Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§7.2: the general form consensus zzz-update is local averaging

Proved
BoydADMM.Consensus.general_z_update_local_average

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

admmconsensusp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-bookp2o-v1

Consider general form consensus (§7.2) with the index map G\mathcal GG, ρ>0\rho>0ρ>0, and assume every global entry is copied at least once: kg=#{(i,j):G(i,j)=g}≥1k_g=\#\{(i,j):\mathcal G(i,j)=g\}\ge1kg​=#{(i,j):G(i,j)=g}≥1 for every ggg. Fix xi∈Rnix_i\in\mathbb R^{n_i}xi​∈Rni​ (standing for xik+1x_i^{k+1}xik+1​) and yi∈Rniy_i\in\mathbb R^{n_i}yi​∈Rni​ (standing for yiky_i^kyik​). Then z∈Rnz\in\mathbb R^nz∈Rn minimizes

∑i=1N(−yiTz~i+(ρ/2)∥xi−z~i∥22)\sum_{i=1}^N\big(-y_i^{T}\tilde z_i+(\rho/2)\|x_i-\tilde z_i\|_2^2\big)i=1∑N​(−yiT​z~i​+(ρ/2)∥xi​−z~i​∥22​)

if and only if, for every global index ggg,

zg=∑G(i,j)=g((xi)j+(1/ρ)(yi)j)∑G(i,j)=g1.z_g=\frac{\sum_{\mathcal G(i,j)=g}\big((x_i)_j+(1/\rho)(y_i)_j\big)}{\sum_{\mathcal G(i,j)=g}1}.zg​=∑G(i,j)=g​1∑G(i,j)=g​((xi​)j​+(1/ρ)(yi​)j​)​.

So each global entry is the average of the local entries xi+(1/ρ)yix_i+(1/\rho)y_ixi​+(1/ρ)yi​ that copy it.

Formalization Note The book writes the sum in the zzz-update as ∑i=1m\sum_{i=1}^m∑i=1m​; mmm is a typo for NNN. The hypothesis kg≥1k_g\ge1kg​≥1 is implicit in the book (otherwise zgz_gzg​ does not appear in the objective and the formula divides by zero). Indices are 0-based in Lean.

Preamble
import Mathlib
import Definitions.Def_BoydADMM_Consensus_Model

open scoped BigOperators InnerProductSpace
Formal statement
namespace BoydADMM.Consensus

/-- §7.2, p. 55: in general form consensus, if every global component is copied at least once
(`k_g ≥ 1` for every `g`), then `z` minimizes the `z`-update objective
`∑_i (−y_i^{kT} z̃_i + (ρ/2)‖x_i^{k+1} − z̃_i‖²)` over `z ∈ ℝⁿ` iff every component is the local
average `z_g = ∑_{G(i,j)=g} ((x_i^{k+1})_j + (1/ρ)(y_i^k)_j) / k_g`. Here `x`, `y` stand for
`x^{k+1}`, `y^k`. -/
theorem general_z_update_local_average {n N : ℕ} {nl : Fin N → ℕ}
    (G : (i : Fin N) → Fin (nl i) → Fin n) (hG : ∀ g, 1 ≤ kg G g) (ρ : ℝ) (hρ : 0 < ρ)
    (x y : (i : Fin N) → EuclideanSpace ℝ (Fin (nl i))) (z : EuclideanSpace ℝ (Fin n)) :
    (∀ z' : EuclideanSpace ℝ (Fin n),
        ∑ i, (-⟪y i, ztil G z i⟫_ℝ + (ρ / 2) * ‖x i - ztil G z i‖ ^ 2) ≤
          ∑ i, (-⟪y i, ztil G z' i⟫_ℝ + (ρ / 2) * ‖x i - ztil G z' i‖ ^ 2)) ↔
      ∀ g, z g = (∑ p ∈ entriesOf G g, (x p.1 p.2 + (1 / ρ) * y p.1 p.2)) / (kg G g : ℝ) := by sorry

end BoydADMM.Consensus
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. 55, §7.2 (z-update of general form consensus ADMM; the book prints ∑_{i=1}^m for ∑_{i=1}^N)
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