Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(48), proof of Lemma 5.1, p. 15 — the monotonicity inequality of A at x_{k+1} and a zero of A + B + C

Proved
FRBSplitting.ThreeOp.eq_48

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

monotone-inclusionoperator-splittingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1resolvent

Let HHH be a real inner product space, A:H⇉HA:H\rightrightarrows HA:H⇉H monotone, and B,C:H→HB,C:H\to HB,C:H→H arbitrary maps. Let λ>0\lambda>0λ>0, let JλAJ_{\lambda A}JλA​ be the resolvent of λA\lambda AλA, and let (xk)k≥−1(x_k)_{k\ge -1}(xk​)k≥−1​ be generated by the three-operator scheme (47),

xk+1=JλA(xk−2λB(xk)+λB(xk−1)−λC(xk))∀k∈N.x_{k+1}=J_{\lambda A}\big(x_k-2\lambda B(x_k)+\lambda B(x_{k-1})-\lambda C(x_k)\big)\qquad\forall k\in\mathbb N.xk+1​=JλA​(xk​−2λB(xk​)+λB(xk−1​)−λC(xk​))∀k∈N.

Let x∈(A+B+C)−1(0)x\in(A+B+C)^{-1}(0)x∈(A+B+C)−1(0), i.e. −(B+C)(x)∈A(x)-(B+C)(x)\in A(x)−(B+C)(x)∈A(x). Then for every k∈Nk\in\mathbb Nk∈N,

0≤⟨xk+1−xk, x−xk+1⟩+λ⟨B(xk)−B(x), x−xk+1⟩+λ⟨B(xk)−B(xk−1), x−xk⟩+λ⟨B(xk)−B(xk−1), xk−xk+1⟩+λ⟨C(xk)−C(x), x−xk⟩+λ⟨C(xk)−C(x), xk−xk+1⟩.\begin{aligned}0\le{}&\langle x_{k+1}-x_k,\,x-x_{k+1}\rangle+\lambda\langle B(x_k)-B(x),\,x-x_{k+1}\rangle\\&+\lambda\langle B(x_k)-B(x_{k-1}),\,x-x_k\rangle+\lambda\langle B(x_k)-B(x_{k-1}),\,x_k-x_{k+1}\rangle\\&+\lambda\langle C(x_k)-C(x),\,x-x_k\rangle+\lambda\langle C(x_k)-C(x),\,x_k-x_{k+1}\rangle.\end{aligned}0≤​⟨xk+1​−xk​,x−xk+1​⟩+λ⟨B(xk​)−B(x),x−xk+1​⟩+λ⟨B(xk​)−B(xk−1​),x−xk​⟩+λ⟨B(xk​)−B(xk−1​),xk​−xk+1​⟩+λ⟨C(xk​)−C(x),x−xk​⟩+λ⟨C(xk​)−C(x),xk​−xk+1​⟩.​

This is the starting inequality of the analysis of the scheme: each of the six terms is then estimated separately, using the monotonicity and Lipschitz continuity of BBB and the cocoercivity of CCC, to obtain the energy inequality of Lemma 5.1.

Formalization Note. The resolvent is a hypothesis: J : ℝ → H → H with IsResolventFamily A J, i.e. γ−1(w−Jγw)∈A(Jγw)\gamma^{-1}(w-J_\gamma w)\in A(J_\gamma w)γ−1(w−Jγ​w)∈A(Jγ​w) for every γ>0\gamma>0γ>0 and www. Indices are shifted by one: Lean's x j is xj−1x_{j-1}xj−1​, so xk+1,xk,xk−1x_{k+1},x_k,x_{k-1}xk+1​,xk​,xk−1​ are x (k+2), x (k+1), x k. The inequality uses only the monotonicity of AAA, so the statement assumes only that (no maximality, no Lipschitz or cocoercivity hypotheses on BBB and CCC); it is therefore slightly more general than the paper's setting.

Preamble
import Mathlib
import Definitions.Def_ThreeOpSplitting_Accel_MonotoneOperators
import Definitions.Def_FRBSplitting_ThreeOp_Setting
open InnerProductSpace ThreeOpSplitting.Accel
Formal statement
namespace FRBSplitting.ThreeOp

theorem eq_48 {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℝ H]
    (A : H → Set H) (B C : H → H) (J : ℝ → H → H) (lam : ℝ)
    (hA : IsMonotoneOp A)
    (hJ : IsResolventFamily A J)
    (hlam0 : 0 < lam)
    (x : ℕ → H) (hrun : IsThreeOpRun J B C lam x)
    (xs : H) (hxs : xs ∈ FRBSplitting.Weak.zeroSet A (fun y => B y + C y)) :
    ∀ k : ℕ, 0 ≤ ⟪x (k + 2) - x (k + 1), xs - x (k + 2)⟫_ℝ
        + lam * ⟪B (x (k + 1)) - B xs, xs - x (k + 2)⟫_ℝ
        + lam * ⟪B (x (k + 1)) - B (x k), xs - x (k + 1)⟫_ℝ
        + lam * ⟪B (x (k + 1)) - B (x k), x (k + 1) - x (k + 2)⟫_ℝ
        + lam * ⟪C (x (k + 1)) - C xs, xs - x (k + 1)⟫_ℝ
        + lam * ⟪C (x (k + 1)) - C xs, x (k + 1) - x (k + 2)⟫_ℝ := by sorry

end FRBSplitting.ThreeOp
Source
Malitsky & Tam, A Forward-Backward Splitting Method for Monotone Inclusions Without Cocoercivity, arXiv:1808.04162v4, p. 15, (48) in the proof of Lemma 5.1
Human review
  • Endorsed by Shuze Chen · Oct 10, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 10, 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