Proof of Theorem 4.4, pp. 15–16 — under Assumption 4.1, the optimal values of (12f) and (13) coincide
ProvedWassDDRO.Extremal.program12f_eq_program13Let , let samples lie in , let , and let be measurable and satisfy Assumption 4.1. Then
equals
that is, the optimal value of program (12f) equals the optimal value of program (13), with the conventions of p. 14 for .
This is the first part of the proof of Theorem 4.4. It is a statement about two finite-dimensional programs and does not involve probability measures; combined with the identity (10) = (12f) from the proof of Theorem 4.2, it gives the first claim of Theorem 4.4.
Formalization Note Both sides are EReal values of programs over feasibility predicates (see the Setting definitions). The measurability hypothesis on is the paper's standing assumption on the pieces (p. 11); it is not used by this finite-dimensional statement but kept for uniformity.
import Mathlib import Definitions.Def_WassDDRO_Extremal_Setting
namespace WassDDRO.Extremal
/-- First part of the proof of Theorem 4.4, pp. 15–16: under Assumption 4.1 the optimal value
of (12f) equals the optimal value of (13), for every ε ≥ 0 (Lagrangian dual (14a), (14b),
minimax (14c), Lemma 4.5, (14d)–(14g), rescaling). -/
theorem program12f_eq_program13 {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E]
[FiniteDimensional ℝ E] [MeasurableSpace E] [BorelSpace E]
{K N : ℕ} (hK : 0 < K) (hN : 0 < N) (Ξ : Set E) (ℓ : Fin K → E → EReal)
(hmeas : ∀ k, Measurable (ℓ k)) (ξhat : Fin N → E) (hξ : ∀ i, ξhat i ∈ Ξ)
(ε : ℝ) (hε : 0 ≤ ε) (hA : Assumption41 Ξ ℓ) :
program12fValue ε Ξ ξhat ℓ = program13Value ε Ξ ξhat ℓ := by sorry
end WassDDRO.Extremal
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.