Footnote 9, p.19 — the nested functionals are monotone and translation equivariant (and map to , with (5.17))
ProvedRiskStopTC.nested_propertiesdynamic-programmingoptimal-stoppingp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1risk-measuresstochastic-programming
Throughout, is a probability space with a filtration , , , and for a fixed . Relations between random variables hold -almost surely.
Let , , be monotone, translation equivariant mappings with (5.17), and let be the nested functionals (5.1). For :
- maps into ;
- is monotone: in implies ;
- is translation equivariant: for , ;
- has the local property: for , .
The proof of Theorem 5.6 applies all four properties to the composed mappings .
Differences from the source:
- generalization: the nested functionals are monotone and translation equivariant → also: ρ_{s,t} maps Z_t into Z_s and satisfies (5.17) for A ∈ F_s; reason: the proof of Theorem 5.6 uses both for ρ_{m,T} (p.21, 'by (5.17)'); dropping clauses 1 and 4 gives the footnote
- encoding: Z_t = L^p(Ω, F_t, P) as classes of a.e.-equal functions; ⪯ and = almost surely → functions Ω → ℝ with an F_t-measurable representative in L^p (IsLp (ℱ t) P p); relations ≤ᵐ[P] and =ᵐ[P]; reason: every class has such a representative; Lean convention; same instances
- encoding: monotone translation equivariant mappings ϱ_t|F_t : Z_{t+1} → Z_t with (5.17), t = 0, …, T − 1 → one family ϱ : ℕ → (Ω → ℝ) → Ω → ℝ with IsStoppingFamily (maps-to, monotone, translation equivariant, (5.17)) on IsLp (ℱ (t+1)) arguments, t < T; reason: values off Z_{t+1} and maps with t ≥ T are never used; monotonicity makes each ϱ_t respect a.e. equality, so it acts on classes; same instances
- encoding: ρ_{s,t} for 0 ≤ s < t ≤ T, and ρ_{T,T}(Z_T) ≡ Z_T (footnote 10) → nested ϱ s t, with ρ_{t,t} = id for every t; reason: the paper fixes ρ_{T,T} = id; the identity at other diagonal indices is used only where the statement reads as the paper's at s = t
Preamble
import Mathlib import Definitions.Def_RiskFunctionals_NestedAVaR_Compositions import Definitions.Def_MultiperiodRisk_Bellman_TestSet import Definitions.Def_RiskStopTC_IsStoppingFamily import Definitions.Def_RiskStopTC_nested open MeasureTheory RiskFunctionals.NestedAVaR
Formal statement
namespace RiskStopTC
/-- **The nested functionals inherit the axioms.** Pichler, Liu & Shapiro, arXiv:1808.10807v3,
footnote 9, p.19: "Recall that the considered nested functionals are assumed to be monotone and
translation equivariant, while can be convex or concave."
For `s ≤ t ≤ T`, `ρ_{s,t} = nested ϱ s t`:
1. maps `Z_t` into `Z_s`;
2. is monotone on `Z_t`;
3. is translation equivariant: `ρ_{s,t}(X + W) = ρ_{s,t}(X) + W` for `X ∈ Z_t`, `W ∈ Z_s`;
4. has the local property (5.17): `ρ_{s,t}(1_A X) = 1_A ρ_{s,t}(X)` for `X ∈ Z_t`, `A ∈ F_s`.
**Formalization Note.** `ρ_{s,s}` is the identity. Clauses 1 and 4 are not in the footnote; the
proof of Theorem 5.6 uses them (p.21, "by (5.17)", applied to the composed mappings). -/
theorem nested_properties {Ω : Type*} {mΩ : MeasurableSpace Ω} (ℱ : Filtration ℕ mΩ) (P : Measure[mΩ] Ω)
[IsProbabilityMeasure P] (p : ENNReal) (hp : 1 ≤ p) (T : ℕ) (hℱ : IsStandardFiltration ℱ T)
(ϱ : ℕ → (Ω → ℝ) → Ω → ℝ) (hϱ : IsStoppingFamily ℱ P p T ϱ)
(s t : ℕ) (hst : s ≤ t) (htT : t ≤ T) :
(∀ X, IsLp (ℱ t) P p X → IsLp (ℱ s) P p (nested ϱ s t X)) ∧
(∀ X X', IsLp (ℱ t) P p X → IsLp (ℱ t) P p X' → X ≤ᵐ[P] X' →
nested ϱ s t X ≤ᵐ[P] nested ϱ s t X') ∧
(∀ X W, IsLp (ℱ t) P p X → IsLp (ℱ s) P p W →
nested ϱ s t (X + W) =ᵐ[P] nested ϱ s t X + W) ∧
(∀ X, IsLp (ℱ t) P p X → ∀ A : Set Ω, MeasurableSet[ℱ s] A →
nested ϱ s t (A.indicator X) =ᵐ[P] A.indicator (nested ϱ s t X)) := by sorry
end RiskStopTC
Source
Pichler, Liu & Shapiro, Risk averse stochastic programming: time consistency and optimal stopping, arXiv:1808.10807v3, footnote 9, p.19
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.