how do genetic algorithms relate to their biological ...biological origins? l separation of genetic...
Post on 28-Jun-2020
6 Views
Preview:
TRANSCRIPT
MIT Design Inquiry
How do genetic algorithms relate to theirbiological origins?
How do they relate to human processes ofdesign?
Why is there power in this metaphor?
How can they be used to extend thecapabilities of the designer?
John GeroProfessor of Design Science
University of SydneyVisiting Professor of Design and Computation
MIT
MIT Design Inquiry
How do genetic algorithms relate to theirbiological origins?
l Separation of genetic material (genotype[representation]) from organism (phenotype[design])
l Expressing genotype as organisml Organism carries genotype and reproduces
genotype using ‘genetic’ processes of crossoverand mutation
l Darwin’s natural selection uses fitnesses oforganisms in their environment to improve the genepool
l GA is a simple model of this process
MIT Design Inquiry
Genetic processesGenetic processes
Crossover Points
A
C
B
D
Parents
Offspring
MIT Design Inquiry
A A B
B B C
A B C
B A B
Parents
Crossover point
Offspring
MIT Design Inquiry
A A B
B B C
A A C
B B B
Parents Offspring
Crossover point
MIT Design Inquiry
How do they relate to human processes ofdesign?
l Can map genetic representation onto acomputational representation of a design; can mapphenotype onto a interpretable view of a design
l Humans work on single or few designs at a time/genetics works on a population of ‘designs’ in parallel
l Humans can be seen to “search” design spaces - thisis one interpretation of what GAs are doing.
MIT Design Inquiry
Why is there power in this metaphor?
l Guaranteed improvement - Darwinianevolution
l Large scale searchl Blind searchl Fitness can be human evaluationl Fitness can change over evolutionary timel Can produce complexityl Can produce unexpected results
MIT Design Inquiry
How can they be used to extend the capabilitiesof the designer?
l Creativity through genetic engineering
l Novel designs through extending geneticcrossover
l Novel designs through different fitnesses
MIT Design Inquiry
Genetic Engineering and CreativeDesign
Genetic Engineering and CreativeDesign
l Background
l genes, genotype, phenotype, fitness
l Connecting genes to performance in fitness
l Emergent gene clusters ˛ evolved genes
MIT Design Inquiry
••
xx
xx
xx
••
“bad”
“good”
“good” genotypes “bad” genotypes
Total Population
MIT Design Inquiry
a a brule 1 a ab
rule 2 aa
brule 3
a abrule 4 a rule 5 a rule 6
a ba
rule 7 ab
arule 8
ba
b
a
MIT Design Inquiry
design 1 design 2 design 3
design 4 design 5 design 6
design 7 design 8 design 9 design 10
{1,12,2,8,5,4,4,2,8,5,7}good
{1,2,1,8,2,8,5,5,6,6,8,1}good
{3,2,2,6,5,8,2,1,4,4,3,1}bad
{6,4,1,2,8,5,4,2,8,5,3,3}good
{3,4,8,2,8,1,6,5,7,3}bad
{2,3,2,3,4,3,5,6,5,1,6,2}neutral
{3,1,8,5,5,6,4,6,1,1,3,3}good
{1,6,4,2,7,3,4,8,6,1,6,2}bad
{6,4,1,2,3,4,5,2,1,7,4}neutral
{2,3,7,5,1,2,8,3,1,6,2,1}bad
Composite building block A{2,8,5}
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
Mondrian
MIT Design Inquiry
Genotype Form
l In form of a tree
l Each node has four variables
l direction of rectangular split (4 values)
l fraction of the split (15 values)
l colour of split area (10 values)
l line width (3 values)
MIT Design Inquiry
Fitnesses for Representation
l offset between actual and required positions ofdissection lines
l number of lines with correct line width,normalised
l number of correct colour panels, normalised
l number of lines assigned, normalised
l number of unassigned lines, normalised
Genetically Engineered Mondrian
Genetically Engineered Mondrian
MIT Design Inquiry
Genetically Engineered Frank Lloyd Wright Windows
MIT Design Inquiry
Flondrians
l Mondrian painting ˛ genetically engineeredgenes: M-genes
l Frank Lloyd Wright windows ˛ geneticallyengineered genes in same representation: F-genes
l “Flondrians” are the genetic product of matingM-genes with F-genes
MIT Design Inquiry
MIT Design Inquiry
MIT Design Inquiry
How Many Designs Are There andWhere Are They?
MIT Design Inquiry
Genetic crossover as an interpolation
g1
g2
gcpc
p2
p1
Genotypic space GPhenotypic space P
C(g1,g2) Æ gcC(p1,p2) Æ pc
MIT Design Inquiry
P+
P
p1 p2
MIT Design Inquiry
Interpolation
MIT Design Inquiry
MIT Design Inquiry
(interactive Genetic Art III)
MIT Design Inquiry
Image detection
(a) (b)
MIT Design Inquiry
Modelling Interest
Berlyne’s model of arousal based on novelty using Wundt curve
HE
DO
NIC
VA
LU
E
NOVELTYNx
Hx
Reward
Punish
0
1
-1
n1 n2
MIT Design Inquiry
Different novelty functions
MIT Design Inquiry
Different novelty preferences
N=0 N=1 N=2 N=3
N=4 N=5 N=6 N=7
N=8 N=9 N=10 N=11
N=12 N=13 N=14 N=15
N=16 N=17 N=18 N=19
top related