Eq. (19a) — for uncorrelated b and c, E(c′Db) = (Ec)′D(Eb)
ProvedChanceDetEquiv.EModel.eq_19achance-constrained-programmingp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-programming
Let and be random vectors in and on a probability space, with integrable components , and integrable products . Suppose and are uncorrelated in the sense that for all (no assumption is made on the components of among themselves, or of ). Then for every matrix ,
This is the step that makes the objective of the E-model deterministic: the expected functional of (18) becomes the bilinear function of the decision rule.
Formalization Note The paper's "b and c are uncorrelated" is stated as the componentwise product rule for expectations, which is exactly what uncorrelatedness of the components means.
Preamble
import Mathlib import Definitions.Def_ChanceDetEquiv_EModel_Model open MeasureTheory ProbabilityTheory Matrix
Formal statement
namespace ChanceDetEquiv.EModel
/-- **Eq. (19a)**, p. 26: if `b` and `c` are uncorrelated (componentwise:
`E(c_j b_k) = E(c_j) E(b_k)`), then `E(c'Db) = (Ec)'D(Eb)` for every decision rule `D`. -/
theorem eq_19a {Ω : Type*} [MeasurableSpace Ω] {P : Measure Ω} [IsProbabilityMeasure P]
{m n : ℕ} (b : Ω → Fin m → ℝ) (c : Ω → Fin n → ℝ)
(hb : ∀ k, Integrable (fun ω => b ω k) P)
(hc : ∀ j, Integrable (fun ω => c ω j) P)
(hcb : ∀ j k, Integrable (fun ω => c ω j * b ω k) P)
(huncorr : ∀ j k, ∫ ω, c ω j * b ω k ∂P = (∫ ω, c ω j ∂P) * (∫ ω, b ω k ∂P))
(D : Matrix (Fin n) (Fin m) ℝ) :
expectedObjective P b c D = detObjective P b c D := by sorry
end ChanceDetEquiv.EModel
Source
Charnes and Cooper, Deterministic Equivalents for Optimizing and Satisficing under Chance Constraints, Oper. Res. 11 (1963), p. 26, Eq. (19a)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.