KSLCore/ksl.simopt.solvers.algorithms/CENormalSampler

CENormalSampler

class CENormalSampler(val problemDefinition: ProblemDefinition, meanSmoother: Double = defaultMeanSmoother, sdSmoother: Double = defaultStdDevSmoother, coefficientOfVariationThreshold: Double = defaultCoefficientOfVariationThreshold, streamNum: Int = 0, val streamProvider: RNStreamProviderIfc = KSLRandom.DefaultRNStreamProvider) : CESamplerIfc(source)

Constructors

CENormalSampler

constructor(problemDefinition: ProblemDefinition, meanSmoother: Double = defaultMeanSmoother, sdSmoother: Double = defaultStdDevSmoother, coefficientOfVariationThreshold: Double = defaultCoefficientOfVariationThreshold, streamNum: Int = 0, streamProvider: RNStreamProviderIfc = KSLRandom.DefaultRNStreamProvider)

Types

Companion

object Companion

Properties

cvThreshold

var cvThreshold: Double

This threshold represents the bound used to consider whether the coefficient of variation for the population parameters has converged.

dimension

open override val dimension: Int

the expected size of the array from sample()

eliteStatistics

val eliteStatistics: List<StatisticIfc>

A list of statistics for the population within the last sample of elites.

initialVariabilityFactor

var initialVariabilityFactor: Double

This can be used to increase/decrease the variability associated with the initial parameter setting. For example, a value of 1.1 increases the starting standard deviation by 10%. The setting must be a positive value. The default value is specified by defaultInitialVariabilityFactor.

means

val means: DoubleArray

A copy of the current mean parameters.

meanSmoother

var meanSmoother: Double

Used to smooth the estimated values of the mean parameters via exponential smoothing.

problemDefinition

open override val problemDefinition: ProblemDefinition

The problem definition associated with the sampling process.

rnStream

open override val rnStream: RNStreamIfc

rnStream provides a reference to the underlying stream of random numbers.

sdSmoother

var sdSmoother: Double

Used to smooth the estimated values of the standard deviation parameters via exponential smoothing.

stdDeviations

val stdDeviations: DoubleArray

A copy of the current standard deviation parameters.

stdDeviationThresholds

val stdDeviationThresholds: DoubleArray

Computes the standard deviation thresholds used to check if the distribution has converged. These values are abs(mean)*cvThreshold

streamNumber

open override val streamNumber: Int

The assigned stream number for the generation process

streamProvider

open override val streamProvider: RNStreamProviderIfc

The stream provider associated with the sampling process.

Functions

hasConverged

open override fun hasConverged(): Boolean

The sampler is considered to be converged if the component standard deviations are less than their standard deviation thresholds based on the coefficient of variation threshold.

initializeParameters

open override fun initializeParameters(values: DoubleArray)

This function sets the initial mean and standard deviation parameters for the sampler. The mean values are specified by the supplied array. The initial standard deviation values are set based on the range of the input values associated with the problem. If the range of possible values is higher, then the initial standard deviation is higher. The range is used to approximate the standard deviation. The initialVariabilityFactor can be used to inflate or deflate this setting as needed.

parameters

open override fun parameters(): DoubleArray

sample

open override fun sample(array: DoubleArray)

Returns a sample from the current cross-entropy distribution.

toString

open override fun toString(): String

updateParameters

open override fun updateParameters(elites: List<DoubleArray>)

The underlying parameters of the sampling mechanism should be updated. Implementors need to handle the edge case of less than 2 elites.