Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 2.2(i) — penalties from generating functions have zero conditional mean under F\mathbb FF

Proved
InfoRelax.IdealPenalty.proposition_2_2_i

by mikedeng1 · Oct 8, 2026 · Mathlib 0df444a (Lean v4.33.1)

conditional-expectationdualitydynamic-programminginformation-relaxationp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1

Let G\mathbb GG be a relaxation of F\mathbb FF, let the action set be countable, and let (w0,…,wT)(w_0,\dots,w_T)(w0​,…,wT​) be generating functions such that each wtw_twt​ depends only on the first t+1t+1t+1 actions, each wt(a)w_t(a)wt​(a) is almost everywhere strongly measurable, and all are bounded almost everywhere by one constant. Define

zt(a)=E[wt(a)∣Gt]−E[wt(a)∣Ft],z(a)=∑t=0Tzt(a).z_t(a)=\mathbb E[w_t(a)\mid\mathcal G_t]-\mathbb E[w_t(a)\mid\mathcal F_t],\qquad z(a)=\sum_{t=0}^T z_t(a).zt​(a)=E[wt​(a)∣Gt​]−E[wt​(a)∣Ft​],z(a)=t=0∑T​zt​(a).

Then E[z(α)]\mathbb E[z(\alpha)]E[z(α)] exists for every policy α\alphaα (so z∈Zz\in\mathcal Zz∈Z), and for every nonanticipative policy αF∈AF\alpha_F\in\mathcal A_{\mathbb F}αF​∈AF​,

E[zt(αF)∣Ft]=0  almost surely, for all t,andE[z(αF)]=0.\mathbb E[z_t(\alpha_F)\mid\mathcal F_t]=0\ \text{ almost surely, for all } t,\qquad\text{and}\qquad\mathbb E[z(\alpha_F)]=0.E[zt​(αF​)∣Ft​]=0  almost surely, for all t,andE[z(αF​)]=0.

In particular zzz is dual feasible, with the defining inequality of ZF\mathcal Z_{\mathbb F}ZF​ holding with equality.

Formalization Note zt(αF)z_t(\alpha_F)zt​(αF​) is the random variable ω↦zt(αF(ω))(ω)\omega\mapsto z_t(\alpha_F(\omega))(\omega)ω↦zt​(αF​(ω))(ω), where each zt(a)z_t(a)zt​(a) is a fixed version of a conditional expectation. The countable action set and the almost-everywhere bound are pinned regularity assumptions.

Preamble
import Mathlib
import Definitions.Def_InfoRelax_IdealPenalty_Framework
import Definitions.Def_InfoRelax_IdealPenalty_RecursiveModel
import Definitions.Def_InfoRelax_IdealPenalty_Penalty

open MeasureTheory
Formal statement
namespace InfoRelax.IdealPenalty

/-- **Proposition 2.2(i) (Constructing Good Penalties)**, Brown, Smith & Sun (2010), Oper. Res.,
DOI 10.1287/opre.1090.0796, p. 5.

Let `𝔾` be a relaxation of `𝔽` and `(w_0, …, w_T)` generating functions, each `w_t` depending only
on `a_0, …, a_t`. Put `z_t(a) = 𝔼[w_t(a) | 𝒢_t] − 𝔼[w_t(a) | 𝓕_t]` and `z(a) = ∑_t z_t(a)`. Then
for all `α_F ∈ 𝒜_𝔽`, `𝔼[z_t(α_F) | 𝓕_t] = 0` for all `t`, and `𝔼[z(α_F)] = 0`.

The first conjunct, `z ∈ 𝒵` (the expectations `𝔼[z(α)]` exist for every policy), is part of
the statement so that (i) yields `z ∈ 𝒵_𝔽`.

**Formalization Note.** `z_t(α_F)` is the random variable `ω ↦ z_t(α_F(ω))(ω)`, where each
`z_t(a)` is a fixed version of a conditional expectation. Pinned regularity: countable action type
with the discrete σ-algebra; the `w_t(a)` a.e. strongly measurable and a.e. bounded by one constant
(`GeneratingFns`). `𝔼[· | 𝓕_t] = 0` holds almost surely. -/
theorem proposition_2_2_i {Ω X : Type*} [mΩ : MeasurableSpace Ω] [MeasurableSpace X]
    [DiscreteMeasurableSpace X] [Countable X] {T : ℕ} (μ : Measure Ω) [IsProbabilityMeasure μ]
    (A : Set (Fin (T + 1) → X)) (𝔽 𝔾 : Filtration (Fin (T + 1)) mΩ)
    (h𝔾 : IsRelaxation 𝔽 𝔾) (w : Fin (T + 1) → (Fin (T + 1) → X) → Ω → ℝ)
    (hw : GeneratingFns μ w) :
    penalty μ 𝔽 𝔾 w ∈ penalties μ A ∧
      ∀ α ∈ adaptedPolicies A 𝔽,
        (∀ t, μ[fun ω => penaltyComp μ 𝔽 𝔾 w t (α ω) ω | 𝔽 t] =ᵐ[μ] 0) ∧
          ∫ ω, penalty μ 𝔽 𝔾 w (α ω) ω ∂μ = 0 := by sorry

end InfoRelax.IdealPenalty
Source
Brown, Smith & Sun, Information Relaxations and Duality in Stochastic Dynamic Programs, Oper. Res. (Articles in Advance, 2010), DOI 10.1287/opre.1090.0796, p. 5, Proposition 2.2(i)
Human review
  • Endorsed by Shuze Chen · Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 8, 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