Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 1 — strong duality for the worst-case expected loss over an optimal-transport ball

Disproved
RWPI.Classif.proposition_1

by mikedeng1 · Oct 5, 2026 · Mathlib 0df444a (Lean v4.33.1)

distributionally-robust-optimizationdualityoptimal-transportp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1

Let Z=Rd×RZ = \mathbb R^d \times \mathbb RZ=Rd×R and let c:Z×Z→[0,∞]c : Z \times Z \to [0,\infty]c:Z×Z→[0,∞] be a lower semicontinuous cost function with c(z,z)=0c(z,z) = 0c(z,z)=0 for every z∈Zz \in Zz∈Z. Let l:Z→[0,∞)l : Z \to [0,\infty)l:Z→[0,∞) be an upper semicontinuous nonnegative loss. Let (X1,Y1),…,(Xn,Yn)∈Z(X_1,Y_1),\dots,(X_n,Y_n) \in Z(X1​,Y1​),…,(Xn​,Yn​)∈Z with n≥1n \ge 1n≥1, let PnP_nPn​ be their empirical distribution, and let δ>0\delta > 0δ>0. For γ≥0\gamma \ge 0γ≥0 put

φγ(Xi,Yi)=sup⁡u∈Rd, v∈R{l(u,v)−γ c((u,v),(Xi,Yi))}.\varphi_\gamma(X_i,Y_i) = \sup_{u \in \mathbb R^d,\ v \in \mathbb R}\big\{ l(u,v) - \gamma\, c\big((u,v),(X_i,Y_i)\big) \big\}.φγ​(Xi​,Yi​)=u∈Rd, v∈Rsup​{l(u,v)−γc((u,v),(Xi​,Yi​))}.

Then

sup⁡P: Dc(P,Pn)≤δEP[l(X,Y)]=min⁡γ≥0{γδ+1n∑i=1nφγ(Xi,Yi)},\sup_{P:\ D_c(P,P_n)\le\delta} \mathbb E_P\big[l(X,Y)\big] = \min_{\gamma \ge 0}\Big\{ \gamma\delta + \frac1n \sum_{i=1}^n \varphi_\gamma(X_i,Y_i) \Big\},P: Dc​(P,Pn​)≤δsup​EP​[l(X,Y)]=γ≥0min​{γδ+n1​i=1∑n​φγ​(Xi​,Yi​)},

and the minimum on the right is attained: there is γ∗≥0\gamma^* \ge 0γ∗≥0 at which the right-hand expression equals the left-hand side, and the left-hand side is at most the right-hand expression for every γ≥0\gamma \ge 0γ≥0. Both sides take values in [0,∞][0,\infty][0,∞].

This strong duality turns the supremum over infinitely many distributions into a one-dimensional minimization, and is the first step in recovering regularized estimators as distributionally robust ones. The paper quotes it from Blanchet & Murthy (Math. Oper. Res. 2019, Theorem 1).

Formalization Note The loss is a fixed function lll (the parameter β\betaβ of the paper plays no role in the identity; the paper's consequence (12) follows by taking the infimum over β\betaβ on both sides). Two restrictions relative to the page are recorded: the loss is nonnegative, so expectations are lower Lebesgue integrals in [0,∞][0,\infty][0,∞], and the radius is strictly positive, as in Blanchet & Murthy's theorem (at δ=0\delta = 0δ=0 the minimum need not be attained). φγ\varphi_\gammaφγ​ is computed in [0,∞][0,\infty][0,∞] with the convention ∞⋅0=0\infty\cdot 0 = 0∞⋅0=0.

Preamble
import Mathlib
import Definitions.Def_WassersteinDRO_Regularization_empiricalDistribution
import Definitions.Def_RWPI_SqrtLasso_worstCase
import Definitions.Def_RWPI_Classif_phi

open MeasureTheory
open scoped ENNReal
Formal statement
namespace RWPI.Classif

/-- Proposition 1, p. 10 (from Blanchet & Murthy, Math. Oper. Res. 2019, Theorem 1): for a
lower semicontinuous cost `c : ℝ^{d+1} × ℝ^{d+1} → [0, ∞]` vanishing on the diagonal, an upper
semicontinuous nonnegative loss `l` and a radius `δ > 0`, the worst-case expected loss over
`{P : D_c(P, P_n) ≤ δ}` equals `min_{γ ≥ 0} { γδ + (1/n) Σᵢ φ_γ(Xᵢ, Yᵢ) }`, the minimum being
attained. -/
theorem proposition_1 {d n : ℕ} (hn : 0 < n) (X : Fin n → Fin d → ℝ) (Y : Fin n → ℝ)
    (c : (Fin d → ℝ) × ℝ → (Fin d → ℝ) × ℝ → ℝ≥0∞)
    (hc : LowerSemicontinuous (Function.uncurry c)) (hc0 : ∀ z, c z z = 0)
    (l : (Fin d → ℝ) × ℝ → ℝ) (hl : UpperSemicontinuous l) (hl0 : ∀ z, 0 ≤ l z)
    (δ : ℝ) (hδ : 0 < δ) :
    ∃ γ : ℝ, 0 ≤ γ ∧
      RWPI.SqrtLasso.worstCase c δ
          (WassersteinDRO.Regularization.empiricalDistribution (fun i => (X i, Y i))) l =
        ENNReal.ofReal (γ * δ) + (n : ℝ≥0∞)⁻¹ * ∑ i, phi c l γ (X i, Y i) ∧
      ∀ γ' : ℝ, 0 ≤ γ' →
        RWPI.SqrtLasso.worstCase c δ
            (WassersteinDRO.Regularization.empiricalDistribution (fun i => (X i, Y i))) l ≤
          ENNReal.ofReal (γ' * δ) + (n : ℝ≥0∞)⁻¹ * ∑ i, phi c l γ' (X i, Y i) := by sorry

end RWPI.Classif
Source
Blanchet, Kang & Murthy, Robust Wasserstein Profile Inference and Applications to Machine Learning, arXiv:1610.05627v4, p. 10, Proposition 1 (from Blanchet & Murthy 2019, Theorem 1)
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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
© 2026 Prove2Me