Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 6.7, p. 64 — Bregman three-point identity

Proved
ModernOnlineLearning.OMD.lemma_6_7

by mikedeng1 · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

bregman-divergenceconvex-optimizationonline-mirror-descentp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1

Let ψ\psiψ be strictly convex on its domain XXX and differentiable on int⁡X\operatorname{int}XintX. For x,y∈int⁡Xx,y\in\operatorname{int}Xx,y∈intX and z∈Xz\in Xz∈X, its Bregman divergence obeys

Bψ(z;x)+Bψ(x;y)−Bψ(z;y)=⟨∇ψ(y)−∇ψ(x),z−x⟩.B_\psi(z;x)+B_\psi(x;y)-B_\psi(z;y)=\langle\nabla\psi(y)-\nabla\psi(x),z-x\rangle.Bψ​(z;x)+Bψ​(x;y)−Bψ​(z;y)=⟨∇ψ(y)−∇ψ(x),z−x⟩.

The identity reorganizes three divergence terms into one dual pairing; it is used in the one-step analysis of mirror descent.

Formalization Note Gradients are Fréchet derivatives in the continuous dual. The statement assumes differentiability on the interior, as Definition 6.4 does; the related published Beck–Teboulle identity assumes the stronger C1C^1C1 condition and therefore is not used as an exact reference.

Preamble
import Mathlib
import Definitions.Def_ModernOnlineLearning_OMD_Defs
Formal statement
namespace ModernOnlineLearning.OMD

/-- Orabona, Lemma 6.7, p. 64, with only the differentiability required by
Definition 6.4. -/
theorem lemma_6_7 {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E]
    [FiniteDimensional ℝ E] (X : Set E) (ψ : E → ℝ)
    (hstrict : StrictConvexOn ℝ X ψ)
    (hdiff : DifferentiableOn ℝ ψ (interior X))
    (x y z : E) (hx : x ∈ interior X) (hy : y ∈ interior X) (hz : z ∈ X) :
    BeckTeboulleMD.EMDA.bregman ψ z x +
      BeckTeboulleMD.EMDA.bregman ψ x y -
      BeckTeboulleMD.EMDA.bregman ψ z y =
        (fderiv ℝ ψ y - fderiv ℝ ψ x) (z - x) := by sorry

end ModernOnlineLearning.OMD
Source
Orabona, arXiv:1912.13213v10, Lemma 6.7, p. 64
Human review
  • Endorsed by Shuze Chen · Oct 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 9, 2026

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

  • Endorsed by iris · Nov 1, 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