libcmaes::CMAParameters< TGenoPheno > Class Template Reference

Parameters for various flavors of the CMA-ES algorithm. More...

#include <cmaparameters.h>

Inheritance diagram for libcmaes::CMAParameters< TGenoPheno >:

Public Member Functions
	CMAParameters (const int &dim, const double *x0, const double &sigma, const int &lambda=-1, const uint64_t &seed=0, const TGenoPheno &gp=TGenoPheno())
	Constructor. More...
	CMAParameters (const std::vector< double > &x0, const double &sigma, const int &lambda=-1, const uint64_t &seed=0, const TGenoPheno &gp=TGenoPheno())
	Constructor. More...
	CMAParameters (const std::vector< double > &x0, const std::vector< double > &sigma, const int &lambda=-1, const std::vector< double > &lbounds=std::vector< double >(), const std::vector< double > &ubounds=std::vector< double >(), const uint64_t &seed=0)
	Constructor. More...
void	initialize_parameters ()
	initialize required parameters based on dim, lambda, x0 and sigma.
void	reset_as_fixed (const int &k)
void	set_noisy ()
	adapt parameters for noisy objective function.
void	set_algo (const int &algo)
	sets the optimization algorithm. Note: overrides Parameters::set_algo More...
void	set_str_algo (const std::string &algo)
	sets the optimization algorithm. More...
double	get_sigma_init () const
	returns initial sigma value More...
void	set_gradient (const bool &gradient)
	activates the gradient injection scheme. If no gradient function is defined, injects a numerical gradient solution instead Note: overrides Parameters::set_gradient More...
void	set_sep ()
	fix parameters for sep-CMA-ES, using only the diagonal of covariance matrix.
bool	is_sep () const
	whether algorithm leverages separability. More...
void	set_vd ()
	activates VD decomposition.
bool	is_vd () const
	whether algorithm uses vd update. More...
void	set_fixed_p (const int &index, const double &value)
	freezes a parameter to a given value in genotype during optimization. Adapts some generic parameters as well. More...
void	unset_fixed_p (const int &index)
	unfreezes a parameter. More...
void	set_restarts (const int &nrestarts)
	sets the maximum number of restarts (applies to IPOP and BIPOP). More...
int	get_restarts () const
	get the number of restarts (applies to IPOP and BIPOP). More...
void	set_lazy_update (const bool &lz)
	sets the lazy update (i.e. updates the eigenvalues every few steps). More...
bool	get_lazy_update ()
	get lazy update status. More...
void	set_elitism (const int &e)
	sets elitism: 0 -> no elitism 1 -> elitism: reinjects the best-ever seen solution 2 -> initial elitism: reinject x0 as long as it is not improved upon 3 -> initial elitism on restart: restart if best encountered solution is not the the final solution and reinjects the best solution until the population has better fitness, in its majority
void	set_stopping_criteria (const int &criteria, const bool &active)
	all stopping criteria are active by default, this allows to control them More...
void	set_tpa (const int &b)
	activates / deactivates two-point adaptation step-size mechanism. Overrides parameters::set_tpa by automatically setting dsigma value. More...
void	set_tpa_dsigma (const double &d)
	sets dsigma value, use with care. More...
Public Member Functions inherited from libcmaes::Parameters< TGenoPheno >
	Parameters ()
	empty constructor.
	Parameters (const int &dim, const double *x0, const int &lambda=-1, const uint64_t &seed=0, const TGenoPheno &gp=GenoPheno< NoBoundStrategy >())
	constructor More...
void	set_x0 (const double &x0)
	sets initial objective function parameter values to x0 across all dimensions More...
void	set_x0 (const double *x0)
	sets initial objective function parameter values to array x0 More...
void	set_x0 (const dVec &x0)
	sets initial objective function parameter values from Eigen vector More...
void	set_x0 (const double &x0min, const double &x0max)
	sets bounds on initial objective function parameter values. Bounds are the same across all dimensions, and initial value is sampled uniformly within these bounds. More...
void	set_x0 (const double *x0min, const double *x0max)
	sets bounds on initial objective function parameter values. Initial value is sampled uniformly within these bounds. More...
void	set_x0 (const std::vector< double > &x0min, const std::vector< double > &x0max)
	sets bounds on initial objective function parameter values. Initial value is sampled uniformly within these bounds. More...
void	set_x0 (const dVec &x0min, const dVec &x0max)
	sets bounds on initial objective function parameter values. Initial value is sampled uniformly within these bounds. More...
dVec	get_x0min () const
	returns lower bound on x0 vector More...
dVec	get_x0max () const
	returns upper bound on x0 vector More...
void	set_fixed_p (const int &index, const double &value)
	freezes a parameter to a given value during optimization. More...
void	unset_fixed_p (const int &index)
	unfreezes a parameter. More...
void	set_max_iter (const int &maxiter)
	sets the maximum number of iterations allowed for the optimization. More...
int	get_max_iter () const
	returns maximum number of iterations More...
void	set_max_fevals (const int &fevals)
	sets the maximum budget of objective function calls allowed for the optimization. More...
int	get_max_fevals () const
	returns maximum budget of objective function calls More...
void	set_ftarget (const double &val)
	sets the objective function target value when known. More...
void	reset_ftarget ()
	resets the objective function target value to its inactive state.
double	get_ftarget () const
	returns objective function target value. More...
void	set_seed (const int &seed)
	sets random generator's seed, 0 is special value to generate random seed. More...
int	get_seed () const
	returns random generator's seed. More...
void	set_ftolerance (const double &v)
	sets function tolerance as stopping criteria for TolHistFun: monitors the difference in function value over iterations and stops optimization when below tolerance. More...
double	get_ftolerance () const
	returns function tolerance More...
void	set_xtolerance (const double &v)
	sets parameter tolerance as stopping criteria for TolX. More...
double	get_xtolerance () const
	returns parameter tolerance More...
int	lambda () const
	returns lambda, number of offsprings per generation More...
int	dim () const
	returns the problem's dimension More...
void	set_quiet (const bool &quiet)
	sets the quiet mode (no output from the library) for the optimization at hand More...
bool	quiet () const
	returns whether the quiet mode is on. More...
void	set_algo (const int &algo)
	sets the optimization algorithm. More...
int	get_algo () const
	returns which algorithm is set for the optimization at hand. More...
void	set_gp (const TGenoPheno &gp)
	sets the genotype/phenotype transform object. More...
TGenoPheno	get_gp () const
	returns the current genotype/phenotype transform object. More...
void	set_fplot (const std::string &fplot)
	sets the output filename (activates the output to file). More...
void	set_full_fplot (const bool &b)
	activates / deactivates the full output (for legacy plotting). More...
std::string	get_fplot () const
	returns the current output filename. More...
void	set_gradient (const bool &gradient)
	activates the gradient injection scheme. If no gradient function is defined, injects a numerical gradient solution instead More...
bool	get_gradient () const
	returns whether the gradient injection scheme is activated. More...
void	set_edm (const bool &edm)
	activates computation of expected distance to minimum when optimization has completed More...
bool	get_edm () const
	returns whether edm is activated. More...
void	set_mt_feval (const bool &mt)
	activate / deactivate the parallel evaluation of objective function More...
bool	get_mt_feval () const
	returns whether the parallel evaluation of objective function is activated More...
void	set_max_hist (const int &m)
	sets maximum history size, allows to keep memory requirements fixed. More...
void	set_maximize (const bool &maximize)
	active internal maximization scheme (simply returns -f instead of f) More...
bool	get_maximize () const
	returns whether the maximization mode is enabled More...
void	set_initial_fvalue (const bool &b)
	whether to compute initial objective function value (i.e. at x0) More...
void	set_uh (const bool &b)
	activates / deactivates uncertainty handling scheme. More...
bool	get_uh () const
	get uncertainty handling status. More...
void	set_tpa (const int &b)
	activates / deactivates two-point adaptation step-size mechanism More...
int	get_tpa () const
	get two-point adapation step-size mechanism status. More...

Friends
class	CMASolutions
template<class U , class V >
class	CMAStrategy
template<class U , class V , class W >
class	ESOStrategy
template<class U >
class	CMAStopCriteria
template<class U , class V >
class	IPOPCMAStrategy
template<class U , class V >
class	BIPOPCMAStrategy
class	CovarianceUpdate
class	ACovarianceUpdate
template<class U >
class	errstats
class	VDCMAUpdate

Additional Inherited Members
Protected Attributes inherited from libcmaes::Parameters< TGenoPheno >
int	_dim
int	_lambda = -1
int	_max_iter = -1
int	_max_fevals = -1
bool	_quiet = true
std::string	_fplot = ""
bool	_full_fplot = false
dVec	_x0min
dVec	_x0max
double	_ftarget = -std::numeric_limits<double>::infinity()
double	_ftolerance = 1e-12
double	_xtol = 1e-12
uint64_t	_seed = 0
int	_algo = 0
bool	_with_gradient =false
bool	_with_edm =false
std::unordered_map< int, double >	_fixed_p
TGenoPheno	_gp
bool	_mt_feval = false
int	_max_hist = -1
bool	_maximize = false
bool	_initial_fvalue = false
bool	_uh = false
double	_rlambda
double	_epsuh = 1e-7
double	_thetauh = 0.2
double	_csuh = 1.0
double	_alphathuh = 1.0
int	_tpa = 1
double	_tpa_csigma = 0.3
Static Protected Attributes inherited from libcmaes::Parameters< TGenoPheno >
static std::map< std::string, int >	_algos = {{"cmaes",0},{"ipop",1},{"bipop",2},{"acmaes",3},{"aipop",4},{"abipop",5},{"sepcmaes",6},{"sepipop",7},{"sepbipop",8},{"sepacmaes",9},{"sepipop",10},{"sepbipop",11},{"vdcma",12},{"vdipopcma",13},{"vdbipopcma",14}}

Detailed Description

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>
class libcmaes::CMAParameters< TGenoPheno >

Parameters for various flavors of the CMA-ES algorithm.

Constructor & Destructor Documentation

template<class TGenoPheno >

libcmaes::CMAParameters< TGenoPheno >::CMAParameters	(	const int &	dim,
		const double *	x0,
		const double &	sigma,
		const int &	lambda = -1,
		const uint64_t &	seed = 0,
		const TGenoPheno &	gp = TGenoPheno()
	)

Constructor.

Parameters

dim	problem dimensions
x0	initial search point
sigma	initial distribution step size (positive, otherwise automatically set)
lambda	number of offsprings sampled at each step
seed	initial random seed, useful for reproducing results (if unspecified, automatically generated from current time)
gp	genotype / phenotype object
sep	whether to use sep-CMA-ES, using diagonal covariance matrix (modifies covariance default learning rate)

template<class TGenoPheno >

libcmaes::CMAParameters< TGenoPheno >::CMAParameters	(	const std::vector< double > &	x0,
		const double &	sigma,
		const int &	lambda = -1,
		const uint64_t &	seed = 0,
		const TGenoPheno &	gp = TGenoPheno()
	)

Constructor.

Parameters

x0	initial search point as vector of problem dimension
sigma	initial distribution step size (positive, otherwise automatically set)
lambda	number of offsprings sampled at each step
seed	initial random seed, useful for reproducing results (if unspecified, automatically generated from current time)
gp	genotype / phenotype object
sep	whether to use sep-CMA-ES, using diagonal covariance matrix (modifies covariance default learning rate)

template<class TGenoPheno >

libcmaes::CMAParameters< TGenoPheno >::CMAParameters	(	const std::vector< double > &	x0,
		const std::vector< double > &	sigma,
		const int &	lambda = -1,
		const std::vector< double > &	lbounds = std::vector<double>(),
		const std::vector< double > &	ubounds = std::vector<double>(),
		const uint64_t &	seed = 0
	)

Constructor.

Parameters

x0	initial search point as vector of problem dimension
sigma	vector of initial distribution step sizes (positive, otherwise automatically set)
lambda	number of offsprings sampled at each step
seed	initial random seed, useful for reproducing results (if unspecified, automatically generated from current time)
gp	genotype / phenotype object
sep	whether to use sep-CMA-ES, using diagonal covariance matrix (modifies covariance default learning rate)

Member Function Documentation

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

bool libcmaes::CMAParameters< TGenoPheno >::get_lazy_update ( )

inline

get lazy update status.

Parameters

whether

lazy update is activated

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

int libcmaes::CMAParameters< TGenoPheno >::get_restarts ( ) const

inline

get the number of restarts (applies to IPOP and BIPOP).

Returns

number of restarts

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

double libcmaes::CMAParameters< TGenoPheno >::get_sigma_init ( ) const

inline

returns initial sigma value

Returns

initial sigma value

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

bool libcmaes::CMAParameters< TGenoPheno >::is_sep ( ) const

inline

whether algorithm leverages separability.

Returns

separability status

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

bool libcmaes::CMAParameters< TGenoPheno >::is_vd ( ) const

inline

whether algorithm uses vd update.

Returns

vd status

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

void libcmaes::CMAParameters< TGenoPheno >::set_algo ( const int &  algo )

inline

sets the optimization algorithm. Note: overrides Parameters::set_algo

Parameters

algo	from CMAES_DEFAULT, IPOP_CMAES, BIPOP_CMAES, aCMAES, aIPOP_CMAES, aBIPOP_CMAES, sepCMAES, sepIPOP_CMAES, sepBIPOP_CMAES, sepaCMAES, sepaIPOP_CMAES, sepaBIPOP_CMAES, VD_CMAES, VD_IPOP_CMAES, VD_BIPOP_CMAES

template<class TGenoPheno >

void libcmaes::CMAParameters< TGenoPheno >::set_fixed_p	(	const int &	index,
		const double &	value
	)

freezes a parameter to a given value in genotype during optimization. Adapts some generic parameters as well.

Parameters

index	dimension index of the parameter to be frozen
value	frozen value of the parameter

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

void libcmaes::CMAParameters< TGenoPheno >::set_gradient ( const bool &  gradient )

inline

activates the gradient injection scheme. If no gradient function is defined, injects a numerical gradient solution instead Note: overrides Parameters::set_gradient

Parameters

gradient

true/false

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

void libcmaes::CMAParameters< TGenoPheno >::set_lazy_update ( const bool &  lz )

inline

sets the lazy update (i.e. updates the eigenvalues every few steps).

Parameters

lz	whether to activate the lazy update

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

void libcmaes::CMAParameters< TGenoPheno >::set_restarts ( const int &  nrestarts )

inline

sets the maximum number of restarts (applies to IPOP and BIPOP).

Parameters

nrestarts

maximum number of restarts

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

void libcmaes::CMAParameters< TGenoPheno >::set_stopping_criteria ( const int &  criteria, const bool &  active  )

inline

all stopping criteria are active by default, this allows to control them

Parameters

criteria	stopping criteria CMAStopCritType, see cmastopcriteria.h
active	whether to activate this criteria

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

void libcmaes::CMAParameters< TGenoPheno >::set_str_algo ( const std::string &  algo )

inline

sets the optimization algorithm.

Parameters

algo	as string from cmaes,ipop,bipop,acmaes,aipop,abipop,sepcmaes,sepipop,sepbipop,sepacmaes,sepaipop,sepabipop,vdcma,vdipopcma,vdbipopcma

template<class TGenoPheno >

void libcmaes::CMAParameters< TGenoPheno >::set_tpa

(

const int &

b

)

activates / deactivates two-point adaptation step-size mechanism. Overrides parameters::set_tpa by automatically setting dsigma value.

Parameters

b	0: no, 1: auto, 2: yes

template<class TGenoPheno = GenoPheno<NoBoundStrategy>>

void libcmaes::CMAParameters< TGenoPheno >::set_tpa_dsigma ( const double &  d )

inline

sets dsigma value, use with care.

Parameters

d

dsigma

template<class TGenoPheno >

void libcmaes::CMAParameters< TGenoPheno >::unset_fixed_p

(

const int &

index

)

unfreezes a parameter.

Parameters

index

dimenion index of the parameter to unfreeze

---

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

• src/cmaparameters.h
• src/cmaparameters.cc