Positive retirement value below the Gittins index
ProvedBanditAlgorithm.gittinsRetirementValue_pos_of_lt_indexbanditsgittins-indexoptimal-stopping
If the charge is strictly below the Gittins index , an epsilon-optimal admissible stopping rule has strictly positive expected discounted net reward. Consequently .
Preamble
import Definitions.Def_GittinsRetirementValue open MeasureTheory ProbabilityTheory ENNReal
Formal statement
theorem BanditAlgorithm.gittinsRetirementValue_pos_of_lt_index
{S : Type*} [MeasurableSpace S]
(P : Kernel S S) [IsMarkovKernel P]
{r : S → ℝ} (hr : Measurable r) {α : ℝ}
(hα0 : 0 < α) (hα1 : α < 1)
(hint : DiscountedRewardIntegrable P r α)
(x : S) (γ : ℝ) (hg : γ < gittinsIndex P r α x) :
0 < gittinsRetirementValue P r α γ x := by
sorrySource
Lattimore and Szepesvári, Bandit Algorithms (CUP 2020), fair-charge characterization Eq. (35.8), ratio characterization Eq. (35.9), printed pp.448--449.