Lemma 9 (corrected) — E[K(θ,z)] ≤ 1 + 2z/(E[D₁]−θ) + 8π² E[(D₁−E[D₁])⁴]/(E[D₁]−θ)⁴
ProvedAIMInventory.UnknownRho.lemma_9_correctedLet be i.i.d. nonnegative demands with finite positive mean and finite fourth moment, . For real and let
be the first time the random walk started at with increments enters . If , then for every
The paper prints this lemma without the leading :
where denotes the fourth central moment. As printed it is false: for deterministic demand , and the walk is positive at time and equals at time , so , while the printed right side is . The printed proof counts the integers with as at most ; the correct count is at most , which gives the statement above.
The lemma makes the first term of Theorem 8 explicit: with a finite fourth moment, the expected hitting time there is finite and depends only on and the moments of demand, so that term is .
Formalization Note. is a Lebesgue integral in and the right side is embedded with ENNReal.ofReal; the right side is positive under the hypotheses. The fourth central moment is a Bochner integral, which is honest because is a hypothesis.
import Mathlib import Definitions.Def_AIMInventory_UnknownRho_Model open MeasureTheory open scoped ENNReal NNReal
namespace AIMInventory.UnknownRho
/-- Lemma 9, p. 13, corrected by the `+1` needed to count the integer
times below `2z/(E[D]-θ)`. The printed bound fails for deterministic demand
and a sufficiently small positive `z`. -/
theorem lemma_9_corrected (P : DemandLaw) (θ z : ℝ)
(hθ : θ < CappedBaseStock.mean P) (hz : 0 < z)
(hfourth : Integrable (fun d : ℝ≥0 => (d : ℝ) ^ 4) P.law) :
expectedK P θ z ≤ ENNReal.ofReal
(1 + 2 * z / (CappedBaseStock.mean P - θ) +
8 * Real.pi ^ 2 * centralFourth P /
(CappedBaseStock.mean P - θ) ^ 4) := by sorry
end AIMInventory.UnknownRho
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.