Proof of Theorem 2 — one bounded interval contains a minimizer of for every
ProvedWorstCaseCVaR.Discrete.reduction_to_intervalLet , the scenarios , the decision and be as in the setting of §2.2, with , and let be any set of probability vectors in . Then there is a nonempty, closed and bounded interval () such that for every the function attains its minimum over at some point of . Consequently, for every ,
and hence .
This reduces the unbounded threshold variable to a compact interval that does not depend on , which is what makes a compact minimax theorem applicable in the proof of Theorem 2.
Formalization Note The statement is existential in : the interval is not fixed in the statement. "Attains its minimum over at a point of " is ∃ α ∈ Set.Icc a b, IsMinOn (fun α' => G … α' π) Set.univ α. The paper's preliminary remark that the bounded set lies in a polytope spanned by finitely many distributions is its route to this fact and is not part of the statement.
import Mathlib import Definitions.Def_WorstCaseCVaR_Discrete_Setting
namespace WorstCaseCVaR.Discrete
/-- Proof of Theorem 2, Zhu & Fukushima (2009), p. 1167: there is one nonempty, closed, bounded
interval `𝒜 = [a, b]` such that, for every `π ∈ 𝒫_π`, `G_β(x, ·, π)` attains its minimum over `ℝ`
at a point of `𝒜`; hence `min_{α ∈ ℝ} G_β(x, α, π) = min_{α ∈ 𝒜} G_β(x, α, π)`. -/
theorem reduction_to_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)) :
∃ a b : ℝ, a ≤ b ∧ ∀ π ∈ P, ∃ α ∈ Set.Icc a b,
IsMinOn (fun α' : ℝ => G f ys β x α' π) Set.univ α := by sorry
end WorstCaseCVaR.Discrete
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.