Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Appendix A, Eq. (181) — Cramér-type bound on P(N(t)>κt)\mathbf P(N(t)>\kappa t)P(N(t)>κt)

Proved
BellWilliams2001.ThresholdPolicy.renewal_upper_deviation

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

large-deviationsp2o-batch-pfp1bp2o-gran-per-chapterp2o-plan-paperp2o-v1renewal-process

Let ζ(1),ζ(2),…\zeta(1),\zeta(2),\dotsζ(1),ζ(2),… be strictly positive independent random variables with ζ(2),ζ(3),…\zeta(2),\zeta(3),\dotsζ(2),ζ(3),… identically distributed, and suppose Λ(l)=log⁡Eelζ(i)<∞\Lambda(l)=\log\mathbf E e^{l\zeta(i)}<\inftyΛ(l)=logEelζ(i)<∞ for lll in a neighbourhood of 000 and i≥2i\ge2i≥2. Let m=Eζ(2)m=\mathbf E\zeta(2)m=Eζ(2), ν=1/m\nu=1/mν=1/m, X(n)=∑i=1nζ(i)X(n)=\sum_{i=1}^n\zeta(i)X(n)=∑i=1n​ζ(i), N(t)=sup⁡{n≥0:X(n)≤t}N(t)=\sup\{n\ge0:X(n)\le t\}N(t)=sup{n≥0:X(n)≤t}, and Λ∗\Lambda^*Λ∗ the Legendre–Fenchel transform of Λ\LambdaΛ. For ε>0\varepsilon>0ε>0, κ=ν+ε\kappa=\nu+\varepsilonκ=ν+ε and t>2/εt>2/\varepsilont>2/ε,

P(N(t)>κt)≤exp⁡(−(κt−1)Λ∗(tκt−1))≤exp⁡(−(κt−1)Λ∗(1ν+12ε)).\mathbf P(N(t)>\kappa t)\le\exp\Big(-(\kappa t-1)\Lambda^*\Big(\frac{t}{\kappa t-1}\Big)\Big)\le\exp\Big(-(\kappa t-1)\Lambda^*\Big(\frac1{\nu+\frac12\varepsilon}\Big)\Big).P(N(t)>κt)≤exp(−(κt−1)Λ∗(κt−1t​))≤exp(−(κt−1)Λ∗(ν+21​ε1​)).

Together with (184) this controls the deviations of the arrival and service processes from their rates on the time scale of the threshold, and it fixes how large the constant ccc in Lr=[clog⁡r]L^r=[c\log r]Lr=[clogr] must be.

Formalization Note The paper's fixed index rrr is suppressed. N(t)N(t)N(t) takes values in N∪{∞}\mathbb N\cup\{\infty\}N∪{∞}, Λ\LambdaΛ and Λ∗\Lambda^*Λ∗ in the extended reals, and exp⁡(−∞)=0\exp(-\infty)=0exp(−∞)=0.

Preamble
import Mathlib
import Definitions.Def_BellWilliams2001_ThresholdPolicy_Paths
import Definitions.Def_BellWilliams2001_ThresholdPolicy_LargeDeviation

open MeasureTheory ProbabilityTheory
open scoped ENNReal
Formal statement
namespace BellWilliams2001.ThresholdPolicy

/-- Appendix A, (181) (p. 644): a Cramér-type upper bound for a delayed renewal process.
`ζ(1), ζ(2), …` (paper's index base; `ζ 0` is unused) are strictly positive independent random
variables, `ζ(2), ζ(3), …` identically distributed, with a finite moment generating function on a
neighbourhood of `0` ((177)); `m = E ζ(2)`, `ν = 1/m`, `X(n) = ∑_{i=1}^n ζ(i)`,
`N(t) = sup{n ≥ 0 : X(n) ≤ t}` ((178)–(179), in `ℕ∞`), `Λ = log E e^{lζ(2)}`, `Λ*` its
Legendre–Fenchel transform (182). For `ε > 0`, `κ = ν + ε` and `t > 2/ε`,
`P(N(t) > κt) ≤ exp(−(κt − 1) Λ*(t/(κt − 1))) ≤ exp(−(κt − 1) Λ*(1/(ν + ε/2)))`.
The paper's fixed index `r` is suppressed. -/
theorem renewal_upper_deviation {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω)
    [IsProbabilityMeasure P] (ζ : ℕ → Ω → ℝ)
    (hmeas : ∀ i, Measurable (ζ i))
    (hpos : ∀ i ω, 1 ≤ i → 0 < ζ i ω)
    (hindep : iIndepFun (fun i : ℕ => ζ (i + 1)) P)
    (hident : ∀ i, 2 ≤ i → IdentDistrib (ζ i) (ζ 2) P P)
    (hmgf : ∃ δ : ℝ, 0 < δ ∧ ∀ l ∈ Set.Ioo (-δ) δ, ∀ i, 2 ≤ i →
      Integrable (fun ω => Real.exp (l * ζ i ω)) P)
    (ν : ℝ) (hν : ν = 1 / ∫ ω, ζ 2 ω ∂P)
    (ε : ℝ) (hε : 0 < ε) (κ : ℝ) (hκ : κ = ν + ε) (t : ℝ) (ht : 2 / ε < t) :
    P {ω | ENNReal.ofReal (κ * t) < (renewalCount (partialSum (fun i => ζ i ω)) t : ℝ≥0∞)} ≤
        EReal.exp (-(((κ * t - 1 : ℝ) : EReal) *
          legendre (logMGF P (ζ 2)) (t / (κ * t - 1)))) ∧
      EReal.exp (-(((κ * t - 1 : ℝ) : EReal) *
          legendre (logMGF P (ζ 2)) (t / (κ * t - 1)))) ≤
        EReal.exp (-(((κ * t - 1 : ℝ) : EReal) *
          legendre (logMGF P (ζ 2)) (1 / (ν + ε / 2)))) := by sorry

end BellWilliams2001.ThresholdPolicy
Source
Bell and Williams, Dynamic scheduling of a system with two parallel servers in heavy traffic with resource pooling, Ann. Appl. Probab. 11 (2001), pp. 644–645, Appendix A, (177)–(179), (181), (182)
Human review
  • Endorsed by Shuze Chen · Oct 4, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 4, 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