Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(3), p. 645 (Chen–Kuhn–Wiesemann, Thm 3) — X_DR(S) = {x ∈ X : ∃ t ≥ 0, r ≥ 0, dist(ξ_i, S(x)) ≥ t − r_i, ϵt ≥ θ + (1/N)Σr_i}

Proved
WassDRCCP.Improved.eq_3_dual_form

by mikedeng1 · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

chance-constraintdistributionally-robustdualityp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1wasserstein

Let EEE be a finite-dimensional real normed space with its Borel σ-algebra, ξ1,…,ξN∈E\xi_1, \dots, \xi_N \in Eξ1​,…,ξN​∈E a sample with N≥1N \ge 1N≥1, ϵ∈(0,1)\epsilon \in (0, 1)ϵ∈(0,1) and θ>0\theta > 0θ>0. Let X⊆RL\mathcal X \subseteq \mathbb R^LX⊆RL and let S\mathcal SS assign to each xxx a set S(x)⊆E\mathcal S(x) \subseteq ES(x)⊆E that, for every x∈Xx \in \mathcal Xx∈X, is open and has nonempty complement. Then

XDR(S)={x∈X:∃ t≥0, r≥0, dist⁡(ξi,S(x))≥t−ri (i∈[N]), ϵt≥θ+1N∑i∈[N]ri},\mathcal X_{\mathrm{DR}}(\mathcal S) = \Big\{x \in \mathcal X : \exists\, t \ge 0,\ r \ge 0,\ \operatorname{dist}(\xi_i, \mathcal S(x)) \ge t - r_i\ (i \in [N]),\ \epsilon t \ge \theta + \tfrac1N \textstyle\sum_{i \in [N]} r_i\Big\},XDR​(S)={x∈X:∃t≥0, r≥0, dist(ξi​,S(x))≥t−ri​ (i∈[N]), ϵt≥θ+N1​∑i∈[N]​ri​},

where XDR(S)\mathcal X_{\mathrm{DR}}(\mathcal S)XDR​(S) is the set of x∈Xx \in \mathcal Xx∈X whose worst-case violation probability sup⁡P∈FN(θ)P[ξ∉S(x)]\sup_{\mathbb P \in \mathcal F_N(\theta)} \mathbb P[\xi \notin \mathcal S(x)]supP∈FN​(θ)​P[ξ∈/S(x)] over the 1-Wasserstein ball of radius θ\thetaθ around the empirical distribution is at most ϵ\epsilonϵ.

This is the result of Chen, Kuhn and Wiesemann (Theorem 3; also Xie, Proposition 1) that the paper cites and on which every one of its exact reformulations rests: it replaces a supremum over probability measures by finitely many scalar constraints. It is related to the strong duality for worst-case probabilities of Blanchet and Murthy, which is on the platform as ModelRiskOT.WorstProb.eq_13.

Formalization Note The worst-case probability is the published worstProb with cost ∥u−v∥\|u - v\|∥u−v∥ around the published empirical distribution. The hypothesis that E∖S(x)E \setminus \mathcal S(x)E∖S(x) is nonempty guards against Mathlib's junk value dist⁡(⋅,∅)=0\operatorname{dist}(\cdot, \emptyset) = 0dist(⋅,∅)=0 (the paper's infimum would be +∞+\infty+∞); without it the right-hand side would wrongly exclude such xxx. N≥1N \ge 1N≥1, ϵ∈(0,1)\epsilon \in (0,1)ϵ∈(0,1) and θ>0\theta > 0θ>0 are the paper's standing assumptions of §§2–4 (p. 645). Compactness of X\mathcal XX is not assumed.

Preamble
import Mathlib
import Definitions.Def_WassDRCCP_Improved_Setting

open MeasureTheory
open scoped ENNReal
Formal statement
namespace WassDRCCP.Improved

/-- Ho-Nguyen, Kılınç-Karzan, Küçükyavuz & Lee (2022), (3), p. 645 (Chen, Kuhn & Wiesemann,
Theorem 3): when `S(x)` is open for each `x ∈ X` and `θ > 0`,
`X_DR(S) = {x ∈ X : ∃ t ≥ 0, r ≥ 0, dist(ξ_i, S(x)) ≥ t − r_i (i ∈ [N]), ϵ t ≥ θ + (1/N) Σ_i r_i}`.
The unsafe set `S(x)ᶜ` is assumed nonempty, so that the distance is finite. -/
theorem eq_3_dual_form
    {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E] [FiniteDimensional ℝ E]
    [MeasurableSpace E] [BorelSpace E] {L N : ℕ}
    (S : (Fin L → ℝ) → Set E) (X : Set (Fin L → ℝ)) (ξ : Fin N → E) (ϵ θ : ℝ)
    (hN : 0 < N) (hϵ : 0 < ϵ) (hϵ1 : ϵ < 1) (hθ : 0 < θ)
    (hS_open : ∀ x ∈ X, IsOpen (S x)) (hS_compl : ∀ x ∈ X, (S x)ᶜ.Nonempty) :
    XDR S X ξ ϵ θ = dualSet S X ξ ϵ θ := by sorry

end WassDRCCP.Improved
Source
Ho-Nguyen, Kılınç-Karzan, Küçükyavuz & Lee, Math. Program. 196 (2022) 641–672, p. 645, (3), citing Chen, Kuhn & Wiesemann [8, Theorem 3]
Human review
  • Endorsed by Shuze Chen · Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 8, 2026

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

  • Endorsed by Community (Bot) · Oct 29, 2026

    Confirmed by the moderator at approval.

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