Selector::ClassificationSelectionMetrics Class Reference

Class to extract metrics from either the agent or the environment. More...

#include <classificationSelectionMetrics.h>

Inheritance diagram for Selector::ClassificationSelectionMetrics:

Public Member Functions
	ClassificationSelectionMetrics ()=default
	Default constructor.
	ClassificationSelectionMetrics (const std::vector< double > &scorePerClass, const std::vector< size_t > &nbEvalPerClass)
	Constructor with score and utility initialization.
virtual const std::vector< double > &	getScorePerClass () const
virtual const std::vector< size_t > &	getNbEvalPerClassPerClass () const
void	initMetrics (const TPG::TPGVertex *agent, const Learn::LearningEnvironment &learningEnvironment) override
	Specialization of the initialisation of the metrics.
void	extractMetricsEpisode (const TPG::TPGVertex *agent, size_t nbStepsExecuted, const Learn::LearningEnvironment &learningEnvironment) override
	Specialization of the extraction of the metrics at the end of an episode.
virtual void	weightedSum (std::shared_ptr< SelectionMetrics > other, size_t nbEvaluation, size_t nbEvaluationOther) override
	Specialization of weightedSum method to add the score per class and nbEvalPerClass.
Public Member Functions inherited from Selector::SelectionMetrics
	SelectionMetrics ()=default
	Default constructor.
virtual	~SelectionMetrics ()=default
	Default destructor.
	SelectionMetrics (double score, double utility=0)
	Constructor with score and utility initialization.
virtual double	getScore () const
virtual double	getUtility () const
virtual void	extractMetricsStep (const TPG::TPGVertex *agent, std::vector< double > actionValues, const Learn::LearningEnvironment &learningEnvironment)
	Extract metrics from the agent in the learning environment.

Protected Attributes
std::vector< double >	scorePerClass
	Vector storing a double score per class (i.e. per Action) of a classification LearningEnvironment.
std::vector< size_t >	nbEvalPerClass
	Vector storing a size_t value per class representing the number of evaluation per class.
Protected Attributes inherited from Selector::SelectionMetrics
double	score = 0
double	utility = 0

Additional Inherited Members
Static Public Member Functions inherited from Selector::SelectionMetrics
template<class T >
static T	weightedSum (T value, T valueOther, size_t nbEvaluation, size_t nbEvaluationOther)
	Perform a weighted sum between 2 values.

Detailed Description

Class to extract metrics from either the agent or the environment.

This metrics can be used to specify the selection of the selector. This class does not implement any metrics, it need to be override by the different selection methods.

Constructor & Destructor Documentation

ClassificationSelectionMetrics()

Selector::ClassificationSelectionMetrics::ClassificationSelectionMetrics ( const std::vector< double > & scorePerClass, const std::vector< size_t > & nbEvalPerClass )

inline

Constructor with score and utility initialization.

Parameters

[in]	scorePerClass	the vector of score obtained by the agent per class.
[in]	nbEvalPerClass	the vector of number of evaluation per class.

Member Function Documentation

extractMetricsEpisode()

void Selector::ClassificationSelectionMetrics::extractMetricsEpisode ( const TPG::TPGVertex * agent, size_t nbStepsExecuted, const Learn::LearningEnvironment & learningEnvironment )

overridevirtual

Specialization of the extraction of the metrics at the end of an episode.

Reimplemented from Selector::SelectionMetrics.

getNbEvalPerClassPerClass()

const std::vector< size_t > & Selector::ClassificationSelectionMetrics::getNbEvalPerClassPerClass ( ) const

virtual

Return the number of evaluation per class of the agent.

getScorePerClass()

const std::vector< double > & Selector::ClassificationSelectionMetrics::getScorePerClass ( ) const

virtual

Return the score per class of the agent.

initMetrics()

void Selector::ClassificationSelectionMetrics::initMetrics ( const TPG::TPGVertex * agent, const Learn::LearningEnvironment & learningEnvironment )

overridevirtual

Specialization of the initialisation of the metrics.

Reimplemented from Selector::SelectionMetrics.

weightedSum()

void Selector::ClassificationSelectionMetrics::weightedSum ( std::shared_ptr< SelectionMetrics > other, size_t nbEvaluation, size_t nbEvaluationOther )

overridevirtual

Specialization of weightedSum method to add the score per class and nbEvalPerClass.

Reimplemented from Selector::SelectionMetrics.

---

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

• D:/a/gegelati/gegelati/gegelatilib/include/selector/classificationSelectionMetrics.h
• D:/a/gegelati/gegelati/gegelatilib/src/selector/classificationSelectionMetrics.cpp