_
	The FMDP to store the learned model.
HashTable< Idx, VarLearnerTable * >	_actionLearners_
bool	_actionReward_
HashTable< Idx, RewardLearnerType * >	_actionRewardLearners_
RewardLearnerType *	_rewardLearner_
const double	_learningThreshold_
const double	_similarityThreshold_

Miscelleanous methods
double	_rmax_
	learnerSize
double	_modaMax_
	learnerSize
Size	size ()
	learnerSize
const IVisitableGraphLearner *	varLearner (Idx actionId, const DiscreteVariable *var) const
	extractCount
virtual double	rMax () const
	learnerSize
virtual double	modaMax () const
	learnerSize

Detailed Description

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>
class gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >

Definition at line 76 of file fmdpLearner.h.

Member Typedef Documentation

◆ RewardLearnerType

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

using gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::RewardLearnerType

private

Initial value:

typename LearnerSelect< LearnerSelection,
                                                     IMDDI< RewardAttributeSelection, true >,
                                                     ITI< RewardAttributeSelection, true > >::type

Definition at line 82 of file fmdpLearner.h.

◆ VariableLearnerType

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

using gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::VariableLearnerType

private

Initial value:

 
typename LearnerSelect< LearnerSelection,
                        IMDDI< VariableAttributeSelection, false >,
                        ITI< VariableAttributeSelection, false > >::type

Definition at line 77 of file fmdpLearner.h.

◆ VarLearnerTable

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

using gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::VarLearnerTable = HashTable< const DiscreteVariable*, VariableLearnerType* >

private

Definition at line 86 of file fmdpLearner.h.

Constructor & Destructor Documentation

◆ FMDPLearner()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::FMDPLearner	(	double	learningThreshold,
		bool	actionReward,
		double	similarityThreshold = 0.05 )

Default constructor.

Definition at line 67 of file fmdpLearner_tpl.h.

                                                           :
      _actionReward_(actionReward), _learningThreshold_(lT), _similarityThreshold_(sT) {
    GUM_CONSTRUCTOR(FMDPLearner);
    _rewardLearner_ = nullptr;
  }

References FMDPLearner(), _actionReward_, _learningThreshold_, _rewardLearner_, and _similarityThreshold_.

Referenced by FMDPLearner(), and ~FMDPLearner().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ ~FMDPLearner()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::~FMDPLearner ( )

Default destructor.

Definition at line 80 of file fmdpLearner_tpl.h.

                     {
    for (auto actionIter = _actionLearners_.beginSafe(); actionIter != _actionLearners_.endSafe();
         ++actionIter) {
      for (auto learnerIter = actionIter.val()->beginSafe();
           learnerIter != actionIter.val()->endSafe();
           ++learnerIter)
        delete learnerIter.val();
      delete actionIter.val();
      if (_actionRewardLearners_.exists(actionIter.key()))
        delete _actionRewardLearners_[actionIter.key()];
    }
 
    if (_rewardLearner_) delete _rewardLearner_;
 
    GUM_DESTRUCTOR(FMDPLearner);
  }

References FMDPLearner(), _actionLearners_, _actionRewardLearners_, and _rewardLearner_.

Here is the call graph for this function:

Member Function Documentation

◆ _instantiateFunctionGraph_() [1/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

MultiDimFunctionGraph< double > * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateFunctionGraph_ ( )

inline

Initializes the learner.

Definition at line 120 of file fmdpLearner.h.

                                                                  {
      return _instantiateFunctionGraph_(Int2Type< LearnerSelection >());
    }

References _instantiateFunctionGraph_().

Referenced by _instantiateFunctionGraph_(), and initialize().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ _instantiateFunctionGraph_() [2/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

MultiDimFunctionGraph< double > * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateFunctionGraph_ ( Int2Type< IMDDILEARNER > )

inline

Initializes the learner.

Definition at line 124 of file fmdpLearner.h.

                                                                                          {
      return MultiDimFunctionGraph< double,
                                    ExactTerminalNodePolicy >::getReducedAndOrderedInstance();
    }

◆ _instantiateFunctionGraph_() [3/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

MultiDimFunctionGraph< double > * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateFunctionGraph_ ( Int2Type< ITILEARNER > )

inline

Initializes the learner.

Definition at line 129 of file fmdpLearner.h.

                                                                                        {
      return MultiDimFunctionGraph< double, ExactTerminalNodePolicy >::getTreeInstance();
    }

References gum::MultiDimFunctionGraph< GUM_SCALAR, TerminalNodePolicy >::getTreeInstance().

Here is the call graph for this function:

◆ _instantiateRewardLearner_() [1/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

RewardLearnerType * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateRewardLearner_	(	MultiDimFunctionGraph< double > *	target,
		gum::VariableSet &	mainVariables )

inline

Initializes the learner.

Definition at line 166 of file fmdpLearner.h.

                                                                                                {
      return _instantiateRewardLearner_(target, mainVariables, Int2Type< LearnerSelection >());
    }

References _instantiateRewardLearner_().

Referenced by _instantiateRewardLearner_(), and initialize().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ _instantiateRewardLearner_() [2/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

RewardLearnerType * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateRewardLearner_	(	MultiDimFunctionGraph< double > *	target,
		gum::VariableSet &	mainVariables,
		Int2Type< IMDDILEARNER >	)

inline

Initializes the learner.

Definition at line 171 of file fmdpLearner.h.

                                                                            {
      return new RewardLearnerType(target,
                                   _learningThreshold_,
                                   _similarityThreshold_,
                                   mainVariables);
    }

References _learningThreshold_, and _similarityThreshold_.

◆ _instantiateRewardLearner_() [3/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

RewardLearnerType * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateRewardLearner_	(	MultiDimFunctionGraph< double > *	target,
		gum::VariableSet &	mainVariables,
		Int2Type< ITILEARNER >	)

inline

Initializes the learner.

Definition at line 180 of file fmdpLearner.h.

                                                                          {
      return new RewardLearnerType(target, _learningThreshold_, mainVariables);
    }

References _learningThreshold_.

◆ _instantiateVarLearner_() [1/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

VariableLearnerType * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateVarLearner_	(	MultiDimFunctionGraph< double > *	target,
		gum::VariableSet &	mainVariables,
		const DiscreteVariable *	learnedVar )

inline

Initializes the learner.

Definition at line 136 of file fmdpLearner.h.

                                                                                              {
      return _instantiateVarLearner_(target,
                                     mainVariables,
                                     learnedVar,
                                     Int2Type< LearnerSelection >());
    }

References _instantiateVarLearner_().

Referenced by _instantiateVarLearner_(), and initialize().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ _instantiateVarLearner_() [2/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

VariableLearnerType * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateVarLearner_	(	MultiDimFunctionGraph< double > *	target,
		gum::VariableSet &	mainVariables,
		const DiscreteVariable *	learnedVar,
		Int2Type< IMDDILEARNER >	)

inline

Initializes the learner.

Definition at line 145 of file fmdpLearner.h.

                                                                           {
      return new VariableLearnerType(target,
                                     _learningThreshold_,
                                     _similarityThreshold_,
                                     mainVariables,
                                     learnedVar);
    }

References _learningThreshold_, and _similarityThreshold_.

◆ _instantiateVarLearner_() [3/3]

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

VariableLearnerType * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_instantiateVarLearner_	(	MultiDimFunctionGraph< double > *	target,
		gum::VariableSet &	mainVariables,
		const DiscreteVariable *	learnedVar,
		Int2Type< ITILEARNER >	)

inline

Initializes the learner.

Definition at line 156 of file fmdpLearner.h.

                                                                         {
      return new VariableLearnerType(target, _learningThreshold_, mainVariables, learnedVar);
    }

References _learningThreshold_.

◆ addObservation()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

bool gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::addObservation	(	Idx	actionId,
		const Observation *	obs )

virtual

Gives to the learner a new transition.

Parameters

actionId	: the action on which the transition was made
obs	: the observed transition

Returns: true if learning this transition implies structural changes (can trigger a new planning)

Implements gum::ILearningStrategy.

Definition at line 160 of file fmdpLearner_tpl.h.

                                                              {
    for (SequenceIteratorSafe< const DiscreteVariable* > varIter = _fmdp_->beginVariables();
         varIter != _fmdp_->endVariables();
         ++varIter) {
      _actionLearners_[actionId]->getWithDefault(*varIter, nullptr)->addObservation(newObs);
      _actionLearners_[actionId]->getWithDefault(*varIter, nullptr)->updateGraph();
    }
 
    if (_actionReward_) {
      _actionRewardLearners_[actionId]->addObservation(newObs);
      _actionRewardLearners_[actionId]->updateGraph();
    } else {
      _rewardLearner_->addObservation(newObs);
      _rewardLearner_->updateGraph();
    }
 
    _rmax_ = _rmax_ < std::abs(newObs->reward()) ? std::abs(newObs->reward()) : _rmax_;
 
    return false;
  }

References _actionLearners_, _actionReward_, _actionRewardLearners_, _fmdp_, _rewardLearner_, _rmax_, and gum::Observation::reward().

Here is the call graph for this function:

◆ initialize()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

void gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::initialize ( FMDP< double > * fmdp )

virtual

Initializes the learner.

Implements gum::ILearningStrategy.

Definition at line 108 of file fmdpLearner_tpl.h.

                                       {
    _fmdp_ = fmdp;
 
    _modaMax_ = 0;
    _rmax_    = 0.0;
 
    gum::VariableSet mainVariables;
    for (auto varIter = _fmdp_->beginVariables(); varIter != _fmdp_->endVariables(); ++varIter) {
      mainVariables.insert(*varIter);
      _modaMax_ = _modaMax_ < (*varIter)->domainSize() ? (*varIter)->domainSize() : _modaMax_;
    }
 
    for (auto actionIter = _fmdp_->beginActions(); actionIter != _fmdp_->endActions();
         ++actionIter) {
      // Adding a Hashtable for the action
      _actionLearners_.insert(*actionIter, new VarLearnerTable());
 
      // Adding a learner for each variable
      for (auto varIter = _fmdp_->beginVariables(); varIter != _fmdp_->endVariables(); ++varIter) {
        MultiDimFunctionGraph< double >* varTrans = _instantiateFunctionGraph_();
        varTrans->setTableName("ACTION : " + _fmdp_->actionName(*actionIter)
                               + " - VARIABLE : " + (*varIter)->name());
        _fmdp_->addTransitionForAction(*actionIter, *varIter, varTrans);
        _actionLearners_[*actionIter]->insert(
            (*varIter),
            _instantiateVarLearner_(varTrans, mainVariables, _fmdp_->main2prime(*varIter)));
      }
 
      if (_actionReward_) {
        MultiDimFunctionGraph< double >* reward = _instantiateFunctionGraph_();
        reward->setTableName("REWARD - ACTION : " + _fmdp_->actionName(*actionIter));
        _fmdp_->addRewardForAction(*actionIter, reward);
        _actionRewardLearners_.insert(*actionIter,
                                      _instantiateRewardLearner_(reward, mainVariables));
      }
    }
 
    if (!_actionReward_) {
      MultiDimFunctionGraph< double >* reward = _instantiateFunctionGraph_();
      reward->setTableName("REWARD");
      _fmdp_->addReward(reward);
      _rewardLearner_ = _instantiateRewardLearner_(reward, mainVariables);
    }
  }

References _actionLearners_, _actionReward_, _actionRewardLearners_, _fmdp_, _instantiateFunctionGraph_(), _instantiateRewardLearner_(), _instantiateVarLearner_(), _modaMax_, _rewardLearner_, _rmax_, gum::Set< Key >::insert(), and gum::MultiDimFunctionGraph< GUM_SCALAR, TerminalNodePolicy >::setTableName().

Here is the call graph for this function:

◆ modaMax()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

virtual double gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::modaMax ( ) const

inlinevirtual

learnerSize

Returns

Implements gum::ILearningStrategy.

Definition at line 244 of file fmdpLearner.h.

244{ return _modaMax_; }

References _modaMax_.

◆ rMax()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

virtual double gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::rMax ( ) const

inlinevirtual

learnerSize

Returns

Implements gum::ILearningStrategy.

Definition at line 238 of file fmdpLearner.h.

238{ return _rmax_; }

References _rmax_.

◆ size()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

Size gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::size ( )

virtual

learnerSize

Returns

Implements gum::ILearningStrategy.

Definition at line 188 of file fmdpLearner_tpl.h.

             {
    Size s = 0;
    for (SequenceIteratorSafe< Idx > actionIter = _fmdp_->beginActions();
         actionIter != _fmdp_->endActions();
         ++actionIter) {
      for (SequenceIteratorSafe< const DiscreteVariable* > varIter = _fmdp_->beginVariables();
           varIter != _fmdp_->endVariables();
           ++varIter)
        s += _actionLearners_[*actionIter]->getWithDefault(*varIter, nullptr)->size();
      if (_actionReward_) s += _actionRewardLearners_[*actionIter]->size();
    }
 
    if (!_actionReward_) s += _rewardLearner_->size();
 
    return s;
  }

References _actionLearners_, _actionReward_, _actionRewardLearners_, _fmdp_, and _rewardLearner_.

◆ updateFMDP()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

void gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::updateFMDP ( )

virtual

Starts an update of datastructure in the associated FMDP.

Implements gum::ILearningStrategy.

Definition at line 212 of file fmdpLearner_tpl.h.

                   {
    for (SequenceIteratorSafe< Idx > actionIter = _fmdp_->beginActions();
         actionIter != _fmdp_->endActions();
         ++actionIter) {
      for (SequenceIteratorSafe< const DiscreteVariable* > varIter = _fmdp_->beginVariables();
           varIter != _fmdp_->endVariables();
           ++varIter)
        _actionLearners_[*actionIter]->getWithDefault(*varIter, nullptr)->updateFunctionGraph();
      if (_actionReward_) _actionRewardLearners_[*actionIter]->updateFunctionGraph();
    }
 
    if (!_actionReward_) _rewardLearner_->updateFunctionGraph();
  }

References _actionLearners_, _actionReward_, _actionRewardLearners_, _fmdp_, and _rewardLearner_.

◆ varLearner()

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

const IVisitableGraphLearner * gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::varLearner	(	Idx	actionId,
		const DiscreteVariable *	var ) const

inlinevirtual

extractCount

Implements gum::ILearningStrategy.

Definition at line 234 of file fmdpLearner.h.

                                                                                              {
      return _actionLearners_[actionId]->getWithDefault(var, nullptr);
    }

References _actionLearners_.

Member Data Documentation

◆ _actionLearners_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

HashTable< Idx, VarLearnerTable* > gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_actionLearners_

private

Definition at line 256 of file fmdpLearner.h.

Referenced by ~FMDPLearner(), addObservation(), initialize(), size(), updateFMDP(), and varLearner().

◆ _actionReward_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

bool gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_actionReward_

private

Definition at line 258 of file fmdpLearner.h.

Referenced by FMDPLearner(), addObservation(), initialize(), size(), and updateFMDP().

◆ _actionRewardLearners_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

HashTable< Idx, RewardLearnerType* > gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_actionRewardLearners_

private

Definition at line 259 of file fmdpLearner.h.

Referenced by ~FMDPLearner(), addObservation(), initialize(), size(), and updateFMDP().

◆ _fmdp_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

FMDP< double >* gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_fmdp_

private

The FMDP to store the learned model.

Definition at line 254 of file fmdpLearner.h.

Referenced by addObservation(), initialize(), size(), and updateFMDP().

◆ _learningThreshold_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

const double gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_learningThreshold_

private

Definition at line 262 of file fmdpLearner.h.

Referenced by FMDPLearner(), _instantiateRewardLearner_(), _instantiateRewardLearner_(), _instantiateVarLearner_(), and _instantiateVarLearner_().

◆ _modaMax_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

double gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_modaMax_

private

learnerSize

Returns

Definition at line 247 of file fmdpLearner.h.

Referenced by initialize(), and modaMax().

◆ _rewardLearner_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

RewardLearnerType* gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_rewardLearner_

private

Definition at line 260 of file fmdpLearner.h.

Referenced by FMDPLearner(), ~FMDPLearner(), addObservation(), initialize(), size(), and updateFMDP().

◆ _rmax_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

double gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_rmax_

private

learnerSize

Returns

Definition at line 241 of file fmdpLearner.h.

Referenced by addObservation(), initialize(), and rMax().

◆ _similarityThreshold_

template<TESTNAME VariableAttributeSelection, TESTNAME RewardAttributeSelection, LEARNERNAME LearnerSelection>

const double gum::FMDPLearner< VariableAttributeSelection, RewardAttributeSelection, LearnerSelection >::_similarityThreshold_

private

Definition at line 263 of file fmdpLearner.h.

Referenced by FMDPLearner(), _instantiateRewardLearner_(), and _instantiateVarLearner_().

The documentation for this class was generated from the following files:

agrum/FMDP/learning/fmdpLearner.h
agrum/FMDP/learning/fmdpLearner_tpl.h

Public Member Functions

Private Types

Private Attributes

Miscelleanous methods

Detailed Description

Member Typedef Documentation

◆ RewardLearnerType

◆ VariableLearnerType

◆ VarLearnerTable

Constructor & Destructor Documentation

◆ FMDPLearner()

◆ ~FMDPLearner()

Member Function Documentation

◆ _instantiateFunctionGraph_() [1/3]

◆ _instantiateFunctionGraph_() [2/3]

◆ _instantiateFunctionGraph_() [3/3]

◆ _instantiateRewardLearner_() [1/3]

◆ _instantiateRewardLearner_() [2/3]

◆ _instantiateRewardLearner_() [3/3]

◆ _instantiateVarLearner_() [1/3]

◆ _instantiateVarLearner_() [2/3]

◆ _instantiateVarLearner_() [3/3]

◆ addObservation()

◆ initialize()

◆ modaMax()

◆ rMax()

◆ size()

◆ updateFMDP()

◆ varLearner()

Member Data Documentation

◆ _actionLearners_

◆ _actionReward_

◆ _actionRewardLearners_

◆ _fmdp_

◆ _learningThreshold_

◆ _modaMax_

◆ _rewardLearner_

◆ _rmax_

◆ _similarityThreshold_