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