Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 2, p. 10 — f_δ : B × ℝ₊ → ℝ ∪ {∞} is proper and convex under Assumptions 1 and 2

Proved
OTDRO.Cvx.dual_proper_convex

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

convex-analysisdistributionally-robust-optimizationdualityoptimal-transportp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let P0P_0P0​ be a Borel probability measure on Rd\mathbb{R}^dRd, let AAA satisfy Assumption 1 and ℓ\ellℓ and P0P_0P0​ satisfy Assumption 2, let δ>0\delta > 0δ>0 and let the decision set B⊆RdB \subseteq \mathbb{R}^dB⊆Rd be convex. Consider the dual objective

fδ(β,λ)=EP0[ℓrob(β,λ;X)],(β,λ)∈B×R+,f_\delta(\beta, \lambda) = E_{P_0}\big[\ell_{rob}(\beta, \lambda; X)\big], \qquad (\beta, \lambda) \in B \times \mathbb{R}_+,fδ​(β,λ)=EP0​​[ℓrob​(β,λ;X)],(β,λ)∈B×R+​,

with values in R∪{∞}\mathbb{R} \cup \{\infty\}R∪{∞}. Then fδf_\deltafδ​ is proper and convex:

  1. fδ(β,λ)>−∞f_\delta(\beta, \lambda) > -\inftyfδ​(β,λ)>−∞ for every (β,λ)∈B×R+(\beta, \lambda) \in B \times \mathbb{R}_+(β,λ)∈B×R+​;
  2. for every β∈B\beta \in Bβ∈B there is λ≥0\lambda \ge 0λ≥0 with fδ(β,λ)<+∞f_\delta(\beta, \lambda) < +\inftyfδ​(β,λ)<+∞;
  3. for (β1,λ1),(β2,λ2)∈B×R+(\beta_1, \lambda_1), (\beta_2, \lambda_2) \in B \times \mathbb{R}_+(β1​,λ1​),(β2​,λ2​)∈B×R+​ and α∈[0,1]\alpha \in [0, 1]α∈[0,1],
fδ(αβ1+(1−α)β2, αλ1+(1−α)λ2)≤αfδ(β1,λ1)+(1−α)fδ(β2,λ2).f_\delta\big(\alpha\beta_1 + (1-\alpha)\beta_2,\ \alpha\lambda_1 + (1-\alpha)\lambda_2\big) \le \alpha f_\delta(\beta_1, \lambda_1) + (1-\alpha) f_\delta(\beta_2, \lambda_2).fδ​(αβ1​+(1−α)β2​, αλ1​+(1−α)λ2​)≤αfδ​(β1​,λ1​)+(1−α)fδ​(β2​,λ2​).

By Theorem 1 of the paper, the distributionally robust objective sup⁡P:Dc(P0,P)≤δEP[ℓ(βTX)]\sup_{P : D_c(P_0, P) \le \delta} E_P[\ell(\beta^{\mathsf T}X)]supP:Dc​(P0​,P)≤δ​EP​[ℓ(βTX)] equals inf⁡λ≥0fδ(β,λ)\inf_{\lambda \ge 0} f_\delta(\beta, \lambda)infλ≥0​fδ​(β,λ), so the robust problem becomes a joint convex minimization of fδf_\deltafδ​ over (β,λ)(\beta, \lambda)(β,λ); this is the basis of the paper's stochastic gradient schemes.

Formalization Note fδf_\deltafδ​ is the EReal-valued integral ∫ℓrob+ dP0−∫ℓrob− dP0\int \ell_{rob}^+ \, dP_0 - \int \ell_{rob}^- \, dP_0∫ℓrob+​dP0​−∫ℓrob−​dP0​. Convexity is the displayed inequality in EReal (with 0⋅(+∞)=00 \cdot (+\infty) = 00⋅(+∞)=0), since ConvexOn does not apply to EReal-valued maps; together with item 1 the right-hand side never involves −∞+∞-\infty + \infty−∞+∞. "Proper" is item 1 together with item 2; item 2 is stated for each β∈B\beta \in Bβ∈B, which is stronger than "finite at some point of B×R+B \times \mathbb{R}_+B×R+​" whenever B≠∅B \neq \emptysetB=∅ and avoids a nonemptiness hypothesis on BBB (for B=∅B = \emptysetB=∅ the statement is vacuous). The convexity of BBB is the paper's standing assumption (p. 1); δ>0\delta > 0δ>0 is the paper's standing assumption on the budget.

Preamble
import Mathlib
import Definitions.Def_OTDRO_Dual_Setting
Formal statement
namespace OTDRO.Cvx

open MeasureTheory

/-- **Theorem 2**, p. 10: under Assumptions 1 and 2, the dual objective
`f_δ(β, λ) = E_{P₀}[ℓ_rob(β, λ; X)]`, a map `B × ℝ₊ → ℝ ∪ {∞}`, is proper and convex:
1. it never takes the value `−∞` on `B × ℝ₊`;
2. for every `β ∈ B` it is finite at some `λ ≥ 0`;
3. for `(βᵢ, λᵢ) ∈ B × ℝ₊` and `α ∈ [0, 1]`,
   `f_δ(αβ₁ + (1 − α)β₂, αλ₁ + (1 − α)λ₂) ≤ α f_δ(β₁, λ₁) + (1 − α) f_δ(β₂, λ₂)`. -/
theorem dual_proper_convex {d : ℕ} (P0 : Measure (EuclideanSpace ℝ (Fin d)))
    [IsProbabilityMeasure P0]
    (A : EuclideanSpace ℝ (Fin d) → Matrix (Fin d) (Fin d) ℝ) (ρmin ρmax : ℝ)
    (hA : OTDRO.Dual.Assumption1 P0 A ρmin ρmax)
    (ℓ : ℝ → ℝ) (δ : ℝ) (hδ : 0 < δ) (B : Set (EuclideanSpace ℝ (Fin d))) (hB : Convex ℝ B)
    (h2 : OTDRO.Dual.Assumption2 P0 ℓ) :
    (∀ β ∈ B, ∀ lam : ℝ, 0 ≤ lam → OTDRO.Dual.fDelta P0 ℓ A δ β lam ≠ ⊥) ∧
    (∀ β ∈ B, ∃ lam : ℝ, 0 ≤ lam ∧ OTDRO.Dual.fDelta P0 ℓ A δ β lam ≠ ⊤) ∧
    (∀ β₁ ∈ B, ∀ β₂ ∈ B, ∀ lam₁ lam₂ : ℝ, 0 ≤ lam₁ → 0 ≤ lam₂ →
      ∀ α : ℝ, 0 ≤ α → α ≤ 1 →
        OTDRO.Dual.fDelta P0 ℓ A δ (α • β₁ + (1 - α) • β₂) (α * lam₁ + (1 - α) * lam₂) ≤
          (α : EReal) * OTDRO.Dual.fDelta P0 ℓ A δ β₁ lam₁ +
            ((1 - α : ℝ) : EReal) * OTDRO.Dual.fDelta P0 ℓ A δ β₂ lam₂) := by sorry

end OTDRO.Cvx
Source
arXiv:1810.02403v3, Theorem 2, p. 10 (proof in §5.1, p. 33)
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).

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