Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Parallel projection ADMM reduces to project-and-average

Proved
BoydADMM.Constrained.parallel_projection_simplification

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

admmconvex-optimizationp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-bookp2o-v1projections

Let N≥1N\ge1N≥1 and ρ>0\rho>0ρ>0, and let A1,…,AN⊆Rn\mathcal A_1,\ldots,\mathcal A_N\subseteq\mathbb R^nA1​,…,AN​⊆Rn be nonempty closed convex sets. Start the parallel projection ADMM iteration from arbitrary z0z^0z0 and ui0u_i^0ui0​. Then the average dual variable vanishes after every update, uˉk+1=0\bar u^{k+1}=0uˉk+1=0 for k≥0k\ge0k≥0, and zk+1=xˉk+1z^{k+1}=\bar x^{k+1}zk+1=xˉk+1 for k≥1k\ge1k≥1. From k≥2k\ge2k≥2 the same run therefore satisfies

xik+1=ΠAi(xˉk−uik),uik+1=uik+(xik+1−xˉk+1)(i=1,…,N).x_i^{k+1}=\Pi_{\mathcal A_i}(\bar x^k-u_i^k),\qquad u_i^{k+1}=u_i^k+(x_i^{k+1}-\bar x^{k+1})\qquad(i=1,\ldots,N).xik+1​=ΠAi​​(xˉk−uik​),uik+1​=uik​+(xik+1​−xˉk+1)(i=1,…,N).

If uˉ0=0\bar u^0=0uˉ0=0, these two simplified updates already hold for every k≥1k\ge1k≥1. Thus the simplified algorithm describes the iterates of the displayed three-update method, with its initial step and index ranges made explicit.

Formalization Note Projection is relational and the theorem ranges over every run; it does not assume the vanishing dual average. The book's intermediate formula prints −zk-z^k−zk where the iteration requires −zk+1-z^{k+1}−zk+1. It also says the first step projects onto the sets Ci\mathcal C_iCi​; these are the Ai\mathcal A_iAi​ of §5.1.2.

Preamble
import Mathlib
import Definitions.Def_BoydADMM_Constrained_Run
Formal statement
namespace BoydADMM.Constrained

/-- §5.1.2, p. 36: every parallel projection ADMM run reduces to project-and-average
after its dual average has vanished. The index ranges keep the arbitrary initial state. -/
theorem parallel_projection_simplification {N n : ℕ} (hN : 0 < N)
    (A : Fin N → Set (Vec n))
    (hAne : ∀ i, (A i).Nonempty) (hAclosed : ∀ i, IsClosed (A i))
    (hAconvex : ∀ i, Convex ℝ (A i))
    (ρ : ℝ) (hρ : 0 < ρ)
    (x : ℕ → Blocks N n) (z : ℕ → Vec n) (u : ℕ → Blocks N n)
    (hrun : IsParallelRun A x z u) :
    (∀ k, avg (u (k + 1)) = 0) ∧
    (∀ k, 1 ≤ k → z (k + 1) = avg (x (k + 1))) ∧
    (∀ k, 2 ≤ k → ∀ i,
      RandomGradFree.Nonsmooth.IsMetricProjection (A i)
        (avg (x k) - u k i) (x (k + 1) i) ∧
      u (k + 1) i = u k i + (x (k + 1) i - avg (x (k + 1)))) ∧
    (avg (u 0) = 0 → ∀ k, 1 ≤ k → ∀ i,
      RandomGradFree.Nonsmooth.IsMetricProjection (A i)
        (avg (x k) - u k i) (x (k + 1) i) ∧
      u (k + 1) i = u k i + (x (k + 1) i - avg (x (k + 1)))) := by sorry

end BoydADMM.Constrained
Source
Boyd, Parikh, Chu, Peleato, Eckstein, Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers, Found. Trends Mach. Learn. 3(1) (2011), p. 36, §5.1.2, parallel-projection simplification; https://doi.org/10.1561/2200000016
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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