Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(6), p. 415 — the average tracker equals the average stochastic gradient: ȳ_k = (1/n) Σᵢ gᵢ(x_{i,k}, ξ_{i,k})

Proved
StochGradTrack.Const.tracking_identity

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

distributed-optimizationgradient-trackingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let W∈Rn×n\mathbf W\in\mathbb R^{n\times n}W∈Rn×n have unit column sums, 1⊤W=1⊤\mathbf 1^\top\mathbf W=\mathbf 1^\top1⊤W=1⊤, and run DSGT (4) with constant stepsize α\alphaα from any deterministic x0\mathbf x_0x0​, with yi,0=gi(xi,0,ξi,0)y_{i,0}=g_i(x_{i,0},\xi_{i,0})yi,0​=gi​(xi,0​,ξi,0​). Then for every k≥0k\ge0k≥0 and every realization of the noise,

yˉk=1n1⊤G(xk,ξk)=1n∑i=1ngi(xi,k,ξi,k).\bar y_k=\frac1n\mathbf 1^\top G(\mathbf x_k,\boldsymbol\xi_k)=\frac1n\sum_{i=1}^n g_i(x_{i,k},\xi_{i,k}).yˉ​k​=n1​1⊤G(xk​,ξk​)=n1​i=1∑n​gi​(xi,k​,ξi,k​).

This is why yi,ky_{i,k}yi,k​ is called a tracker: the network average of the trackers equals the average of the current stochastic gradients, so once the trackers agree each of them approximately follows the averaged gradient.

Formalization Note. The identity is pathwise and uses only 1⊤W=1⊤\mathbf 1^\top\mathbf W=\mathbf 1^\top1⊤W=1⊤ from Assumptions 1–4, which are therefore not hypotheses here. The page writes yi,0=g(xi,0,ξi,0)y_{i,0}=g(x_{i,0},\xi_{i,0})yi,0​=g(xi,0​,ξi,0​) without the index iii; the recursion (4) and this statement use gig_igi​.

Preamble
import Mathlib
import Definitions.Def_StochGradTrack_Const_Model
Formal statement
namespace StochGradTrack.Const

open MeasureTheory ProbabilityTheory Matrix

theorem tracking_identity {n p m : ℕ} {Ω : Type*} (α : ℝ) (W : Matrix (Fin n) (Fin n) ℝ)
    (g : Fin n → E p → E m → E p) (ξ : ℕ → Fin n → Ω → E m) (x0 : Stack n p)
    (hcol : ∀ j, ∑ i, W i j = 1) :
    ∀ k ω, avg (ys (fun _ => α) W g ξ x0 k ω) = (n : ℝ)⁻¹ • ∑ i, g i (xs (fun _ => α) W g ξ x0 k ω i) (ξ k i ω) := by sorry

end StochGradTrack.Const
Source
Pu & Nedić, Math. Program. 187 (2021), (6), p. 415
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).

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