You are on page 1of 26

)(Genetic Algorithms

. .

...
)(Presentation Overview

.
.
.
.
.
.

...
)(General Introduction

) (1975
) (John Holland ).(University of Michigan

) (Search ) (Optimization.
) (Genetic
.
) (Binary
).(Continues

) (Parallelism .

.

...
)(Biological Introduction

) (Gene

) (Chromosome
).(DNA

DNA



.


)(Genetic Code
.

...
)(Biological Introduction



) (George Mendel ).(1865
)...(Allele
) (dominate
).(recessive
) (Mutation
.
...

) (mitosis
) (Chromosomes .
) (meiosis

).(kinetochore

...
)(Biological Introduction

) (Mendel
) (evolution
.
) (Charles Darwin
) (Natural Selection
...

.

...
)(Main Algorithm



...



.

...
)(Binary Genetic Programming

)(Binary

.


...

...
) (Selecting the Variables and the Cost Function


) (Chromosome ...
p1,p2,pn ...
]Chromosome = [p1,p2,,pn

) (fitness
.
Cost = f(chromosome) = ...
)f(p1,p2,,pn
.
... ...
]chromosome = [ x, y , z , w
...
2x
)6 y 3 ( z w

2x < 4

y + 2z 6

644.5

12.33

= )f ( x, y, z , w

...
)(Variable Encoding and Decoding


) (Binary Digits
.

).(Decoding
)(Quantization
.

.
.
:

chromosome = [1111001001
14243 00101111
1424
3...0000101001
] 14243
genen

gene 2

gene1

...
)(Variable Encoding and Decoding

pn plo
=
):(Encoding
phi plo
m 1

} gene[ p]2 p
p =1

pnorm

gene[m] = round{ pnorm 2

):(Decoding
N gene

m
) ( M +1
gene
[
m
]
2
+
2

m =1

= pquant

qn = pquant ( phi plo ) + plo


:
pnorm plo . phi .
] gene[m pquant pn pnorm

...
)(The Population


.


.


.

...
)(Natural Selection


.

.

.
...

30
%20 6
.

...
)(Selection


.
...
) (Tournaments

.
...
.
) (Wheel of Fortune

.

...
)(Mating


) (meiosis .
)(Crossing Over
) (kinetochore .

.
...
.
00001111 :

00000000 :
11110000 :

11111111 :
:
01010101 :

01010110 :
10001110 :

10001101 :

...
)(Mutations


.
) (Global Optimization
).(Local Optimization
.


) (.
.

# mutations = ( N pop 1) N bits

...
)(Elitism



) (.
).(Elitism

.

.
.

...
)(Civil Engineering Applications

...1
11000 .
.
.

3000

4000

280

400

50

100


.
.

...
)(Civil Engineering Applications

...
)50 = (Population
)30 = (Generations
)60% = (Crossover Rate
)4 = (Bit Length
:
)11000 = (MAXc

I1 = 3000

C1 = 4000

n1

I2 = 280

C2 = 400

n2

I3 = 50

C3 = 100

n3

...
)(Civil Engineering Applications

...

...

Value = n1 i1 + n2 i2 + n3 i3

Cost = n1 c1 + n2 c2 + n3 c3 MAX c

n1, n2, n3
value .MAXc=11000
...
Value

1 + ( MAX c Cost ) 2

= Fitness

) (Fitness
) (Value ) (Cost )(MAXc
.

...
)(Civil Engineering Applications

...

4.
1111 .15

.
.

..

.101100101001 ...
...
n1=1011=11
n2=0010=2
n3=1001=9

...
)(Civil Engineering Applications

50 .
)
(... 30 ) 30(.

0

0.2

0.2

8060

n1=2

n2=7

n3=2

8060

11000

...
)(Civil Engineering Applications

...2





.

:

1.7+3.4*r : ) (r

.

...
)(Civil Engineering Applications

...

...
) (xn,yn .
) (xfs,yfs .
) (wn
.
...
)(Population
= .20
) Mating
.10 = (Chromosomes
)0.2 = (Mutation Rate
).25 = (Generations

100

) cost = wn ( xn x fs ) + ( yn y fs
2

n =1

...
)(Civil Engineering Applications

...3
) (

:

...
(References)

Matthew Wall - Overview of Genetic Algorithms


Mechanical Engineering Department Massachusetts
Institute of Technology (MIT).
Biles, J.A. 1994. GenJam: A GA for generating jazz
solos. Proc. Int. Computer Music Conf, San Francisco.
Angeline, P.J. 1996. Evolving Fractal Movies. Proc. 1st
An. Conf. on Genetic Programming, MIT Press,
Cambridge.

You might also like