(3.3), p. 8 —
ProvedScenarioExact.PartOne.prob_S_eq_integralLet be a probability space and consider the uncertain program (1.1) with , convex closed and convex closed , such that every finite instance has a unique solution , and assume the problem is fully-supported. Let be the distribution of the violation of , as in (3.1). Then for every and every set of cardinality ,
where is the set of multi-extractions whose support constraints have exactly the indexes in .
Since the sets partition up to a null set, (3.3) yields the moment identities (3.4) that determine .
Formalization Note. The integral is the lower Lebesgue integral of (as an extended nonnegative real) over against the law (violationLaw); is concentrated on , so this is the paper's , atom at included. The statement is for every of cardinality , as printed. No measurability of is assumed: the left side is the outer measure, and the paper's "measurability ... is assumed for granted" (p. 4) is replaced by the joint measurability of and the measurability of the solution maps. Existence and uniqueness are assumed for every number of constraints, with the solution family θs; the nonempty-interior clause of Assumption 1 is not used and not assumed. Indexes are 0-based.
import Mathlib import Definitions.Def_ScenarioApproach_Generalization_violation import Definitions.Def_ScenarioApproach_Generalization_scenarioProgram import Definitions.Def_ScenarioApproach_Generalization_supportConstraint import Definitions.Def_ScenarioExact_PartOne_Basic open MeasureTheory ScenarioApproach.Generalization open scoped ENNReal
namespace ScenarioExact.PartOne
/-- (3.3), p. 8: for a fully-supported problem, `m ≥ d` and every index set `𝓘` of
cardinality `d`, `ℙ^m{S_𝓘} = ∫_0^1 (1 − α)^{m−d} F(dα)`, with `F` the law (3.1). -/
theorem prob_S_eq_integral
{d : ℕ} {Δ : Type*} [MeasurableSpace Δ]
(P : Measure Δ) [IsProbabilityMeasure P]
(c : EuclideanSpace ℝ (Fin d)) (Θ : Set (EuclideanSpace ℝ (Fin d)))
(Θδ : Δ → Set (EuclideanSpace ℝ (Fin d)))
-- Assumption 1 / (1.1): `𝒳` and every `𝒳_δ` are convex and closed
(hΘ_convex : Convex ℝ Θ) (hΘ_closed : IsClosed Θ)
(hΘδ_convex : ∀ δ, Convex ℝ (Θδ δ)) (hΘδ_closed : ∀ δ, IsClosed (Θδ δ))
-- Assumption 1: for every number `k` of constraints and every sample, `P_k` has a unique solution
(hexu : ∀ (k : ℕ) (ω : Fin k → Δ), ∃! θ, IsSolution c Θ Θδ ω θ)
-- measurability ("assumed for granted", p. 4): the constraint relation is jointly measurable
(hmeas : MeasurableSet {p : EuclideanSpace ℝ (Fin d) × Δ | p.1 ∈ Θδ p.2})
-- `θs k ω` is the solution `x*_k` of `P_k` with the sampled constraints `ω`; measurable in `ω`
(θs : (k : ℕ) → (Fin k → Δ) → EuclideanSpace ℝ (Fin d))
(hθs : ∀ (k : ℕ) (ω : Fin k → Δ), IsSolution c Θ Θδ ω (θs k ω))
(hθs_meas : ∀ k, Measurable (θs k))
-- Definition 2.3: the problem is fully-supported
(hfs : FullySupported c Θ Θδ P)
{m : ℕ} (hm : d ≤ m) (I : Finset (Fin m)) (hI : I.card = d) :
Measure.pi (fun _ : Fin m => P) (S c Θ Θδ θs I) =
∫⁻ α in Set.Icc (0 : ℝ) 1, ENNReal.ofReal ((1 - α) ^ (m - d)) ∂(violationLaw P Θδ θs) := by sorry
end ScenarioExact.PartOne
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.