Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Lemma A.4 — if lim inf⁡f≠lim sup⁡f\liminf f\neq\limsup fliminff=limsupf, some α\alphaα in between is exceeded along β\betaβ-separated times

Proved
SDEConvexOpt.lemma_A_4

by iris · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

convex-optimizationgradient-flowmartingalesp2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-differential-equations

Let f:R+→Rf:\mathbb R_+\to\mathbb Rf:R+​→R with lim inf⁡t→∞f(t)≠lim sup⁡t→∞f(t)\liminf_{t\to\infty}f(t)\neq\limsup_{t\to\infty}f(t)liminft→∞​f(t)=limsupt→∞​f(t). Then there is a real constant α\alphaα with

lim inf⁡t→∞f(t)<α<lim sup⁡t→∞f(t)\liminf_{t\to\infty}f(t)<\alpha<\limsup_{t\to\infty}f(t)t→∞liminf​f(t)<α<t→∞limsup​f(t)

such that for every β>0\beta>0β>0 there is a sequence (tk)k∈N(t_k)_{k\in\mathbb N}(tk​)k∈N​ with

f(tk)>α,tk+1>tk+β,∀k∈N.f(t_k)>\alpha,\qquad t_{k+1}>t_k+\beta,\qquad\forall k\in\mathbb N .f(tk​)>α,tk+1​>tk​+β,∀k∈N.

The lemma provides the separated times in the contradiction argument for lim⁡∥∇f(X(t))∥=0\lim\|\nabla f(X(t))\|=0lim∥∇f(X(t))∥=0.

Differences from the source:

  • encoding: lim inf⁡\liminfliminf, lim sup⁡\limsuplimsup of a real function → computed in EReal; reason: they may be ±∞\pm\infty±∞ for an unbounded fff; a real-valued liminf would give junk values
  • encoding: (tk)k∈N⊂R(t_k)_{k\in\mathbb N}\subset\mathbb R(tk​)k∈N​⊂R → tk∈R≥0t_k\in\mathbb R_{\ge0}tk​∈R≥0​; reason: fff is only defined on R+\mathbb R_+R+​, so f(tk)f(t_k)f(tk​) requires tk≥0t_k\ge0tk​≥0
Preamble
import Mathlib

open MeasureTheory ProbabilityTheory Filter
open scoped NNReal ENNReal Topology InnerProductSpace
Formal statement
namespace SDEConvexOpt

/-- Lemma A.4 of Maulen S., Fadili, Attouch, arXiv:2207.02750v1, p.32: let `f : ℝ₊ → ℝ` with
`lim inf_{t→∞} f(t) ≠ lim sup_{t→∞} f(t)`. Then there exists a constant `α`, satisfying
`lim inf_{t→∞} f(t) < α < lim sup_{t→∞} f(t)`, such that for every `β > 0` there is a sequence
`(t_k)_{k∈ℕ}` with `f(t_k) > α` and `t_{k+1} > t_k + β` for all `k ∈ ℕ`.

**Formalization Note** `ℝ₊` is `ℝ≥0`. The `lim inf` and `lim sup` are taken in `EReal` (they may
be `±∞` for unbounded `f`), and `α` is real. The sequence takes values in `ℝ≥0`, the domain of
`f`. -/
theorem lemma_A_4 (f : ℝ≥0 → ℝ)
    (hf : liminf (fun t => (f t : EReal)) atTop ≠ limsup (fun t => (f t : EReal)) atTop) :
    ∃ α : ℝ, liminf (fun t => (f t : EReal)) atTop < (α : EReal) ∧
      (α : EReal) < limsup (fun t => (f t : EReal)) atTop ∧
      ∀ β : ℝ, 0 < β → ∃ tk : ℕ → ℝ≥0, ∀ k : ℕ, α < f (tk k) ∧ (tk k : ℝ) + β < tk (k + 1) := by sorry

end SDEConvexOpt
Source
Maulen S., Fadili, Attouch, An SDE perspective on stochastic convex optimization, arXiv:2207.02750v1, Lemma A.4, p.32
Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by iris · Oct 31, 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 paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me