Proof of Theorem 4, p. 939 — the optimal sales in (3) are a = min{d, v₀ − v₁}, so κ̄_t(v, ζ | d) = q̂(v₀ − v₁ − d) + γ f̄_{t+1}(v₊)
ProvedZipkinLostSales.Variability.kappa_fbar_eqConsider the lost-sales model of Zipkin (2008) in the transformed state (nonnegative vectors , with ), with lead time , costs , discount factor , and one-period demand law , a probability measure on with finite mean. Let be an optimal cost function of the recursion (any number of periods to go), and let be the value of program (4) with continuation .
Then for every , every and every demand , the choice (sell as much as possible) is optimal in program (3), that is, is optimal in (4), and
This identity is what makes the decision-before-demand problem (5), , agree with the recursion (1). The paper remarks that it is "not hard, but also not essential for this argument"; it is the bridge between the recursion and the program in which the joint convexity of Theorem 11's proof is stated.
Formalization Note The optimal cost is indexed by the number of periods to go ("for all " becomes "for all "); kappa is program (4) as an infimum over . The conditions , read the paper's "unit cost" and "discount rate"; finite mean of demand is not in the paper and is added so that is finite. The statement is for the model's own : it is false for an arbitrary continuation function.
import Mathlib import Definitions.Def_ZipkinLostSales_Variability_Model open MeasureTheory
namespace ZipkinLostSales.Variability
/-- Proof of Theorem 4 (Zipkin 2008, p. 939): in program (3)/(4) with continuation `f̄_{t+1}` the
optimal sales are `a = min{d, v₀ − v₁}`, so that `κ̄(v, ζ | d) = q̂(v₀ − v₁ − d) + γ f̄_{t+1}(v₊)`
with `v₊` the transition of p. 938. Stated for one demand law `μ` and every number of periods
to go `k`. -/
theorem kappa_fbar_eq {L : ℕ} (hL : 0 < L) (c hh p γ : ℝ) (hc : 0 ≤ c) (hhh : 0 ≤ hh)
(hp : 0 ≤ p) (hγ0 : 0 < γ) (hγ1 : γ ≤ 1) (μ : Measure ℝ) [IsProbabilityMeasure μ]
(hμ : μ (Set.Iio 0) = 0) (hμint : Integrable id μ) :
∀ k : ℕ, ∀ v ∈ ZipkinLostSales.LNatural.V L, ∀ ζ : ℝ, ζ ≤ 0 → ∀ d : ℝ, 0 ≤ d →
ZipkinLostSales.LNatural.kappa hh p γ (ZipkinLostSales.LNatural.fbar c hh p γ μ k) d v ζ
= ZipkinLostSales.LNatural.qhat hh p (ZipkinLostSales.LNatural.vext v 0 - ZipkinLostSales.LNatural.vext v 1 - d) + γ * ZipkinLostSales.LNatural.fbar c hh p γ μ k (ZipkinLostSales.LNatural.next v d ζ) := by sorry
end ZipkinLostSales.Variability
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.