Proof of Theorem 2 — Lemma 1 on :
ProvedWorstCaseCVaR.Discrete.lemma_1_on_intervalcvarminimaxp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let , the scenarios , the decision and be as in the setting of §2.2, with . Let be a nonempty compact convex set of probability vectors and let with be a nonempty closed bounded interval. Then has a saddle point :
so that
all extrema being attained.
This is the exchange of minimization and maximization on the compact product , the central step of the proof of Theorem 2.
Formalization Note The interval is arbitrary (), not the specific interval produced by the reduction step; the statement is the saddle-point form of the attained min–max equality.
Preamble
import Mathlib import Definitions.Def_WorstCaseCVaR_Discrete_Setting
Formal statement
namespace WorstCaseCVaR.Discrete
/-- Proof of Theorem 2, Zhu & Fukushima (2009), p. 1167 (Lemma 1 applied on `𝒜 × 𝒫_π`): for every
nonempty closed bounded interval `𝒜 = [a, b]` and every nonempty compact convex set `𝒫_π` of
probability vectors, `G_β(x, ·, ·)` has a saddle point `(α₀, π₀)` on `𝒜 × 𝒫_π`; equivalently,
`max_{π ∈ 𝒫_π} min_{α ∈ 𝒜} G_β(x, α, π) = min_{α ∈ 𝒜} max_{π ∈ 𝒫_π} G_β(x, α, π)` with all extrema
attained. -/
theorem lemma_1_on_interval {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)
(hcpt : IsCompact P) (hcvx : Convex ℝ P) (a b : ℝ) (hab : a ≤ b) :
∃ α₀ ∈ Set.Icc a b, ∃ π₀ ∈ P, ∀ α ∈ Set.Icc a b, ∀ π ∈ P,
G f ys β x α₀ π ≤ G f ys β x α₀ π₀ ∧ G f ys β x α₀ π₀ ≤ G f ys β x α π₀ := by sorry
end WorstCaseCVaR.Discrete
Source
Zhu & Fukushima, Worst-Case Conditional Value-at-Risk with Application to Robust Portfolio Management, Oper. Res. 57(5), 2009, p. 1167, proof of Theorem 2 ("By Lemma 1, we get …")
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.