Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 4.2: martingale noise with bounded qqq-th moments and ∑γn1+q/2<∞\sum\gamma_n^{1+q/2}<\infty∑γn1+q/2​<∞ satisfies A1 almost surely

Proved
StochApproxDyn.MartingaleNoise.robbinsMonro_A1_ae

by mikedeng1 · Oct 4, 2026 · Mathlib 0df444a (Lean v4.33.1)

martingalesp2o-batch-pfp1ap2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-approximation

Let (Ω,F,P)(\Omega,\mathcal F,P)(Ω,F,P) be a probability space with a nondecreasing sequence {Fn}\{\mathcal F_n\}{Fn​} of sub-σ\sigmaσ-algebras, F:Rd→RdF:\mathbb R^d\to\mathbb R^dF:Rd→Rd continuous, and {γn}\{\gamma_n\}{γn​} a deterministic step sequence (γn≥0\gamma_n\ge0γn​≥0, ∑γn=∞\sum\gamma_n=\infty∑γn​=∞, γn→0\gamma_n\to0γn​→0). Let {xn}\{x_n\}{xn​}, given by

xn+1−xn=γn+1(F(xn)+Un+1),x_{n+1}-x_n=\gamma_{n+1}\big(F(x_n)+U_{n+1}\big),xn+1​−xn​=γn+1​(F(xn​)+Un+1​),

be a Robbins–Monro algorithm: UnU_nUn​ is Fn\mathcal F_nFn​-measurable and E(Un+1∣Fn)=0E(U_{n+1}\mid\mathcal F_n)=0E(Un+1​∣Fn​)=0. Suppose that for some q≥2q\ge2q≥2

sup⁡nE(∥Un+1∥q)<∞and∑nγn1+q/2<∞.\sup_nE\big(\|U_{n+1}\|^q\big)<\infty\qquad\text{and}\qquad\sum_n\gamma_n^{1+q/2}<\infty .nsup​E(∥Un+1​∥q)<∞andn∑​γn1+q/2​<∞.

Then assumption A1 holds with probability 111: for PPP-almost every ω\omegaω, for every T>0T>0T>0,

lim⁡n→∞sup⁡{∥∑i=nk−1γi+1Ui+1(ω)∥: k=n+1,…,m(τn+T)}=0andlim⁡t→∞Δω(t,T)=0,\lim_{n\to\infty}\sup\Big\{\Big\|\sum_{i=n}^{k-1}\gamma_{i+1}U_{i+1}(\omega)\Big\|:\ k=n+1,\dots,m(\tau_n+T)\Big\}=0\qquad\text{and}\qquad\lim_{t\to\infty}\Delta_\omega(t,T)=0,n→∞lim​sup{​i=n∑k−1​γi+1​Ui+1​(ω)​: k=n+1,…,m(τn​+T)}=0andt→∞lim​Δω​(t,T)=0,

where Δω(t,T)=sup⁡0≤h≤T∥∫tt+hUˉω(s) ds∥\Delta_\omega(t,T)=\sup_{0\le h\le T}\|\int_t^{t+h}\bar U_\omega(s)\,ds\|Δω​(t,T)=sup0≤h≤T​∥∫tt+h​Uˉω​(s)ds∥.

Combined with Proposition 4.1, this shows that the interpolated process of a Robbins–Monro algorithm with bounded iterates is almost surely an asymptotic pseudotrajectory of the flow of FFF.

Formalization Note The exceptional null set is chosen once for all T>0T>0T>0 (the almost-sure quantifier is outside "for all TTT"). Both forms of A1 that the paper calls equivalent are concluded. The moment hypothesis is a uniform bound on integrals in [0,∞][0,\infty][0,∞]; it implies integrability of every Un+1U_{n+1}Un+1​, which the Robbins–Monro definition also requires. qqq is a real number, powers are real powers of nonnegative numbers, and γ0\gamma_0γ0​, U0U_0U0​ are unused.

Preamble
import Mathlib
import Definitions.Def_StochApproxDyn_MartingaleNoise_Interpolation
import Definitions.Def_StochApproxDyn_MartingaleNoise_AssumptionA1
import Definitions.Def_StochApproxDyn_MartingaleNoise_RobbinsMonro

open MeasureTheory ProbabilityTheory Filter Topology
open scoped ENNReal
Formal statement
namespace StochApproxDyn.MartingaleNoise

/-- Benaïm 1999, Proposition 4.2, p. 15: let `{x_n}` given by (7) be a Robbins–Monro algorithm.
Suppose that for some `q ≥ 2`, `sup_n E(‖U_{n+1}‖^q) < ∞` and `∑_n γ_n^{1+q/2} < ∞`. Then
assumption A1 holds with probability 1: for `P`-almost every `ω`, the realised noise
sequence satisfies A1 for every `T > 0` (in both of its stated forms). -/
theorem robbinsMonro_A1_ae {d : ℕ} {Ω : Type*} {m0 : MeasurableSpace Ω}
    (P : Measure Ω) [IsProbabilityMeasure P] (ℱ : Filtration ℕ m0)
    (F : EuclideanSpace ℝ (Fin d) → EuclideanSpace ℝ (Fin d)) (γ : ℕ → ℝ)
    (x U : ℕ → Ω → EuclideanSpace ℝ (Fin d)) (hRM : IsRobbinsMonro P ℱ F γ x U)
    (q : ℝ) (hq : 2 ≤ q)
    (hmom : ∃ M : ℝ≥0∞, M < ⊤ ∧ ∀ n : ℕ, ∫⁻ ω, ‖U (n + 1) ω‖ₑ ^ q ∂P ≤ M)
    (hsum : Summable (fun n : ℕ => γ (n + 1) ^ (1 + q / 2))) :
    ∀ᵐ ω ∂P, IsA1 γ (fun n => U n ω) ∧ IsA1Delta γ (fun n => U n ω) := by sorry

end StochApproxDyn.MartingaleNoise
Source
Benaïm, Dynamics of Stochastic Approximation Algorithms, Séminaire de Probabilités XXXIII, LNM 1709 (1999), DOI 10.1007/BFb0096509, p. 15 (PDF p. 16), Proposition 4.2
Human review
  • Endorsed by Shuze Chen · Oct 5, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 5, 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