Proposition 1 (iii), p. 6 — σ²_x̂(x*_min) ≤ liminf s²_n(x*_n) ≤ limsup s²_n(x*_n) ≤ σ²_x̂(x*_max), w.p.1
ProvedSolutionQuality.SRP.prop1_iiiAssume (A1)–(A3), let , and let be i.i.d. as ; let be optimal solutions of (SP). Let and be optimal solutions of (SP) with the smallest and the largest variance . Then, with probability one,
The sample variance computed at the SAA minimizer is asymptotically bracketed by the smallest and largest variances over the optimal set; when is a singleton it is a consistent estimator of . This is the variance ingredient of the validity proofs of Theorems 2–4.
Formalization Note. The chain of liminf and limsup is stated in its equivalent -form, which avoids the junk value of real liminf on unbounded sequences: almost surely, for every , eventually . and are given points with their defining argmin/argmax properties over .
import Mathlib import Definitions.Def_SolutionQuality_SRP_Setting
namespace SolutionQuality.SRP
open MeasureTheory ProbabilityTheory Filter
theorem prop1_iii {d : ℕ} {Ξ : Type*} [MeasurableSpace Ξ]
(μ : Measure Ξ) [IsProbabilityMeasure μ]
(f : E d → Ξ → ℝ) (hfm : ∀ x, Measurable (f x))
(X : Set (E d)) (hA : Assumptions μ f X)
{Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(ξ : ℕ → Ω → Ξ) (hξ : IsIIDSample P μ ξ)
(xhat : E d) (hxhat : xhat ∈ X)
(xn : ℕ → Ω → E d) (hxn : IsSAAMinimizerSeq P f X ξ xn)
(xmin xmax : E d)
(hxmin : xmin ∈ optSet μ f X ∧ ∀ x ∈ optSet μ f X, gapVar μ f xhat xmin ≤ gapVar μ f xhat x)
(hxmax : xmax ∈ optSet μ f X ∧ ∀ x ∈ optSet μ f X, gapVar μ f xhat x ≤ gapVar μ f xhat xmax) :
∀ᵐ ω ∂P, ∀ ε > 0, ∀ᶠ n in atTop,
gapVar μ f xhat xmin - ε ≤ sampleVar f xhat (fun i => ξ i ω) n (xn n ω) ∧
sampleVar f xhat (fun i => ξ i ω) n (xn n ω) ≤ gapVar μ f xhat xmax + ε := by sorry
end SolutionQuality.SRP
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.