and is convex
ProvedSherbrooke.lowerHull_le_and_convexdiscrete-convexityinventorymarginal-analysismetric
Let be bounded below and let be its convexified function (the lower boundary of the convex hull of ). Then
- for every , and
- is convex on : for every ,
These are the two properties Sherbrooke's definition of asserts ("lies on the boundary of the convex hull", "decreased to the point of convexity", "") and that the proof of the Appendix Theorem uses ("Since the function is convex").
Differences from the source:
- generalization: Ξ_i monotonic decreasing (standing hypothesis) → hypothesis dropped; reason: this step does not use it; the statement specializes to the source's setting by adding it
- encoding: Ξ_i′ defined on the boundary of the convex hull of Ξ_i with Ξ_i′ ≤ Ξ_i → lowerHull Ξ m, the supremum of a m + b over all affine minorants k ↦ a k + b of Ξ on ℕ; reason: on ℕ this is the same function: the lower boundary of the convex hull of the points (k, Ξ(k)), equal to the greatest discretely convex minorant
- encoding: finite lower bound → BddBelow (Set.range Ξ); reason: same instances
- encoding: Ξ_i′ convex → forward differences of Ξ′ nondecreasing on ℕ; reason: discrete convexity of a sequence; same instances
Preamble
import Mathlib import Definitions.Def_Sherbrooke_lowerHull
Formal statement
namespace Sherbrooke
/-- **M1: `Ξ′` lies below `Ξ` and is convex.** Sherbrooke, *METRIC*, Operations Research 16
(1968), unnumbered, pp. 135–136: "Let us define a new function Ξ_i′(m) which lies on the boundary
of the convex hull of Ξ_i(m). In other words at any point where Ξ_i(m) is not convex the function
is decreased to the point of convexity in forming the new function Ξ_i′(m)." Used in the proof of
the Appendix Theorem, p. 140: "Since the function Ξ_i′ is convex".
If `Ξ : ℕ → ℝ` is bounded below, then `lowerHull Ξ ≤ Ξ` pointwise and `lowerHull Ξ` is
discretely convex: its forward differences are nondecreasing.
**Formalization Note.** Discrete convexity on `ℕ` is
`Ξ′(m+1) − Ξ′(m) ≤ Ξ′(m+2) − Ξ′(m+1)` for all `m`. The paper's standing hypothesis that `Ξ` is
monotonic decreasing is not needed for this step and is dropped (a generalization). -/
theorem lowerHull_le_and_convex (Ξ : ℕ → ℝ) (hbdd : BddBelow (Set.range Ξ)) :
(∀ m : ℕ, lowerHull Ξ m ≤ Ξ m) ∧
∀ m : ℕ, lowerHull Ξ (m + 1) - lowerHull Ξ m ≤
lowerHull Ξ (m + 2) - lowerHull Ξ (m + 1) := by sorry
end Sherbrooke
Source
Sherbrooke, METRIC: A Multi-Echelon Technique for Recoverable Item Control, Operations Research 16 (1968), doi:10.1287/opre.16.1.122, unnumbered, pp. 135–136 (Ξ_i′ on the boundary of the convex hull, decreased to the point of convexity); used in the Appendix, proof of the Theorem, p. 140