normal approximation of geometric poisson functionals · 1. chaos expansion of poisson functionals...
TRANSCRIPT
![Page 1: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/1.jpg)
Gunter LastInstitut fur StochastikKarlsruher Institut fur Technologie
Normal approximationof geometric Poisson functionals
Gunter Last (Karlsruhe)joint work with
Daniel Hug, Giovanni Peccati, Matthias Schulte
presented at the
Simons Workshop on Stochastic Geometry and Point ProcessesUT Austin
May 5-8, 2014
![Page 2: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/2.jpg)
1. Chaos expansion of Poisson functionals
Setting
η is a Poisson process on some measurable space (X,X ) withintensity measure λ. This is a random element in the space Nof all integer-valued σ-finite measures on X, equipped with theusual σ-field (and distribution Πλ) with the following twoproperties
The random variables η(B1), . . . , η(Bm) are stochasticallyindependent whenever B1, . . . ,Bm are measurable andpairwise disjoint.
P(η(B) = k) =λ(B)k
k !exp[−λ(B)], k ∈ N0,B ∈ X ,
where∞ke−∞ := 0 for all k ∈ N0.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 3: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/3.jpg)
Definition (Difference operator)
For a measurable function f : N→ R and x ∈ X we define afunction Dx f : N→ R by
Dx f (µ) := f (µ+ δx )− f (µ).
For x1, . . . , xn ∈ X we define Dnx1,...,xn f : N→ R inductively by
Dnx1,...,xn f := D1
x1Dn−1
x2,...,xn f ,
where D1 := D and D0f = f .
Lemma
For any f ∈ L2(Pη)
Tnf (x1, . . . , xn) := EDnx1,...,xn f (η),
defines a function Tnf ∈ L2s(λn).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 4: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/4.jpg)
Definition
Let n ∈ N and g ∈ L2(λn). Then In(g) denotes the multipleWiener-Ito integral of g w.r.t. the compensated Poisson processη − λ. For c ∈ R let I0(c) := c. These integrals have theproperties
EIn(g)In(h) = n!〈g, h〉n, n ∈ N0,
EIm(g)In(h) = 0, m 6= n.
Here
g(x1, . . . , xn) :=1n!
∑π∈Σn
g(xπ(1), . . . , xπ(n))
denotes the symmetrization of g.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 5: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/5.jpg)
Definition
Let L2η denote the space of all random variables F ∈ L2(P) such
that F = f (η) P-almost surely, for some measurable function(representative) f : N→ R.
Theorem (Wiener ’38; Ito ’56; Y. Ito ’88; L. and Penrose ’11)
For any F ∈ L2η there are uniquely determined fn ∈ L2
s(λn) suchthat P-a.s.
F = EF +∞∑
n=1
In(fn).
Moreover, we have fn = 1n!Tnf , where f is a representative of F .
Gunter Last Normal approximation of geometric Poisson functionals
![Page 6: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/6.jpg)
2. Malliavin operators
Definition
Let F ∈ L2η have representative f . Define DxF := Dx f (η) for
x ∈ X, and, more generally Dnx1,...,xnF := Dn
x1,...,xn f (η) for anyn ∈ N and x1, . . . , xn ∈ X. The mapping(ω, x1, . . . , xn) 7→ Dn
x1,...,xnF (ω) is denoted by DnF (or by DF inthe case n = 1).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 7: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/7.jpg)
Theorem (Y. Ito ’88; Nualart and Vives ’90; L. and Penrose ’11)
Suppose F = EF +∑∞
n=1 In(fn) ∈ L2η. Then DF ∈ L2(P⊗ λ) iff F
is in dom D (the domain of the Malliavin difference operator),that is
∞∑n=1
nn!
∫f 2n dλn <∞.
In this case we have P-a.s. and for λ-a.e. x ∈ X that
DxF =∞∑
n=1
nIn−1(fn(x , ·)).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 8: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/8.jpg)
Definition
Let H ∈ L2η(P⊗ λ). Define h0(x) := EH(x) and
hn(x , x1, . . . , xn) := 1n!EDn
x1,...,xnH(x) and assume that H is indom δ (the domain of the operator δ), that is
∞∑n=0
(n + 1)!
∫h2
n dλn+1 <∞.
Then the Kabanov-Skorohod integral δ(H) of H is given by
δ(H) :=∞∑
n=0
In+1(hn).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 9: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/9.jpg)
Theorem (Nualart and Vives ’90)
Let F ∈ dom D and H ∈ dom δ. Then
E∫
(DxF )H(x)λ(dx) = EFδ(H).
Theorem (Picard ’96; L. and Penrose ’11)
Let H ∈ L1η(P⊗ λ) ∩ L2
η(P⊗ λ) be in the domain of δ and haverepresentative h. Then
δ(H) =
∫h(η − δx , x) η(dx)−
∫h(η, x)λ(dx) P-a.s.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 10: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/10.jpg)
Definition
The domain dom L of the Ornstein-Uhlenbeck generator L isgiven by all F ∈ L2
η satisfying
∞∑n=1
n2n!‖fn‖2n <∞.
For F ∈ dom L one defines
LF := −∞∑
n=1
nIn(fn).
The (pseudo) inverse L−1 of L is given by
L−1F := −∞∑
n=1
1n
In(fn).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 11: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/11.jpg)
3. Normal approximation: General results
Definition
The Wasserstein distance between the laws of two randomvariables Y1,Y2 is defined as
dW (Y1,Y2) = suph∈Lip(1)
|Eh(Y1)− Eh(Y2)|.
Theorem (Peccati, Sole, Taqqu and Utzet ’10)
Let F ∈ dom D such that EF = 0 and N be standard normal.Then
dW (F ,N) ≤ E∣∣∣1− ∫ (DxF )(−DxL−1F )λ(dx)
∣∣∣+ E
∫(DxF )2|DxL−1F |λ(dx).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 12: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/12.jpg)
Definition
The Kolmogorov distance between the laws of two randomvariables Y1,Y2 is defined by
dK (Y1,Y2) = supx∈R|P(Y1 ≤ x)− P(Y2 ≤ x)|.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 13: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/13.jpg)
Theorem (Schulte ’12)
For F ∈ dom D and N standard normal
dK (F ,N) ≤[E(
1−∫
(DxF )(−DxL−1F )λ(dx)
)2]1/2
+ 2c(F )
[E∫
(DxF )2(DxL−1F )2λ(dx)
]1/2
+ supt∈R
E∫
Dx1{F > t}(DxF )|DxL−1F |λ(dx),
where
c(F ) =
[E∫
(DxF )4λ(dx)
]1/2
+
[E∫
(DxF )2(DyF )2λ2(d(x , y))
]1/4 ((EF 4)1/4 + 1
).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 14: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/14.jpg)
Theorem (Hug, L. and Schulte ’13)
Let F ∈ L2η have the chaos expansion
F = EF +∞∑
n=1
In(fn)
and assume that Var F > 0. Assume that there are a > 0 andb ≥ 1 such that∫
|(fm ⊗ fm ⊗ fn ⊗ fn)σ|dλ|σ| ≤a bm+n
(m!)2(n!)2
for all σ ∈ Πmn. Let N be a standard Gaussian random variable.Then, under an additional integrability assumption,
dW
(F − EF√Var F
,N)≤ 2
152
∞∑n=1
n17/2 bn
bn/14c!
√a
Var F.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 15: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/15.jpg)
Theorem (L., Peccati and Schulte ’14)
Let F ∈ dom D be such that EF = 0 and Var F = 1, and let Nbe standard Gaussian. Then,
dW (F ,N) ≤ γ1 + γ2 + γ3,
where
γ21 := 16
∫ [E(Dx1F )2(Dx2F )2]1/2[E(D2
x1,x3F )2(D2
x2,x3F )2]1/2 dλ3,
γ22 :=
∫E(D2
x1,x3F )2(D2
x2,x3F )2 dλ3,
γ3 :=
∫E|DxF |3 λ(dx).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 16: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/16.jpg)
Theorem (L., Peccati and Schulte ’14)
For F ,G ∈ dom D with EF = EG = 0, we have
E(Cov(F ,G)−
∫(DxF )(−DxL−1G)λ(dx)
)2
≤ 3∫ [
E(D2x1,x3
F )2(D2x2,x3
F )2]1/2[E(Dx1G)2(Dx2G)2]1/2 dλ3
+
∫ [E(Dx1F )2(Dx2F )2]1/2[E(D2
x1,x3G)2(D2
x2,x3G)2]1/2 dλ3
+
∫ [E(D2
x1,x3F )2(D2
x2,x3F )2]1/2[E(D2
x1,x3G)2(D2
x2,x3G)2]1/2 dλ3.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 17: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/17.jpg)
4. The Boolean model
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
Gunter Last Normal approximation of geometric Poisson functionals
![Page 18: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/18.jpg)
4. The Boolean model
Gunter Last Normal approximation of geometric Poisson functionals
![Page 19: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/19.jpg)
4. The Boolean model
Gunter Last Normal approximation of geometric Poisson functionals
![Page 20: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/20.jpg)
Setting
η is a Poisson process on Kd (the space of convex bodies) withintensity measure Λ of the translation invariant form
Λ(·) = γ
∫∫1{K + x ∈ ·}dx Q(dK ),
where γ > 0, and Q is a probability measure on Kd withQ({∅}) = 0 and∫
Vd (K + C)Q(dK ) <∞, C ⊂ Rd compact.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 21: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/21.jpg)
Definition
The Boolean model (based on η) is the random closed set
Z :=⋃
K∈ηK .
Remark
The Boolean model is stationary, that is
Z + x d= Z , x ∈ Rd .
Remark
The intersection Z ∩W of the Boolean model Z with a convexset W ∈ Kd belongs to the convex ring Rd , that is, Z ∩W is afinite union of convex bodies.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 22: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/22.jpg)
4.1 Mean values
Theorem
Let ψ : Rd → R be measurable, additive, translation invariantand locally bounded. Let W ∈ Kd with Vd (W ) > 0. Then thelimit
δψ := limr→∞
Eψ(Z ∩ rW )
Vd (rW )
exists and is given by
δψ = E[ψ(Z ∩ [0,1]d )− ψ(Z ∩ ∂+[0,1]d )],
where ∂+[0,1]d is the upper right boundary of the unit cube[0,1]d . Moreover, due to ergodicity, there is a pathwise versionof this convergence.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 23: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/23.jpg)
Example
The intrinsic volumes V0, . . . ,Vd have all properties required bythe theorem. For K ∈ Kd , they are defined by the Steinerformula
Vd (K + rBd ) =d∑
j=0
r jκjVd−j(K ), r ≥ 0,
where κj is the volume of the Euclidean unit ball in Rj . Theintrinsic volumes can be additively extended to the convex ring.
Definition
Let Z0 be typical grain, that is, a random closed set withdistribution Q and define
vi := EVi(Z0) =
∫Vi(K )Q(dK ), i = 0, . . . ,d ,
Gunter Last Normal approximation of geometric Poisson functionals
![Page 24: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/24.jpg)
Example
The volume fraction of Z is defined by
p := EVd (Z ∩ [0,1]d ) = P(0 ∈ Z )
and given by the formula
p = 1− exp[−γvd ].
Note that
Eλd (Z ∩ B) = pλd (B), B ∈ B(Rd ),
that is δVd = p.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 25: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/25.jpg)
Example
For any W ∈ Kd ,
EVd−1(Z ∩W ) = Vd (W )(1− p)γvd−1 + Vd−1(W )p.
Therefore the surface density δVd−1 of Z is given by the formula
δVd−1 = (1− p)γvd−1.
Definition
The Boolean model is called isotropic if Q is invariant underrotations or, equivalently, if
Z d= ρZ
for all rotations ρ : Rd → Rd .
Gunter Last Normal approximation of geometric Poisson functionals
![Page 26: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/26.jpg)
Theorem (Miles ’76, Davy ’78)
Assume that Z is isotropic and let j ∈ {0, . . . ,d}. Then
EVj(Z ∩W )− Vj(W ) = −(1− p)d∑
k=j
Vk (W )Pj,k (γvj , . . . , γvd−1)
for any W ∈ Kd , where the polynomials Pj,k are defined below.In particular
δVj = −(1− p)Pj,d (γvj , . . . , γvd−1).
Remark
The first formula can be extended to more general additivefunctionals.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 27: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/27.jpg)
Definition
For j ∈ {0, . . . ,d − 1} and k ∈ {j , . . . ,d} define a polynomialPj,k on Rd−j of degree k − j by
Pj,k (tj , . . . , td−1) := 1{k = j}
+ ckj
k−j∑s=1
(−1)s
s!
d−1∑m1,...,ms=j
m1+...+ms=sd+j−k
s∏i=1
cmid tmi ,
whereck
j :=k !κk
j!κj.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 28: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/28.jpg)
4.2 Covariance structure
Definition
For p ≥ 1 the integrability assumption IA(p) holds if
EVi(Z0)p <∞, i = 0, . . . ,d ,
Assumption
IA(2) is assumed to hold throughout the rest of this section.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 29: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/29.jpg)
Definition
LetCW (x) := Vd (W ∩ (W + x)), x ∈ Rd ,
be the covariogram of W ∈ Kd and
Cd (x) := EVd (Z0 ∩ (Z0 + x)), x ∈ Rd ,
the mean covariogram of the typical grain.
Theorem
We have
P(0 ∈ Z , x ∈ Z )− p2 = (1− p)2(eγCd (x) − 1), x ∈ Rd ,
Var(Vd (Z ∩W )) = (1− p)2∫
CW (x)(eγCd (x) − 1
)dx , W ∈ Kd .
Gunter Last Normal approximation of geometric Poisson functionals
![Page 30: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/30.jpg)
Definition
Let W ∈ Kd satisfy Vd (W ) > 0. The asymptotic covariances ofthe intrinsic volumes are defined by
σi,j := limr→∞
Cov(Vi(Z ∩ rW ),Vj(Z ∩ rW ))
Vd (rW ), i , j = 0, . . . ,d .
Example
As it is well-known that
limr(W )→∞
CW (x)
Vd (W )= 1,
where r(W ) denotes the inradius of W , it follows that
σd ,d = (1− p)2∫ (
eγCd (x) − 1)dx .
Gunter Last Normal approximation of geometric Poisson functionals
![Page 31: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/31.jpg)
Theorem (Hug, L. and Schulte ’13)
The asymptotic covariances σi,j exist. Moreover, there is aconstant c > 0 depending only on the dimension and Λ, suchthat ∣∣∣∣Cov(Vi(Z ∩ rW ),Vj(Z ∩ rW ))
Vd (rW )− σi,j
∣∣∣∣ ≤ cr(W )
.
Moreover, this rate of convergence is optimal.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 32: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/32.jpg)
Main ideas of the Proof:The Fock space representation of Poisson functionals (cf. L.and Penrose ’11) gives for any F ,G ∈ L2
η, that
Cov(F ,G)
=∞∑
n=1
1n!
∫(EDn
K1,...,KnF )(EDn
K1,...,KnG) Λn(d(K1, . . . ,Kn)).
For an additive functional ψ : Rd → R define
fψ,W (µ) := ψ(Z (µ) ∩W ),
where Z (µ) is the union of all grains charged by the countingmeasure µ. Then
DnK1,...,Kn
fψ,W (µ)
= (−1)n(ψ(Z (µ) ∩ K1 ∩ . . . ∩ Kn ∩W )− ψ(K1 ∩ . . . ∩ Kn ∩W )).
Gunter Last Normal approximation of geometric Poisson functionals
![Page 33: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/33.jpg)
One can prove that for all n ∈ N and K1, . . . ,Kn ∈ Kd
|EDnK1,...,Kn
fψ,W (η)| ≤ β(ψ)d∑
i=0
Vi(K1 ∩ . . . ∩ Kn ∩W ),
where the constant β(ψ) does only depend on ψ,Λ and thedimension. Applying the (new) integral geometric inequalities∫
Vk (A ∩ (K + x))dx ≤ β1
d∑i=0
Vi(A)d∑
r=k
Vr (K ), A ∈ Kd ,
where β1 depends only on the dimension, and∫ d∑k=0
Vk (A ∩ K1 ∩ . . . ∩ Kn) Λn(d(K1, . . . ,Kn)) ≤ αnd∑
k=0
Vk (A),
where α := γ(d + 1)β1∑d
i=0 EVi(Z0) allows to use dominatedconvergence to derive the result.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 34: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/34.jpg)
4.3 Normal approximation
Theorem (Hug, L. and Schulte ’13)
Assume that IA(4) holds. Let W ∈ Kd with r(W ) ≥ 1,i ∈ {0, . . . ,d}, and N be a centred Gaussian random variable.Then, for all W ∈ Kd with sufficiently large inradius,
dW(Var(Vi(Z ∩W ))−1(Vi(Z ∩W )− EVi(Z ∩W )),N
)≤ cVd (W )−1/2,
where the constant c > 0 depends only on Λ and thedimension. If only IA(2) holds, then we still have convergencein distribution.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 35: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/35.jpg)
Main ideas of the Proof:Let
fn(K1, . . . ,Kn)
:=(−1)n
n!
(EVi(Z ∩ K1 ∩ · · · ∩ Kn ∩W )− Vi(K1 ∩ · · · ∩ Kn ∩W )
)be the kernels of the chaos expansion of Vi(Z ∩W ). Bound∫
|(fm ⊗ fm ⊗ fn ⊗ fn)σ|dΛ|σ|
using the previous integral geometric inequalities and apply oneof the previous theorems.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 36: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/36.jpg)
Remark
The above result remains true for any additive, locally boundedand measurable (but not necessarily translation invariant)functional ψ, provided that the variance ofψ(Z ∩W )/Vd (W )−1/2 does not degenerate for large W .
Theorem (Hug, L. and Schulte ’13)
If the typical grain Z0 has nonempty interior with positiveprobability, then the covariance matrix (σi,j) is positive definite.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 37: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/37.jpg)
5. Poisson-Voronoi tessellation
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
Gunter Last Normal approximation of geometric Poisson functionals
![Page 38: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/38.jpg)
5. Poisson-Voronoi tessellation
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
Gunter Last Normal approximation of geometric Poisson functionals
![Page 39: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/39.jpg)
Setting
η is a stationary Poisson process on Rd with unit intensity.
Definition
The Poisson-Voronoi tessellation is the collection of all cells
C(x , η) = {y ∈ Rd : ‖x − y‖ ≤ ‖z − y‖, z ∈ η}, x ∈ η.
For k ∈ {0, . . . ,d} let X k denote the system of all k -faces of thetessellation.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 40: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/40.jpg)
Theorem (Avram and Bertsimas ’93, Penrose and Yukich’05, L.,Peccati and Schulte ’14)
Fix W ∈ Kd and let
V (k ,i)r :=
∑G∈X
Vi(G ∩ rW ), r > 0,
where k ∈ {0, . . . ,d}, i ∈ {0, . . . ,min{k ,d − 1}}. There areconstants ck ,i , such that
dW
(V (k ,i)
r − EV (k ,i)r√
Var V (k ,i)r
,N)≤ ck ,i r−d/2, r ≥ 1.
A similar result holds for the Kolmogorov distance.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 41: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/41.jpg)
Main ideas of the Proof: Use the stabilizing properties of thePoisson Voronoi tessellation to show that
supr≥1
supx
E∣∣DxV (k ,i)
r∣∣5 + sup
r≥1supx ,y
E∣∣D2
x ,yV (k ,i)r
∣∣5 <∞and, for q := 1/20,∫
rWP(DxV (k ,i)
r 6= 0)qdx ≤ crd ,
∫rW
(∫rW
P(D2x ,yV (k ,i)
r 6= 0)qdx
)2
dy ≤ crd .
Gunter Last Normal approximation of geometric Poisson functionals
![Page 42: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/42.jpg)
Show with other methods that
lim infr→∞
r−d Var V (k ,i)r > 0.
Combine this with (a consequence of) the second orderPoincare inequality to conclude the result.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 43: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/43.jpg)
6. References
Avram, F. and Bertsimas, D. (1993). On central limittheorems in geometrical probability. Ann. Appl. Probab. 3,1033-1046.Chatterjee, S. (2009). Fluctuations of eigenvalues andsecond order Poincare inequalities. Probab. TheoryRelated Fields 143, 1-40.Hug, D., Last, G. and Schulte, M. (2013). Second orderproperties and central limit theorems for Boolean models.arXiv: 1308.6519.Ito, K. (1956). Spectral type of the shift transformation ofdifferential processes with stationary increments. Trans.Amer. Math. Soc. 81, 253-263.Ito, Y. (1988). Generalized Poisson functionals. Probab.Th. Rel. Fields 77, 1-28.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 44: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/44.jpg)
Kabanov, Y.M. (1975). On extended stochastic integrals.Theory Probab. Appl. 20, 710-722.Last, G., Peccati, G. and Schulte, M. (2014). Normalapproximation on Poisson spaces: Mehler’s formula,second order Poincare inequalities and stabilization.arXiv:1401.7568Last, G. and Penrose, M.D. (2011). Fock spacerepresentation, chaos expansion and covarianceinequalities for general Poisson processes. Probab. Th.Rel. Fields 150, 663-690.Nualart, D. and Vives, J. (1990). Anticipative calculus forthe Poisson process based on the Fock space. In:Seminaire de Probabilites XXIV, Lecture Notes in Math.,1426, 154-165.
Gunter Last Normal approximation of geometric Poisson functionals
![Page 45: Normal approximation of geometric Poisson functionals · 1. Chaos expansion of Poisson functionals Setting is aPoisson processon some measurable space (X;X) with intensity measure](https://reader033.vdocuments.site/reader033/viewer/2022060305/5f09398f7e708231d425d0f4/html5/thumbnails/45.jpg)
Peccati, G., Sole, J.L., Taqqu, M.S. and Utzet, F. (2010).Stein’s method and normal approximation of Poissonfunctionals. Ann. Probab. 38, 443-478.Peccati, G. and Zheng, C. (2010). Multi-dimensionalGaussian fluctuations on the Poisson space. Electron. J.Probab. 15, 1487-1527.Penrose, M.D. and Yukich, J.E. (2005). Normalapproximation in geometric probability. Barbour, A. and L.Chen (eds.). Stein’s method and applications. SingaporeUniversity Press.Schulte, M. (2012). Normal approximation of Poissonfunctionals in Kolmogorov distance. arXiv: 1206.3967.Wiener, N. (1938). The homogeneous chaos. Am. J. Math.60, 897-936.Wu, L. (2000). A new modified logarithmic Sobolevinequality for Poisson point processes and severalapplications. Probab. Theory Related Fields 118, 427-438.
Gunter Last Normal approximation of geometric Poisson functionals