Problem 10.1: is convex in , and it has a minimizer
ProvedInventoryControl.cs_total_cost_convexLet satisfy the fractile condition (10.8), with , and . Then the total cost of Eq. (10.9) is a convex function of the echelon position on , and, when , it attains its minimum at some .
Convexity is the content of the book's Problem 10.1: is convex exactly because minimizes the convex function , and convexity survives the expectation over and the addition of the linear term . For the minimum exists because in both directions: like as and like as .
The guard is needed. When and , is strictly above for every , because is strictly convex with its unique minimum at and is unbounded, and it tends to as . So there is no minimizer. This is the upstream mirror of the book's remark that forces (p. 196): with free upstream holding the optimal runs off to infinity. Convexity is what the book uses to justify searching for a local minimum only, and is the echelon order-up-to level of installation 2.
import Definitions.Def_InventoryControl_clarkScarf
namespace InventoryControl
theorem cs_total_cost_convex (e1 e2 b1 mu sigma : ℝ) (he1 : 0 ≤ e1) (he2 : 0 ≤ e2) (hb : 0 < b1)
(hs : 0 < sigma) (L1 L2 : ℕ) (S1 : ℝ)
(hS1 : ProbabilityTheory.cdf (ProbabilityTheory.gaussianReal 0 1)
((S1 - (L1 + 1) * mu) / (Real.sqrt (L1 + 1) * sigma)) = (e2 + b1) / (e1 + e2 + b1)) :
ConvexOn ℝ Set.univ (csTotalCost e1 e2 b1 mu sigma L1 L2 S1)
∧ (0 < e2 → ∃ S2 : ℝ, ∀ y2 : ℝ, csTotalCost e1 e2 b1 mu sigma L1 L2 S1 S2
≤ csTotalCost e1 e2 b1 mu sigma L1 L2 S1 y2) := by sorry
end InventoryControl
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.