The beauty of

Fractals

Ananda Dasgupta1

1 Indian Institute of Science, Education and Research

Kolkata

MC121
What are fractals?

I The term fractal was coined in 1975 by Benoit Mandelbrot.

What are fractals?

I The term fractal was coined in 1975 by Benoit Mandelbrot.

I Since then, there have been many attempts to define them ...
What are fractals?

I The term fractal was coined in 1975 by Benoit Mandelbrot.

I Since then, there have been many attempts to define them ...
I “a rough or fragmented geometric shape that can be split into
parts, each of which is (at least approximately) a reduced-size
copy of itself”
What are fractals?

I The term fractal was coined in 1975 by Benoit Mandelbrot.

I Since then, there have been many attempts to define them ...
I “a rough or fragmented geometric shape that can be split into
parts, each of which is (at least approximately) a reduced-size
copy of itself”
I “a geometric pattern that is repeated at every scale and so
cannot be represented by classical geometry ”
What are fractals?

I The term fractal was coined in 1975 by Benoit Mandelbrot.

I Since then, there have been many attempts to define them ...
I “a rough or fragmented geometric shape that can be split into
parts, each of which is (at least approximately) a reduced-size
copy of itself”
I “a geometric pattern that is repeated at every scale and so
cannot be represented by classical geometry ”
I “An object with a fractal dimension”
What are fractals?

I The term fractal was coined in 1975 by Benoit Mandelbrot.

I Since then, there have been many attempts to define them ...
I “a rough or fragmented geometric shape that can be split into
parts, each of which is (at least approximately) a reduced-size
copy of itself”
I “a geometric pattern that is repeated at every scale and so
cannot be represented by classical geometry ”
I “An object with a fractal dimension”
I Most definitions fall short by one way or the other...
What are fractals?

I The term fractal was coined in 1975 by Benoit Mandelbrot.

I Since then, there have been many attempts to define them ...
I “a rough or fragmented geometric shape that can be split into
parts, each of which is (at least approximately) a reduced-size
copy of itself”
I “a geometric pattern that is repeated at every scale and so
cannot be represented by classical geometry ”
I “An object with a fractal dimension”
I Most definitions fall short by one way or the other...
I Maybe, its best to start with a few examples.
The Koch Curve

I Start with a line segment.

The Koch Curve

I Start with a line segment.

I This is the level 0 Koch curve.
The Koch Curve

I Start with a line segment.

I This is the level 0 Koch curve.
I Trisect it.
The Koch Curve

I Start with a line segment.

I This is the level 0 Koch curve.
I Trisect it.
I Replace the middle segment by the
other two sides of an equilateral
triangle.
The Koch Curve

I Start with a line segment.

I This is the level 0 Koch curve.
I Trisect it.
I Replace the middle segment by the
other two sides of an equilateral
triangle.
I This gives us the level 1 Koch curve.
The Koch Curve

I Start with a line segment.

I This is the level 0 Koch curve.
I Trisect it.
I Replace the middle segment by the
other two sides of an equilateral
triangle.
I This gives us the level 1 Koch curve.
I Repeat this again ...
The Koch Curve

I Start with a line segment.

I This is the level 0 Koch curve.
I Trisect it.
I Replace the middle segment by the
other two sides of an equilateral
triangle.
I This gives us the level 1 Koch curve.
I Repeat this again ...
I and again ...
The Koch Curve

I Start with a line segment.

I This is the level 0 Koch curve.
I Trisect it.
I Replace the middle segment by the
other two sides of an equilateral
triangle.
I This gives us the level 1 Koch curve.
I Repeat this again ...
I and again ...
I and again ...
The Koch Curve

I Start with a line segment.

I This is the level 0 Koch curve.
I Trisect it.
I Replace the middle segment by the
other two sides of an equilateral
triangle.
I This gives us the level 1 Koch curve.
I Repeat this again ...
I and again ...
I and again ...
I The final Koch curve is the limit
obtained after an infinite number of
iterations!
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
I If you magnify a third of the curve ...
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
I If you magnify a third of the curve ...
I You get the whole curve back!
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
I If you magnify a third of the curve ...
I You get the whole curve back!
I The Koch curve is exactly self similar!
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
I If you magnify a third of the curve ...
I You get the whole curve back!
I The Koch curve is exactly self similar!
I The length grows on each iteration
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
I If you magnify a third of the curve ...
I You get the whole curve back!
I The Koch curve is exactly self similar!
I The length grows on each iteration
4
I ... by a factor of .
3
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
I If you magnify a third of the curve ...
I You get the whole curve back!
I The Koch curve is exactly self similar!
I The length grows on each iteration
4
I ... by a factor of .
3
I ...
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
I If you magnify a third of the curve ...
I You get the whole curve back!
I The Koch curve is exactly self similar!
I The length grows on each iteration
4
I ... by a factor of .
3
I ...
I ...
The Koch Curve

The Koch curve is a fascinating object!

I It is continuous everywhere ...
I ... but differentiable nowhere!
I If you magnify a third of the curve ...
I You get the whole curve back!
I The Koch curve is exactly self similar!
I The length grows on each iteration
4
I ... by a factor of .
3
I ...
I ...
I The length of the final curve is
infinite!!
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
I then its length is 1!
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
I then its length is 1!
1
I Use a ruler units long ...
3
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
I then its length is 1!
1
I Use a ruler units long ...
3
I its length is now 4!
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
I then its length is 1!
1
I Use a ruler units long ...
3
I its length is now 4!
I Keep on shrinking the ruler by a factor
1
of
3
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
I then its length is 1!
1
I Use a ruler units long ...
3
I its length is now 4!
I Keep on shrinking the ruler by a factor
1
of
3
I Measured length grows 4 times on
every step!
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
I then its length is 1!
1
I Use a ruler units long ...
3
I its length is now 4!
I Keep on shrinking the ruler by a factor
1
of
3
I Measured length grows 4 times on
every step!
I ...
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
I then its length is 1!
1
I Use a ruler units long ...
3
I its length is now 4!
I Keep on shrinking the ruler by a factor
1
of
3
I Measured length grows 4 times on
every step!
I ...
I ...
But how long is the Koch curve?
I The numerical length of a curve
depends on the size of your ruler!
I If your ruler is 1 unit long ...
I then its length is 1!
1
I Use a ruler units long ...
3
I its length is now 4!
I Keep on shrinking the ruler by a factor
1
of
3
I Measured length grows 4 times on
every step!
I ...
I ...
I And this continues, no matter how
small the ruler becomes!
Fractional dimension??!!

I For ordinary curves, this factor should

be 3 = 31 !
Fractional dimension??!!

I For ordinary curves, this factor should

be 3 = 31 !
I For plane figures,
Fractional dimension??!!

I For ordinary curves, this factor should

be 3 = 31 !
I For plane figures,
I this factor should be 9 = 32
Fractional dimension??!!

I For ordinary curves, this factor should

be 3 = 31 !
I For plane figures,
I this factor should be 9 = 32
I For solid bodies, it is 27 = 33
Fractional dimension??!!

I For ordinary curves, this factor should

be 3 = 31 !
I For plane figures,
I this factor should be 9 = 32
I For solid bodies, it is 27 = 33
I For the Koch curve, we have 4 = 3D
Fractional dimension??!!

I For ordinary curves, this factor should

be 3 = 31 !
I For plane figures,
I this factor should be 9 = 32
I For solid bodies, it is 27 = 33
I For the Koch curve, we have 4 = 3D
I where
log 4
D= = 1.261859507
log 3
is the fractal dimension of the Koch
curve.
Cantor’s dust

I Start with the interval [0, 1].

Cantor’s dust

I Start with the interval [0, 1].

I Drop the middle third.
Cantor’s dust

I Start with the interval [0, 1].

I Drop the middle third.
I Then drop the middle third of of the
remaining intervals.
Cantor’s dust

I Start with the interval [0, 1].

I Drop the middle third.
I Then drop the middle third of of the
remaining intervals.
I Continue ...
Cantor’s dust

I Start with the interval [0, 1].

I Drop the middle third.
I Then drop the middle third of of the
remaining intervals.
I Continue ...
I ...
Cantor’s dust

I Start with the interval [0, 1].

I Drop the middle third.
I Then drop the middle third of of the
remaining intervals.
I Continue ...
I ...
I After infinite iterations you get the
Cantor set.
Cantor’s dust

I Start with the interval [0, 1].

I Drop the middle third.
I Then drop the middle third of of the
remaining intervals.
I Continue ...
I ...
I After infinite iterations you get the
Cantor set.
I AKA the Cantor dust
Cantor’s dust
I First you throw away a length of 31 .
Cantor’s dust
I First you throw away a length of 31 .
2
I Next time 32
.
Cantor’s dust
I First you throw away a length of 31 .
2
I Next time 32
.
22
I Next ... 33
Cantor’s dust
I First you throw away a length of 31 .
2
I Next time 32
.
22
I Next ... 33
I Continue ...
Cantor’s dust
I First you throw away a length of 31 .
2
I Next time 32
.
22
I Next ... 33
I Continue ...
I In all, you throw away :

1 2 22
+ 2 + 3 + ...
3 3 3
Cantor’s dust
I First you throw away a length of 31 .
2
I Next time 32
.
22
I Next ... 33
I Continue ...
I In all, you throw away :

1 2 22
+ 2 + 3 + ...
3 3 3
I The infinite G.P. adds up to
1 1
× 2
=1
3 1− 3
Cantor’s dust
I First you throw away a length of 31 .
2
I Next time 32
.
22
I Next ... 33
I Continue ...
I In all, you throw away :

1 2 22
+ 2 + 3 + ...
3 3 3
I The infinite G.P. adds up to
1 1
× 2
=1
3 1− 3

I The dust has measure 0.

Cantor’s dust
I First you throw away a length of 31 .
2
I Next time 32
.
22
I Next ... 33
I Continue ...
I In all, you throw away :

1 2 22
+ 2 + 3 + ...
3 3 3
I The infinite G.P. adds up to
1 1
× 2
=1
3 1− 3

I The dust has measure 0.

I Yet it has an uncountably infinite
number of points.
“Dimension” of the Cantor dust

I If you have a ruler of unit length

“Dimension” of the Cantor dust

I If you have a ruler of unit length

I You can cover the dust with one ruler.
“Dimension” of the Cantor dust

I If you have a ruler of unit length

I You can cover the dust with one ruler.
1
I With a ruler of length 3
“Dimension” of the Cantor dust

I If you have a ruler of unit length

I You can cover the dust with one ruler.
1
I With a ruler of length 3
I You need only 2!
“Dimension” of the Cantor dust

I If you have a ruler of unit length

I You can cover the dust with one ruler.
1
I With a ruler of length 3
I You need only 2!
1
I For length 32
“Dimension” of the Cantor dust

I If you have a ruler of unit length

I You can cover the dust with one ruler.
1
I With a ruler of length 3
I You need only 2!
1
I For length 32
I you need 4
“Dimension” of the Cantor dust

I If you have a ruler of unit length

I You can cover the dust with one ruler.
1
I With a ruler of length 3
I You need only 2!
1
I For length 32
I you need 4
I The Cantor dust has dimension
log 2
D=
log 3
Siérpinski triangle

I Start with an equilateral triangle

Siérpinski triangle

I Start with an equilateral triangle

I Drop the middle triangle
Siérpinski triangle

I Start with an equilateral triangle

I Drop the middle triangle
I Continue dropping the middle triangles
Siérpinski triangle

I Start with an equilateral triangle

I Drop the middle triangle
I Continue dropping the middle triangles
I ...
Siérpinski triangle

I Start with an equilateral triangle

I Drop the middle triangle
I Continue dropping the middle triangles
I ...
I ...
Siérpinski triangle

I Start with an equilateral triangle

I Drop the middle triangle
I Continue dropping the middle triangles
I ...
I ...
I After infinite iterations,
Siérpinski triangle

I Start with an equilateral triangle

I Drop the middle triangle
I Continue dropping the middle triangles
I ...
I ...
I After infinite iterations,
I You get the Sierpinski triangle
Siérpinski triangle

I Start with an equilateral triangle

I Drop the middle triangle
I Continue dropping the middle triangles
I ...
I ...
I After infinite iterations,
I You get the Sierpinski triangle
I A fractal with dimension
log 3
D= = 1.584962501
log 2
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
I Let’s begin with ’F’ and play a little
game.
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
I Let’s begin with ’F’ and play a little
game.
I At each stage, change
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
I Let’s begin with ’F’ and play a little
game.
I At each stage, change
I ’F’ → ’FGF’
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
I Let’s begin with ’F’ and play a little
game.
I At each stage, change
I ’F’ → ’FGF’
I ’G’ → ’GGG’
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
I Let’s begin with ’F’ and play a little
game.
I At each stage, change
I ’F’ → ’FGF’
I ’G’ → ’GGG’
I Stage 1 : ’FGF’
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
I Let’s begin with ’F’ and play a little
game.
I At each stage, change
I ’F’ → ’FGF’
I ’G’ → ’GGG’
I Stage 1 : ’FGF’
I Stage 2 : ’FGFGGGFGF’
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
I Let’s begin with ’F’ and play a little
game.
I At each stage, change
I ’F’ → ’FGF’
I ’G’ → ’GGG’
I Stage 1 : ’FGF’
I Stage 2 : ’FGFGGGFGF’
I Stage 3 :
’FGFGGGFGFGGGGGGGGGFGFGGGFGF’
Cantor’s dust - a different approach

I Imagine an alphabet that has only two

letters
I ’F’ for forward
I ’G’ for gap
I Let’s begin with ’F’ and play a little
game.
I At each stage, change
I ’F’ → ’FGF’
I ’G’ → ’GGG’
I Stage 1 : ’FGF’
I Stage 2 : ’FGFGGGFGF’
I Stage 3 :
’FGFGGGFGFGGGGGGGGGFGFGGGFGF’
What about the Koch triangle?

I Here we need an alphabet with 3

letters
What about the Koch triangle?

I Here we need an alphabet with 3

letters
I ’F’ for going forward
What about the Koch triangle?

I Here we need an alphabet with 3

letters
I ’F’ for going forward
I ’L’ for turning left (through 60◦ )
What about the Koch triangle?

I Here we need an alphabet with 3

letters
I ’F’ for going forward
I ’L’ for turning left (through 60◦ )
I ’R’ for turning right (through 60◦ )
What about the Koch triangle?

I Here we need an alphabet with 3

letters
I ’F’ for going forward
I ’L’ for turning left (through 60◦ )
I ’R’ for turning right (through 60◦ )
I We start with a ’F’
What about the Koch triangle?

I Here we need an alphabet with 3

letters
I ’F’ for going forward
I ’L’ for turning left (through 60◦ )
I ’R’ for turning right (through 60◦ )
I We start with a ’F’
I In the next step we have ’FLFRRFLF’
What about the Koch triangle?

I Here we need an alphabet with 3

letters
I ’F’ for going forward
I ’L’ for turning left (through 60◦ )
I ’R’ for turning right (through 60◦ )
I We start with a ’F’
I In the next step we have ’FLFRRFLF’
I Repeat, replacing each ’F’ of the last
stage by ’FLFRRFLF’
What about the Koch triangle?

I Here we need an alphabet with 3

letters
I ’F’ for going forward
I ’L’ for turning left (through 60◦ )
I ’R’ for turning right (through 60◦ )
I We start with a ’F’
I In the next step we have ’FLFRRFLF’
I Repeat, replacing each ’F’ of the last
stage by ’FLFRRFLF’
I
’FLFRRFLFLFLFRRFLFRRFLFRRFLFLFLFRRFLF’
What about the Koch triangle?

I Here we need an alphabet with 3

letters
I ’F’ for going forward
I ’L’ for turning left (through 60◦ )
I ’R’ for turning right (through 60◦ )
I We start with a ’F’
I In the next step we have ’FLFRRFLF’
I Repeat, replacing each ’F’ of the last
stage by ’FLFRRFLF’
I
’FLFRRFLFLFLFRRFLFRRFLFRRFLFLFLFRRFLF’
I ...
L-Systems

I The string rewriting method discussed

above gives rise to L-systems.
L-Systems

I The string rewriting method discussed

above gives rise to L-systems.
I First described by Hungarian
theoretical botanist Aristid
Lindenmeyer.
L-Systems

I The string rewriting method discussed

above gives rise to L-systems.
I First described by Hungarian
theoretical botanist Aristid
Lindenmeyer.
I He wanted to describe the way plants
and cells grow.
L-Systems

I The string rewriting method discussed

above gives rise to L-systems.
I First described by Hungarian
theoretical botanist Aristid
Lindenmeyer.
I He wanted to describe the way plants
and cells grow.
I His clue : branches are often very
similar to trees!
L-Systems

I The string rewriting method discussed

above gives rise to L-systems.
I First described by Hungarian
theoretical botanist Aristid
Lindenmeyer.
I He wanted to describe the way plants
and cells grow.
I His clue : branches are often very
similar to trees!
I L-systems with suitable rules can
generate amazingly lifelike structures.
Michael Barnsley’s amazing copy machine

I Let’s begin with a triangle.

Michael Barnsley’s amazing copy machine

I Let’s begin with a triangle.

I Make three copies, reduced by a factor
of two.
Michael Barnsley’s amazing copy machine

I Let’s begin with a triangle.

I Make three copies, reduced by a factor
of two.
I Shift two of the copies, as shown.
Michael Barnsley’s amazing copy machine

I Let’s begin with a triangle.

I Make three copies, reduced by a factor
of two.
I Shift two of the copies, as shown.
I Repeat ...
Michael Barnsley’s amazing copy machine

I Let’s begin with a triangle.

I Make three copies, reduced by a factor
of two.
I Shift two of the copies, as shown.
I Repeat ...
I ...
Michael Barnsley’s amazing copy machine

I Let’s begin with a triangle.

I Make three copies, reduced by a factor
of two.
I Shift two of the copies, as shown.
I Repeat ...
I ...
I infinite iterations of the “copy and
shift” routine gives the Siérpinski
triangle!
What if you’d started with some other shape?
What if you’d started with some other shape?
What if you’d started with some other shape?
What if you’d started with some other shape?
What if you’d started with some other shape?
What if you’d started with some other shape?

The Siérpinski traingle is an attractor of the “copy and shift”

iterations!
Iterated function systems
I Start with S ⊆ [0, 1] × [0, 1]
Iterated function systems
I Start with S ⊆ [0, 1] × [0, 1]
I Carry out the set of functions
x y 
f1 : (x, y ) 7→ ,
2 2 
x +1 y
f2 : (x, y ) →
7 ,
2 2
!
x + 12 y + 1
f3 : (x, y ) 7→ ,
2 2
on every element of S.
Iterated function systems
I Start with S ⊆ [0, 1] × [0, 1]
I Carry out the set of functions
x y 
f1 : (x, y ) 7→ ,
2 2 
x +1 y
f2 : (x, y ) →
7 ,
2 2
!
x + 12 y + 1
f3 : (x, y ) 7→ ,
2 2
on every element of S.
I Form S 0 = f1 (S) ∪ f2 (S) ∪ f3 (S).
Iterated function systems
I Start with S ⊆ [0, 1] × [0, 1]
I Carry out the set of functions
x y 
f1 : (x, y ) 7→ ,
2 2 
x +1 y
f2 : (x, y ) →
7 ,
2 2
!
x + 12 y + 1
f3 : (x, y ) 7→ ,
2 2
on every element of S.
I Form S 0 = f1 (S) ∪ f2 (S) ∪ f3 (S).
I Repeat, taking S 0 as S, forever!
Iterated function systems
I Start with S ⊆ [0, 1] × [0, 1]
I Carry out the set of functions
x y 
f1 : (x, y ) 7→ ,
2 2 
x +1 y
f2 : (x, y ) →
7 ,
2 2
!
x + 12 y + 1
f3 : (x, y ) 7→ ,
2 2
on every element of S.
I Form S 0 = f1 (S) ∪ f2 (S) ∪ f3 (S).
I Repeat, taking S 0 as S, forever!
I Since each map is a contraction, the iterations will converge
to a final shape!
Iterated function systems
I Start with S ⊆ [0, 1] × [0, 1]
I Carry out the set of functions
x y 
f1 : (x, y ) 7→ ,
2 2 
x +1 y
f2 : (x, y ) →
7 ,
2 2
!
x + 12 y + 1
f3 : (x, y ) 7→ ,
2 2
on every element of S.
I Form S 0 = f1 (S) ∪ f2 (S) ∪ f3 (S).
I Repeat, taking S 0 as S, forever!
I Since each map is a contraction, the iterations will converge
to a final shape!
I Which is none other than the Siérpinski triangle for this
iterated function system!
Iterated function systems
I Start with S ⊆ [0, 1] × [0, 1]
I Carry out the set of functions
x y 
f1 : (x, y ) 7→ ,
2 2 
x +1 y
f2 : (x, y ) →
7 ,
2 2
!
x + 12 y + 1
f3 : (x, y ) 7→ ,
2 2
on every element of S.
I Form S 0 = f1 (S) ∪ f2 (S) ∪ f3 (S).
I Repeat, taking S 0 as S, forever!
I Since each map is a contraction, the iterations will converge
to a final shape!
I Which is none other than the Siérpinski triangle for this
iterated function system!
I Change the IFS to get other attractors.
Iterated function systems - Random version

I Iterating over shapes is computation intensive!

Iterated function systems - Random version

I Iterating over shapes is computation intensive!

I It is better to have an algorithm that works on points.
Iterated function systems - Random version

I Iterating over shapes is computation intensive!

I It is better to have an algorithm that works on points.
I We can play the chaos game!!
Iterated function systems - Random version

I Iterating over shapes is computation intensive!

I It is better to have an algorithm that works on points.
I We can play the chaos game!!
I Start with a random point.
Iterated function systems - Random version

I Iterating over shapes is computation intensive!

I It is better to have an algorithm that works on points.
I We can play the chaos game!!
I Start with a random point.
I Apply one of the functions in the IFS set chosen at random.
Iterated function systems - Random version

I Iterating over shapes is computation intensive!

I It is better to have an algorithm that works on points.
I We can play the chaos game!!
I Start with a random point.
I Apply one of the functions in the IFS set chosen at random.
I The probabaility of a given function being chosen is
proportional to the determinant of the transformation.
Iterated function systems - Random version

I Iterating over shapes is computation intensive!

I It is better to have an algorithm that works on points.
I We can play the chaos game!!
I Start with a random point.
I Apply one of the functions in the IFS set chosen at random.
I The probabaility of a given function being chosen is
proportional to the determinant of the transformation.
I If the determinant is zero for a transformation, a small
number is taken for the corresponding probability.
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
I ...
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
I ...
I Plot the points, discarding the first
few!
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
I ...
I Plot the points, discarding the first
few!
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
I ...
I Plot the points, discarding the first
few!
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
I ...
I Plot the points, discarding the first
few!
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
I ...
I Plot the points, discarding the first
few!
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
I ...
I Plot the points, discarding the first
few!
The chaos game

I Mark off three points A = (0, 0),

B = (1, 0) and C = (0.5, 1).
I Choose a point P ∈ [1, 1].
I Roll a dice.
I If the dice shows
I 1 or 2 : move to the midpoint of AP
I 3 or 4 : move to the midpoint of BP
I 5 or 6 : move to the midpoint of CP
I Repeat .. lots of times!
I ...
I ...
I Plot the points, discarding the first
few!
The chaos game
A more mathematical description

Iterate with one of the three maps

x y 
f1 : (x, y ) 7→ ,
2 2 
x +1 y
f2 : (x, y ) 7→ ,
2 2
!
x + 12 y + 1
f3 : (x, y ) 7→ ,
2 2

chosen randomly with equal probability.

The chaos game

Plot of
I the first 1000 points.
The chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
The chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
The chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
I the first 15000 points.
The chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
I the first 15000 points.
I the first 20000 points.
The chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
I the first 15000 points.
I the first 20000 points.
I the first 25000 points.
Another chaos game

Consider the IFS made from the following functions :

 
19.05x + 0.72y + 1.86 −0.15x + 16.9y − .28
f1 : (x, y ) 7→ ,
5.63x + 2.01y + 20.0 5.63x + 2.01y + 20.0
 
0.2x + 4.4y + 7.5 −0.3x − 4.4y − 10.4
f2 : (x, y ) 7→ ,
0.2x + 8.8y + 15.4 0.2x + 8.8y + 15.4
 
96.5x + 35.2y + 5.8 −131.4x − 6.5y + 19.1
f3 : (x, y ) 7→ ,
134.8x + 30.7y + 7.5 134.8x + 30.7y + 7.5
 
−32.5x + 5.81y − 2.9 122.9x − 0.1y − 19.9
f1 : (x, y ) 7→ ,
−128.1x − 24.3y − 5.8 −128.1x − 24.3y − 5.8

with odds of 60 : 1 : 20 : 19
Another chaos game

Plot of
I the first 1000 points.
Another chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
Another chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
Another chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
I the first 15000 points.
Another chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
I the first 15000 points.
I the first 20000 points.
Another chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
I the first 15000 points.
I the first 20000 points.
I the first 25000 points.
Another chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
I the first 15000 points.
I the first 20000 points.
I the first 25000 points.
Another chaos game

Plot of
I the first 1000 points.
I the first 5000 points.
I the first 10000 points.
I the first 15000 points.
I the first 20000 points.
I the first 25000 points.
References

I “The algorithmic beauty of plants” by A. Lindenmeyer and P.

Prusinkiewicz - available freely from
http://algorithmicbotany.org/papers/#webdocs
I “Fractals everywhere” by M. Barnsley.