kumulant

kumulant/com.eignex.kumulant.bandit.univariate/EpsilonGreedySpec

EpsilonGreedySpec

@Serializable

@SerialName(value = "EpsilonGreedy")

data class EpsilonGreedySpec(val epsilon: Double = 0.1, val priorMean: Double = 0.0, val priorWeight: Double = 0.02, val priorSquaredDeviations: Double = 0.02) : BanditPolicySpec<WeightedVarianceResult> (source)

Spec for EpsilonGreedy.

Constructors

EpsilonGreedySpec

constructor(epsilon: Double = 0.1, priorMean: Double = 0.0, priorWeight: Double = 0.02, priorSquaredDeviations: Double = 0.02)

Properties

val epsilon: Double

Probability of exploring uniformly.

val priorMean: Double

Per-arm prior on the running reward mean.

priorSquaredDeviations

val priorSquaredDeviations: Double

Prior on Sum (x - mean)^2 * w.

val priorWeight: Double

Per-arm prior pseudo-count.

Functions

fun <R : Result> BanditPolicySpec<R>.materialize(): BanditPolicy<R>

Build a live BanditPolicy from its spec.

EpsilonGreedySpec

constructor(epsilon: Double = 0.1, priorMean: Double = 0.0, priorWeight: Double = 0.02, priorSquaredDeviations: Double = 0.02)(source)

epsilon

val epsilon: Double(source)

Probability of exploring uniformly.

priorMean

val priorMean: Double(source)

Per-arm prior on the running reward mean.

priorSquaredDeviations

val priorSquaredDeviations: Double(source)

Prior on Sum (x - mean)^2 * w.

priorWeight

val priorWeight: Double(source)

Per-arm prior pseudo-count.