Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 3.2 a), (3.16), p. 17 — the RSGF method bounds (1/L)E‖∇f(x_R)‖² for nonconvex f

Proved
GhadimiLan.RSGF.theorem_3_2_a

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

complexitynonconvex-optimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-approximationzeroth-order-optimization

Consider the stochastic program f∗=inf⁡x∈Rn{f(x):=E[F(x,ξ)]}f^* = \inf_{x\in\mathbb R^n} \{ f(x) := \mathbb E[F(x,\xi)] \}f∗=infx∈Rn​{f(x):=E[F(x,ξ)]} (3.1), where ξ∼P\xi \sim Pξ∼P. Assume:

  1. F(⋅,ξ)∈CL1,1(Rn)F(\cdot,\xi) \in \mathcal C^{1,1}_L(\mathbb R^n)F(⋅,ξ)∈CL1,1​(Rn) almost surely, with L>0L > 0L>0;
  2. Assumption A1 for G(x,ξ)=∇xF(x,ξ)G(x,\xi) = \nabla_x F(x,\xi)G(x,ξ)=∇x​F(x,ξ), with variance bound σ2\sigma^2σ2;
  3. Assumption A3;
  4. fff is bounded below.

Run the randomized stochastic gradient free (RSGF) method (3.12)–(3.13) from x1x_1x1​. It uses smoothing parameter μ>0\mu > 0μ>0, iteration limit N≥1N \ge 1N≥1, and stepsizes with 0<γk<1/[2(n+4)L]0 < \gamma_k < 1/[2(n+4)L]0<γk​<1/[2(n+4)L]. At each step it draws a fresh sample ξk\xi_kξk​ and a fresh standard Gaussian direction uku_kuk​, all mutually independent. Its output is xRx_RxR​, where the random index RRR is independent of the samples and has probability mass function

PR(k)=γk−2L(n+4)γk2∑j=1N[γj−2L(n+4)γj2],k=1,…,N.(3.15)P_R(k) = \frac{\gamma_k - 2L(n+4)\gamma_k^2}{\sum_{j=1}^N [\gamma_j - 2L(n+4)\gamma_j^2]}, \qquad k = 1,\dots,N. \tag{3.15}PR​(k)=∑j=1N​[γj​−2L(n+4)γj2​]γk​−2L(n+4)γk2​​,k=1,…,N.(3.15)

Then ∥∇f(xR)∥2\|\nabla f(x_R)\|^2∥∇f(xR​)∥2 is integrable and

1L E[∥∇f(xR)∥2]≤1∑k=1N[γk−2L(n+4)γk2][Df2+2μ2(n+4)(1+L(n+4)2∑k=1N(γk4+Lγk2))+2(n+4)σ2∑k=1Nγk2],\frac1L\,\mathbb E\big[\|\nabla f(x_R)\|^2\big] \le \frac{1}{\sum_{k=1}^N [\gamma_k - 2L(n+4)\gamma_k^2]}\Big[D_f^2 + 2\mu^2(n+4)\Big(1 + L(n+4)^2\sum_{k=1}^N\Big(\frac{\gamma_k}{4} + L\gamma_k^2\Big)\Big) + 2(n+4)\sigma^2\sum_{k=1}^N\gamma_k^2\Big],L1​E[∥∇f(xR​)∥2]≤∑k=1N​[γk​−2L(n+4)γk2​]1​[Df2​+2μ2(n+4)(1+L(n+4)2k=1∑N​(4γk​​+Lγk2​))+2(n+4)σ2k=1∑N​γk2​],

where Df2=2(f(x1)−f∗)/LD_f^2 = 2(f(x_1)-f^*)/LDf2​=2(f(x1​)−f∗)/L (2.5), and the expectation is over RRR, ξ[N]\xi_{[N]}ξ[N]​ and u[N]u_{[N]}u[N]​.

This is the main convergence guarantee of the paper for nonconvex stochastic optimization with zeroth-order (function value) information only. With constant stepsizes it gives the O(n/ϵ2)O(n/\epsilon^2)O(n/ϵ2) complexity of Corollary 3.3.

Formalization Note Rn\mathbb R^nRn is EuclideanSpace ℝ (Fin n), and the constants (n+4)(n+4)(n+4) use the same nnn. The smoothing parameter μ\muμ is μs, and μ is the probability measure. The value f∗f^*f∗ is the infimum of fff (IsGLB), and Df2D_f^2Df2​ is written out as 2(f(x1)−f∗)/L2(f(x_1)-f^*)/L2(f(x1​)−f∗)/L. The following are disclosed additions that make the paper's implicit setting explicit:

  • the samples are measurable and mutually independent, and RRR is independent of them;
  • FFF is jointly measurable and F(x,⋅)F(x,\cdot)F(x,⋅), ∇xF(x,⋅)\nabla_xF(x,\cdot)∇x​F(x,⋅) are integrable;
  • the stepsizes are positive.

The integrability of ∥∇f(xR)∥2\|\nabla f(x_R)\|^2∥∇f(xR​)∥2 is part of the conclusion.

Preamble
import Mathlib
import Definitions.Def_GhadimiLan_RSGF_Model
import Definitions.Def_GhadimiLan_RSGF_Method

open MeasureTheory ProbabilityTheory
Formal statement
namespace GhadimiLan.RSGF

/-- Theorem 3.2 a), Eq. (3.16) (Ghadimi & Lan, arXiv:1309.5549v1, pp. 16–17). Under the standing
assumptions of §3 (A1 for `G = ∇_x F`, A3, `F(·, ξ) ∈ C^{1,1}_L` a.s.), with `f = objective P F`
bounded below with infimum `fstar`, stepsizes `0 < γ_k < 1/[2(n+4)L]`, smoothing parameter
`μs > 0` (the paper's `µ`; `μ` is the probability measure) and the random output index `R` with
the pmf (3.15), drawn independently of the samples, the RSGF output satisfies
`(1/L) E‖∇f(x_R)‖² ≤ [D_f² + 2µ²(n+4)(1 + L(n+4)² Σ(γ_k/4 + Lγ_k²)) + 2(n+4)σ² Σγ_k²] / Σ[γ_k − 2L(n+4)γ_k²]`
with `D_f² = 2(f(x_1) − f*)/L` (2.5). The integrability of `‖∇f(x_R)‖²` is part of the claim. -/
theorem theorem_3_2_a {n : ℕ} {Ω Ξ : Type*} [MeasurableSpace Ω] [MeasurableSpace Ξ]
    (μ : Measure Ω) [IsProbabilityMeasure μ] (P : Measure Ξ) [IsProbabilityMeasure P]
    (F : EuclideanSpace ℝ (Fin n) → Ξ → ℝ) (L σ μs : ℝ) (hL : 0 < L) (hσ : 0 ≤ σ) (hμs : 0 < μs)
    (hF : SZOAssumptions P F L σ)
    (fstar : ℝ) (hfstar : IsGLB (Set.range (objective P F)) fstar)
    (N : ℕ) (hN : 1 ≤ N) (γ : ℕ → ℝ)
    (hγ : ∀ k ∈ Finset.Icc 1 N, 0 < γ k ∧ γ k < 1 / (2 * ((n : ℝ) + 4) * L))
    (ξ : ℕ → Ω → Ξ) (u : ℕ → Ω → EuclideanSpace ℝ (Fin n)) (hs : IsSZOSampling μ P ξ u)
    (x1 : EuclideanSpace ℝ (Fin n)) (x : ℕ → Ω → EuclideanSpace ℝ (Fin n))
    (hx : IsRSGFRun F μs γ x1 ξ u x)
    (R : Ω → ℕ) (hR : IsOutputIndex μ ξ u N (rsgfPMF L n γ N) R) :
    Integrable (fun ω => ‖gradient (objective P F) (x (R ω) ω)‖ ^ 2) μ ∧
      1 / L * ∫ ω, ‖gradient (objective P F) (x (R ω) ω)‖ ^ 2 ∂μ ≤
        1 / (∑ k ∈ Finset.Icc 1 N, (γ k - 2 * L * ((n : ℝ) + 4) * γ k ^ 2)) *
          (2 * (objective P F x1 - fstar) / L
            + 2 * μs ^ 2 * ((n : ℝ) + 4) *
                (1 + L * ((n : ℝ) + 4) ^ 2 * ∑ k ∈ Finset.Icc 1 N, (γ k / 4 + L * γ k ^ 2))
            + 2 * ((n : ℝ) + 4) * σ ^ 2 * ∑ k ∈ Finset.Icc 1 N, γ k ^ 2) := by sorry

end GhadimiLan.RSGF
Source
Ghadimi & Lan, arXiv:1309.5549v1, Theorem 3.2 a), Eqs. (3.15)–(3.16), pp. 16–17; D_f from Eq. (2.5), p. 6
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