Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 1(ii), p. 12 β€” joint convexity of the relative entropy I(β„™β€²,β„™) on 𝒫 Γ— 𝒫

Proved
DROOptimal.Predictor.proposition_1_ii

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

convex-analysisinformation-theoryp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1relative-entropy

Let P\mathcal PP be the probability simplex on Ξ={1,…,d}\Xi=\{1,\dots,d\}Ξ={1,…,d} and I(Pβ€²,P)∈[0,∞]I(\mathbb P',\mathbb P)\in[0,\infty]I(Pβ€²,P)∈[0,∞] the relative entropy of Pβ€²\mathbb P'Pβ€² with respect to P\mathbb PP (Definition 5, with pβ€²log⁑(pβ€²/0)=+∞p'\log(p'/0)=+\inftypβ€²log(pβ€²/0)=+∞ for pβ€²>0p'>0pβ€²>0). For all pairs (P1β€²,P1),(P2β€²,P2)∈PΓ—P(\mathbb P_1',\mathbb P_1),(\mathbb P_2',\mathbb P_2)\in\mathcal P\times\mathcal P(P1′​,P1​),(P2′​,P2​)∈PΓ—P and every λ∈[0,1]\lambda\in[0,1]λ∈[0,1],

I((1βˆ’Ξ»)P1β€²+Ξ»P2β€²,Β (1βˆ’Ξ»)P1+Ξ»P2)≀(1βˆ’Ξ») I(P1β€²,P1)+λ I(P2β€²,P2).I\big((1-\lambda)\mathbb P_1'+\lambda\mathbb P_2',\ (1-\lambda)\mathbb P_1+\lambda\mathbb P_2\big)\le(1-\lambda)\,I(\mathbb P_1',\mathbb P_1)+\lambda\, I(\mathbb P_2',\mathbb P_2).I((1βˆ’Ξ»)P1′​+Ξ»P2′​,Β (1βˆ’Ξ»)P1​+Ξ»P2​)≀(1βˆ’Ξ»)I(P1′​,P1​)+Ξ»I(P2′​,P2​).

The proof of Theorem 4 uses this with P1β€²=P2β€²\mathbb P_1'=\mathbb P_2'P1′​=P2′​, to move a model along a segment while keeping its relative entropy distance from the observed frequencies below rrr.

Formalization Note Both sides are computed in EReal, where 0β‹…(+∞)=00\cdot(+\infty)=00β‹…(+∞)=0; so at Ξ»=0\lambda=0Ξ»=0 (or Ξ»=1\lambda=1Ξ»=1) an infinite term multiplied by zero does not contribute, as in the usual convention of convex analysis.

Preamble
import Mathlib
import Definitions.Def_DROOptimal_Predictor_Setting
import Definitions.Def_DROOptimal_Predictor_Problem5
Formal statement
namespace DROOptimal.Predictor

/-- Proposition 1(ii) (Convexity), p. 12: for all pairs (ℙ′₁, ℙ₁), (β„™β€²β‚‚, β„™β‚‚) ∈ 𝒫 Γ— 𝒫 and
λ ∈ [0, 1],
I((1 βˆ’ Ξ»)ℙ′₁ + Ξ»β„™β€²β‚‚, (1 βˆ’ Ξ»)ℙ₁ + Ξ»β„™β‚‚) ≀ (1 βˆ’ Ξ») I(ℙ′₁, ℙ₁) + Ξ» I(β„™β€²β‚‚, β„™β‚‚),
in `EReal` (where 0 · ∞ = 0). -/
theorem proposition_1_ii {d : β„•} (β„™'₁ ℙ₁ β„™'β‚‚ β„™β‚‚ : Ξ” d) (t : ℝ) (ht : t ∈ Set.Icc (0 : ℝ) 1) :
    relEntropy (mix t ht β„™'₁ β„™'β‚‚) (mix t ht ℙ₁ β„™β‚‚) ≀
      ((1 - t : ℝ) : EReal) * relEntropy β„™'₁ ℙ₁ + ((t : ℝ) : EReal) * relEntropy β„™'β‚‚ β„™β‚‚ := by sorry

end DROOptimal.Predictor
Source
Van Parys, Mohajerin Esfahani & Kuhn, From Data to Decisions: Distributionally Robust Optimization is Optimal, arXiv:1704.04118v3, p. 12, Proposition 1(ii)
Human review
  • Endorsed by Shuze Chen Β· Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 Β· 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
Β© 2026 Prove2Me