3. Motivation : Poisson polytopes
1
2
3
4
5
6
0
1 23456
η ξ(η)
Question
What happens when the number of points goes to infinity ?
2/25 Institut Mines-Telecom Functional Poisson convergence
4. Poisson point process
Hypothesis
Points are distributed to a Poisson process of control measure tK
Definition
The number of points is a Poisson rv (t K(Sd−1))
Given the number of points, they are independently drawn
with distribution K
3/25 Institut Mines-Telecom Functional Poisson convergence
5. Rescaling : γ = 2/(d − 1)
Campbell-Mecke formula
E
x1,··· ,xk ∈ω
(k)
=
f (x) = tk
f (x1, · · · , xk)K(dx1, · · · , dxk)
4/25 Institut Mines-Telecom Functional Poisson convergence
6. Rescaling : γ = 2/(d − 1)
Campbell-Mecke formula
E
x1,··· ,xk ∈ω
(k)
=
f (x) = tk
f (x1, · · · , xk)K(dx1, · · · , dxk)
Mean number of points (after rescaling)
1
2
E
x=y∈ω
1 tγx−tγy ≤β =
t2
2 Sd−1⊗Sd−1
1 x−y ≤t−γβdxdy
4/25 Institut Mines-Telecom Functional Poisson convergence
7. Rescaling : γ = 2/(d − 1)
Campbell-Mecke formula
E
x1,··· ,xk ∈ω
(k)
=
f (x) = tk
f (x1, · · · , xk)K(dx1, · · · , dxk)
Mean number of points (after rescaling)
1
2
E
x=y∈ω
1 tγx−tγy ≤β =
t2
2 Sd−1⊗Sd−1
1 x−y ≤t−γβdxdy
Geometry
Vd−1(Sd−1
∩Bd
βt−γ (y)) = κd−1(βt−γ
)d−1
+
(d − 1)κd−1
2
(βt−γ
)d
+O(t−γ(
4/25 Institut Mines-Telecom Functional Poisson convergence
8. Rescaling : γ = 2/(d − 1)
Campbell-Mecke formula
E
x1,··· ,xk ∈ω
(k)
=
f (x) = tk
f (x1, · · · , xk)K(dx1, · · · , dxk)
Mean number of points (after rescaling)
1
2
E
x=y∈ω
1 tγx−tγy ≤β =
t2
2 Sd−1⊗Sd−1
1 x−y ≤t−γβdxdy
Geometry
Vd−1(Sd−1
∩Bd
βt−γ (y)) = κd−1(βt−γ
)d−1
+
(d − 1)κd−1
2
(βt−γ
)d
+O(t−γ(
4/25 Institut Mines-Telecom Functional Poisson convergence
10. Schulte-Th¨ale (2012) based on Peccati (2011)
P(t2/(d−1)
Tm(ξ) > x) − e−βx(d−1)
m−1
i=0
(βxd−1)i
i!
≤ Cx t− min(1/2,2/(d−1))
What about speed of convergence as a process ?
5/25 Institut Mines-Telecom Functional Poisson convergence
11. Configuration space
Definition
A configuration is a locally finite set of particles on a Polish spaceY
f dω =
x∈ω
f (x)
6/25 Institut Mines-Telecom Functional Poisson convergence
12. Configuration space
Definition
A configuration is a locally finite set of particles on a Polish spaceY
f dω =
x∈ω
f (x)
Vague topology
ωn
vaguely
−−−−→ ω ⇐⇒ f dωn
n→∞
−−−→ f dω
for all f continuous with compact support from Y to R
6/25 Institut Mines-Telecom Functional Poisson convergence
13. Configuration space
Definition
A configuration is a locally finite set of particles on a Polish spaceY
f dω =
x∈ω
f (x)
Vague topology
ωn
vaguely
−−−−→ ω ⇐⇒ f dωn
n→∞
−−−→ f dω
for all f continuous with compact support from Y to R
Be careful !
δn
vaguely
−−−−→ ∅
6/25 Institut Mines-Telecom Functional Poisson convergence
14. Convergence in configuration space
Theorem (DST)
dR(Pn, Q)
n→∞
−−−→ 0 =⇒ Pn
distr.
−−−→ Q
Convergence in NY
7/25 Institut Mines-Telecom Functional Poisson convergence
15. Example
Our settings
Pt : PPP of intensity tK on C ⊂ X
f : dom f = Ck/Sk −→ Y
Definition
T(
x∈η
δx ) =
(x1,··· ,xk )∈ηk
=
δt2/d f (x1,··· ,xk ) := ξ(η)
f ∗K : image measure of (tK)k by f
M: intensity of the target Poisson PP
Example
8/25 Institut Mines-Telecom Functional Poisson convergence
16. Main result
Theorem (Two moments are sufficient)
sup
F∈TV–Lip1
E F PPP(M) − E F T∗
(PPP(tK)
≤ distTV(L, M) + 2(var ξ(Y) − Eξ(Y))
9/25 Institut Mines-Telecom Functional Poisson convergence
17. Stein method
The main tool
Construct a Markov process (X(s), s ≥ 0)
with values in configuration space
ergodic with PPM(M) as invariant distribution
X(s)
distr.
−−−→ PPM(M)
for all initial condition X(0)
for which PPM(M) is a stationary distribution
X(0)
distr.
= PPM(M) =⇒ X(s)
distr.
= PPM(M), ∀s > 0
10/25 Institut Mines-Telecom Functional Poisson convergence
18. In one picture
PPP(M) PPP(M)
T#(PPP(tK))
P#
s (PPP(M))
P#
s (T#(PPP(tK)))
11/25 Institut Mines-Telecom Functional Poisson convergence
19. Realization of a Glauber process
Y
Time0
X(s)
sS1 S2
S1, S2, · · · : Poisson process of intensity M(Y) ds
Lifetimes : Exponential rv of param. 1
Remark : Nb of particles ∼ M/M/∞
12/25 Institut Mines-Telecom Functional Poisson convergence
24. Stein representation formula
Definition
PtF(η) = E[F(X(t)) | X(0) = η]
Fundamental Lemma
F(ω)πM(dω) − EF(ξ(η)) = E
∞
0
LPsF(ξ(η))ds
14/25 Institut Mines-Telecom Functional Poisson convergence
25. What we have to compute
Distance representation
dR(PPP(M), T#
(PPP(tK)))
= sup
F∈TV–Lip1
E
∞
0 Y
[PsF(ξ(η) + δy ) − PsF(ξ(η))] M(dy) ds
+ E
∞
0 y∈ξ(η)
[PsF(ξ(η) − δy ) − PtsF(ξ(η))] ds
15/25 Institut Mines-Telecom Functional Poisson convergence
26. Transformation of the stochastic integral
Bis repetita
dR(PPP(M), ξ(η))
= sup
F∈TV–Lip1
E
∞
0 Y
[PtF(ξ(η) + δy ) − PtF(ξ(η))] M(dy) dt
+ E
∞
0 y∈ξ(η)
[PtF(ξ(η) − δy ) − PtF(ξ(η))] dt
16/25 Institut Mines-Telecom Functional Poisson convergence
27. Mecke formula
Mecke formula ⇐⇒ IPP
E
y∈ζ
f (y, ζ) =
Y
Ef (y, ζ + δy ) M dy)
is equivalent to
E
Y
Dy U(ζ)f (y, ζ) M(dy) = E U(ζ)
Y
f (y, ζ)(dζ(y) − M(dy))
17/25 Institut Mines-Telecom Functional Poisson convergence
29. A bit of geometry
η + δx + δy ξ(η + δx + δy )
f (x,y)
19/25 Institut Mines-Telecom Functional Poisson convergence
30. Last step
dR(PPP(M), ξ(η))
= sup
F∈TV–Lip1
∞
0 Y
Eζ [PtF(ζ + δy ) − PtF(ζ)] (M − f ∗
Kk
)(dy) dt
+Remainder
20/25 Institut Mines-Telecom Functional Poisson convergence
31. A key property (on the target space)
Definition
Dx F(ζ) = F(ζ + δx ) − F(ζ)
21/25 Institut Mines-Telecom Functional Poisson convergence
32. A key property (on the target space)
Definition
Dx F(ζ) = F(ζ + δx ) − F(ζ)
Intertwining property
For the Glauber point process
Dx PtF(ζ) = e−t
PtDx F(ζ)
21/25 Institut Mines-Telecom Functional Poisson convergence
33. Consequence
∞
0 Y
Eζ [PtF(ζ + δy ) − PtF(ζ)] (M − f ∗
Kk
)(dy) dt
=
∞
0 Y
Eζ[Dy PtF(ζ)] (M − f ∗
Kk
)(dy) dt
=
∞
0
e−t
Y
Eζ[PtDy F(ζ)] (M − f ∗
Kk
)(dy) dt
≤
Y
|M − f ∗
Kk
|(dy)
22/25 Institut Mines-Telecom Functional Poisson convergence
36. Compound Poisson approximation
The process
L
(b)
t (µ) =
1
2
(x,y)∈µ2
t,=
x − y b
1 x−y ≤δt
Theorem (Reitzner-Schlte-Thle (2013) w.o. conv. rate)
Assume
t2δb
t
t→∞
−−−→ λ
25/25 Institut Mines-Telecom Functional Poisson convergence
37. Compound Poisson approximation
The process
L
(b)
t (µ) =
1
2
(x,y)∈µ2
t,=
x − y b
1 x−y ≤δt
Theorem (Reitzner-Schlte-Thle (2013) w.o. conv. rate)
Assume
t2δb
t
t→∞
−−−→ λ
N ∼ Poisson(κd λ/2)
25/25 Institut Mines-Telecom Functional Poisson convergence
38. Compound Poisson approximation
The process
L
(b)
t (µ) =
1
2
(x,y)∈µ2
t,=
x − y b
1 x−y ≤δt
Theorem (Reitzner-Schlte-Thle (2013) w.o. conv. rate)
Assume
t2δb
t
t→∞
−−−→ λ
N ∼ Poisson(κd λ/2)
(Xi , i ≥ 1) iid, uniform in Bd (λ1/d )
25/25 Institut Mines-Telecom Functional Poisson convergence
39. Compound Poisson approximation
The process
L
(b)
t (µ) =
1
2
(x,y)∈µ2
t,=
x − y b
1 x−y ≤δt
Theorem (Reitzner-Schlte-Thle (2013) w.o. conv. rate)
Assume
t2δb
t
t→∞
−−−→ λ
N ∼ Poisson(κd λ/2)
(Xi , i ≥ 1) iid, uniform in Bd (λ1/d )
Then
dTV(t2b/d
L
(b)
t ,
N
j=1
Xj
b
) ≤ c(|t2
δb
t − λ| + t− min(2/d,1)
)
25/25 Institut Mines-Telecom Functional Poisson convergence