Proof of Theorem 2 (via the proof of Theorem 1) — min-max inequality for every
ProvedWorstCaseCVaR.Discrete.minmax_inequalityLet , the scenarios , the decision and , , be as in the setting of §2.2, with , and let be a nonempty set of probability vectors in . Then for every
Equivalently,
This is the elementary min-max inequality (weak duality) on the unrestricted threshold line ; together with the saddle point on it closes the gap between and in the proof of Theorem 2. The paper prints it in the proof of Theorem 1 for the mixture function and invokes it for through "a further discussion similar to the proof of Theorem 1".
Formalization Note The inequality is stated pointwise in , which is equivalent to the inf form. The suprema are real sSups of images; for a nonempty set of probability vectors and both images are nonempty and bounded above, so they are true suprema. Compactness of is not needed for this inequality and is not assumed.
import Mathlib import Definitions.Def_WorstCaseCVaR_Discrete_Setting
namespace WorstCaseCVaR.Discrete
/-- The min-max inequality used in the proof of Theorem 2, Zhu & Fukushima (2009), p. 1167 (stated
there for Theorem 1's `H_β`): `inf_{α ∈ ℝ} max_{π ∈ 𝒫_π} G_β(x, α, π) ≥ sup_{π ∈ 𝒫_π} min_{α ∈ ℝ}
G_β(x, α, π) = WCVaR_β(x)`, in the equivalent form: for every `α`,
`WCVaR_β(x) ≤ max_{π ∈ 𝒫_π} G_β(x, α, π)`. -/
theorem minmax_inequality {n m S : ℕ} (f : (Fin n → ℝ) → (Fin m → ℝ) → ℝ)
(ys : Fin S → Fin m → ℝ) (x : Fin n → ℝ) (β : ℝ) (hβ0 : 0 < β) (hβ1 : β < 1)
(P : Set (Fin S → ℝ)) (hP : P ⊆ stdSimplex ℝ (Fin S)) (hne : P.Nonempty) :
∀ α : ℝ, wcvar f ys β x P ≤ sSup ((fun π => G f ys β x α π) '' P) := by sorry
end WorstCaseCVaR.Discrete
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.