Selector::MapElites::MapElitesDescriptor Class Reference

Abstract class representing a descriptor for Map Elites selection. More...

#include <mapElitesDescriptor.h>

Inheritance diagram for Selector::MapElites::MapElitesDescriptor:

Public Member Functions
virtual bool	isInit () const
	method that return if the initDescriptor method has been executed
virtual void	initDescriptor (const TPG::TPGGraph &graph, const Learn::LearningEnvironment &learningEnvironment)=0
	initiate the descriptor attribute (such as nbDescriptors and min/max ranges)
virtual std::string	getName () const =0
	abstract method returning the name of the descriptor.
virtual std::pair< double, double >	getMinAndMaxRange () const
	Get the min and max range of the descriptor.
virtual size_t	getNbDescriptors () const
	Get the number of the descriptor.
virtual void	extractMetricsStep (std::vector< double > &metrics, const TPG::TPGVertex *agent, std::vector< double > actionValues, const Learn::LearningEnvironment &learningEnvironment) const
	Extract the metrics at each step of the evaluation.
virtual void	extractMetricsEpisode (std::vector< double > &metrics, const TPG::TPGVertex *agent, size_t nbStepsExecuted, const Learn::LearningEnvironment &learningEnvironment) const
	Extract the metrics at the end of the evaluation episode.

Protected Attributes
size_t	nbDescriptors
	Number of descriptor.
double	minRange
	Minimum value in the archive.
double	maxRange
	Maximum value in the archive.
bool	init = false

Detailed Description

Abstract class representing a descriptor for Map Elites selection.

This class need to be inherrit to create specific descriptors. Example of descriptors are given in selector/mapElites/mapElitesDefaultDescriptors

Member Function Documentation

extractMetricsEpisode()

virtual void Selector::MapElites::MapElitesDescriptor::extractMetricsEpisode ( std::vector< double > & metrics, const TPG::TPGVertex * agent, size_t nbStepsExecuted, const Learn::LearningEnvironment & learningEnvironment ) const

inlinevirtual

Extract the metrics at the end of the evaluation episode.

Parameters

[in]	metrics	vector of current extracted metrics
[in]	agent	the TPGVertex being evaluated.
[in]	nbStepsExecuted	number of steps executed at the end of the episode.
[in]	learningEnvironment	the learning environment used to get the metrics.

Reimplemented in Selector::MapElites::DefaultDescriptors::ActionValues.

extractMetricsStep()

virtual void Selector::MapElites::MapElitesDescriptor::extractMetricsStep ( std::vector< double > & metrics, const TPG::TPGVertex * agent, std::vector< double > actionValues, const Learn::LearningEnvironment & learningEnvironment ) const

inlinevirtual

Extract the metrics at each step of the evaluation.

Parameters

[in]	metrics	vector of current extracted metrics
[in]	agent	the TPGVertex being evaluated.
[in]	actionValues	the action values taken by the agent at this step.
[in]	learningEnvironment	the learning environment used to get the metrics.

Reimplemented in Selector::MapElites::DefaultDescriptors::ActionValues.

getName()

virtual std::string Selector::MapElites::MapElitesDescriptor::getName ( ) const

pure virtual

abstract method returning the name of the descriptor.

Implemented in Selector::MapElites::DefaultDescriptors::ActionValues.

initDescriptor()

virtual void Selector::MapElites::MapElitesDescriptor::initDescriptor ( const TPG::TPGGraph & graph, const Learn::LearningEnvironment & learningEnvironment )

pure virtual

initiate the descriptor attribute (such as nbDescriptors and min/max ranges)

Parameters

[in]	graph	current graph
[in]	learningEnvironment	the learning environment used

Implemented in Selector::MapElites::DefaultDescriptors::ActionValues.

Member Data Documentation

init

bool Selector::MapElites::MapElitesDescriptor::init = false

protected

Boolean value to check if the initDescriptor method has been executed.

---

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

• D:/a/gegelati/gegelati/gegelatilib/include/selector/mapElites/mapElitesDescriptor.h