Basic Applications

1
Basic Applications
2
E D C D E E E
hold D D D hold E G G hold
3
Marry Had a Little Lamb

• use GAs to sing!?!?
• e.g. transform notes into binary
• add characters for half nodes, etc.
• two approaches for cost function
• computer
• human

4
Objective Cost Function

• cost ?n (answern-guessn)
• population 316 48
• keep 24 (recombine 1/2 the parents)
• mutation rate 0.05

5
(No Transcript)
6
Subjective Cost Function

• listen to only the first 7 notes
• user gives a cost 0-100 (0 - exact match)
• GA uses elitism, but cost is not monotonic

7
(No Transcript)
8
Creating Music using GAs

• create random music
• use humans to evaluate it

9
Genetic Art

• use some mathematical function to create initial
  population
• define mathematical function for crosover
• use human to select best results

10
(No Transcript)
11
Initial Fractals
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
Word Guess

• GA needs to guess a word
• a1, b2, ...
• ? (guessn-answern)2

16
(No Transcript)
17
(No Transcript)
18
Word guess

• different cost function
• correct letter 0
• incorrect letter 1
• cost number of incorrect letters
• with new cost function Colorado in 17 itterations

19
(No Transcript)
20
GA Parameters

• Colorado
• population size 32
• keep 16
• mutation rate 0.04

21
Locating an Emergency Response Unit
22
ERU

• response time 1.73.4r, where r is distance
• cost ? fi di, where fi is frequency and di is
  distance

23
(No Transcript)
24
(No Transcript)
25
GA parameters

• binary and continuous
• pop 10
• mutation rate 0.2
• figure shows average result for 20 runs
• binary 1 iteration slower

26
(No Transcript)
27
The evolution of horses

• want to use GAs to predicate how horses will
  evolve
• idea each horse has a value ? adapti wi
• natural selection nature tries to optimize this
  value
• start with 20 random horses
• predicate how they will involve based on
  importance of different characteristics

28
The Evolution of Horses

• each horse has a bunch of characteristics
• breed
• color
• hooves
• Length of Mane and Tale
• Fight/flee
• Socks
• Face
• Eyes
• Water requirements
• Running

29
Horses

• we are also interested in the environment of the
  horse
• Desert
• Plains
• Dry Mountains
• Northern Tundra
• Pine Forest
• Outback

30
Parameters

• Horses adjust to environment, i.e. there is an
  adaptation factor
• Different traits have different importance
  weights in different environments
• Two types of environments natural and breeding

31
Horses

• e.g. desert (.5 .6 .1 .1 .3 .4 .1 .1 .8 .6 .9 .5)
• cost function for each horse
• ? adapti wi
• encode traits in binary
• mutation rate 9
• pairing by rank
• run for 50 generations
• tables list characteristics of horse with the
  lower cost

32
(No Transcript)
33
Horse Results

• Less important characteristics varied widely
  between horses
• Was the experiment realistic
• no, e.g. much more blue eyed horses than in real
  life
• human preferences don't much with survivor
  characteristics

34
Horses - Experiment 2

• Play with color genes
• two white genes
• if both positive - horse dies
• gray gene
• if one of the two gray genes is positive, then
  the horse is gray if no white gene
• if at least one of the black gene is positive,
  then the horse can be black
• if both are negative, then the horse will be red
• use those complicated rules to code the cost
  functions

35
Experiment 2 - cont'd

• Add probability of gene appearing
• cost function ? adapt_i p_i

36
(No Transcript)