Finite-sum compensated step integrals
DefinitionKLSModular_LevyStochCalc_Poisson_CompensatedDensityStepIntegralVendored from the LevyStochCalc library (https://github.com/cgarryZA/LevyStochCalc, commit 984dbf0b790031eac05d56fe5a84b65e15e1e932). stepIntegral is the finite sum of the compensated integrals of a family of SimplePredictable integrands, for a probability measure and sigma-finite mark measure . Under strong measurability of the coefficients at their partition endpoints and an IsPoissonFiltration hypothesis, martingale_stepIntegral_compensated states that this finite-sum process is a martingale. compensated_inter_add_diff_ae gives almost-everywhere additivity of compensated masses over and when both sets are measurable and the reference intensity of is finite, with the required toReal conversions.
Original module imports
import LevyStochCalc.Poisson.CompensatedDensityMarkApprox
Original declaration excerpt
Shown in the original module context.
noncomputable def stepIntegral
{P : Measure Ω} [IsProbabilityMeasure P]
{ν : Measure E} [SigmaFinite ν]
(N : LevyStochCalc.Poisson.PoissonRandomMeasure P ν)
{T : ℝ} {k : ℕ} (Φ : Fin k → SimplePredictable Ω E ν T) (t : ℝ) (ω : Ω) : ℝ :=
∑ j, simpleIntegral N (Φ j) t ω
Module navigation
Direct project imports: LevyStochCalc/Poisson/CompensatedDensityMarkApprox.
Project result: The KLS conjecture.
Original module · LevyStochCalc.Poisson.CompensatedDensityStepIntegral · Frozen source 408915f.
import Definitions.Def_KLSModular_LevyStochCalc_Poisson_CompensatedDensityMarkApprox
/-
Copyright (c) 2026 Christian Garry. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Christian Garry
-/
-- Vendored from https://github.com/cgarryZA/LevyStochCalc at commit 984dbf0b790031eac05d56fe5a84b65e15e1e932 (upstream path LevyStochCalc/Poisson/CompensatedDensityStepIntegral.lean; Apache-2.0; provenance in formalization/vendor/vendor-lock.json of the KLS repository)
open MeasureTheory ProbabilityTheory
open scoped NNReal ENNReal
namespace LevyStochCalc
namespace Poisson
namespace Compensated
universe u v
variable {Ω : Type u} [MeasurableSpace Ω]
variable {E : Type v} [MeasurableSpace E]
/-- The compensated integral of a **finite family** of simple predictable
integrands: `∑ⱼ ∫ φⱼ dÑ`. -/
noncomputable def stepIntegral
{P : Measure Ω} [IsProbabilityMeasure P]
{ν : Measure E} [SigmaFinite ν]
(N : LevyStochCalc.Poisson.PoissonRandomMeasure P ν)
{T : ℝ} {k : ℕ} (Φ : Fin k → SimplePredictable Ω E ν T) (t : ℝ) (ω : Ω) : ℝ :=
∑ j, simpleIntegral N (Φ j) t ω
/-- A finite family of adapted simple predictables integrates to a martingale on the
natural filtration (the finite sum of the per-piece compensated martingales). -/
lemma martingale_stepIntegral_compensated
{P : Measure Ω} [IsProbabilityMeasure P]
{ν : Measure E} [SigmaFinite ν]
(N : LevyStochCalc.Poisson.PoissonRandomMeasure P ν)
(ℱ : Filtration ℝ ‹MeasurableSpace Ω›) (hℱ : IsPoissonFiltration N ℱ)
{T : ℝ} {k : ℕ} (Φ : Fin k → SimplePredictable Ω E ν T)
(h_adapt : ∀ j : Fin k, ∀ i : Fin (Φ j).N,
@MeasureTheory.StronglyMeasurable Ω ℝ _
(ℱ ((Φ j).partition i.castSucc))
((Φ j).ξ i)) :
MeasureTheory.Martingale (fun t : ℝ => stepIntegral N Φ t)
ℱ P := by
have hfun : (fun t : ℝ => stepIntegral N Φ t)
= ∑ j : Fin k, (fun t : ℝ => simpleIntegral N (Φ j) t) := by
funext t ω
simp only [stepIntegral, Finset.sum_apply]
rw [hfun]
have hmart : ∀ s : Finset (Fin k),
MeasureTheory.Martingale (∑ j ∈ s, fun t : ℝ => simpleIntegral N (Φ j) t)
ℱ P := by
intro s
induction s using Finset.induction with
| empty =>
simp only [Finset.sum_empty]
exact MeasureTheory.martingale_zero ℝ _ P
| insert j s hj ih =>
rw [Finset.sum_insert hj]
exact (martingale_simpleIntegral_compensated N ℱ hℱ (Φ j) (h_adapt j)).add ih
exact hmart Finset.univ
/-- **Compensated additivity over `inter`/`diff`** (a.e.). For measurable `B` with
finite intensity and measurable `C`, `Ñ(B) = Ñ(B ∩ C) + Ñ(B ∖ C)` a.e. (where the
`ℕ`-valued count `N(·,B)` is finite). Measure additivity (`measure_inter_add_diff`)
in `toReal`. -/
lemma compensated_inter_add_diff_ae
{P : Measure Ω} [IsProbabilityMeasure P]
{ν : Measure E} [SigmaFinite ν]
(N : LevyStochCalc.Poisson.PoissonRandomMeasure P ν)
{B C : Set (ℝ × E)} (hB : MeasurableSet B) (hC : MeasurableSet C)
(hfin : LevyStochCalc.Poisson.referenceIntensity ν B ≠ ⊤) :
(fun ω => N.compensated B ω)
=ᵐ[P] (fun ω => N.compensated (B ∩ C) ω + N.compensated (B \ C) ω) := by
filter_upwards [N.integer_valued hB hfin] with ω hω
obtain ⟨n, hn⟩ := hω
have hBfin : N.N ω B ≠ ⊤ := by rw [hn]; exact ENNReal.natCast_ne_top n
have hint_ne : N.N ω (B ∩ C) ≠ ⊤ :=
ne_top_of_le_ne_top hBfin (measure_mono Set.inter_subset_left)
have hdiff_ne : N.N ω (B \ C) ≠ ⊤ :=
ne_top_of_le_ne_top hBfin (measure_mono Set.sdiff_subset)
have hrefint : LevyStochCalc.Poisson.referenceIntensity ν (B ∩ C) ≠ ⊤ :=
ne_top_of_le_ne_top hfin (measure_mono Set.inter_subset_left)
have hrefdiff : LevyStochCalc.Poisson.referenceIntensity ν (B \ C) ≠ ⊤ :=
ne_top_of_le_ne_top hfin (measure_mono Set.sdiff_subset)
simp only [LevyStochCalc.Poisson.PoissonRandomMeasure.compensated]
rw [show N.N ω B = N.N ω (B ∩ C) + N.N ω (B \ C) from
(measure_inter_add_sdiff (μ := N.N ω) B hC).symm,
show LevyStochCalc.Poisson.referenceIntensity ν B
= LevyStochCalc.Poisson.referenceIntensity ν (B ∩ C)
+ LevyStochCalc.Poisson.referenceIntensity ν (B \ C) from
(measure_inter_add_sdiff (μ := LevyStochCalc.Poisson.referenceIntensity ν) B hC).symm,
ENNReal.toReal_add hint_ne hdiff_ne, ENNReal.toReal_add hrefint hrefdiff]
ring
end Compensated
end Poisson
end LevyStochCalc