kumulant

TopTwoThompsonSpec

@Serializable
@SerialName(value = "TopTwoThompson")
data class TopTwoThompsonSpec<R : Result>(val nbrArms: Int, val policy: ThompsonSamplingSpec<R>, val beta: Double = 0.5, val maxResamples: Int = 32) : UnivariateBanditSpec(source)

Spec for TopTwoThompsonBandit.

Constructors

TopTwoThompsonSpec

constructor(nbrArms: Int, policy: ThompsonSamplingSpec<R>, beta: Double = 0.5, maxResamples: Int = 32)(source)

Properties

beta

Probability of playing the round's top sample.

maxResamples

Cap on the resample loop when searching for the second arm.

nbrArms

Number of arms in the population.

policy

Thompson sampling policy carrying the per-arm posterior.