Theorem 1 — two (s, S) policies with equal average cost below have equal average cost everywhere
ProvedVeinottWagnerSS.Selection.theorem1_aCost_eq_of_eq_belowinventorymarkov-chainp2o-batch-p200bp2o-gran-per-chapterp2o-plan-paperp2o-v1ss-policy
Let and consider two stationary policies and (so , ) with . Let be the equivalent average cost per period from the starting stock . If
then
The theorem reduces the comparison of two policies on all starting stocks to the comparison on the finitely many stocks that matter in Step iii of the paper's algorithm, and it is one of the two ingredients of Theorem 2.
Formalization Note The model's standing assumptions (convexity and coercivity of , , finite mean demand) are fields of Model; the statement is kept under them as in the paper, although it does not need them.
Preamble
import Mathlib import Definitions.Def_VeinottWagnerSS_Selection_Model
Formal statement
namespace VeinottWagnerSS.Selection
/-- Veinott & Wagner (1965), Theorem 1, p. 542: if `0 ≤ α < 1`, if the two `(s, S)` policies
`(s, S)`, `(s', S')` satisfy `s ≤ s'`, and if `a_α(x | s, S) = a_α(x | s', S')` for every
`x < s'`, then `a_α(x | s, S) = a_α(x | s', S')` for all `x`. -/
theorem theorem1_aCost_eq_of_eq_below (M : Model) (α : ℝ) (hα0 : 0 ≤ α) (hα1 : α < 1)
(s S s' S' : ℤ) (hsS : s ≤ S) (hsS' : s' ≤ S') (hss' : s ≤ s')
(heq : ∀ x : ℤ, x < s' → aCost M α s S x = aCost M α s' S' x) :
∀ x : ℤ, aCost M α s S x = aCost M α s' S' x := by sorry
end VeinottWagnerSS.Selection
Source
Veinott & Wagner, Computing Optimal (s, S) Inventory Policies, Management Sci. 11 (1965), p. 542, Theorem 1 (proof: Appendix §3, pp. 550-551)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.