Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 5 — Gyöngy–Krylov maximal inequality for dominated processes

Proved
SabanisEuler.UniformRate.maximal_inequality

by mikedeng1 · Sep 30, 2026 · Mathlib 0df444a (Lean v4.33.1)

maximal-inequalityp2o-batch-p200bp2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-processesstopping-time

Let (Ω,F,{Ft}t≥0,P)(\Omega,\mathcal F,\{\mathcal F_t\}_{t\ge0},P)(Ω,F,{Ft​}t≥0​,P) be a filtered probability space with right-continuous filtration, and let T∈[0,∞)T\in[0,\infty)T∈[0,∞). Let f=(ft)t∈[0,T]f=(f_t)_{t\in[0,T]}f=(ft​)t∈[0,T]​ and g=(gt)t∈[0,T]g=(g_t)_{t\in[0,T]}g=(gt​)t∈[0,T]​ be nonnegative, {Ft}\{\mathcal F_t\}{Ft​}-adapted processes with continuous paths on [0,T][0,T][0,T]. Suppose that for every constant c>0c>0c>0 and every stopping time τ≤T\tau\le Tτ≤T,

E[fτ1{g0≤c}]≤E[gτ1{g0≤c}].\mathbb E\big[f_\tau\mathbb 1_{\{g_0\le c\}}\big]\le\mathbb E\big[g_\tau\mathbb 1_{\{g_0\le c\}}\big].E[fτ​1{g0​≤c}​]≤E[gτ​1{g0​≤c}​].

Then for every stopping time τ≤T\tau\le Tτ≤T and every γ∈(0,1)\gamma\in(0,1)γ∈(0,1),

E[sup⁡t≤τftγ]≤2−γ1−γ E[sup⁡t≤τgtγ].\mathbb E\Big[\sup_{t\le\tau}f_t^{\gamma}\Big]\le\frac{2-\gamma}{1-\gamma}\,\mathbb E\Big[\sup_{t\le\tau}g_t^{\gamma}\Big].E[t≤τsup​ftγ​]≤1−γ2−γ​E[t≤τsup​gtγ​].

This is a Lenglart-type domination inequality: a bound on expectations at stopping times upgrades to a bound on the expected running maximum, at the price of a fractional power γ<1\gamma<1γ<1. In the proof of Theorem 3 it turns an Itô-formula estimate for E[(φ(τ)∣X(τ)−Xn(τ)∣2)p/2]\mathbb E[(\varphi(\tau)|X(\tau)-X_n(\tau)|^2)^{p/2}]E[(φ(τ)∣X(τ)−Xn​(τ)∣2)p/2] at stopping times into a bound on the expected supremum over [0,T][0,T][0,T].

Formalization Note Nonnegativity is built into the value type [0,∞)[0,\infty)[0,∞); expectations are Lebesgue integrals in [0,∞][0,\infty][0,∞], so both sides may be +∞+\infty+∞. Stopping times take values in [0,∞][0,\infty][0,∞] and are required to satisfy τ≤T\tau\le Tτ≤T everywhere. Nothing is required of f,gf,gf,g after TTT. The process ggg is not assumed nondecreasing, and the right-hand side is E[sup⁡t≤τgtγ]\mathbb E[\sup_{t\le\tau}g_t^{\gamma}]E[supt≤τ​gtγ​], as printed. The right-continuity of the filtration is the paper's standing usual condition; completeness of the filtration is not imposed.

Preamble
import Mathlib

open MeasureTheory ProbabilityTheory Filter
open scoped ENNReal NNReal Topology BigOperators
Formal statement
namespace SabanisEuler.UniformRate

/-- Sabanis (2016), arXiv:1308.1796v4, p. 19, Lemma 5 (the Gyöngy–Krylov / Lenglart-type
maximal inequality). Let `T ∈ [0, ∞)` and let `f = (f_t)_{t ∈ [0, T]}`,
`g = (g_t)_{t ∈ [0, T]}` be nonnegative `ℱ`-adapted processes with continuous paths on
`[0, T]` such that, for every constant `c > 0` and every stopping time `τ ≤ T`,
`𝔼[f_τ 𝟙_{g₀ ≤ c}] ≤ 𝔼[g_τ 𝟙_{g₀ ≤ c}]`. Then for every stopping time `τ ≤ T` and every
`γ ∈ (0, 1)`,
`𝔼[sup_{t ≤ τ} f_t^γ] ≤ (2 - γ)/(1 - γ) · 𝔼[sup_{t ≤ τ} g_t^γ]`.
Nonnegativity is built into the value type `ℝ≥0`; expectations are lower Lebesgue integrals
in `[0, ∞]`. Nothing is required of `f`, `g` after `T`. `g` is not assumed nondecreasing,
and the right-hand side is `sup_{t ≤ τ} g_t^γ`, as printed. -/
theorem maximal_inequality
    {Ω : Type*} [mΩ : MeasurableSpace Ω]
    (P : Measure Ω) [IsProbabilityMeasure P]
    (ℱ : Filtration ℝ≥0 mΩ) [ℱ.IsRightContinuous]
    (T : ℝ≥0)
    (f g : ℝ≥0 → Ω → ℝ≥0)
    (hf_cont : ∀ ω, ContinuousOn (fun t => f t ω) (Set.Icc 0 T))
    (hg_cont : ∀ ω, ContinuousOn (fun t => g t ω) (Set.Icc 0 T))
    (hf_adapt : ∀ t ≤ T, Measurable[ℱ t] (f t))
    (hg_adapt : ∀ t ≤ T, Measurable[ℱ t] (g t))
    (hdom : ∀ c : ℝ≥0, 0 < c → ∀ τ : Ω → WithTop ℝ≥0, IsStoppingTime ℱ τ →
      (∀ ω, τ ω ≤ (T : WithTop ℝ≥0)) →
      ∫⁻ ω, {ω | g 0 ω ≤ c}.indicator (fun ω => ((stoppedValue f τ ω : ℝ≥0) : ℝ≥0∞)) ω ∂P
        ≤ ∫⁻ ω, {ω | g 0 ω ≤ c}.indicator (fun ω => ((stoppedValue g τ ω : ℝ≥0) : ℝ≥0∞)) ω ∂P) :
    ∀ τ : Ω → WithTop ℝ≥0, IsStoppingTime ℱ τ → (∀ ω, τ ω ≤ (T : WithTop ℝ≥0)) →
      ∀ γ : ℝ, 0 < γ → γ < 1 →
        ∫⁻ ω, (⨆ (t : ℝ≥0) (_ : (t : WithTop ℝ≥0) ≤ τ ω), (f t ω : ℝ≥0∞) ^ γ) ∂P
          ≤ ENNReal.ofReal ((2 - γ) / (1 - γ)) *
            ∫⁻ ω, (⨆ (t : ℝ≥0) (_ : (t : WithTop ℝ≥0) ≤ τ ω), (g t ω : ℝ≥0∞) ^ γ) ∂P := by sorry

end SabanisEuler.UniformRate
Source
Sabanis, Euler approximations with varying coefficients, arXiv:1308.1796v4, p. 19, Lemma 5 (citing Gyöngy–Krylov)
Human review
  • Endorsed by Shuze Chen · Oct 1, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 1, 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