Você está na página 1de 13

Outline

Random Graph theory

Inhomogeneous Brownian motion


Rajkumar Saha
City University of New York

May 18, 2016

1 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

2 / 13

Random Graph theory

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

Brownian Motion
Brownian Motion
Definition (in low dimention case): A real-valued stochastic
process {B(t) : t 0} is called a (linear) Brownian Motion
with started in x R if the following holds :
i B(x) = x
ii the process has independent increament, i.e. for all time
0 6 t1 6 t2 6, 6 tn the increment
B(tn ) B(tn1 ), , B(t2 ) B(t1 ) are independent random
variables.
iii for all t 0 and h > 0 the increaments B(t + h) B(t) are
normally distributed with expectation zero and varience h.
iv the function t B(t) is continuous. Inaddition, we say that
{B(t) : t 0} is a standard Brownian Motion if if x = 0

3 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

Brownian motion with drift and scaling


Relation to standard Brownian motion:
Suppose that Z = Zt : t [0, ) is a standard Brownian motion,
and that R and (0, ). Let Xt = t + Zt for t [0, ).
Then X = Xt : t [0, ) is a Brownian motion with drift
parameter and scale parameter .

4 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

Graphs of the Brownian motion on one dimention:

5 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

Critical Random Gaphs


Defination:
The random graph G(n, p(edge) = p(n)) is obtained from the
complete graph on n vertices, by independently retaining each
edge with probability p(n) and deleting with probability
1 q(n) and the n asymtotic component sizes are the
classical object of study. A fundamental result going back to
Erdo 00 s and Re 0 nyi says that when p(n) = c/n for fixed
0 < c < be independent observations such that E (Yi ) = 0,
then
i for c < 1 the largest component size is (log (n))
ii for c > 1 the largest component has size (n) while the
second largest component size is (log (n)).
iii for c = 1 the largest and second largest components both
have size (n2/3 ).
6 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

Critical Random Gaphs of Erdo 00 s and Re 0 nyi n=50:

7 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

The breadth-first walk


Defination: Consider a graph on vertices (1, 2, , n). We
shall specify the breadth-first ordering (v (1), , v (n)) of the
vertices, and an associated integer-valued sequence
(z(i); o i n) we shall call breath-first walk.
For example, each element exists at a certain level (or depth) in
the tree:

8 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

A Non-Uniform graph Model


Underlying idea:
For a positive real vector X = (x1 , x2 , ......, xn ) q > 0 define a
random graph W(x, q) on a vertices {1, 2, , n} as follows.
Each pair (i, j) of vertices is an edge with probability
1 exp(qxi xj ), independently for distinct pairs.Pinterpret xi
as the size of vertex i P
and say a component C = iC xi .
Given X, define r = i xi r , r 1

9 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

Graphs of the Brownian motion on one dimention:

10 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

Multiplicative Coalescent
There is a natural process describing W(x, q) as q varies, which we
now describe in somewhat different notation.Fix x l 2 For each
pair i < j, create a exponential (rate 1) r.v. ij , independent from
different pairs. Given t, cosider the graph where there exist an
edge (i, j) if ij txi xj : this is a construction of W(x, t),
simultaneously for all 0 t < . Let Xi (x, t) be the size of the
i 0 th largest component of this W(x, t) , and let
X (x, t) = Xi (x, t), i 1 Picture the typical state y = (yi ) as a
collection of cluster of sizes y1 , y2 , an initial vector xof finite
length, X (x, t) is a continuous time finite-state(the state-space
depending on the initial x)Markov chain whose dynamics are
described by
for each pair of clusters of sizes (x, y ) they merge at rate xy into a
cluster of size x + y . The natural state space for this process is l 2 .
11 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

Examples
Example:1 Define the (coalescent Kernel, K (x, y ), a typical
kernel considered in the applied science literature is:
K (x, y ) = (x 1/3 + y 1/3 )(x 1/3 + y 1/3 )

12 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Outline

Random Graph theory

R-code:

13 / 13

Rajkumar Saha

Inhomogeneous Brownian motion

Você também pode gostar