Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§A.5 Step 2, p. 23 — with y0=0y^0=0y0=0: Q(θ0)=3σ2/(4L2)+∥x0−x∗∥2Q(\theta^0)=3\sigma^2/(4L^2)+\|x^0-x^*\|^2Q(θ0)=3σ2/(4L2)+∥x0−x∗∥2

Proved
SAG.SmallStep.lyapunov_init

by mikedeng1 · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

lyapunov-functionp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1

Let n≥1n\ge1n≥1, L>0L>0L>0, let f1,…,fn:Rp→Rf_1,\dots,f_n:\mathbb R^p\to\mathbb Rf1​,…,fn​:Rp→R be differentiable with gradients fi′f'_ifi′​, and let x∗x^*x∗ be a minimizer of g=1n∑ifig=\frac1n\sum_if_ig=n1​∑i​fi​. Let QQQ be the Lyapunov function of §A.5 with α=12nL\alpha=\frac1{2nL}α=2nL1​ and put σ2=1n∑i=1n∥fi′(x∗)∥2\sigma^2=\frac1n\sum_{i=1}^n\|f'_i(x^*)\|^2σ2=n1​∑i=1n​∥fi′​(x∗)∥2. For the initial state θ0=(0,…,0,x0)\theta^0=(0,\dots,0,x^0)θ0=(0,…,0,x0) with all yi0=0y^0_i=0yi0​=0,

Q(θ0)=3σ24L2+∥x0−x∗∥2.Q(\theta^0)=\frac{3\sigma^2}{4L^2}+\|x^0-x^*\|^2 .Q(θ0)=4L23σ2​+∥x0−x∗∥2.

This evaluates the starting value of the Lyapunov function and produces the constant 9σ2/(4L2)9\sigma^2/(4L^2)9σ2/(4L2) of Proposition 1.

Formalization Note The page's computation uses ∑ifi′(x∗)=0\sum_if'_i(x^*)=0∑i​fi′​(x∗)=0, which follows from x∗x^*x∗ minimizing ggg; the theorem assumes the minimality, not the gradient identity. (The page's intermediate display writes (1−2n)αn2\frac{(1-2n)\alpha}{n^2}n2(1−2n)α​ where the definition of AAA gives (1−2n)α2n2\frac{(1-2n)\alpha^2}{n^2}n2(1−2n)α2​; that term vanishes at y0=0y^0=0y0=0.)

Preamble
import Mathlib
import Definitions.Def_SAGA_Convex_finiteSum
import Definitions.Def_SAG_SmallStep_blockForm
import Definitions.Def_SAG_SmallStep_lyapunov
Formal statement
namespace SAG.SmallStep

/-- §A.5 Step 2 (arXiv:1202.6258v4, p. 23): with `α = 1/(2nL)` and the table initialized to
`y⁰ = 0`, `Q(θ⁰) = 3σ²/(4L²) + ‖x⁰ − x*‖²`, where `σ² = (1/n) ∑ᵢ ‖f'ᵢ(x*)‖²` and `x*` minimizes
`g`. -/
theorem lyapunov_init {p n : ℕ} (hn : 0 < n)
    (f : Fin n → EuclideanSpace ℝ (Fin p) → ℝ)
    (f' : Fin n → EuclideanSpace ℝ (Fin p) → EuclideanSpace ℝ (Fin p)) (L : ℝ) (hL : 0 < L)
    (hgrad : ∀ i x, HasGradientAt (f i) (f' i x) x)
    (xstar : EuclideanSpace ℝ (Fin p))
    (hmin : ∀ x, SAGA.Convex.fAvg f xstar ≤ SAGA.Convex.fAvg f x)
    (x0 : EuclideanSpace ℝ (Fin p)) :
    lyapunov f' (1 / (2 * (n : ℝ) * L)) xstar (fun _ => 0, x0)
      = 3 * ((1 / (n : ℝ)) * ∑ i, ‖f' i xstar‖ ^ 2) / (4 * L ^ 2) + ‖x0 - xstar‖ ^ 2 := by sorry

end SAG.SmallStep
Source
Le Roux, Schmidt & Bach, A Stochastic Gradient Method with an Exponential Convergence Rate for Finite Training Sets, arXiv:1202.6258v4, p. 23, §A.5 Step 2, "Initializing all the y_i^0 to 0, we get"
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 7, 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