Class to represent the Novelty Search Algorithm. More...

#include <NoveltySearch.h>

Inheritance diagram for NoveltySearch< S >:

Public Member Functions
	NoveltySearch (unique_ptr< Distance< float >> dist, const float &threshold=2000, const float &finalThresh=0.0001, const int &k=15)
	Creates a new Novelty Search instance. More...

Front< S >	getResults () override
	Returns the results obtained by the Novelty Search in a Front object. More...

vector< S >	run (vector< S > &population, const Problem< S > *problem) override
	Novelty Search Algorithm It looks for novelty using the genotypes of the individuals in the population. Computes the sparseness of each individual in the noveltyArchive and population. More...

virtual void	cmpFinals (vector< S > &population, const Problem< S > *problem=nullptr)
	Compares the individuals in the population against the neighbors inside the archive of final Ss. If the score is good enough the individual will be included inside the archive of final solutions. More...

virtual void	insertIntoArchive (const S &solution)
	Method to insert a new individual into the noveltyArchive of novelty solutions. More...

const vector< S > &	getArchive () const

virtual json	to_json ()

float	getThreshold () const

float	getK () const

float	getFinalThresh () const

Public Member Functions inherited from Search< S >
	Search ()=default
	Constructs a default Search.

Protected Member Functions
virtual vector< Descriptor >	beforeRun (const vector< S > &population)
	Performs computational work necessary for running the NS This method creates a combined population using the individuals from the NS archive and the population. The resulting vector contains the features of each individual. More...

virtual vector< Descriptor >	beforeCmpFinals (const vector< S > &population)

virtual void	insertFinal (const S &solution)
	Method to insert a new individual into the final set of solutions The default behaviour is to get all the variables of the solution. More...

Protected Attributes
unique_ptr< Distance< float > >	distance

vector< S >	noveltyArchive

vector< S >	finalSs

vector< Descriptor >	finalSsDesc

float	threshold

float	finalSThreshold

int	k

Detailed Description

template<typename S>
class NoveltySearch< S >

Class to represent the Novelty Search Algorithm.

Template Parameters

P
S

Constructor & Destructor Documentation

◆ NoveltySearch()

template<typename S >

NoveltySearch< S >::NoveltySearch	(	unique_ptr< Distance< float >>	dist,
		const float &	threshold = `2000`,
		const float &	finalThresh = `0.0001`,
		const int &	k = `15`
	)

explicit

Creates a new Novelty Search instance.

Template Parameters

P
S

Parameters

threshold
dist
k

Member Function Documentation

◆ beforeRun()

template<typename S >

vector< Descriptor > NoveltySearch< S >::beforeRun ( const vector< S > & population )

protectedvirtual

Performs computational work necessary for running the NS This method creates a combined population using the individuals from the NS archive and the population. The resulting vector contains the features of each individual.

Template Parameters

MS

Parameters

population

Reimplemented in NSPerformance< S >, and NSFeatures< S >.

◆ cmpFinals()

template<typename S >

void NoveltySearch< S >::cmpFinals	(	vector< S > &	population,
		const Problem< S > *	problem = `nullptr`
	)

virtual

Compares the individuals in the population against the neighbors inside the archive of final Ss. If the score is good enough the individual will be included inside the archive of final solutions.

Template Parameters

Problem
S
T

Parameters

population

◆ getResults()

template<typename S >

Front< S > NoveltySearch< S >::getResults

overridevirtual

Returns the results obtained by the Novelty Search in a Front object.

Template Parameters

Problem
S
T

Returns

Implements Search< S >.

◆ insertFinal()

template<typename S >

void NoveltySearch< S >::insertFinal ( const S & solution )

protectedvirtual

Method to insert a new individual into the final set of solutions The default behaviour is to get all the variables of the solution.

Template Parameters

Problem
S
T

Parameters

S

Reimplemented in NSFeatures< S >, and NSPerformance< S >.

◆ insertIntoArchive()

template<typename S >

void NoveltySearch< S >::insertIntoArchive ( const S & solution )

virtual

Method to insert a new individual into the noveltyArchive of novelty solutions.

Template Parameters

Problem
S
T

Parameters

S

◆ run()

template<typename S >

vector< S > NoveltySearch< S >::run	(	vector< S > &	population,
		const Problem< S > *	problem
	)

overridevirtual

Novelty Search Algorithm It looks for novelty using the genotypes of the individuals in the population. Computes the sparseness of each individual in the noveltyArchive and population.

Template Parameters

Problem
S
T

Parameters

population
problem

Returns

Implements Search< S >.

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

include/dignea/searches/NoveltySearch.h

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

template<typename S> class NoveltySearch< S >

Constructor & Destructor Documentation

◆ NoveltySearch()

Member Function Documentation

◆ beforeRun()

◆ cmpFinals()

◆ getResults()

◆ insertFinal()

◆ insertIntoArchive()

◆ run()

template<typename S>
class NoveltySearch< S >