Theorem 2.4 — Jensen lower bound for the expected recourse
ProvedKallMayerSLP.Recourse.theorem_2_4Let be fixed, let be affine in as in (2.1), let the cost be deterministic, and let be an integrable random vector in with distribution and mean . Assume complete recourse (Assumption 2.1) and (Assumption 2.2). Then for every the expected recourse is a finite integral, is finite, and
The bound replaces the distribution by the one-point distribution at its mean; it is the starting point of the successive discrete approximation (bounding) methods of the chapter.
Formalization Note is the Bochner integral and the published KallMayer.Recourse.ExpectedRecourse with constant cost (the field q0). The conclusion asserts integrability of the recourse value, finiteness of the left-hand side, and the inequality between real numbers.
Differences from the source:
- encoding: probability space (Ω, ℱ, P) with ℱ = σ(ξ) and the induced distribution ℙ_ξ on ℝʳ → a probability measure μ = ℙ_ξ on
Fin r → ℝ(canonical space, ξ = id); reason: the statement involves ξ only through ℙ_ξ; same instances - encoding: T(ξ), h(ξ), q(ξ) affine in ξ by (2.1), W(ξ) ≡ W fixed →
RandomDatawith fields T0, Tj, h0, hj, q0, qj and one fixed matrix W; reason: notation only - encoding: recourse value Q(x; T(ξ), h(ξ), q(ξ)) as an inf with values ±∞ → the published
KallMayer.Recourse.PointwiseRecourse, anERealinfimum (+∞ if infeasible, −∞ if unbounded); reason: the book's convention of p.200; same values - encoding: 𝒬(x) = 𝔼_ξ[Q(x; T(ξ), h(ξ))] → the real part of Q(x; ξ) is integrable (with measurability), and 𝒬 = the published
KallMayer.Recourse.ExpectedRecourse(Bochner integral of the real part); reason: under Assumption 2.1 and the ξ-independent Assumption 2.2 the recourse LP is primal and dual feasible for every ξ, so Q(x; ξ) is finite everywhere and its real part is Q itself - encoding: the inequality (2.14) between extended values → Q(x; T(ξ̄), h(ξ̄)) ≠ ⊥, ≠ ⊤, and the inequality between its real part and 𝒬(x); reason: the left-hand side is finite for the same reason, so no infinite value is collapsed
- encoding: q(ξ) ≡ q deterministic → the field
D.q0as constant cost; reason: same instances - generalization: standing assumption ξ ∈ L² (p.199) → ξ integrable (
Integrable id μ); reason: the text before the theorem (p.211) notes that Cor. 2.1 gives finiteness of 𝒬 for ξ ∈ L¹ in this deterministic-q setting; L² ⊂ L¹ on a probability space, so this specializes to the book's statement
import Mathlib import Definitions.Def_KallMayer_Recourse_CompleteRecourse import Definitions.Def_KallMayer_Recourse_PointwiseRecourse import Definitions.Def_KallMayer_Recourse_ExpectedRecourse import Definitions.Def_KallMayerSLP_Recourse_RandomData open MeasureTheory open KallMayer.Recourse (CompleteRecourse PointwiseRecourse ExpectedRecourse) open Matrix
namespace KallMayerSLP.Recourse
/-- Theorem 2.4 (Kall & Mayer, Stochastic Linear Programming (2005), Chapter 3, Theorem 2.4, p.211,
PDF 220): given the Assumptions 2.1 and 2.2, with `ξ̄ = 𝔼_ξ[ξ]`, the expected recourse
`𝒬(x) = 𝔼_ξ[Q(x; T(ξ), h(ξ))]` is bounded below due to `Q(x; T(ξ̄), h(ξ̄)) ≤ 𝒬(x)` (2.14).
Context (p.210): `q(ξ) ≡ q` deterministic, Assumption 2.2 reading `{u | Wᵀu ≤ q} ≠ ∅`.
**Formalization Note.** `μ = ℙ_ξ` on `ℝʳ`, `ξ̄ = ∫ ξ dμ`; the deterministic cost is `D.q0`. `ξ` is only
assumed integrable (`ξ ∈ L¹`, which Cor. 2.1 shows suffices, p.211) instead of `ξ ∈ L²`. The
conclusion asserts, for every `x`, that `Q(x; ξ)` has an integrable real part (so `𝒬(x)` is a genuine
finite integral), that `Q(x; T(ξ̄), h(ξ̄))` is finite, and the inequality (2.14) between the real
numbers. -/
theorem theorem_2_4 {m₂ n₁ n₂ r : ℕ} (W : Matrix (Fin m₂) (Fin n₂) ℝ) (D : RandomData m₂ n₁ n₂ r)
(μ : Measure (Fin r → ℝ)) [IsProbabilityMeasure μ] (hξ : Integrable id μ)
(hW : CompleteRecourse W)
(hdual : ∃ u : Fin m₂ → ℝ, W.transpose *ᵥ u ≤ D.q0)
(x : Fin n₁ → ℝ) :
Integrable (fun ξ => (PointwiseRecourse W (D.T ξ) (D.h ξ) D.q0 x).toReal) μ ∧
PointwiseRecourse W (D.T (∫ ξ, ξ ∂μ)) (D.h (∫ ξ, ξ ∂μ)) D.q0 x ≠ ⊥ ∧
PointwiseRecourse W (D.T (∫ ξ, ξ ∂μ)) (D.h (∫ ξ, ξ ∂μ)) D.q0 x ≠ ⊤ ∧
(PointwiseRecourse W (D.T (∫ ξ, ξ ∂μ)) (D.h (∫ ξ, ξ ∂μ)) D.q0 x).toReal ≤
ExpectedRecourse μ W D.T D.h (fun _ => D.q0) x := by sorry
end KallMayerSLP.Recourse
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.