eoEsMutate< EOT > Class Template Reference

ES-style mutation in the large. More...

#include <eoEsMutate.h>

Inheritance diagram for eoEsMutate< EOT >:

List of all members.

Public Types
typedef EOT::Fitness	FitT
	Fitness-type.
Public Member Functions
	eoEsMutate (eoEsMutationInit &_init, eoRealVectorBounds &_bounds)
	Initialization.
virtual	~eoEsMutate ()
	Virtual Destructor.
virtual std::string	className () const
	Classname.
virtual bool	operator() (eoEsSimple< FitT > &_eo)
	Mutate eoEsSimple.
virtual bool	operator() (eoEsStdev< FitT > &_eo)
	Standard mutation in ES.
virtual bool	operator() (eoEsFull< FitT > &_eo)
	Correlated mutations in ES.
Private Member Functions
void	init (eoEsSimple< FitT >, eoEsMutationInit &_init)
	Initialization of simple ES.
void	init (eoEsStdev< FitT >, eoEsMutationInit &_init)
	Initialization of standard ES.
void	init (eoEsFull< FitT >, eoEsMutationInit &_init)
	Initialization of full ES.
Private Attributes
double	TauLcl
	Local factor for mutation of std deviations.
double	TauGlb
	Global factor for mutation of std deviations.
double	TauBeta
	Factor for mutation of correlation parameters.
eoRealVectorBounds &	bounds
	Bounds of parameters.
Static Private Attributes
static const double	stdev_eps = 1.0e-40
	Minimum stdev.

---

Detailed Description

template<class EOT>
class eoEsMutate< EOT >

ES-style mutation in the large.

Obviously, valid only for eoES*. It is currently valid for three types of ES chromosomes:

• eoEsSimple: Exactly one stdandard-deviation
• eoEsStdev: As many standard deviations as object variables
• eoEsFull: The whole guacemole: correlations, stdevs and object variables

Each of these three variant has it's own operator() in eoEsMutate and intialization is also split into three cases (that share some commonalities)

Definition at line 61 of file eoEsMutate.h.

---

Constructor & Destructor Documentation

template<class EOT>

eoEsMutate< EOT >::eoEsMutate	(	eoEsMutationInit &	_init,
		eoRealVectorBounds &	_bounds
	)		[inline]

Initialization.

Parameters:

_init	Proxy class for initializating the three parameters eoEsMutate needs
_bounds	Bounds for the objective variables

Definition at line 75 of file eoEsMutate.h.

References eoEsMutate< EOT >::init().

---

Member Function Documentation

template<class EOT>

virtual std::string eoEsMutate< EOT >::className

(

void

)

const [inline, virtual]

Classname.

Inherited from eoObject

See also:

eoObject

Returns:

Name of class.

Reimplemented from eoMonOp< EOT >.

Definition at line 91 of file eoEsMutate.h.

template<class EOT>

virtual bool eoEsMutate< EOT >::operator()

(

eoEsSimple< FitT > &

_eo

)

[inline, virtual]

Mutate eoEsSimple.

Parameters:

_eo	Individual to mutate.

Definition at line 98 of file eoEsMutate.h.

References eoEsMutate< EOT >::bounds, eoRealBaseVectorBounds::foldsInBounds(), eoRng::normal(), eoEsMutate< EOT >::stdev_eps, and eoEsMutate< EOT >::TauLcl.

template<class EOT>

virtual bool eoEsMutate< EOT >::operator()

(

eoEsStdev< FitT > &

_eo

)

[inline, virtual]

Standard mutation in ES.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts. Standard mutation of object variables and standard deviations in ESs.

If there are fewer different standard deviations available than the dimension of the objective function requires, the last standard deviation is responsible for ALL remaining object variables.

Parameters:

_eo	Individual to mutate.

See also:

Schwefel 1977: Numerische Optimierung von Computer-Modellen mittels der Evolutionsstrategie, pp. 165 ff.

Definition at line 129 of file eoEsMutate.h.

References eoEsMutate< EOT >::bounds, eoRealBaseVectorBounds::foldsInBounds(), eoRng::normal(), eoEsMutate< EOT >::stdev_eps, eoEsMutate< EOT >::TauGlb, and eoEsMutate< EOT >::TauLcl.

template<class EOT>

virtual bool eoEsMutate< EOT >::operator()

(

eoEsFull< FitT > &

_eo

)

[inline, virtual]

Correlated mutations in ES.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts. Mutation of object variables, standard deviations, and their correlations in ESs.

Parameters:

_eo	Individual to mutate.

See also:

• H.-P. Schwefel: Internal Report of KFA Juelich, KFA-STE-IB-3/80, p. 43, 1980.
• G. Rudolph: Globale Optimierung mit parallelen Evolutionsstrategien, Diploma Thesis, University of Dortmund, 1990.

Definition at line 160 of file eoEsMutate.h.

References eoEsMutate< EOT >::bounds, eoRealBaseVectorBounds::foldsInBounds(), eoRng::normal(), eoEsMutate< EOT >::stdev_eps, eoEsMutate< EOT >::TauBeta, eoEsMutate< EOT >::TauGlb, and eoEsMutate< EOT >::TauLcl.

template<class EOT>

void eoEsMutate< EOT >::init	(	eoEsStdev< FitT >	,
		eoEsMutationInit &	_init
	)		[inline, private]

Initialization of standard ES.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Definition at line 229 of file eoEsMutate.h.

References eoEsMutate< EOT >::bounds, eoEsMutationInit::TauGlb(), eoEsMutate< EOT >::TauGlb, eoEsMutationInit::TauLcl(), and eoEsMutate< EOT >::TauLcl.

template<class EOT>

void eoEsMutate< EOT >::init	(	eoEsFull< FitT >	,
		eoEsMutationInit &	_init
	)		[inline, private]

Initialization of full ES.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Definition at line 245 of file eoEsMutate.h.

References eoEsMutate< EOT >::init(), eoEsMutationInit::TauBeta(), eoEsMutate< EOT >::TauBeta, eoEsMutate< EOT >::TauGlb, and eoEsMutate< EOT >::TauLcl.

---

Member Data Documentation

template<class EOT>

const double eoEsMutate< EOT >::stdev_eps = 1.0e-40 [static, private]

Minimum stdev.

If you let the step-size go to 0, self-adaptation stops, therefore we give a lower bound. The actual value used is somewhat arbitrary and the is no theoretical reasoning known for it (Sep 2005).

The code that we have in EO is a port from a C code that Thomas Bäck kindly donated to the community some years ago. It has been modified by Marc Schoenauer for inclusion in EvolC, than by Maarten Keijzer into EO. The exact value was adjusted based on practice.

Removing this doesn't work well, but it was never tried to figure out what the best value would be.

Definition at line 279 of file eoEsMutate.h.

Referenced by eoEsMutate< EOT >::operator()().

---

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

• eoEsMutate.h