Proof of Theorem 3.1, pp. 22–23 — the instance is symmetric about (1, (1/2, …, 1/2)) and satisfies (2.1)
ProvedRobustPower.CostGap.instance_symmetricLet , let be a probability space and let be measurable, with range exactly the cube , and such that the coordinates are independent and uniformly distributed on (the law of under is the product of the uniform distributions on ). Put for every , so that the uncertainty set is . Then
In particular the scenario at the point of symmetry has and , so the measure satisfies the paper's condition (2.1). This shows that the large gap of Theorem 3.1 occurs under the symmetry hypotheses that suffice, for right-hand-side uncertainty alone, to bound the gap by 2.
Formalization Note The uncertainty set lives in , and symmetry is Definition 1.2 in that product group; the expectation of is the Bochner integral of the vector-valued map .
import Mathlib import Definitions.Def_RobustPower_StochGap_SymmetricSets open MeasureTheory
namespace RobustPower.CostGap
/-- Proof of Theorem 3.1, pp. 22–23: the uncertainty set `{(b(ω), d(ω))} = {1} × [0, 1]ⁿ` is
symmetric about `(1, (1/2, …, 1/2))`, and `E_μ[d(ω)] = (1/2, …, 1/2)`, so (2.1) holds. -/
theorem instance_symmetric {n : ℕ} (hn : 0 < n)
{Ω : Type*} [MeasurableSpace Ω] (μ : Measure Ω) [IsProbabilityMeasure μ]
(d : Ω → Fin n → ℝ) (hd : Measurable d) (hrange : Set.range d = Set.Icc 0 1)
(hlaw : μ.map d = Measure.pi (fun _ : Fin n => volume.restrict (Set.Icc (0 : ℝ) 1))) :
RobustPower.StochGap.IsSymmetricAbout (Set.range fun ω => ((fun _ => 1 : Fin 1 → ℝ), d ω))
((fun _ => 1 : Fin 1 → ℝ), (fun _ => 1 / 2 : Fin n → ℝ)) ∧
∫ ω, d ω ∂μ = (fun _ => 1 / 2 : Fin n → ℝ) := by sorry
end RobustPower.CostGap
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.