Beruflich Dokumente
Kultur Dokumente
Study the dynamic behavior of fractals Understand the stochastic method Assignment of an address to the points of an attractor A Examination of the address space with dynamic system theory Chaotic properties of the address space can be transfered to the attractor
Christoph Traxler
Given: IFS {X; f1,f2,...,fn} with the attractor A = f1(A) ... fn(A), decomposed in n subsets Definition of the address of a point x A: x fa(A) address of x: a x fafb(A) address of x: ab ... Infinite many indices are necessary to identify a point x A exactly
Christoph Traxler
4.1
f1f2(A) 2
2 ... 12 ...
3
21 ...
22 ...
33 f1f3 31 32 13
4
23
11 12 21 22
Christoph Traxler
4.2
11213
Christoph Traxler 5
Address Space
!
Def.: The metric address space (n,dS) for an IFS {X; f1,f2,...,fn} is defined by n = {1 2 ... i ... | i {1 ... n} } and the Symbol Metric
d s ( , ) =
i =1
i i
(n + 1)i
4.3
Def.: The mapping :n A is defined as () = f1f2 ... fm ... (x) = aA, n, x arbitrary () is independent of the starting point x is a continuous mapping, similar addresses correspond to nearby points
! !
Christoph Traxler
Address Space
!
The metric spaces (n, dS) and (A,d) have the same properties Thus the dynamic behaviour of A can be analyzed by examining the dynamic behaviour of n -1(a) = { n: () = a}, a A, set of all addresses for a point a
Christoph Traxler
4.4
If each point of A has a unique address, then A is called totally disconnected The Sierpinski gasket consists of infinite many branching (touching) points, thus only corner points have a unique address
Totally disconnected
Christoph Traxler
Just touching
9
Overlapping
Dynamic Systems
!
Def.: Let (X,d) be a metric space and f: XX a function, then {X, f} is called dynamic system Def.: The sequence {fn(x)} = {x,f(x),f2(x),...} is called orbit of x X Example: (n, ) is a dynamic system, : nn is called shift function and defined as (1 2 3 4 ...) = 2 3 4 ...
10
Christoph Traxler
4.5
An IFS {X; f1,f2,...,fn} defines a special dynamic system {H(X), W}, where W is the Hutchinson operator The most interesting dynamic systems operate with non linear functions (Julia sets, Mandelbrot set, strange attractors)
Christoph Traxler
11
Dynamic Systems
!
Orbit of {R,f}: the fixpoints are the intersection points of the median with the graph
Christoph Traxler
12
4.6
Def.: {X,f} is a dynamic system, a point p X is called periodic point if there exists a number n > 0, so that fn(p) = p {p, f(p),...,fn(p)} is called cycle of p and n is called cycle length or period of p p periodic in {X,f} p is fixpoint of {X,fn} Minimal period of p = min{n | fn(p) = p }
! !
Christoph Traxler
13
Fixpoints
!
Def.: A fixpoint x=f(x) of {X,f} is called attractive, if there exists an > 0, so that f is a contraction mapping in B(x,) repelling, if there exists an > 0, s>1, so that d(x,f(y)) sd(x,y), yB(x,)
f f f x f f x f
attractive point
Christoph Traxler 14
repelling point
4.7
The dynamic system {H(X),W} of an IFS has exactly one fixpoint, no periodic points, all orbits converge to A Usually dynamic systems have several fixpoints A periodic point p with period n is called attractive (repelling), if p is an attractive (repelling) fixpoint of {X,fn}
Christoph Traxler
15
Fixpoints
!
attractive point
repelling point
Christoph Traxler 16
4.8
repelling point
repelling point
Christoph Traxler 17
Def.: {X; f1,f2,...,fn} is an IFS, the shift transformation S: A A is defined as S(a) = fi-1(a), a fi(A) {A, S} is a dynamic system fm(A) fn(A) S(a) is ambiguous a fm(A) fn(A) Orbits of points from A can be examined in {A ,S} (backward orbits)
18
Christoph Traxler
4.9
A
S
19
The knowledge of the dynamic behavior of {n,} can be applied to {A,S} as well It is much easier to examine the properties of {n,} than of {A,S} {n,} contains no geometric but only topological information about {A,S}
Christoph Traxler
20
4.10
Example:
a5 = a 6
-1
a1
a0
a3
Christoph Traxler
21
Examples:
4.11
Def.: A dynamic system {X,f} is called chaotic if: (1) (2) (3) {X,f} is transitive {X,f} is sensitive with respect to the starting conditions The set of periodic orbits of f is dense in X
23
Christoph Traxler
Def.: {X,f} is transitive, if there exists a number n for the open sets U, V X so that U fn (V) Orbits of points taken from an arbitrary small subset reach every part of X
U V
Christoph Traxler 24
4.12
{n,} is transitive:
! V open set V exists a
combination of symbol
25
Christoph Traxler
{n,} is transitive:
! B(,) = 12 ... m12 ...,
i {1,,n}
Christoph Traxler
26
4.13
Def.: {X,f} is sensitive towards starting conditions if there exists a number d > 0, so that x X and B(x,), > 0, y B(x,) and n > 0, so that d(fn(x), fn(y)) > d Nearby orbits move away from each other
B(x,) fn(y) fn(x)
27
y x
Christoph Traxler
{n,} is sensitive:
! = 12 ... m1... ! = 12 ... m2 ... ! dS(,) 1/(n+1)m ! dS(m(),m()) 1/(n+1) ! and are the starting conditions for
m () = 1... m () = 2 ...
Christoph Traxler
28
4.14
Def.: (X,d) is a metric space, a set S X is called dense in X, if for each point xX there exists a sequence {sn} in S, with lim sn=x (X is called closure of S) Example: The set of rational numbers Q is dense in R
Christoph Traxler
29
s1 = 1 s2 = 12 sm = 12 ... m
Christoph Traxler 30
4.15
Christoph Traxler
31
{n,} is a chaotic dynamic system, because: (1) It is transitive (2) It is sensitive towards starting conditions (3) Set of periodic orbits of is dense in n
{n,} chaotic {A,S} chaotic
32
Christoph Traxler
4.16
The attractor A of an IFS {X; f1,f2,...,fn} is approximated by a random sequence of points xn = fi(xn-1), i {1,...,n} Example: n=2, Random sequence 112 ...1: Point x0 x1=f1(x0) x2=f1(x1) x3=f2(x2) xmax Address 1 11 211 max The sequence {xmax,...,x3,x2,x1,x0} is an orbit of the chaotic system {A,S}
33
Christoph Traxler
Orbit {Tn()} converges to periodic repelling point Orbit {Tn()} comes closer to a periodic repelling point from which it moves away New points of A are always generated
! not periodic:
!
Christoph Traxler
34
4.17
How frequent are starting points with a too short period ? U(p) - number of periodic orbits with minimal period p
U ( p) = (n p
k divides p
k U (k )) / p
Christoph Traxler
35
Example {2,} :
! U(1) = 2 ! U(2) = 1 ! U(3) = 2 ! U(10) = 99 ! U(15) = 2182 ! U(20) = 52377 ! U(p) is a fast increasing function
fixpoints 1, 2 fixpoint 12
Christoph Traxler
36
4.18
If the starting point is periodic, then its period is very long with high probability Orbits of chaotic dynamic systems are distributed among the whole attractor
! {x, f1(x), f2f1(x), f2f1f1(x), ...} x A, covers A ! {p, f1(p), f2f1(p), f2f1f1(p), ...} p A, converges
Christoph Traxler
37
Conclusion
!
Examination of the chaotic behavior of fractal attractors of an IFS {X; f1,f2,...,fn} Introduction of the address space n and the dynamic system {n,} Relation between {n,} and {A,S} chaotic properties of {n,} can be transfered to {A,S}
Christoph Traxler
38
4.19
The stochastic method can be analyzed with (backward) orbits in {A,S} {A,S} is chaotic orbit is distributed among the whole attractor with very high probability The random orbit generates a good approximation of A
Christoph Traxler
39
4.20