Serialized Form

Package fr.inria.optimization.cmaes

Class fr.inria.optimization.cmaes.CMAEvolutionStrategy extends java.lang.Object implements Serializable

serialVersionUID: 2918241407634253526L

Serialized Fields

versionNumber

java.lang.String versionNumber

options

CMAOptions options

options that can be changed (fields can be assigned) at any time to control the running behavior

sp

CMAParameters sp

parameters

CMAParameters parameters

strategy parameters that can be set initially

stopConditions

CMAEvolutionStrategy.StopCondition stopConditions

permits access to whether and which termination conditions were satisfied

N

int N

seed

long seed

rand

java.util.Random rand

math

fr.inria.optimization.cmaes.CMAEvolutionStrategy.MyMath math

axisratio

double axisratio

counteval

long counteval

countiter

long countiter

bestever_eval

long bestever_eval

bestever_x

double[] bestever_x

bestever_fit

double bestever_fit

sigma

double sigma

typicalX

double[] typicalX

initialX

double[] initialX

LBound

double[] LBound

UBound

double[] UBound

xmean

double[] xmean

xmean_fit

double xmean_fit

pc

double[] pc

ps

double[] ps

C

double[][] C

maxsqrtdiagC

double maxsqrtdiagC

minsqrtdiagC

double minsqrtdiagC

B

double[][] B

diagD

double[] diagD

flgdiag

boolean flgdiag

startsigma

double[] startsigma

maxstartsigma

double maxstartsigma

minstartsigma

double minstartsigma

iniphase

boolean iniphase

state

double state

state (postconditions): -1 not yet initialized 0 initialized init() 0.5 reSizePopulation 1 samplePopulation, sampleSingle, reSampleSingle 2.5 updateSingle 3 updateDistribution

citerlastwritten

long citerlastwritten

countwritten

long countwritten

lockDimension

int lockDimension

mode

int mode

SINGLE_MODE

int SINGLE_MODE

PARALLEL_MODE

int PARALLEL_MODE

countCupdatesSinceEigenupdate

long countCupdatesSinceEigenupdate

fit

fr.inria.optimization.cmaes.CMAEvolutionStrategy.FitnessCollector fit

recentFunctionValue

double recentFunctionValue

recentMaxFunctionValue

double recentMaxFunctionValue

recentMinFunctionValue

double recentMinFunctionValue

idxRecentOffspring

int idxRecentOffspring

arx

double[][] arx

population

double[][] population

recent population, no idea whether this is useful to be public

xold

double[] xold

BDz

double[] BDz

artmp

double[] artmp

propertiesFileName

java.lang.String propertiesFileName

properties

java.util.Properties properties

timings

fr.inria.optimization.cmaes.CMAEvolutionStrategy.Timing timings

fileswritten

java.lang.String[] fileswritten

Class fr.inria.optimization.cmaes.CMAEvolutionStrategy.CMAException extends java.lang.RuntimeException implements Serializable

serialVersionUID: 1L

Class fr.inria.optimization.cmaes.CMAOptions extends java.lang.Object implements Serializable

serialVersionUID: 2255162105325585121L

Serialized Fields

diagonalCovarianceMatrix

long diagonalCovarianceMatrix

number of initial iterations with diagonal covariance matrix, where 1 means always. Default is diagonalCovarianceMatrix=0, but this will presumably change in future. As long as iterations<=diagonalCovarianceMatrix the internal time complexity is linear in the search space dimensionality (memory requirements remain quadratic).

lowerStandardDeviations

double[] lowerStandardDeviations

lower bound for standard deviations (step sizes). The Array can be of any length. The i-th entry corresponds to the i-th variable. If length<dim the last entry is recycled for all remaining variables. Zero entries mean, naturally, no lower bound.

CAVE: there is an interference with stopTolX (and stopTolXFactor): if lowerStdDev is larger than stopTolX, the termination criterion can never be satisfied.

Example:

 CMAEvolutionStrategy es = new CMAEvolutionStrategy(); 
 es.options.lowerStandardDeviations = new double[]{1e-4,1e-8}; // 1e-8 for all but first variable
 

See Also:

CMAOptions.stopTolX, CMAOptions.stopTolXFactor

upperStandardDeviations

double[] upperStandardDeviations

upper bound for standard deviations (step lengths). Zero entries mean no upper bound. Be aware of the interference with option stopTolUpXFactor.

See Also:

CMAOptions.lowerStandardDeviations, CMAOptions.stopTolUpXFactor

stopFitness

double stopFitness

stop if function value drops below the target function value stopFitness. Default = Double.MIN_VALUE

stopTolFun

double stopTolFun

stop if the maximum function value difference of all iteration-best solutions of the last 10 + 30*N/lambda iterations and all solutions of the recent iteration become <= stopTolFun. Default = 1e-12.

stopTolFunHist

double stopTolFunHist

stop if the maximum function value difference of all iteration-best solutions of the last 10 + 30*N/lambda iterations become smaller than stopTolFunHist. Default = 1e-13. The measured objective function value differences do not include repair penalties.

stopTolX

double stopTolX

stop if search steps become smaller than stopTolX. Default = 0

stopTolXFactor

double stopTolXFactor

stop if search steps become smaller than stopTolXFactor * initial step size. Default = 1e-11.

stopTolUpXFactor

double stopTolUpXFactor

stop if search steps become larger than stopTolUpXFactor * initial step size. Default = 1e3. When this termination criterion applies on a static objective function, the initial step-size was chosen far too small (or divergent behavior is observed).

stopMaxFunEvals

long stopMaxFunEvals

stop if the number of objective function evaluations exceed stopMaxFunEvals

stopMaxIter

long stopMaxIter

stop if the number of iterations (generations) exceed stopMaxIter

stopnow

boolean stopnow

if true stopping message "Manual:..." is generated

verbosity

int verbosity

determines whether CMA says hello after initialization.

See Also:

CMAEvolutionStrategy.helloWorld()

outputFileNamesPrefix

java.lang.String outputFileNamesPrefix

Output files written will have the names outputFileNamesPrefix*.dat

writeDisplayToFile

int writeDisplayToFile

if chosen > 0 the console output from functions print... is saved additionally into a file, by default outcmaesdisp.dat

maxTimeFractionForEigendecomposition

double maxTimeFractionForEigendecomposition

only for >= 1 results are always exactly reproducible, as otherwise the update of the eigensystem is conducted depending on time measurements, defaut is 0.2

maxTimeFractionForWriteToDefaultFiles

double maxTimeFractionForWriteToDefaultFiles

default is 0.1

checkEigenSystem

int checkEigenSystem

checks eigendecomposition mainly for debugging purpose, default is 0==no-check; the function checkEigenSystem requires O(N^3) operations.

Class fr.inria.optimization.cmaes.CMAParameters extends java.lang.Object implements Serializable

serialVersionUID: -1305062342816588003L

Serialized Fields

supplemented

int supplemented

locked

int locked

lambda

int lambda

mu

int mu

mucov

double mucov

mueff

double mueff

weights

double[] weights

damps

double damps

cs

double cs

cc

double cc

ccov

double ccov

ccovsep

double ccovsep

chiN

double chiN

flgLambdaChanged

int flgLambdaChanged

recombinationType

CMAParameters.RecombinationType recombinationType

Class fr.inria.optimization.cmaes.CMASolution extends java.lang.Object implements Serializable

serialVersionUID: 6257830429350615236L

Serialized Fields

functionValue

double functionValue

objective function value of x

x

double[] x

argument to objective function to be optimized

evaluation

long evaluation

count when the solution was evaluated