Proof of Theorem 5, p. 33 — K(ν, δ_{y′}) = ∫ k(y, y′) dν(y)
ProvedOTSensitivity.MaxFun.transportCost_diracdirac-measureoptimal-transportp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a Polish space, continuous with , and a probability measure on . For every point , the optimal transport cost from to the Dirac mass is
The only coupling of with is , so the infimum in (1) is taken over one plan. The identity turns into , which is how the proof of Theorem 5 identifies the maximum, and it is the same computation as (69), p. 36.
Formalization Note is the published RWPI.SqrtLasso.transportCost, with the marginal as first argument and the cost evaluated as , ; is not assumed symmetric. Values lie in , so both sides may be .
Preamble
import Mathlib import Definitions.Def_RWPI_SqrtLasso_transportCost import Definitions.Def_OTSensitivity_MaxFun_Setting open MeasureTheory ProbabilityTheory open scoped ENNReal
Formal statement
namespace OTSensitivity.MaxFun
/-- Proof of Theorem 5, p. 33 (and (69), p. 36): the only coupling of `ν` with a Dirac mass
`δ_{y'}` is `ν × δ_{y'}`, so `K(ν, δ_{y'}) = ∫ k(y, y') dν(y)`. -/
theorem transportCost_dirac {𝒴 : Type*} [TopologicalSpace 𝒴] [PolishSpace 𝒴] [MeasurableSpace 𝒴]
[BorelSpace 𝒴] (k : 𝒴 → 𝒴 → ℝ≥0∞) (hk : IsCost k)
(ν : Measure 𝒴) [IsProbabilityMeasure ν] (y' : 𝒴) :
RWPI.SqrtLasso.transportCost k ν (Measure.dirac y') = ∫⁻ y, k y y' ∂ν := by sorry
end OTSensitivity.MaxFun
Source
Borgonovo, Figalli, Plischke & Savaré, Global Sensitivity Analysis via Optimal Transport, authors' manuscript of 7 March 2024 (Management Science, doi:10.1287/mnsc.2023.01796), Appendix A, proof of Theorem 5, display K(ν, δ_{f(x)}) = ∫ k(y, f(x)) dν(y), p. 33; see also (69), p. 36
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.