Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(2.11), p. 7 — Σ(γ_k − Lγ_k²/2)E‖∇f(x_k)‖² ≤ f(x_1) − f* + (Lσ²/2)Σγ_k²

Proved
GhadimiLan.RSG.eq_2_11

by mikedeng1 · Oct 8, 2026 · Mathlib 0df444a (Lean v4.33.1)

nonconvex-optimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1sgdstochastic-approximation

Let f∈CL1,1(Rn)f\in\mathcal C^{1,1}_L(\mathbb R^n)f∈CL1,1​(Rn) with L>0L>0L>0 be bounded below with f∗=inf⁡xf(x)f^*=\inf_x f(x)f∗=infx​f(x). Run the RSG method from x1x_1x1​ with a Borel oracle GGG satisfying Assumption A1 (noise level σ\sigmaσ), and stepsizes with 0<γk<2/L0<\gamma_k<2/L0<γk​<2/L for k=1,…,Nk=1,\dots,Nk=1,…,N, where N≥1N\ge1N≥1. Then each ∥∇f(xk)∥2\|\nabla f(x_k)\|^2∥∇f(xk​)∥2, k=1,…,Nk=1,\dots,Nk=1,…,N, is integrable and

∑k=1N(γk−L2γk2) E∥∇f(xk)∥2≤f(x1)−f∗+Lσ22∑k=1Nγk2.\sum_{k=1}^N\Big(\gamma_k-\frac L2\gamma_k^2\Big)\,\mathbb E\|\nabla f(x_k)\|^2\le f(x_1)-f^*+\frac{L\sigma^2}{2}\sum_{k=1}^N\gamma_k^2.k=1∑N​(γk​−2L​γk2​)E∥∇f(xk​)∥2≤f(x1​)−f∗+2Lσ2​k=1∑N​γk2​.

This is the averaged form of (2.9): together with the weighting identity for the random output index it yields Theorem 2.1 a).

Formalization Note The stepsize positivity 0<γk0<\gamma_k0<γk​ is added to the paper's γk<2/L\gamma_k<2/Lγk​<2/L, as everywhere in this mission. The expectation is over the noise.

Preamble
import Mathlib
import Definitions.Def_ConvexOptAlg_SmoothGD_Defs
import Definitions.Def_GhadimiLan_RSG_Model
open MeasureTheory ProbabilityTheory
open scoped InnerProductSpace
Formal statement
namespace GhadimiLan.RSG

/-- Eq. (2.11) (Ghadimi & Lan, arXiv:1309.5549v1, proof of Theorem 2.1, p. 7): for
`f ∈ C^{1,1}_L(ℝⁿ)` (`L > 0`) bounded below with `f* = inf f`, stepsizes `0 < γ_k < 2/L`
(`k = 1, …, N`, `N ≥ 1`) and an RSG run under Assumption A1, each `‖∇f(x_k)‖²` is integrable
and `Σ_{k=1}^N (γ_k − (L/2)γ_k²) E‖∇f(x_k)‖² ≤ f(x_1) − f* + (Lσ²/2) Σ_{k=1}^N γ_k²`. -/
theorem eq_2_11 {n : ℕ} (f : E n → ℝ) (g : E n → E n) (L : ℝ)
    (hf : ConvexOptAlg.SmoothGD.IsBetaSmooth f g L) (hL : 0 < L)
    (fstar : ℝ) (hfstar : IsGLB (Set.range f) fstar)
    {Ξ : Type*} [MeasurableSpace Ξ] (G : E n → Ξ → E n) (hG : Measurable (Function.uncurry G))
    {Ω : Type*} [MeasurableSpace Ω] (μ : Measure Ω) [IsProbabilityMeasure μ]
    (ℱ : Filtration ℕ ‹MeasurableSpace Ω›) (ξ : ℕ → Ω → Ξ) (σ : ℝ)
    (N : ℕ) (hN : 1 ≤ N) (γ : ℕ → ℝ) (hγ : ∀ k ∈ Finset.Icc 1 N, 0 < γ k ∧ γ k < 2 / L)
    (x1 : E n) (x : ℕ → Ω → E n) (hx : IsRSGRun G γ x1 ξ x)
    (hA1 : AssumptionA1 μ ℱ g G ξ x σ) :
    (∀ k ∈ Finset.Icc 1 N, Integrable (fun ω => ‖g (x k ω)‖ ^ 2) μ) ∧
    ∑ k ∈ Finset.Icc 1 N, (γ k - L / 2 * γ k ^ 2) * ∫ ω, ‖g (x k ω)‖ ^ 2 ∂μ ≤
      f x1 - fstar + L * σ ^ 2 / 2 * ∑ k ∈ Finset.Icc 1 N, γ k ^ 2 := by sorry

end GhadimiLan.RSG
Source
Ghadimi & Lan, arXiv:1309.5549v1, proof of Theorem 2.1, Eq. (2.11), p. 7
Human review
  • Endorsed by Shuze Chen · Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 8, 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