introduction to evolutionary computing

Upload: whiteblueching

Post on 03-Jun-2018

224 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 Introduction to Evolutionary Computing

    1/29

    C H A P T E R 1

    INTRODUCTION

  • 8/12/2019 Introduction to Evolutionary Computing

    2/29

    LECTURER

    Dr Zeratul Izzah Mohd Yusoh

    Industrial Computing

    Leel !" #I $ing

    %& ''1 &(%%

  • 8/12/2019 Introduction to Evolutionary Computing

    3/29

    Positioning o) EC and the *asi+ EC metaphor

    Histori+al perspe+tie

    ,iologi+al inspiration-

    Dar$inian eolution theor. /simpli)ied0

    2eneti+s /simpli)ied0

    Motiation )or EC

    3hat +an EC do- e4amples o) appli+ation areas

    CONTENTS

  • 8/12/2019 Introduction to Evolutionary Computing

    4/29

    Borg Vogons

    Biotop

    Art

    Life Sciences Social Sciences

    Mathematics Physics

    Software Engineering

    Neural Nets Evolutionary Computing Fuy Systems

    Computational !ntelligence etc

    Computer Science etc

    E"act Sciences etc

    Science Politics Sports etc

    Society Stones # Seas etc

    Earth etc

    $niverse

    %ou are here

    POSITIONING OF EC

  • 8/12/2019 Introduction to Evolutionary Computing

    5/29

    POSITIONING OF EC

    EC is part o) +omputer s+ien+e

    EC is not part o) li)e s+ien+es5*iolog.

    ,iolog. deliered inspiration and terminolog.

    EC +an *e applied in *iologi+al resear+h

  • 8/12/2019 Introduction to Evolutionary Computing

    6/29

    POSITIONING OF EA

  • 8/12/2019 Introduction to Evolutionary Computing

    7/29

    SUBCATEGORIES OF EA

  • 8/12/2019 Introduction to Evolutionary Computing

    8/29

    THE MAIN EVOLUTIONARY COMPUTINGMETAPHOR

    EVOLUTION

    Enironment

    Indiidual

    6itness

    PROBLEM SOLVING

    Pro*lem

    Candidate 7olution

    8ualit.

    Quality

    chance for seeding new solutions

    Fitness chances for survival and reproduction

  • 8/12/2019 Introduction to Evolutionary Computing

    9/29

    BRIEF HISTORY 1: THE ANCESTORS

    9 1:;

  • 8/12/2019 Introduction to Evolutionary Computing

    10/29

    BRIEF HISTORY 2: THE RISE OF EC

    1985: first international conference (I!"#

    199$: first international conference in %urope (&&'#

    199): first scientific % *ournal (+I, &ress#

    199-: launch of %uropean % .esearch etwor/ %voet

  • 8/12/2019 Introduction to Evolutionary Computing

    11/29

    DARWINIAN EVOLUTION 1:SURVIVAL OF THE FITTEST

    All enironments hae )inite resour+es/i@e@" +an onl. support a limited num*er o) indiiduals

    Li)e)orms hae *asi+ instin+t5 li)e+.+les geared to$ards

    reprodu+tion There)ore some Bind o) sele+tion is ineita*le

    Those indiiduals that +ompete )or the resour+es most

    e))e+tiel. hae in+reased +han+e o) reprodu+tion ote- )itness in natural eolution is a deried" se+ondar.

    measure" i@e@" $e /humans assign a high )itness to

    indiiduals $ith man. o))spring

  • 8/12/2019 Introduction to Evolutionary Computing

    12/29

    DARWINIAN EVOLUTION:SUMMARY

    Population +onsists o) dierse set o) indiiduals

    Com*inations o) traits that are *etter adapted tendto in+rease representation in population

    Indiiduals are =units o) sele+tion>

    ariations o++ur through random +hanges .ielding+onstant sour+e o) diersit." +oupled $ith sele+tion

    means that-

    Population is the =unit o) eolution> ote the a*sen+e o) =guiding )or+e>

  • 8/12/2019 Introduction to Evolutionary Computing

    13/29

    ADAPTIVE LANDSCAPE METAPHOR(WRIGHT, 1932)

    9 Can enisage population $ith n traits as e4isting in

    a n+1dimensional spa+e /lands+ape $ith height+orresponding to )itness

    9 Ea+h di))erent indiidual /phenot.pe represents asingle point on the lands+ape

    9 Population is there)ore a =+loud> o) points" moingon the lands+ape oer time as it eoles adaptation

  • 8/12/2019 Introduction to Evolutionary Computing

    14/29

    EXAMPLE WITH TWO TRAITS

  • 8/12/2019 Introduction to Evolutionary Computing

    15/29

    NATURAL GENETICS

    The in)ormation reFuired to *uild a liing organism is

    +oded in the DA o) that organism 2enot.pe /DA inside determines phenot.pe

    2enes phenot.pi+ traits is a +omple4 mapping

    ne gene ma. a))e+t man. traits /pleiotrop. Man. genes ma. a))e+t one trait /pol.gen.

    7mall +hanges in the genot.pe lead to small+hanges in the organism /e@g@" height" hair +olour

  • 8/12/2019 Introduction to Evolutionary Computing

    16/29

    GENES AND THE GENOME

    2enes are en+oded in strands o) DA +alled

    +hromosomes

    In most +ells" there are t$o +opies o) ea+h

    +hromosome /diploid.

    The +omplete geneti+ material in an indiidualGsgenot.pe is +alled the 2enome

    3ithin a spe+ies" most o) the geneti+ material is the

    same

  • 8/12/2019 Introduction to Evolutionary Computing

    17/29

    EXAMPLE: HOMO SAPIENS

    Human DA is organised into +hromosomes

    Human *od. +ells +ontains !' pairs o)+hromosomes $hi+h together de)ine the ph.si+alattri*utes o) the indiidual-

  • 8/12/2019 Introduction to Evolutionary Computing

    18/29

    REPRODUCTIVE CELLS

    2ametes /sperm and egg +ells +ontain !'

    indiidual +hromosomes rather than !' pairs Cells $ith onl. one +op. o) ea+h +hromosome are

    +alled Haploid

    2ametes are )ormed *. a spe+ial )orm o) +ellsplitting +alled meiosis

    During meiosis the pairs o) +hromosome undergo anoperation +alled crossing-over

  • 8/12/2019 Introduction to Evolutionary Computing

    19/29

    CROSSINGOVER DURING MEIOSIS

    Chromosome pairs align and dupli+ate Inner pairs linB at a centromere and s$ap partso) themseles

    0utcoe is one copy of aternal2paternalchroosoe plus two entirely new co3inations

    "fter crossing4over one of each pair goes into each

    gaete

  • 8/12/2019 Introduction to Evolutionary Computing

    20/29

    FERTILISATION

    'per cell fro Father %gg cell fro +other

    ew person cell (ygote#

  • 8/12/2019 Introduction to Evolutionary Computing

    21/29

    AFTER FERTILISATION

    e$ z.gote rapidl. diides et+ +reating man. +ells

    all $ith the same geneti+ +ontents Although all +ells +ontain the same genes"

    depending on" )or e4ample $here the. are in the

    organism" the. $ill *ehae di))erentl. This pro+ess o) di))erential *ehaiour during

    deelopment is +alled ontogenesis

    All o) this uses" and is +ontrolled *." the same

    me+hanism )or de+oding the genes in DA

  • 8/12/2019 Introduction to Evolutionary Computing

    22/29

    MUTATION

    ++asionall. some o) the geneti+ material +hanges

    er. slightl. during this pro+ess /repli+ation error This means that the +hild might hae geneti+

    material in)ormation not inherited )rom either parent

    This +an *e +atastrophi+- o))spring in not ia*le /most liBel.

    neutral- ne$ )eature not in)luen+es )itness

    adantageous- strong ne$ )eature o++urs

    Redundan+. in the geneti+ +ode )orms a good$a. o) error +he+Bing

  • 8/12/2019 Introduction to Evolutionary Computing

    23/29

    MOTIVATIONS FOR EC: 1

    ature has al$a.s sered as a sour+e o)

    inspiration )or engineers and s+ientists

    The *est pro*lem soler Bno$n in nature is-

    the /human *rain that +reated =the $heel" e$ YorB"

    $ars and so on> /a)ter Douglas AdamsG Hit+hHiBers2uide

    the eolution me+hanism that +reated the human *rain/a)ter Dar$inGs rigin o) 7pe+ies

    Ans$er 1 neuro+omputing Ans$er ! eolutionar. +omputing

  • 8/12/2019 Introduction to Evolutionary Computing

    24/29

    PROBLEM TYPE 1 : OPTIMISATION

    3e hae a model o) our s.stem and seeB inputsthat gie us a spe+i)ied goal

    e6g6

    7 tie ta3les for university call center or hospital

    7 design specifications etc etc

  • 8/12/2019 Introduction to Evolutionary Computing

    25/29

    0ptiisation eaple 1: niversity tieta3ling

    %norously 3ig search space

    ,ieta3les ust 3e good

    ;!ood< is defined 3y a nu3er

    of copeting criteria

    ,ieta3les ust 3e feasi3le

    =ast a*ority of search space

    is infeasi3le

  • 8/12/2019 Introduction to Evolutionary Computing

    26/29

    PROBLEM TYPES 2: MODELLING

    3e hae +orresponding sets o) inputs outputsand seeB model that deliers +orre+t output )oreer. Bno$n input

    %volutionary achine learning

  • 8/12/2019 Introduction to Evolutionary Computing

    27/29

    +odelling eaple: loan applicant crediti3ility

    >ritish 3an/ evolvedcredita3ility odel to predict

    loan paying 3ehavior of new

    applicants

    %volving: prediction odels

    Fitness: odel accuracy onhistorical data

  • 8/12/2019 Introduction to Evolutionary Computing

    28/29

    PROBLEM TYPE 3: SIMULATION

    3e hae a gien model and $ish to Bno$ theoutputs that arise under di))erent input +onditions

    0ften used to answer ;what4if< ?uestions in evolving

    dynaic environents

    e6g6 %volutionary econoics "rtificial @ife

  • 8/12/2019 Introduction to Evolutionary Computing

    29/29

    SIMULATION EXAMPLE:EVOLVING ARTIFICIAL SOCIETIES

    7imulating trade" e+onomi++ompetition" et+@ to +ali*ratemodels

    se models to optimisestrategies and poli+ies

    Eolutionar. e+onom.

    7urial o) the )ittest isuniersal /*ig5small )ish