KSLCore/ksl.app.config.optimization/CESamplerSpec

CESamplerSpec

@Serializable

Cross-entropy sampler selection.

The CE algorithm parameterizes a sampling distribution and updates it each iteration based on the elite sample. This sealed type lets future sampler implementations be added without breaking existing SolverSpec.CrossEntropy documents. Currently only the multivariate normal sampler is exposed.

Inheritors

Normal

Constructors

CESamplerSpec

protected constructor()

Types

Normal

@Serializable

@SerialName(value = "normal")

data class Normal(val meanSmoother: Double = 0.85, val sdSmoother: Double = 0.85, val coefficientOfVariationThreshold: Double = 0.03) : CESamplerSpec

Multivariate-normal cross-entropy sampler; mirrors ksl.simopt.solvers.algorithms.CENormalSampler.