§3.1, p. 588 — strong duality for (11)/(12): Ẑ(x) = Ẑ_D(x) under Q ≻ μμ′
ProvedMinimaxSLP.RhsSDP.strong_dualityLet for , let satisfy complete recourse (Assumption 2) and (Assumption 3 at the constant ), and let be symmetric with (Assumption 4). Then for every the worst-case expected disutility
is a finite real number which is the least upper bound of the expected disutilities over and also the greatest lower bound of the objective values of the dual (12); in particular
The paper states (12) as "the equivalent dual problem under the strong duality condition" and uses in the last line of the proof of Theorem 3.2.
Formalization Note "The strong duality condition" is read as Assumption 4, the interior condition (written (Q - vecMulVec μ μ).PosDef). The IsLUB/IsGLB form rules out the junk values of real sSup/sInf; attainment of the dual minimum is not claimed.
import Mathlib import Definitions.Def_MomentDRO_Conf_Setting import Definitions.Def_WorstCaseVaR_KnownMoments_Basic import Definitions.Def_MinimaxSLP_RhsSDP_Model
namespace MinimaxSLP.RhsSDP
open MeasureTheory Matrix
/-- Strong duality between the moment problem (11) and its dual (12), §3.1, p. 588 ("under the
strong duality condition, the equivalent dual problem is (12)"; used as "from the strong duality
assumption, we have `Ẑ_D(x) = Ẑ(x)`", p. 590). The strong duality condition is read as
Assumption 4, `Q ≻ μμ′`. Under `α_k ≥ 0`, complete recourse (Assumption 2), `{π : W′π ≤ q} ≠ ∅`
(Assumption 3 at the constant `q`) and Assumption 4, for every `x` the real number `Ẑ(x)` is the
least upper bound of `{E_P[𝕌(𝒬(h̃, x))] : P ∈ 𝒫}` and the greatest lower bound of the
objective values `Q · Y + μ′y + y₀` of (12), and `Ẑ(x) = Ẑ_D(x)`. -/
theorem strong_duality {r d n K : ℕ} [NeZero K]
(W : Matrix (Fin r) (Fin d) ℝ) (T : Matrix (Fin r) (Fin n) ℝ) (q : Fin d → ℝ)
(α β : Fin K → ℝ) (μ : Fin r → ℝ) (Q : Matrix (Fin r) (Fin r) ℝ)
(hα : ∀ k, 0 ≤ α k)
(hA2 : ∀ z : Fin r → ℝ, ∃ w : Fin d → ℝ, 0 ≤ w ∧ W *ᵥ w = z)
(hA3 : ∃ π : Fin r → ℝ, Wᵀ *ᵥ π ≤ q)
(hQ : Q.IsSymm) (hA4 : (Q - Matrix.vecMulVec μ μ).PosDef)
(x : Fin n → ℝ) :
IsLUB (primalValues W T q α β μ Q x) (Zhat W T q α β μ Q x) ∧
IsGLB (dualObj μ Q '' dualFeasible12 W T q α β x) (Zhat W T q α β μ Q x) ∧
Zhat W T q α β μ Q x = ZhatD W T q α β μ Q x := by sorry
end MinimaxSLP.RhsSDP
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.