[go: up one dir, main page]
Scribd
Upload
15 views7 pages

N - D - Fields Intro To Analysis On Them

The document discusses (N,d) Gaussian random fields, which are collections of multivariate, mean zero continuous stochastic processes. These fields can be either isotropic or anisotropic. Anisotropic fields have different properties in different directions. Some examples of (N,d) fields discussed include the Brownian sheet and fractional Brownian motion sheet. Time-anisotropic and space-anisotropic fields are also introduced, which satisfy certain operator self-similarity properties. Finally, fully anisotropic (N,d) fields that are both time and space anisotropic are discussed.

Uploaded by

Jihad Nasser
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
15 views7 pages

N - D - Fields Intro To Analysis On Them

The document discusses (N,d) Gaussian random fields, which are collections of multivariate, mean zero continuous stochastic processes. These fields can be either isotropic or anisotropic. Anisotropic fields have different properties in different directions. Some examples of (N,d) fields discussed include the Brownian sheet and fractional Brownian motion sheet. Time-anisotropic and space-anisotropic fields are also introduced, which satisfy certain operator self-similarity properties. Finally, fully anisotropic (N,d) fields that are both time and space anisotropic are discussed.

Uploaded by

Jihad Nasser
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

(N,d) Gaussian Fields

Jihad Nasser
MSU
Original author and Mentor: Professor Yimin Xiao

September 25, 2016

1/7

Random Gaussian Fields

A random Gaussian Field(RGF) is a way of collecting together a class of


multivariate, centered(mean 0) continuous stochastic processes.
They arise naturally in nature, through stochastic partial differential
equations.
Instead of isotropic Gaussian Random Fields we assume there is a
collection of i.i.d random variables {ij } (1 i j N) describing
changes in any direction.
We denote a general (N,d) GRF with values in R N by
N } = {X (t), ..., X (t)}
X={X (t) : t R+
1
d
These are called anisotropic Random Gaussian Fields

September 25, 2016

2/7

Space-anisotropic fields.
Each Coordinate Process Xj ,j=1,2,....,d ; a continuous stochastic process
Such that Xj : R N R N is continuous and j Let:
j (t, h) = E[(Xj (t + h) Xj (t))2 ] t, h R N
Many probabilistic and analytical properties of X are obtained through
asymptotic behavior of j .
As
|h| 0
it follows
j (t, h) 0 t R N

September 25, 2016

3/7

Space-anisotropic Fields
Then as |h| 0 if
i2 (t, h) 0 with different rates for i=1,2,...
We say X is Space-anisotropic
A class of fields giving rise to (N,d) GRFs are operator self-similar fields.
These fields satisfy the property:
dxd matrix operator- D=(dij ) such that for every c > 0
d

{X (ct) : t R N } = {c D X (t) : t R N }
c : R N R N is the linear operator
cD =

X
(lnc)n D n
n=0

n!

September 25, 2016

4/7

EX
The Brownian Sheet, an (N,d) field.
B = {B1 (t), ..., Bd (t)}; Bj (t) 7 R N
which is well understood due to the independence properties of Brownian
Motion
Fix the index vector H = (H1 , ..., HN ) (0, 1)N
We get the (1,N) fractional Brownian Motion Sheet
B H = {B1H (t), ..., BdH (t)}
With the Covariance function
1
E[B H (t)B H (s)] = ij (|t|2H |s|2H |s t|2H )
2
September 25, 2016

5/7

Time-anisotropic Fields

Let X={X (t) : t R N } be a Random Gaussian Field.


We say X is time-anisotropic if there an NxN matrix operator- K , such
that:
d

{X (c K t) : t R N } = {cX (t) : t R N }
The (N,1) fractional Brownian Motion Sheet is operator self-similar:
E[W0H (t)W0H (s)] =

N
Y
1

2
j=1

(|t|2Hj |s|2Hj |t s|2Hj )

September 25, 2016

6/7

(N,d) fields

Consider X a Gaussian Random field both time and space anisotropic.


d

{X (c K t) : t R N } = {c D X (t) : t R d }
Let X0 = {X0 (t) : t R N } be a real-valued centered Gaussian Field.
We create the field (X1 , ..., Xd ) of independent copies of X0 .
Let X = {X (t) : t R N } = {X1 , ..., Xd }.
We deduce many sample path behaviors from the function
[1] 2 (s, t) = E[(X0 (s) X0 (t))2 ] s, t R N

September 25, 2016

7/7

You might also like