Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(3.18) and p. 13, proof of Theorem 3.2 — P(ℳ_N^c) ≤ Σ_i P(f̂_N(x_i) ≥ f(x_i)+ε) + Σ_j P(f̂_N(x_j) ≤ f(x_j)−ε)

Proved
SAARate.Polyhedral.union_bound_p13

by mikedeng1 · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

large-deviationsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1sample-average-approximationstochastic-programming

Let fff and f^N\hat f_Nf^​N​ be the true and sample average functions of the mission's setting, built on a sample ω1,ω2,…\omega^1,\omega^2,\dotsω1,ω2,… defined on a probability space (S,Q)(S,Q)(S,Q). Let AAA and ANA_NAN​ be their optimal sets on Θ\ThetaΘ, and let

MN:={ AN is nonempty and forms a face of A }.(3.16)\mathcal M_N:=\{\,A_N\text{ is nonempty and forms a face of }A\,\}.\tag{3.16}MN​:={AN​ is nonempty and forms a face of A}.(3.16)

Let x1,…,xqx_1,\dots,x_qx1​,…,xq​ and ℓ\ellℓ be such that, for every sample path and every NNN, condition (2.13) f^N(xi)<f^N(xj)\hat f_N(x_i)<\hat f_N(x_j)f^​N​(xi​)<f^​N​(xj​) (i≤ℓ<ji\le\ell<ji≤ℓ<j) implies MN\mathcal M_NMN​, and such that f(xi)<f(xj)f(x_i)<f(x_j)f(xi​)<f(xj​) for all i≤ℓ<ji\le\ell<ji≤ℓ<j. Then there is ε>0\varepsilon>0ε>0, not depending on NNN, with

Q(MNc)≤∑i=1ℓQ(f^N(xi)≥f(xi)+ε)+∑j=ℓ+1qQ(f^N(xj)≤f(xj)−ε)for every N.Q(\mathcal M_N^c)\le\sum_{i=1}^{\ell}Q\big(\hat f_N(x_i)\ge f(x_i)+\varepsilon\big)+\sum_{j=\ell+1}^{q}Q\big(\hat f_N(x_j)\le f(x_j)-\varepsilon\big)\quad\text{for every }N .Q(MNc​)≤i=1∑ℓ​Q(f^​N​(xi​)≥f(xi​)+ε)+j=ℓ+1∑q​Q(f^​N​(xj​)≤f(xj​)−ε)for every N.

The step combines (3.18), MNc⊆{∃ i≤ℓ<j: f^N(xi)≥f^N(xj)}\mathcal M_N^c\subseteq\{\exists\, i\le\ell<j:\ \hat f_N(x_i)\ge\hat f_N(x_j)\}MNc​⊆{∃i≤ℓ<j: f^​N​(xi​)≥f^​N​(xj​)}, with a union bound. It reduces the rate of Theorem 3.2 to the deviations of finitely many one-dimensional sample means.

Formalization Note The points are given through the hypotheses that Lemma 2.4 (c) provides. The standing assumptions (i)–(iv) are not needed for this step and are not assumed. MNc\mathcal M_N^cMNc​ is not shown to be measurable, so Q(MNc)Q(\mathcal M_N^c)Q(MNc​) is its outer measure; an upper bound on it is the stronger statement. Indices are 0-based in Lean.

Preamble
import Mathlib
import Definitions.Def_SAARate_Polyhedral_Setting

open MeasureTheory ProbabilityTheory
Formal statement
namespace SAARate.Polyhedral

/-- Proof of Theorem 3.2, (3.18) and p. 13: given points `x_1, …, x_q` for which condition (2.13)
forces `A_N` to be a nonempty face of `A` (Lemma 2.4 (c)) and `f(x_i) < f(x_j)` for
`i ≤ ℓ < j`, there is `ε > 0` (independent of `N`) with
`P(ℳ_N^c) ≤ Σ_{i ≤ ℓ} P(f̂_N(x_i) ≥ f(x_i) + ε) + Σ_{j > ℓ} P(f̂_N(x_j) ≤ f(x_j) − ε)`.
Indices are 0-based. Failure events are measured by the outer measure `Q`. -/
theorem union_bound_p13 {m : ℕ} {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω)
    (h : SAARate.Sharp.E m → Ω → ℝ) (Θ : Set (SAARate.Sharp.E m))
    {S : Type*} [MeasurableSpace S] (Q : Measure S) (ω : ℕ → S → Ω)
    (q ℓ : ℕ) (x : Fin q → SAARate.Sharp.E m)
    (hc : ∀ (s : ℕ → Ω) (N : ℕ),
      (∀ i j : Fin q, i.val < ℓ → ℓ ≤ j.val → SAARate.Sharp.saaObj h s N (x i) < SAARate.Sharp.saaObj h s N (x j)) →
        (SAARate.Sharp.argminOn (SAARate.Sharp.saaObj h s N) Θ).Nonempty ∧
          IsFace (SAARate.Sharp.argminOn (SAARate.Sharp.expectedObj P h) Θ) (SAARate.Sharp.argminOn (SAARate.Sharp.saaObj h s N) Θ))
    (hgap : ∀ i j : Fin q, i.val < ℓ → ℓ ≤ j.val →
      SAARate.Sharp.expectedObj P h (x i) < SAARate.Sharp.expectedObj P h (x j)) :
    ∃ ε > 0, ∀ N : ℕ,
      Q {s | ¬ ((SAARate.Sharp.argminOn (SAARate.Sharp.saaObj h (fun k => ω k s) N) Θ).Nonempty ∧
          IsFace (SAARate.Sharp.argminOn (SAARate.Sharp.expectedObj P h) Θ) (SAARate.Sharp.argminOn (SAARate.Sharp.saaObj h (fun k => ω k s) N) Θ))} ≤
        (∑ i ∈ Finset.univ.filter (fun i : Fin q => i.val < ℓ),
            Q {s | SAARate.Sharp.expectedObj P h (x i) + ε ≤ SAARate.Sharp.saaObj h (fun k => ω k s) N (x i)}) +
          ∑ j ∈ Finset.univ.filter (fun j : Fin q => ℓ ≤ j.val),
            Q {s | SAARate.Sharp.saaObj h (fun k => ω k s) N (x j) ≤ SAARate.Sharp.expectedObj P h (x j) - ε} := by sorry

end SAARate.Polyhedral
Source
Shapiro & Homem-de-Mello, On Rate of Convergence of Optimal Solutions of Monte Carlo Approximations of Stochastic Programs, preprint (SPEPS copy, edoc.hu-berlin.de), pp. 12–13, proof of Theorem 3.2, (3.16), (3.18) and the display following it
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 7, 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