153 Pages
English

Clustering percolation and directionally convex ordering of point processes

Gain access to the library to view online
Learn more

Description

Clustering, percolation and directionally convex ordering of point processes B. B?aszczyszyn ENS/INRIA, Paris, France blaszczy joint work with D. Yogeshwaran Summer Academy on Stochastic Analysis, Modelling and Simulation of Complex Structures Sollerhaus, Hirschegg/Kleinwalsertal, 11–24 September 2011 – p. 1

  • stochastic analysis

  • blaszczy joint

  • usual probabilistic

  • b?aszczyszyn ens

  • complex structures

  • mean measure

  • euclidean space


Subjects

Informations

Published by
Reads 10
Language English
Document size 4 MB

Exrait

Clustering, percolation and directionally
convex ordering of point processes
B. Błaszczyszyn
ENS/INRIA, Paris, France
www.di.ens.fr/ blaszczy
joint work with D. Yogeshwaran
Summer Academy on Stochastic Analysis, Modelling
and Simulation of Complex Structures
¨
Sollerhaus, Hirschegg/Kleinwalsertal, 11–24 September 2011
– p. 1Point process
Point process: random, locally finite, “pattern of points” Φ
in some spaceE.
A realization of Φ
2
onE =R .
– p. 2Point process; cont’d
Usual probabilistic formalism:
Φ is a measurable mapping from a probability space
(Ω,A, P) to a measurable spaceM “of point patterns”,
d
say, on Euclidean spaceE =R of dimensiond≥ 1.
– p. 3Point process; cont’d
Usual probabilistic formalism:
Φ is a measurable mapping from a probability space
(Ω,A, P) to a measurable spaceM “of point patterns”,
d
say, on Euclidean spaceE =R of dimensiond≥ 1.
A point pattern is considered as a counting measure; its
points are atoms of this measure. Hence
Φ(B) = (random) number of points of Φ in setB
for every measurable (Borel) subsetB⊂E.
– p. 3Point process; cont’d
Usual probabilistic formalism:
Φ is a measurable mapping from a probability space
(Ω,A, P) to a measurable spaceM “of point patterns”,
d
say, on Euclidean spaceE =R of dimensiond≥ 1.
A point pattern is considered as a counting measure; its
points are atoms of this measure. Hence
Φ(B) = (random) number of points of Φ in setB
for every measurable (Borel) subsetB⊂E.
Mean measure of Φ:
E(Φ(B)) = expected number of points of Φ inB.
– p. 3Clustering of points
Clustering in a point pattern roughly means that the points
lie in clusters (groups) with the clusters being spaced out.
– p. 4Clustering of points
Clustering in a point pattern roughly means that the points
lie in clusters (groups) with the clusters being spaced out.
How to compare clustering properties of two point
processes (pp) Φ , Φ having “on average” the same
1 2
number of points per unit of space?
– p. 4Clustering of points
Clustering in a point pattern roughly means that the points
lie in clusters (groups) with the clusters being spaced out.
How to compare clustering properties of two point
processes (pp) Φ , Φ having “on average” the same
1 2
number of points per unit of space?
More precisely, having the same mean measure:
E(Φ (B)) = E(Φ (B)) for allB⊂E.
1 2
– p. 4Stochastic comparison of point processes
But how do we compare random objects (their distributions)?
– p. 5Stochastic comparison of point processes
But how do we compare random objects (their distributions)?
→ stochastic orderings (to be explained).
– p. 5