§2.4, pp. 8–9 —
ProvedModelRiskOT.WorstProb.sup_indicatoroptimal-transportp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1worst-case-probability
Let be a Polish space, let satisfy (A1), let be nonempty and closed, and . With ,
This identifies the function of the dual problem when ; with it, the dual of the worst-case probability problem becomes the univariate problem (13).
Formalization Note The supremum is a real supremum (iSup in ); the family is bounded above by and nonempty, so it is the true supremum. Only and from (A1) are used.
Preamble
import Mathlib import Definitions.Def_ModelRiskOT_WorstProb_WorstCaseProb open MeasureTheory open scoped ENNReal
Formal statement
namespace ModelRiskOT.WorstProb
/-- §2.4, pp. 8–9: for `λ ≥ 0` and a nonempty closed set `A` in a Polish space,
`sup_{y ∈ S} {1_A(y) − λ c(x, y)} = (1 − λ c(x, A))⁺`, where `c(x, A) = inf_{y ∈ A} c(x, y)`. -/
theorem sup_indicator {S : Type*} [TopologicalSpace S] [PolishSpace S] [MeasurableSpace S]
[BorelSpace S] (c : S → S → ℝ) (hc : ModelRiskOT.Duality.AssumptionA1 c)
(A : Set S) (hA : IsClosed A) (hAne : A.Nonempty) (lam : ℝ) (hlam : 0 ≤ lam) (x : S) :
⨆ y : S, (A.indicator (fun _ => (1 : ℝ)) y - lam * c x y) =
max (1 - lam * costToSet c A x) 0 := by sorry
end ModelRiskOT.WorstProb
Source
Blanchet & Murthy, Quantifying Distributional Model Risk via Optimal Transport, arXiv:1604.01446v2, pp. 8–9, §2.4, display after (12)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.