Proposition 2.2(i) — penalties from generating functions have zero conditional mean under
ProvedInfoRelax.IdealPenalty.proposition_2_2_iconditional-expectationdualitydynamic-programminginformation-relaxationp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a relaxation of , let the action set be countable, and let be generating functions such that each depends only on the first actions, each is almost everywhere strongly measurable, and all are bounded almost everywhere by one constant. Define
Then exists for every policy (so ), and for every nonanticipative policy ,
In particular is dual feasible, with the defining inequality of holding with equality.
Formalization Note is the random variable , where each 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
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.