Theorem 2.7: Poisson stochastic-integral criterion, characteristic function and moments
ProvedKyprianou.poissonRandomMeasureIntegralLawsThe Poisson random sum is almost surely absolutely convergent iff min(1,|f|) is intensity-integrable, with the exact characteristic-function, mean and second-moment formulae under their stated conditions.
import Mathlib import Definitions.Def_LeanLevy_RandomMeasure_PoissonRandomMeasure import Definitions.Def_LeanLevy_RandomMeasure_PoissonPointFamily import Definitions.Def_KLSModular_LevyStochCalc_Poisson_RandomMeasure open MeasureTheory ProbabilityTheory open scoped NNReal ENNReal set_option autoImplicit false
namespace Kyprianou
theorem poissonRandomMeasureIntegralLaws
{Ω E : Type} [MeasurableSpace Ω] [MeasurableSpace E]
[Nonempty E] (P : Measure Ω) [IsProbabilityMeasure P]
(m : Measure E) [SigmaFinite m]
(K : ℕ → Ω → ℕ) (Y : ℕ → ℕ → Ω → E)
(hd : ProbabilityTheory.IsPoissonPointFamily K Y m P)
(f : E → ℝ) (hf : Measurable f) :
((∀ᵐ ω ∂P,
Integrable f (ProbabilityTheory.poissonRandomMeasure K Y ω)) ↔
Integrable (fun x => min (1 : ℝ) |f x|) m) ∧
(Integrable (fun x => min (1 : ℝ) |f x|) m →
∀ ξ : ℝ,
charFun (P.map (fun ω =>
∫ x, f x ∂(ProbabilityTheory.poissonRandomMeasure K Y ω))) ξ =
Complex.exp
(∫ x, (Complex.exp (Complex.I *
(((ξ * f x : ℝ) : ℂ))) - 1) ∂m)) ∧
(Integrable f m →
(∫ ω, (∫ x, f x ∂
(ProbabilityTheory.poissonRandomMeasure K Y ω)) ∂P) =
∫ x, f x ∂m) ∧
((Integrable f m ∧ MemLp f 2 m) →
(∫ ω, (∫ x, f x ∂
(ProbabilityTheory.poissonRandomMeasure K Y ω)) ^ 2 ∂P) =
(∫ x, f x ^ 2 ∂m) + (∫ x, f x ∂m) ^ 2)
:= by sorry
end KyprianouRead-back
What the Lean code literally says, in plain math · opencode-go/muse-spark-1.3-contributor
The theorem Kyprianou.poissonRandomMeasureIntegralLaws states the following.
With implicit type parameters Ω E : Type, with installed local typeclass instances MeasurableSpace Ω, MeasurableSpace E, Nonempty E, for explicit P : Measure Ω with installed local instance IsProbabilityMeasure P, for explicit m : Measure E with installed local instance SigmaFinite m, for explicit K : ℕ → Ω → ℕ and explicit Y : ℕ → ℕ → Ω → E, for explicit witness hd : ProbabilityTheory.IsPoissonPointFamily K Y m P, for explicit f : E → ℝ with explicit witness hf : Measurable f, the following four-fold conjunction holds, where ∧ is logical conjunction, ↔ is logical equivalence, → is logical implication, ∀ ξ : ℝ is universal quantification over reals, and ∀ᵐ ω ∂P is almost-everywhere quantification over ω : Ω with respect to P:
First, ∀ᵐ ω ∂P, Integrable f (ProbabilityTheory.poissonRandomMeasure K Y ω) holds if and only if Integrable (fun x => min (1 : ℝ) |f x|) m holds, where in the first term ω : Ω is bound and the named imported term ProbabilityTheory.poissonRandomMeasure K Y ω supplies the measure with respect to which f is integrable, and in the second term x : E is bound and the function maps x to the real minimum of 1 and the absolute value |f x|.
Second, if Integrable (fun x => min (1 : ℝ) |f x|) m holds then for all ξ : ℝ the named imported term charFun applied to the named imported term P.map (fun ω => ∫ x, f x ∂(ProbabilityTheory.poissonRandomMeasure K Y ω)) and to ξ equals the named imported term Complex.exp applied to ∫ x, (Complex.exp (Complex.I * (((ξ * f x : ℝ) : ℂ))) - 1) ∂m, where in the pushforward the bound ω : Ω maps to the integral over bound x : E of f x with respect to ProbabilityTheory.poissonRandomMeasure K Y ω, and in the exponent the bound x : E maps to Complex.exp of Complex.I multiplied by the complex coercion of the real product ξ * f x, minus 1, integrated with respect to m.
Third, if Integrable f m holds then ∫ ω, (∫ x, f x ∂(ProbabilityTheory.poissonRandomMeasure K Y ω)) ∂P equals ∫ x, f x ∂m, where the left side is the integral over bound ω : Ω with respect to P of the inner integral over bound x : E of f x with respect to ProbabilityTheory.poissonRandomMeasure K Y ω, and the right side is the integral over bound x : E of f x with respect to m.
Fourth, if both Integrable f m and the named imported term MemLp f 2 m hold then ∫ ω, (∫ x, f x ∂(ProbabilityTheory.poissonRandomMeasure K Y ω)) ^ 2 ∂P equals (∫ x, f x ^ 2 ∂m) + (∫ x, f x ∂m) ^ 2, where the left side is the integral over bound ω : Ω with respect to P of the square of the inner integral, and the right side is the sum of the integral over bound x : E with respect to m of the square f x ^ 2 and the square of the integral over bound x : E with respect to m of f x.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.