math / data /lean /OAI /Probability /FreeIsing /Brownian /IntegralProcess.lean
introvoyz045's picture
Migrated from GitHub
4a56cd6 verified
Raw History Blame Contribute Delete
6.62 kB
import OAI.Probability.FreeIsing.Brownian.FiltrationIntegralProperties
import OAI.Probability.FreeIsing.Stochastic.StoppedBrownianIntegral
namespace OAI
noncomputable section
open MeasureTheory ProbabilityTheory Filter
open scoped NNReal Topology
namespace Problem367.BrownianIntegralProcess
variable {Ξ© : Type*} [mΞ© : MeasurableSpace Ξ©] {P : Measure Ξ©}
theorem meanSquare_condExp_sub_le [IsFiniteMeasure P]
(m : MeasurableSpace Ξ©) {X Y : Ξ© β†’ ℝ}
(hX : MemLp X 2 P) (hY : MemLp Y 2 P) :
(∫ Ο‰, (P[X | m] Ο‰ - P[Y | m] Ο‰) ^ 2 βˆ‚P) ≀
∫ Ο‰, (X Ο‰ - Y Ο‰) ^ 2 βˆ‚P := by
have hsub := condExp_sub (hX.integrable (by norm_num))
(hY.integrable (by norm_num)) m
calc
_ = ∫ Ο‰, (P[X - Y | m] Ο‰) ^ 2 βˆ‚P := by
apply integral_congr_ae
filter_upwards [hsub] with ω hω
change P[X - Y | m] ω = P[X | m] ω - P[Y | m] ω at hω
rw [hω]
_ ≀ _ := by
have hbound := integral_norm_condExp_rpow_le (m := m)
(by norm_num : (1 : ℝ) ≀ 2)
(f := X - Y) (by
simpa only [Real.rpow_two, Real.norm_eq_abs, sq_abs, Pi.sub_apply] using
(hX.sub hY).integrable_sq)
simpa only [Real.rpow_two, Real.norm_eq_abs, sq_abs, Pi.sub_apply] using hbound
theorem finite_stoppedSum_eq_condExp [IsFiniteMeasure P]
{F : Filtration ℝβ‰₯0 mΞ©} {B : ℝβ‰₯0 β†’ Ξ© β†’ ℝ}
(hB : Martingale B F P) (s : Finset β„•) (a b : β„• β†’ ℝβ‰₯0)
(H : β„• β†’ Ξ© β†’ ℝ) (hab : βˆ€ i ∈ s, a i ≀ b i)
(hH : βˆ€ i ∈ s, StronglyMeasurable[F (a i)] (H i))
(C : β„• β†’ ℝ) (hb : βˆ€ i ∈ s, βˆ€α΅ Ο‰ βˆ‚P, β€–H i Ο‰β€– ≀ C i) (t : ℝβ‰₯0) :
P[(fun Ο‰ => βˆ‘ i ∈ s, H i Ο‰ * (B (b i) Ο‰ - B (a i) Ο‰)) | F t] =ᡐ[P]
fun Ο‰ => βˆ‘ i ∈ s, StoppedSimpleIntegral.intervalSum B (H i) (a i) (b i) t Ο‰ := by
have hint (i : β„•) (hi : i ∈ s) :
Integrable (fun Ο‰ => H i Ο‰ * (B (b i) Ο‰ - B (a i) Ο‰)) P :=
((hB.integrable (b i)).sub (hB.integrable (a i))).bdd_mul
((hH i hi).mono (F.le (a i))).aestronglyMeasurable (hb i hi)
have hsum := condExp_finsetSum hint (F t)
have hsumFun : (βˆ‘ i ∈ s, fun Ο‰ => H i Ο‰ * (B (b i) Ο‰ - B (a i) Ο‰)) =
(fun Ο‰ => βˆ‘ i ∈ s, H i Ο‰ * (B (b i) Ο‰ - B (a i) Ο‰)) := by
funext Ο‰
simp only [Finset.sum_apply]
rw [hsumFun] at hsum
have heach : βˆ€α΅ Ο‰ βˆ‚P, βˆ€ i ∈ s,
P[(fun Ο‰ => H i Ο‰ * (B (b i) Ο‰ - B (a i) Ο‰)) | F t] Ο‰ =
StoppedSimpleIntegral.intervalSum B (H i) (a i) (b i) t Ο‰ :=
s.eventually_all.mpr (fun i hi =>
StoppedSimpleIntegral.intervalSum_eq_condExp_terminal hB (hab i hi) (hH i hi) (hb i hi) t)
filter_upwards [hsum, heach] with Ο‰ hsum heach
simp only [Finset.sum_apply] at hsum
rw [hsum]
exact Finset.sum_congr rfl (fun i hi => heach i hi)
def stoppedUniformSum (B H : ℝβ‰₯0 β†’ Ξ© β†’ ℝ) (T : ℝβ‰₯0) (n : β„•)
(t : ℝβ‰₯0) (Ο‰ : Ξ©) : ℝ :=
βˆ‘ i ∈ Finset.range n, StoppedSimpleIntegral.intervalSum B
(H (BrownianGrid.uniformGrid T n i)) (BrownianGrid.uniformGrid T n i)
(BrownianGrid.uniformGrid T n (i + 1)) t Ο‰
theorem stoppedUniformSum_eq_condExp
{B H : ℝβ‰₯0 β†’ Ξ© β†’ ℝ} {F : Filtration ℝβ‰₯0 mΞ©}
(hB : IsPreBrownianReal B P) (hAd : StronglyAdapted F B)
(hind : βˆ€ s t, s ≀ t β†’ Indep
(MeasurableSpace.comap (fun Ο‰ => B t Ο‰ - B s Ο‰) inferInstance) (F s) P)
(hH : StronglyAdapted F H) {C : ℝ}
(hb : βˆ€α΅ Ο‰ βˆ‚P, βˆ€ t, |H t Ο‰| ≀ C) (T : ℝβ‰₯0) (n : β„•) (t : ℝβ‰₯0) :
P[BrownianGrid.uniformSum B H T n | F t] =ᡐ[P] stoppedUniformSum B H T n t := by
let : IsProbabilityMeasure P := hB.isGaussianProcess.isProbabilityMeasure
apply finite_stoppedSum_eq_condExp (BrownianFiltration.martingale_of_independent_past hB hAd hind)
(Finset.range n) (BrownianGrid.uniformGrid T n)
(fun i => BrownianGrid.uniformGrid T n (i + 1))
(fun i => H (BrownianGrid.uniformGrid T n i))
(fun i _ => BrownianGrid.uniformGrid_mono T n (Nat.le_succ i))
(fun i _ => hH _) (fun _ => C)
intro i hi
filter_upwards [hb] with ω hω
simpa only [Real.norm_eq_abs] using hω (BrownianGrid.uniformGrid T n i)
theorem exists_stopped_uniform_integral_process
{B H : ℝβ‰₯0 β†’ Ξ© β†’ ℝ} {F : Filtration ℝβ‰₯0 mΞ©}
(hB : IsPreBrownianReal B P) (hAd : StronglyAdapted F B)
(hind : βˆ€ s t, s ≀ t β†’ Indep
(MeasurableSpace.comap (fun Ο‰ => B t Ο‰ - B s Ο‰) inferInstance) (F s) P)
(hH : StronglyAdapted F H)
(hc : βˆ€α΅ Ο‰ βˆ‚P, Continuous (fun t => H t Ο‰))
{C : ℝ} (hC : 0 ≀ C) (hb : βˆ€α΅ Ο‰ βˆ‚P, βˆ€ t, |H t Ο‰| ≀ C) (T : ℝβ‰₯0) :
βˆƒ M : ℝβ‰₯0 β†’ Ξ© β†’ ℝ, Martingale M F P ∧
(βˆ€ t, MemLp (M t) 2 P) ∧
(βˆ€ t, (∫ Ο‰, M t Ο‰ βˆ‚P) = 0) ∧
(βˆ€ t, T ≀ t β†’ M t = M T) ∧
(∫ Ο‰, (M T Ο‰) ^ 2 βˆ‚P) =
(∫ t in (0 : ℝ)..(T : ℝ), ∫ Ο‰, (H t.toNNReal Ο‰) ^ 2 βˆ‚P) ∧
(βˆ€ t, Tendsto (fun n => ∫ Ο‰,
(stoppedUniformSum B H T n t Ο‰ - M t Ο‰) ^ 2 βˆ‚P) atTop (𝓝 0)) := by
let : IsProbabilityMeasure P := hB.isGaussianProcess.isProbabilityMeasure
obtain ⟨I, hIm, hI2, hIzero, hIiso, hIlim⟩ :=
BrownianFiltrationGrid.exists_terminal_uniform_integral_isometry
hB hAd hind H hH hc hC hb T
let M : ℝβ‰₯0 β†’ Ξ© β†’ ℝ := fun t => P[I | F t]
have hIint : Integrable I P := hI2.integrable (by norm_num)
have hafter (t : ℝβ‰₯0) (ht : T ≀ t) : M t = I :=
condExp_of_stronglyMeasurable (F.le t) (hIm.mono (F.mono ht)) hIint
have hH2 (t : ℝβ‰₯0) : MemLp (H t) 2 P := by
apply MemLp.of_bound ((hH t).mono (F.le t)).aestronglyMeasurable C
filter_upwards [hb] with ω hω
simpa only [Real.norm_eq_abs] using hω t
refine ⟨M, martingale_condExp I F P, ?_, ?_, ?_, ?_, ?_⟩
Β· intro t
exact hI2.condExp (by norm_num)
Β· intro t
exact (integral_condExp (F.le t)).trans hIzero
Β· intro t ht
exact (hafter t ht).trans (hafter T le_rfl).symm
Β· rw [hafter T le_rfl]
exact hIiso
Β· intro t
apply squeeze_zero (fun n => integral_nonneg (fun Ο‰ => sq_nonneg _)) ?_ hIlim
intro n
calc
_ = ∫ Ο‰, (P[BrownianGrid.uniformSum B H T n | F t] Ο‰ - P[I | F t] Ο‰) ^ 2 βˆ‚P := by
apply integral_congr_ae
filter_upwards [stoppedUniformSum_eq_condExp hB hAd hind hH hb T n t] with ω hω
change (stoppedUniformSum B H T n t Ο‰ - P[I | F t] Ο‰) ^ 2 = _
rw [hω]
_ ≀ _ := meanSquare_condExp_sub_le (F t)
(BrownianFiltrationGrid.uniformSum_memLp hB hAd hind H hH hH2 T n) hI2
end Problem367.BrownianIntegralProcess
end
end OAI