Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 5 β€” ΞΎ^K(Y, X|𝓕) ≀ 𝕄^K[Y], with equality iff Y = f(X) a.s. for an 𝓕-measurable f

Proved
OTSensitivity.MaxFun.theorem_5

by mikedeng1 Β· Oct 14, 2026 Β· Mathlib 0df444a (Lean v4.33.1)

functional-dependenceoptimal-transportp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1sensitivity-analysis

Let (Ξ©,B,P)(\Omega,\mathcal B,P)(Ξ©,B,P) be a probability space, X\mathsf XX and Y\mathsf YY Polish spaces, and X:Ξ©β†’XX:\Omega\to\mathsf XX:Ξ©β†’X, Y:Ξ©β†’YY:\Omega\to\mathsf YY:Ξ©β†’Y measurable, with laws ΞΌ\muΞΌ and Ξ½\nuΞ½. Let k:YΓ—Yβ†’[0,+∞]k:\mathsf Y\times\mathsf Y\to[0,+\infty]k:YΓ—Yβ†’[0,+∞] be continuous with k(y,yβ€²)=0β€…β€ŠβŸΊβ€…β€Šy=yβ€²k(y,y')=0\iff y=y'k(y,yβ€²)=0⟺y=yβ€², and assume

MK[Y]=∫Y∫Yk(y,yβ€²) dΞ½(yβ€²) dΞ½(y)<+∞.\mathbb M^K[Y]=\int_{\mathsf Y}\int_{\mathsf Y}k(y,y')\,d\nu(y')\,d\nu(y)<+\infty.MK[Y]=∫Yβ€‹βˆ«Y​k(y,yβ€²)dΞ½(yβ€²)dΞ½(y)<+∞.

Let F\mathcal FF be any sub-Οƒ\sigmaΟƒ-algebra of the Borel sets of X\mathsf XX and (Ξ½xF)x(\nu^{\mathcal F}_x)_x(Ξ½xF​)x​ any version of the conditional distribution of YYY given (X,F)(X,\mathcal F)(X,F), and let ΞΎK(Y,X∣F)=∫XK(Ξ½,Ξ½xF) dΞΌ(x)\xi^K(Y,X\mid\mathcal F)=\int_{\mathsf X}K(\nu,\nu^{\mathcal F}_x)\,d\mu(x)ΞΎK(Y,X∣F)=∫X​K(Ξ½,Ξ½xF​)dΞΌ(x). Then

  1. ΞΎK(Y,X∣F)≀MK[Y]\xi^K(Y,X\mid\mathcal F)\le\mathbb M^K[Y]ΞΎK(Y,X∣F)≀MK[Y] (equation (20)); in particular ΞΎK(Y,X∣F)\xi^K(Y,X\mid\mathcal F)ΞΎK(Y,X∣F) is finite;
  2. ΞΎK(Y,X∣F)=MK[Y]\xi^K(Y,X\mid\mathcal F)=\mathbb M^K[Y]ΞΎK(Y,X∣F)=MK[Y] if and only if YYY is functionally dependent on XXX through F\mathcal FF, that is, Y=f(X)Y=f(X)Y=f(X) PPP-almost surely for some F\mathcal FF-measurable map f:Xβ†’Yf:\mathsf X\to\mathsf Yf:Xβ†’Y.

This is the max-functionality property (RΓ©nyi's Postulate E) of OT-based sensitivity measures: the measure is bounded by an explicit constant of the output's law, and reaches it exactly when the information carried by (X,F)(X,\mathcal F)(X,F) determines the output.

Formalization Note X\mathsf XX and Y\mathsf YY are Polish (the paper's supports in RnX\mathbb R^{n_X}RnX​, RnY\mathbb R^{n_Y}RnY​ are such spaces); all values lie in [0,+∞][0,+\infty][0,+∞]. "The assumptions of Lemma 4" are the continuity of kkk, the standing assumption k(y,yβ€²)=0β€…β€ŠβŸΊβ€…β€Šy=yβ€²k(y,y')=0\iff y=y'k(y,yβ€²)=0⟺y=yβ€² of Β§3.1, and K(Ξ½Γ—Ξ½)=MK[Y]<∞\mathcal K(\nu\times\nu)=\mathbb M^K[Y]<\inftyK(Ξ½Γ—Ξ½)=MK[Y]<∞; the finiteness is assumed for the whole theorem, as on the page. The statement holds for every sub-Οƒ\sigmaΟƒ-algebra F\mathcal FF and every version of the conditional distribution (any kernel satisfying IsCondDistrib). KKK is the published RWPI.SqrtLasso.transportCost.

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

/-- Theorem 5, p. 10 (max-functionality): under the assumptions of Lemma 4 (`k` continuous,
vanishing exactly on the diagonal, `𝕄^K[Y] < ∞`), for every sub-Οƒ-algebra `𝓕` of `𝓑(𝒳)` and
every version `ΞΊ` of the conditional distribution of `Y` given `(X, 𝓕)`,
`ΞΎ^K(Y, X | 𝓕) ≀ 𝕄^K[Y]` (20), and equality holds if and only if `Y = f(X)` `P`-a.e. for some
`𝓕`-measurable `f : 𝒳 β†’ 𝒴`. -/
theorem theorem_5 {Ξ© 𝒳 𝒴 : Type*} [MeasurableSpace Ξ©]
    [TopologicalSpace 𝒳] [PolishSpace 𝒳] [m𝒳 : MeasurableSpace 𝒳] [BorelSpace 𝒳]
    [TopologicalSpace 𝒴] [PolishSpace 𝒴] [MeasurableSpace 𝒴] [BorelSpace 𝒴]
    (P : Measure Ξ©) [IsProbabilityMeasure P]
    (X : Ξ© β†’ 𝒳) (Y : Ξ© β†’ 𝒴) (hX : Measurable X) (hY : Measurable Y)
    (k : 𝒴 β†’ 𝒴 β†’ ℝβ‰₯0∞) (hk : IsCost k)
    (hMK : MK k (P.map Y) < ⊀)
    (F : MeasurableSpace 𝒳) (hF : F ≀ m𝒳) (ΞΊ : Kernel 𝒳 𝒴) (hΞΊ : IsCondDistrib P X Y F ΞΊ) :
    xi k P X Y ΞΊ ≀ MK k (P.map Y) ∧
      (xi k P X Y ΞΊ = MK k (P.map Y) ↔ IsFunctionallyDependent P X Y F) := by sorry

end OTSensitivity.MaxFun
Source
Borgonovo, Figalli, Plischke & SavarΓ©, Global Sensitivity Analysis via Optimal Transport, authors' manuscript of 7 March 2024 (Management Science, doi:10.1287/mnsc.2023.01796), Theorem 5 and (20), p. 10; proof pp. 33–34
Human review
  • Endorsed by Shuze Chen Β· Oct 14, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 Β· Oct 14, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
Β© 2026 Prove2Me