Вы находитесь на странице: 1из 40

Fractal Geometry,

Graph and Tree


Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal Geometry, Graph and Tree
Constructions
Masters Thesis Presentation
Tommy Lofstedt
Department of Mathematics and Mathematical Statistics
University of Umea
tommy@cs.umu.se
February 8, 2008
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Outline of presentation
Problem Description
Fractal Geometry
Graph-Directed Constructions
Tree Constructions
Equivalence of Graph and Tree Constructions
Equivalence for Union
Results and Conclusions
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Problem description

Previous research

Goals

Purpose

Methods

Graph-directed Constructions

Tree Constructions
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry
2400 km
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry
2800 km
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry
3450 km
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry What is a fractal?
Description:
A fractal is a rough or fragmented geometric shape
that can be subdivided in parts, each of which is
(at least approximately) a reduced/size copy of the
whole.
Benot Mandelbrot, 1975
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry What is a fractal?
Denition

The set has ne structure, it has details on arbitrary


scales.

The set is too irregular to be described with classical


euclidean geometry, both locally and globally.

The set has some form of self-similarity, this could be


approximate or statistical self-similarity.

The Hausdor dimension of the set is strictly greater


than its Topological dimension.

The set has a very simple denition, i.e. it can be


dened recursively.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry What is a fractal?
Example (The Cantor set)

E
0
:

E
1
:

E
2
:

. . .

:
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry What is a fractal?
Example (The von Koch curve)

E
0
:

E
1
:

E
2
:

. . .

:
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Similarity dimension

Magnify a unit line segment twice

Magnify a unit square twice

Magnify a unit cube twice


In general we have m
D
= N. Solving for D gives the
dimension as:
D =
log N
log m
.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Similarity dimension

What happens with the more complex examples?

The Cantor set gives, in E


k+1
, two copies of E
k
of size
1
3
. Thus, D =
log 2
log 1/
1
3
= 0.6309 . . .

The von Koch curve gives, in E


k+1
, four copies of E
k
of
size
1
3
. Thus, D =
log 4
log 1/
1
3
= 1.2618 . . .

Roughly, the dimension of a set describes how much


space the set lls.

However, this method is not general.


Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Hausdor dimension
The Hausdor dimension:

The most general notion

Dened for all sets in metric spaces

Requires some measure theory


The notion of a measure:

Ascribe a numerical size to a set

A function : S [0, ]

() = 0

If A B then (A) (B)

i =1
A
i
)

i =1
(A
i
)
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Hausdor dimension
The Hausdor measure has all these properties, and is
dened as:
Denition
Let F be a subset of a metric space, e.g. R
n
, and s R
+
,
then for any > 0 we let:
H
s

(F) = inf
_

i =1
diam(U
i
)
s
: {U
i
} is a cover of F
_
When decreases, H
s

(F) increases, and approaches a limit


as 0. We write:
H
s
(F) = lim
0
H
s

(F).
We call H
s
(F) the s-dimensional Hausdor measure of F.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Hausdor dimension
If H
s
(F) < then H
t
(F) = 0 when t > s. There is a
critical value of s at which H
s
(F) jumps from to 0. This
value is known as the Hausdor dimension of F, and is
denoted dim
H
F.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Box-counting dimension
The Box-counting dimension:

The most common denition in practical use

Easy to calculate mathematically

Easy to estimate empirically


Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Box-counting dimension
Note that:

The number of line segments of length that are


needed to cover a line of length l is l /

The number of squares with side length that are


needed to cover a square with area A is A/
2

The number of cubes with side length that are needed


to cover a cube with volume V are V/
3

The dimension of the object we try to cover is the


power of the side length, , of the box used
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Box-counting dimension
Let the number of boxes needed to cover a set F be N

(F).
Following the discussion above, the number of boxes needed
to cover the object should be proportional to the box size:

(F)
C

s
, 0

lim
0
N

(F)

s
= C.

lim
0
(log N

(F) + s log ) = log C.

s = lim
0
log N

(F) log C
log
= lim
0
log N

(F)
log
.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Fractal geometry Box-counting dimension
More formally:
Denition
The lower and upper Box-counting dimensions of a set F are
dened as
dim
B
F = liminf
0
log N

(F)
log
dim
B
F = limsup
0
log N

(F)
log
,
respectively. If their values are equal, we refer to the
common value as the Box-counting dimension of F
dim
B
F = lim
0
log N

(F)
log
.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Graph-Directed Constructions
The Cantor set, C, is constructed using two similarity
transformations, S
1
(x) =
x
3
and S
2
(x) =
x
3
+
2
3
. Then
C = S
1
(C) S
2
(C) and
H
s
(C) = H
s
(S
1
(C)) + H
s
(S
2
(C)).
By the scaling property of the Hausdor measure we have
H
s
(C) =
_
1
3
_
s
H
s
(C) +
_
1
3
_
s
H
s
(C) = 2
_
1
3
_
s
H
s
(C).
This means that
1 = 2
_
1
3
_
s
and therefore
s =
log 2
log 3
,
the Hausdor dimension of C, the Cantor set.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Graph-Directed Constructions
U
0
V
0
U
1
V
1
U
2
V
2
U
3
V
3
U
4
V
4
U V
Consider a case where
U = S
1
(U) S
2
(V)
and
V = T
1
(U) T
2
(V).
The Hausdor measure of U and V is then
H
s
(U) = H
s
(S
1
(U)) + H
s
(S
2
(V))
and
H
s
(V) = H
s
(T
1
(U)) + H
s
(T
2
(V)).
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Graph-Directed Constructions
By the scaling property of the Hausdor measure we get
H
s
(U) = r
s
1
H
s
(U) + r
s
2
H
s
(V)
and
H
s
(V) = r
s
3
H
s
(U) + r
s
4
H
s
(V).
This is a linear relationship, and therefore we can write the
above as
v =
_
H
s
(U)
H
s
(V)
_
, M =
_
r
s
1
r
s
2
r
s
3
r
s
4
_
,
which gives the neat matrix equation
v = Mv.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Graph-Directed Constructions

Generalizes to directed multigraphs, G = (V, E)

Hausdor and Box-counting dimensions can be found


directly

Each vertex, v V, corresponds to a metric space

Each edge corresponds to a similarity S


e
, with
contraction r
e

The dimension of such graph-directed sets is given by


an the n n adjacency matrix with elements
A
(s)
i ,j
=

eE
i ,j
r
s
e
.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Graph-Directed Constructions
A set that is described by the graph
has adjacency matrix
A
(s)
=
_
1
2
s
1
4
s
1
2
s
3
4
s
_
.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Graph-Directed Constructions
The spectral radius of A
(s)
, is the largest positive eigenvalue
of the matrix:
(A
(s)
) = max
1i n

i
.
As we saw before, the Hausdor dimensions of the sets is the
eigenvector corresponding to eigenvalue 1.
It turns out that the value of s for which we have spectral
radius 1 and a corresponding eigenvector with all positive
elements is the Hausdor dimension of the underlying set,
i.e.
dim
H
F = s.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Graph-Directed Constructions
This is formalized in the following theorems:
Theorem
Let E
1
, . . . , E
n
be a family of graph-directed sets, and
{T
(i ,j )
}, be a strongly connected similarities without
overlaps. Then there is a number s such that
dim
H
E
i
= dim
B
E
i
= dim
B
E
i
= s and 0 < H
s
(E
i
) < for
all i = 1, . . . , n. Also, s is the unique number satisfying
(A
(s)
) = 1.
Theorem
Each graph-directed construction has dimension
s = max{s
H
: H SC(G)}, where s
H
is the unique number
such that (H
(s
H
)
) = 1. The construction object, F, has
positive and -nite H
s
measure. This number s is such
that dim
H
F = dim
B
F = dim
B
F = s.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Tree Constructions

An alphabet of N letters is denoted as


E = {0, . . . , N 1}

A string is a series of letters from E, e.g. 10110

The length of the string and is written as ||

There is a unique string of length 0, denoted by

n is the prex of of length n

Write if || || and =

Let [] be the set of all strings with prex

The edges of a tree are labeled with the symbols of the


alphabet

The strings are paths in the tree and is the root

Let E
(n)
be the tree of all string of length n

Let E
()
be the tree of all nite strings

Let E
()
be the set of all innite strings
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Tree Constructions
Let real numbers w

be given for each node of the tree


E
()
, such that
w

> w

when <
and
lim
||
w

= 0 for E
()
.
A metric is now dened as follows:

If = then (, ) = 0

If = , then (, ) = w

, where is the longest


common prex of and
Then is a metric on E
()
such that diam[] = w

for all .
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Tree Constructions
With this metric, we have the following properties:

The space E
()
is a metric space

The space E
()
is complete

The space E
()
is separable

[] is an open ball in E
()

{[] : E
()
} is a countable base for the open sets
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Tree Constructions
If we have several recurrent subsets, as in the graph-directed
case, we can let one tree represent each set, called path
spaces.
Metrics are dened for each path space, so that
diam[] = w

, as above. The numbers w

are dened as
w

v
= q
v
,
w
e
= r
e
w

,
with r
e
being the contraction ratio of a similarity
corresponding to an edge e between nodes in the tree and
where q
v
is the diameter of the tree rooted at v.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Tree Constructions
If we select the q
v
such that
(diam[])
s
=

e
(diam[e])
s
, (1)
for some s, then there exists a measure on each of the
spaces satisfying ([]) = (diam[])
s
for all E
()
v
.
Theorem
If is a metric on E
()
v
and s > 0 satisfy
([]) = (diam[])
s
for all E
()
v
, then (B) = H
s
(B)
for all Borel sets B E
()
v
.
Thus, H
s
(E
()
v
) = (E
()
v
) = q
s
v
and since 0 < q
v
< we
have dim
H
E
()
v
= s.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Equivalence of Graph and Tree Constructions
Rearranging Equation 1 we see that
q
s
u
=

eE
uv
v a tree
r
s
e
q
s
v
,
for all trees u.
Expanding the above equation for each tree v
1
, . . . , v
n
we get
q
s
v
1
= r
s
e
v
1
v
1
q
s
v
1
+ + r
s
e
v
1
v
n
q
s
v
n
.
.
.
q
s
v
n
= r
s
e
v
n
v
1
q
s
v
1
+ + r
s
e
v
n
v
n
q
s
v
n
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Equivalence of Graph and Tree Constructions
Rewrite this in matrix form and we get
_
_
_
_
_
q
s
v
1
q
s
v
1
.
.
.
q
s
v
1
_
_
_
_
_
=
_
_
_
_
_
_
r
s
e
v
1
v
1
r
s
e
v
1
v
2
. . . r
s
e
v
1
v
n
r
s
e
v
2
v
1
r
s
e
v
2
v
2
. . . r
s
e
v
2
v
n
.
.
.
.
.
.
.
.
.
.
.
.
r
s
e
v
n
v
1
r
s
e
v
n
v
2
. . . r
s
e
v
n
v
n
_
_
_
_
_
_
_
_
_
_
_
q
s
v
1
q
s
v
1
.
.
.
q
s
v
1
_
_
_
_
_
.
Hence, by the above theorem, we can rewrite the equation as
_
_
_
_
_
H
s
(v
1
)
H
s
(v
2
)
.
.
.
H
s
(v
n
)
_
_
_
_
_
=
_
_
_
_
_
_
r
s
e
v
1
v
1
r
s
e
v
1
v
2
. . . r
s
e
v
1
v
n
r
s
e
v
2
v
1
r
s
e
v
2
v
2
. . . r
s
e
v
2
v
n
.
.
.
.
.
.
.
.
.
.
.
.
r
s
e
v
n
v
1
r
s
e
v
n
v
2
. . . r
s
e
v
n
v
n
_
_
_
_
_
_
_
_
_
_
_
H
s
(v
1
)
H
s
(v
2
)
.
.
.
H
s
(v
n
)
_
_
_
_
_
.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Equivalence of Graph and Tree Constructions
That equation can in turn be rewritten as
v = A
(s)
v,
where
A
(s)
=
_
_
_
_
_
_
r
s
e
v
1
v
1
r
s
e
v
1
v
2
. . . r
s
e
v
1
v
n
r
s
e
v
2
v
1
r
s
e
v
2
v
2
. . . r
s
e
v
2
v
n
.
.
.
.
.
.
.
.
.
.
.
.
r
s
e
v
n
v
1
r
s
e
v
n
v
2
. . . r
s
e
v
n
v
n
_
_
_
_
_
_
, v =
_
_
_
_
_
_
H
s
(E
()
v
1
)
H
s
(E
()
v
2
)
.
.
.
H
s
(E
()
v
n
)
_
_
_
_
_
_
.
We note that v is an eigenvector of A
(s)
with eigenvalue 1,
and conclude that this is equivalent to the graph-directed
solution we saw before.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Equivalence for Union
The Hausdor and upper Box-dimensions are nitely stable:
Theorem
dim
k
_
i =1
F
i
= max
1i k
dimF
i
,
for any nite collection of sets {F
1
, . . . , F
k
}.
The Hausdor dimension is also countably stable, i.e.:
Theorem
If F
1
, F
2
, . . . is a countable sequence of sets, then
dim
H

_
i =1
F
i
= sup
1i <
dim
H
F
i
.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Equivalence for Union
This can be done equally well for with Graph-Directed
Constructions (and with Tree Constructions). Consider the
union of two graphs G
1
= (V
1
, E
1
) and G
2
= (V
2
, E
2
). This
is
G = G
1
G
2
= (V
1
V
2
, E
1
E
2
) = (V, E).
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Equivalence for Union
We have the following theorem:
Theorem
Let G
i
be graph-directed constructions. For the union graph
G =
n
_
i =1
G
i
= (
n
_
i =1
V
i
,
n
_
i =1
E
i
)
it holds that
s = max{s
H
: H SC(G)},
where s
H
is the unique number such that (H
(s
H
)
) = 1. The
construction object, F, has positive and -nite H
s
measure.
The number s is such that dim
H
F = dim
B
F = dim
B
F = s.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Equivalence for Union
We can trivially deduce the following corollary:
Corollary
Let two graphs, G
1
and G
2
be represented as adjacency
matrices, A
1
and A
2
, respectively. The union graph,
G
1
G
2
, then has the adjacency matrix
A =
_
A
1
0
0 A
2
_
,
i.e., the block-diagonal matrix with A
1
and A
2
on the
diagonal. Let
A
(s)
=
_
A
(s)
1
0
0 A
(s)
2
_
.
Then the value of s for which (A
(s)
) = 1 is such that
dim
H
F = dim
B
F = dim
B
F = s.
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
Results and Conclusions

Classical theory is just a special case of the


Graph-Directed Constructions

Result for graph union is never seen elsewhere

Tree Constructions are equivalent to Graph-Directed


Constructions

Are there equally simple results for intersection,


product, etc?

There is a very close relationship between fractal


geometry, graph theory and linear algebra. Can results
from these areas be used in fractal geometry?

Using multifractal theory is the natural next step to


describe measures with Graph-Directed constructions
Fractal Geometry,
Graph and Tree
Constructions
Tommy Lofstedt
Outline
Problem
Description
Fractal Geometry
Graph-Directed
Constructions
Tree Constructions
Equivalence of
Graph and Tree
Constructions
Equivalence for
Union
Results and
Conclusions
The End
Thank you for listening!

Вам также может понравиться