Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 25 — Distributionally robust MMSE estimator

Disproved
WassersteinDRO.Shrinkage.distributionally_robust_mmse_estimator

by mikedeng1 · Sep 21, 2026 · Mathlib 0df444a (Lean v4.33.1)

machine-learningoperations-researchrobust-optimization

If Σ^≻0\hat\Sigma \succ 0Σ^≻0, then the estimation problem (35) is equivalent to the nonlinear convex SDP (36): its optimal value equals the SDP's optimal value. Moreover, if S⋆S^\starS⋆ is optimal in (36) with Syy⋆S^\star_{yy}Syy⋆​ invertible, then the affine function ψ⋆(y)=Sxy⋆(Syy⋆)−1(y−μ^y)+μ^x\psi^\star(y) = S^\star_{xy}(S^\star_{yy})^{-1}(y-\hat\mu_y)+\hat\mu_xψ⋆(y)=Sxy⋆​(Syy⋆​)−1(y−μ^​y​)+μ^​x​ attains the outer infimum of (35), i.e. is a distributionally robust MMSE estimator.

Preamble
import Mathlib
import Definitions.Def_WassersteinDRO_Shrinkage_IsElliptical
import Definitions.Def_WassersteinDRO_Shrinkage_robustMMSEValue
import Definitions.Def_WassersteinDRO_Shrinkage_sdpValue
import Definitions.Def_WassersteinDRO_Shrinkage_sdpFeasibleSet
import Definitions.Def_WassersteinDRO_Shrinkage_sdpObjective
import Definitions.Def_WassersteinDRO_Shrinkage_estimationLoss
import Definitions.Def_WassersteinDRO_Shrinkage_ambiguitySet
import Definitions.Def_WassersteinDRO_Shrinkage_nominalRisk
import Definitions.Def_WassersteinDRO_Shrinkage_affineEstimator

open MeasureTheory
Formal statement
namespace WassersteinDRO.Shrinkage

/-- Theorem 25 (Distributionally robust MMSE estimator), Kuhn et al. 2019, p. 29 — the goal
theorem: if `Σ̂ ≻ 0`, then the estimation problem (35) is equivalent to the nonlinear convex
SDP (36) — `robustMMSEValue ε PN = sdpValue ε Σ̂ λmin`, `λmin` the least eigenvalue of `Σ̂` —
and if `S⋆` is optimal in (36) with `S⋆_yy` invertible (an added hypothesis not explicit on the
page; see `MODERATION_NOTES.md`, pitfall 5), then the affine function
`ψ⋆(y) = S⋆_xy(S⋆_yy)⁻¹(y-µ̂_y) + µ̂_x` attains the outer infimum of (35), i.e. is a
distributionally robust MMSE estimator. -/
theorem distributionally_robust_mmse_estimator {mx my : ℕ}
    (ε : ℝ) (hε : 0 < ε)
    (SigmaHat : Matrix (Fin mx ⊕ Fin my) (Fin mx ⊕ Fin my) ℝ)
    (hSigmaHatSymm : SigmaHat.IsHermitian) (hSigmaHatPD : SigmaHat.PosDef)
    (lambdaMin : ℝ) (hLambdaMinLB : ∀ i, lambdaMin ≤ hSigmaHatSymm.eigenvalues i)
    (hLambdaMinAttained : ∃ i, lambdaMin = hSigmaHatSymm.eigenvalues i)
    (muHatX : EuclideanSpace ℝ (Fin mx)) (muHatY : EuclideanSpace ℝ (Fin my))
    (muHat : EuclideanSpace ℝ (Fin mx ⊕ Fin my))
    (hmuHatX : ∀ i : Fin mx, muHat (Sum.inl i) = muHatX i)
    (hmuHatY : ∀ i : Fin my, muHat (Sum.inr i) = muHatY i)
    (g : ℝ → ℝ) (PN : Measure (EuclideanSpace ℝ (Fin mx ⊕ Fin my)))
    (hElliptical : IsElliptical PN g muHat SigmaHat) :
    robustMMSEValue ε PN = sdpValue ε SigmaHat lambdaMin ∧
      ∀ S : Matrix (Fin mx ⊕ Fin my) (Fin mx ⊕ Fin my) ℝ,
        S ∈ sdpFeasibleSet ε SigmaHat lambdaMin →
        (sdpObjective S : EReal) = sdpValue ε SigmaHat lambdaMin →
        S.toBlocks₂₂.PosDef →
        ⨆ (Q : Measure (EuclideanSpace ℝ (Fin mx ⊕ Fin my)))
            (_ : Q ∈ ambiguitySet ε 2 Set.univ PN)
            (_ : Integrable
              (estimationLoss (affineEstimator S.toBlocks₁₂ S.toBlocks₂₂ muHatX muHatY)) Q),
          (nominalRisk Q
              (estimationLoss (affineEstimator S.toBlocks₁₂ S.toBlocks₂₂ muHatX muHatY)) :
              EReal) =
          robustMMSEValue ε PN := by sorry

end WassersteinDRO.Shrinkage
Source
Kuhn, Mohajerin Esfahani, Nguyen & Shafieezadeh-Abadeh, Wasserstein Distributionally Robust Optimization, INFORMS TutORials 2019, Theorem 25, p. 29
Human review
  • Endorsed by Shuze Chen · Sep 28, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Sep 28, 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