Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 4 — near-optimal plans πn∈Φμ,δ\pi_n \in \Phi_{\mu,\delta}πn​∈Φμ,δ​ with πn(Cn)≥1−1/n\pi_n(C_n) \ge 1 - 1/nπn​(Cn​)≥1−1/n and πn(S×A)≥I−2/n\pi_n(S \times A) \ge I - 2/nπn​(S×A)≥I−2/n

Proved
ModelRiskOT.WorstProb.lemma_4

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

optimal-transportp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1worst-case-probability

Let SSS be a Polish space, ccc a cost satisfying (A1), μ\muμ a probability measure on SSS, δ>0\delta > 0δ>0, and A⊆SA \subseteq SA⊆S a nonempty closed set. Let I=sup⁡{π(S×A):π∈Φμ,δ}I = \sup\{\pi(S \times A) : \pi \in \Phi_{\mu,\delta}\}I=sup{π(S×A):π∈Φμ,δ​} and suppose λ∗∈[0,∞)\lambda^* \in [0,\infty)λ∗∈[0,∞) attains the infimum in (13). Then there exist probability measures πn∈Φμ,δ\pi_n \in \Phi_{\mu,\delta}πn​∈Φμ,δ​, n>1n > 1n>1, such that for every n>1n > 1n>1:

  1. πn(Cn)≥1−1/n\pi_n(C_n) \ge 1 - 1/nπn​(Cn​)≥1−1/n;
  2. ∫(S×S)∖Cnc dπn=0\displaystyle\int_{(S \times S)\setminus C_n} c\,d\pi_n = 0∫(S×S)∖Cn​​cdπn​=0;
  3. I(πn):=πn(S×A)≥I−2/nI(\pi_n) := \pi_n(S \times A) \ge I - 2/nI(πn​):=πn​(S×A)≥I−2/n.

Here CnC_nCn​ is the set of §2.4.1. The plans πn\pi_nπn​ replace a primal optimizer, which need not exist: they are nearly optimal and transport mass only along the near-optimal moves recorded in CnC_nCn​, and Lemma 2 and Theorem 3 are proved by passing to the limit along them.

Formalization Note The sequence is indexed by n∈Nn \in \mathbb Nn∈N and the properties are required for n>1n > 1n>1. Item 3 is written I≤πn(S×A)+2/nI \le \pi_n(S \times A) + 2/nI≤πn​(S×A)+2/n in [0,∞][0,\infty][0,∞], so that truncated subtraction plays no role. πn(Cn)\pi_n(C_n)πn​(Cn​) is the outer measure and the integral over the complement is a lower integral over the restricted measure; CnC_nCn​ is universally measurable, so these are the completion's values.

Preamble
import Mathlib
import Definitions.Def_ModelRiskOT_WorstProb_Cset

open MeasureTheory
open scoped ENNReal
Formal statement
namespace ModelRiskOT.WorstProb

/-- Lemma 4, p. 11. Under (A1), for a nonempty closed `A` and `λ* ∈ [0, ∞)` attaining the infimum in
(13), there are `π_n ∈ Φ_{μ,δ}` (`n > 1`) with `π_n(C_n) ≥ 1 − 1/n`, `∫_{(S×S) \ C_n} c dπ_n = 0`
and `I(π_n) = π_n(S × A) ≥ I − 2/n`, where `I = sup {π(S × A) : π ∈ Φ_{μ,δ}}`.
The last inequality is written `I ≤ π_n(S × A) + 2/n` in `[0, ∞]`. -/
theorem lemma_4 {S : Type*} [TopologicalSpace S] [PolishSpace S] [MeasurableSpace S]
    [BorelSpace S] (c : S → S → ℝ) (hc : ModelRiskOT.Duality.AssumptionA1 c) (μ : Measure S) [IsProbabilityMeasure μ]
    (δ : ℝ) (hδ : 0 < δ) (A : Set S) (hA : IsClosed A) (hAne : A.Nonempty)
    (lamStar : ℝ) (hlam : AttainsInf13 c μ δ A lamStar) :
    ∃ πs : ℕ → Measure (S × S), ∀ n : ℕ, 1 < n →
      πs n ∈ Phi c μ δ ∧
      ENNReal.ofReal (1 - 1 / (n : ℝ)) ≤ πs n (Cset c A lamStar n) ∧
      ∫⁻ p in (Cset c A lamStar n)ᶜ, ENNReal.ofReal (c p.1 p.2) ∂(πs n) = 0 ∧
      worstProbCoupling c μ δ A ≤ πs n (Set.univ ×ˢ A) + ENNReal.ofReal (2 / (n : ℝ)) := by sorry

end ModelRiskOT.WorstProb
Source
Blanchet & Murthy, Quantifying Distributional Model Risk via Optimal Transport, arXiv:1604.01446v2, p. 11, Lemma 4
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