Integrability of stopped discounted rewards
ProvedBanditAlgorithm.integrable_discountedStoppedSumbanditsintegrabilityoptimal-stopping
Under Assumption 35.6, every discounted reward sum stopped at an adapted stopping time is integrable. Pathwise, its absolute value is dominated by the full discounted absolute-reward series.
Preamble
import Mathlib.MeasureTheory.Constructions.Polish.Basic import Mathlib.Analysis.Normed.Group.InfiniteSum import Mathlib.Analysis.SpecificLimits.Normed import Definitions.Def_GittinsIndex open MeasureTheory ProbabilityTheory ENNReal
Formal statement
theorem BanditAlgorithm.integrable_discountedStoppedSum
{S : Type*} [MeasurableSpace S]
(P : Kernel S S) [IsMarkovKernel P]
{r : S → ℝ} (hr : Measurable r) {α : ℝ} (hα0 : 0 ≤ α)
(hint : DiscountedRewardIntegrable P r α) (x : S)
{τ : (ℕ → S) → ℕ∞} (hτ : IsTrajStoppingTime τ) :
Integrable (discountedStoppedSum α r τ) (markovChainMeasure P x) := by
sorrySource
Lattimore and Szepesvári, Bandit Algorithms (CUP 2020), Assumption 35.6 and the discounted retirement game, printed pp.448--449.