Implements the CMA-ES. More...

#include <shark/Algorithms/DirectSearch/CMA.h>

Inheritance diagram for shark::CMA:

Public Types
enum	RecombinationType { EQUAL = 0, LINEAR = 1, SUPERLINEAR = 2 }
	Models the recombination type. More...

Public Types inherited from shark::AbstractSingleObjectiveOptimizer< RealVector >
typedef base_type::SearchPointType	SearchPointType

typedef base_type::SolutionType	SolutionType

typedef base_type::ResultType	ResultType

typedef base_type::ObjectiveFunctionType	ObjectiveFunctionType

Public Types inherited from shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >
enum	Feature
	Models features that the optimizer requires from the objective function. More...

typedef RealVector	SearchPointType

typedef double	ResultType

typedef SingleObjectiveResultSet< RealVector >	SolutionType

typedef AbstractObjectiveFunction< RealVector, ResultType >	ObjectiveFunctionType

typedef TypedFlags< Feature >	Features

typedef TypedFeatureNotAvailableException< Feature >	FeatureNotAvailableException

Public Member Functions
SHARK_EXPORT_SYMBOL	CMA (random::rng_type &rng=random::globalRng)
	Default c'tor. More...

std::string	name () const
	From INameable: return the class name. More...

SHARK_EXPORT_SYMBOL void	read (InArchive &archive)
	Read the component from the supplied archive. More...

SHARK_EXPORT_SYMBOL void	write (OutArchive &archive) const
	Write the component to the supplied archive. More...

SHARK_EXPORT_SYMBOL void	init (ObjectiveFunctionType const &function, SearchPointType const &p)
	Initializes the algorithm for the supplied objective function. More...

SHARK_EXPORT_SYMBOL void	init (ObjectiveFunctionType const &function, SearchPointType const &initialSearchPoint, std::size_t lambda, std::size_t mu, double initialSigma, const boost::optional< RealMatrix > &initialCovarianceMatrix=boost::optional< RealMatrix >())
	Initializes the algorithm for the supplied objective function. More...

SHARK_EXPORT_SYMBOL void	step (ObjectiveFunctionType const &function)
	Executes one iteration of the algorithm. More...

void	setInitialSigma (double initSigma)
	sets the initial step length sigma More...

double	sigma () const
	Accesses the current step size. More...

RealVector const &	mean () const
	Accesses the current population mean. More...

RealVector const &	weights () const
	Accesses the current weighting vector. More...

RealVector const &	evolutionPath () const
	Accesses the evolution path for the covariance matrix update. More...

RealVector const &	evolutionPathSigma () const
	Accesses the evolution path for the step size update. More...

RealMatrix const &	covarianceMatrix () const
	Accesses the covariance matrix of the normal distribution used for generating offspring individuals. More...

RecombinationType	recombinationType () const
	Accesses the recombination type. More...

RecombinationType &	recombinationType ()
	Returns a mutable reference to the recombination type. More...

const double &	lowerBound () const
	Returns a const reference to the lower bound on sigma times smalles eigenvalue. More...

void	setLowerBound (double lowerBound)
	Set the lower bound on sigma times smalles eigenvalue. More...

std::size_t	mu () const
	Returns the size of the parent population \(\mu\). More...

void	setMu (std::size_t mu)
	Sets the number of selected samples. More...

void	setLambda (std::size_t lambda)
	Sets the number of sampled points. More...

std::size_t	lambda () const
	Returns a immutable reference to the size of the offspring population \(\mu\). More...

RealMatrix const &	eigenVectors () const
	Returns eigenvectors of covariance matrix (not considering step size) More...

RealVector const &	eigenValues () const
	Returns a eigenvectors of covariance matrix (not considering step size) More...

double	condition () const
	Returns condition of covariance matrix. More...

std::size_t	numberOfEvaluations () const
	Returns how often a point is evaluated. More...

Public Member Functions inherited from shark::AbstractSingleObjectiveOptimizer< RealVector >
std::size_t	numInitPoints () const
	By default most single objective optimizers only require a single point. More...

virtual void	init (ObjectiveFunctionType const &function, std::vector< SearchPointType > const &initPoints)
	Initialize the optimizer for the supplied objective function using a set of initialisation points. More...

virtual const SolutionType &	solution () const
	returns the current solution of the optimizer More...

Public Member Functions inherited from shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >
const Features &	features () const

virtual void	updateFeatures ()

bool	requiresValue () const

bool	requiresFirstDerivative () const

bool	requiresSecondDerivative () const

bool	canSolveConstrained () const

bool	requiresClosestFeasible () const

virtual	~AbstractOptimizer ()

virtual void	init (ObjectiveFunctionType const &function)
	Initialize the optimizer for the supplied objective function. More...

Public Member Functions inherited from shark::INameable
virtual	~INameable ()

Public Member Functions inherited from shark::ISerializable
virtual	~ISerializable ()
	Virtual d'tor. More...

void	load (InArchive &archive, unsigned int version)
	Versioned loading of components, calls read(...). More...

void	save (OutArchive &archive, unsigned int version) const
	Versioned storing of components, calls write(...). More...

	BOOST_SERIALIZATION_SPLIT_MEMBER ()

Static Public Member Functions
static SHARK_EXPORT_SYMBOL std::size_t	suggestLambda (std::size_t dimension)
	Calculates lambda for the supplied dimensionality n. More...

static SHARK_EXPORT_SYMBOL std::size_t	suggestMu (std::size_t lambda, RecombinationType recomb=SUPERLINEAR)
	Calculates mu for the supplied lambda and the recombination strategy. More...

Protected Types
typedef Individual< RealVector, double, RealVector >	IndividualType
	The type of individual used for the CMA. More...

Protected Member Functions
SHARK_EXPORT_SYMBOL std::vector< IndividualType >	generateOffspring () const
	Samples lambda individuals from the search distribution. More...

SHARK_EXPORT_SYMBOL void	updatePopulation (std::vector< IndividualType > const &offspring)
	Updates the strategy parameters based on the supplied offspring population. More...

SHARK_EXPORT_SYMBOL void	doInit (std::vector< SearchPointType > const &points, std::vector< ResultType > const &functionValues, std::size_t lambda, std::size_t mu, double initialSigma)

Protected Member Functions inherited from shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >
void	checkFeatures (ObjectiveFunctionType const &objectiveFunction)
	Convenience function that checks whether the features of the supplied objective function match with the required features of the optimizer. More...

Additional Inherited Members
Protected Attributes inherited from shark::AbstractSingleObjectiveOptimizer< RealVector >
SolutionType	m_best
	Current solution of the optimizer. More...

Protected Attributes inherited from shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >
Features	m_features

Detailed Description

Implements the CMA-ES.

The algorithm is described in

Hansen, N., S. Kern (2004). Evaluating the CMA Evolution Strategy on Multimodal Test Functions. In Proceedings of the Eighth International Conference on Parallel Problem Solving from Nature (PPSN VIII), pp. 282-291, LNCS, Springer-Verlag

For noisy function, noise handling is supported using the noise level detection algorithm described in Hansen, N., et al. "A method for handling uncertainty in evolutionary optimization with an application to feedback control of combustion." IEEE Transactions on Evolutionary Computation 13.1 (2009): 180-197. Our implementation varies in small details, e.g. instead of the average rank the rank of the average function value is used for updating the strategy parameters which ensures asymptotic unbiasedness. We further do not have an upper bound on the number of reevaluations for the same reason.

Definition at line 65 of file CMA.h.

Member Typedef Documentation

◆ IndividualType

typedef Individual<RealVector, double, RealVector> shark::CMA::IndividualType

protected

The type of individual used for the CMA.

Definition at line 210 of file CMA.h.

Member Enumeration Documentation

◆ RecombinationType

enum shark::CMA::RecombinationType

Models the recombination type.

Enumerator
EQUAL
LINEAR
SUPERLINEAR

Definition at line 69 of file CMA.h.

Constructor & Destructor Documentation

◆ CMA()

SHARK_EXPORT_SYMBOL shark::CMA::CMA ( random::rng_type & rng = random::globalRng )

Default c'tor.

Member Function Documentation

◆ condition()

double shark::CMA::condition ( ) const

inline

Returns condition of covariance matrix.

Definition at line 197 of file CMA.h.

References shark::MultiVariateNormalDistribution::eigenValues(), and eigenValues().

Referenced by main().

◆ covarianceMatrix()

RealMatrix const& shark::CMA::covarianceMatrix ( ) const

inline

Accesses the covariance matrix of the normal distribution used for generating offspring individuals.

Definition at line 141 of file CMA.h.

References shark::MultiVariateNormalDistribution::covarianceMatrix().

◆ doInit()

SHARK_EXPORT_SYMBOL void shark::CMA::doInit	(	std::vector< SearchPointType > const &	points,
		std::vector< ResultType > const &	functionValues,
		std::size_t	lambda,
		std::size_t	mu,
		double	initialSigma
	)

protected

◆ eigenValues()

RealVector const& shark::CMA::eigenValues ( ) const

inline

Returns a eigenvectors of covariance matrix (not considering step size)

Definition at line 192 of file CMA.h.

References shark::MultiVariateNormalDistribution::eigenValues().

Referenced by condition(), and main().

◆ eigenVectors()

RealMatrix const& shark::CMA::eigenVectors ( ) const

inline

Returns eigenvectors of covariance matrix (not considering step size)

Definition at line 187 of file CMA.h.

References shark::MultiVariateNormalDistribution::eigenVectors().

Referenced by main().

◆ evolutionPath()

RealVector const& shark::CMA::evolutionPath ( ) const

inline

Accesses the evolution path for the covariance matrix update.

Definition at line 131 of file CMA.h.

◆ evolutionPathSigma()

RealVector const& shark::CMA::evolutionPathSigma ( ) const

inline

Accesses the evolution path for the step size update.

Definition at line 136 of file CMA.h.

◆ generateOffspring()

SHARK_EXPORT_SYMBOL std::vector<IndividualType> shark::CMA::generateOffspring ( ) const

protected

Samples lambda individuals from the search distribution.

◆ init() [1/2]

SHARK_EXPORT_SYMBOL void shark::CMA::init	(	ObjectiveFunctionType const &	function,
		SearchPointType const &	p
	)

virtual

Initializes the algorithm for the supplied objective function.

Implements shark::AbstractSingleObjectiveOptimizer< RealVector >.

Referenced by main(), and name().

◆ init() [2/2]

SHARK_EXPORT_SYMBOL void shark::CMA::init	(	ObjectiveFunctionType const &	function,
		SearchPointType const &	initialSearchPoint,
		std::size_t	lambda,
		std::size_t	mu,
		double	initialSigma,
		const boost::optional< RealMatrix > &	initialCovarianceMatrix = `boost::optional< RealMatrix >()`
	)

Initializes the algorithm for the supplied objective function.

◆ lambda()

std::size_t shark::CMA::lambda ( ) const

inline

Returns a immutable reference to the size of the offspring population \(\mu\).

Definition at line 182 of file CMA.h.

Referenced by name(), and setLambda().

◆ lowerBound()

const double& shark::CMA::lowerBound ( ) const

inline

Returns a const reference to the lower bound on sigma times smalles eigenvalue.

Definition at line 156 of file CMA.h.

Referenced by setLowerBound().

◆ mean()

RealVector const& shark::CMA::mean ( ) const

inline

Accesses the current population mean.

Definition at line 121 of file CMA.h.

Referenced by main().

◆ mu()

std::size_t shark::CMA::mu ( ) const

inline

Returns the size of the parent population \(\mu\).

Definition at line 166 of file CMA.h.

Referenced by name(), and setMu().

◆ name()

std::string shark::CMA::name ( ) const

inlinevirtual

From INameable: return the class name.

Reimplemented from shark::INameable.

Definition at line 79 of file CMA.h.

References init(), lambda(), mu(), read(), SHARK_EXPORT_SYMBOL, step(), suggestLambda(), suggestMu(), SUPERLINEAR, and write().

◆ numberOfEvaluations()

std::size_t shark::CMA::numberOfEvaluations ( ) const

inline

Returns how often a point is evaluated.

Definition at line 203 of file CMA.h.

◆ read()

SHARK_EXPORT_SYMBOL void shark::CMA::read ( InArchive & archive )

virtual

Read the component from the supplied archive.

Parameters

[in,out] archive The archive to read from.

Reimplemented from shark::ISerializable.

Referenced by name().

◆ recombinationType() [1/2]

RecombinationType shark::CMA::recombinationType ( ) const

inline

Accesses the recombination type.

Definition at line 146 of file CMA.h.

◆ recombinationType() [2/2]

RecombinationType& shark::CMA::recombinationType ( )

inline

Returns a mutable reference to the recombination type.

Definition at line 151 of file CMA.h.

◆ setInitialSigma()

void shark::CMA::setInitialSigma ( double initSigma )

inline

sets the initial step length sigma

It is by default <=0 which means that sigma =1/sqrt(numVariables)

Definition at line 111 of file CMA.h.

Referenced by main().

◆ setLambda()

void shark::CMA::setLambda ( std::size_t lambda )

inline

Sets the number of sampled points.

Definition at line 176 of file CMA.h.

References lambda().

◆ setLowerBound()

void shark::CMA::setLowerBound ( double lowerBound )

inline

Set the lower bound on sigma times smalles eigenvalue.

Definition at line 161 of file CMA.h.

References lowerBound().

◆ setMu()

void shark::CMA::setMu ( std::size_t mu )

inline

Sets the number of selected samples.

Definition at line 171 of file CMA.h.

References mu().

◆ sigma()

double shark::CMA::sigma ( ) const

inline

Accesses the current step size.

Definition at line 116 of file CMA.h.

Referenced by main().

◆ step()

SHARK_EXPORT_SYMBOL void shark::CMA::step ( ObjectiveFunctionType const & function )

virtual

Executes one iteration of the algorithm.

Implements shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >.

Referenced by main(), and name().

◆ suggestLambda()

static SHARK_EXPORT_SYMBOL std::size_t shark::CMA::suggestLambda ( std::size_t dimension )

static

Calculates lambda for the supplied dimensionality n.

Referenced by name().

◆ suggestMu()

static SHARK_EXPORT_SYMBOL std::size_t shark::CMA::suggestMu	(	std::size_t	lambda,
		RecombinationType	recomb = `SUPERLINEAR`
	)

static

Calculates mu for the supplied lambda and the recombination strategy.

Referenced by name().

◆ updatePopulation()

SHARK_EXPORT_SYMBOL void shark::CMA::updatePopulation ( std::vector< IndividualType > const & offspring )

protected

Updates the strategy parameters based on the supplied offspring population.

◆ weights()

RealVector const& shark::CMA::weights ( ) const

inline

Accesses the current weighting vector.

Definition at line 126 of file CMA.h.

◆ write()

SHARK_EXPORT_SYMBOL void shark::CMA::write ( OutArchive & archive ) const

virtual

Write the component to the supplied archive.

Parameters

[in,out] archive The archive to write to.

Reimplemented from shark::ISerializable.

Referenced by name().

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

include/shark/Algorithms/DirectSearch/CMA.h

Public Types

Public Member Functions

Static Public Member Functions

Protected Types

Protected Member Functions

Additional Inherited Members

Detailed Description

Member Typedef Documentation

◆ IndividualType

Member Enumeration Documentation

◆ RecombinationType

Constructor & Destructor Documentation

◆ CMA()

Member Function Documentation

◆ condition()

◆ covarianceMatrix()

◆ doInit()

◆ eigenValues()

◆ eigenVectors()

◆ evolutionPath()

◆ evolutionPathSigma()

◆ generateOffspring()

◆ init() [1/2]

◆ init() [2/2]

◆ lambda()

◆ lowerBound()

◆ mean()

◆ mu()

◆ name()

◆ numberOfEvaluations()

◆ read()

◆ recombinationType() [1/2]

◆ recombinationType() [2/2]

◆ setInitialSigma()

◆ setLambda()

◆ setLowerBound()

◆ setMu()

◆ sigma()

◆ step()

◆ suggestLambda()

◆ suggestMu()

◆ updatePopulation()

◆ weights()

◆ write()