Lemma 7, p. 33 — lower and upper bounds on a dual minimizer
ProvedOTDRO.StrongCvx.dual_optimizer_boundsdistributionally-robust-optimizationoptimal-transportp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1strong-convexity
Assume the paper's Assumptions 1–3 and . For , let minimize the extended-real dual objective over . Write . Then
These are the paper's and ; Proposition 1 makes them uniform over a compact decision set.
Formalization Note Membership is stated separately because Lean's IsMinOn predicate alone does not include membership in the minimizing set.
Preamble
import Mathlib import Definitions.Def_OTDRO_StrongCvx_Assumptions
Formal statement
namespace OTDRO.StrongCvx
open MeasureTheory
/-- Lemma 7, p. 33: explicit lower and upper bounds for any nonnegative
dual minimizer. -/
theorem dual_optimizer_bounds {d : ℕ}
(P0 : Measure (EuclideanSpace ℝ (Fin d))) [IsProbabilityMeasure P0]
(A : EuclideanSpace ℝ (Fin d) → Matrix (Fin d) (Fin d) ℝ)
(ρmin ρmax : ℝ) (hA : OTDRO.Dual.Assumption1 P0 A ρmin ρmax)
(ℓ : ℝ → ℝ) (h2 : OTDRO.Dual.Assumption2 P0 ℓ)
(δ : ℝ) (hδ : 0 < δ) (B : Set (EuclideanSpace ℝ (Fin d)))
(hB : Convex ℝ B) (M : ℝ) (h3 : Assumption3 P0 ℓ B M)
(β : EuclideanSpace ℝ (Fin d)) (hβB : β ∈ B)
(lamStar : ℝ) (hlam : 0 ≤ lamStar)
(hmin : IsMinOn (fun lam => OTDRO.Dual.fDelta P0 ℓ A δ β lam) (Set.Ici 0) lamStar) :
(1 / 2) / Real.sqrt ρmax * ‖β‖ *
Real.sqrt (∫ x, (deriv ℓ (inner ℝ β x)) ^ 2 ∂P0) ≤ lamStar ∧
lamStar ≤ (1 / Real.sqrt ρmin) * ‖β‖ *
Real.sqrt (∫ x, (deriv ℓ (inner ℝ β x)) ^ 2 ∂P0) +
(1 / 2) * Real.sqrt δ * M / ρmin * ‖β‖ ^ 2 := by sorry
end OTDRO.StrongCvx
Source
Blanchet, Murthy & Zhang, arXiv:1810.02403v3, Lemma 7, p. 33; proof pp. 33–34
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.