hewmen brain hy

hey heqe ecmbeu os heeon

wfth hemen0y

sapse
Axon

call body (so wa)

Ldendriteg
Some- whee cell haweles wteel
Senclaitey ohee
kolii ch
herve is connecel to
Cansiey impy o
cell body
heeuor
Axon

proceiy
elen
Equbat biolosic e heen Arifeil hewyn
Cele
dancrie)
Oern
Soma hot ip
Axon
 Netoosl
Apptieation of Neual
) Air traffe Conyol Uuld be aatadl wotth lecationpl4ihud
irection specd Cocb reolor bup takbe ay ipat to
hetwok- Tho outp tV l0oud be
ingtrction in reypos to eah plip
Ahimat behabior ypredatorlpreynearolationghip
2) be cyele my
utckle for cnysis by hatwodk
3.) Giminal Seuleneiny Codd he preticacl
time cutaul Oay npat Hegultig
9) Direct mafi adwisor t d uye haLel hatocvk. e their
databay to d d e wich Cytamer to he teuqeted.
S) Employee hiring teuld be ostimi zad, if neuad net work
wene able topredict wich job applicunt
which applicant Loould skoo
best fob penFormane.
6) Betting on horse races , Stock
stook motle et
Cbulol be bced oh predl etion
F) Praud detection regaucly cneat Cuas fngne or t
Cold be

)
wng
Echo paterny. Arom SoNAR, RAbAR 4 hasyati'c incnht
Cauld be uyed to predid teir terely
) Muie Composition hoy ben trled iy hewd hatuu.The
hot wor is trai'nad to reogue pa ean ln piteh Ateupo
f cetein husic L Hhen
River voaBer leves Could
be predu ad hetyed on upsry
4 time md lscation o each report'
heporls
 Melulloch pits
Eauliest hera
Mp hauwn
necon mat
Binoy
oeigcts usocatta with

Y= wt Myws
P.

Actieeton Auctur hee

esxhbitow bd ho inhbity

trarnig a stm
ANDNNT ying Mp
ing Mp Neuun

1
 omuz d-
< ( 9-

(o'o)
("o)
(o1)
(')
hiun my)Qy (or

|-7m
(o'o)
(o'1)

my J-

|
 This is Sinuplest Fom d NN a
Feeforsad Netooxk , clred on
irfumatien only flo uy
multiple hidaleu hodey
neues haclecot ds
Oke aection

Backptogation
Architecture
Input laye hiadan layee
Vu

Em
Jm
Vom
Oop

tyaining vecto r ; , --- *a)

Pascmaters 2- i/p ve cto r (E, t) sts tn)
t - tatget
eror at o/p t
ndden wit Zj
Hate
cit
bËay on hidden
Voj J

Wdden
 procedwue - 1) iitiali 2e webt
feed fonwa
Bocle Propogatiun
) Uptatiun oeigta by

peogation hatuok is odel ot mus ti laye

Lck htileg Peedfowwad kheoun,
descent qhadiet hayd x tewnny ule
bekpabpsq ation sule
inlnaize tota Squ
supevised leeun?ry
fotluws

Trainng algo-îibn mal Hondon ols.

hitialize weignt toCancila'n s flse alo step 3 to 10
Step) while
Sie eacht a y do step 4 tog
Shep(a) for traa
vecievey Hhe sitmal uit to4 all t t
Lunit vecievey
Input tLnit
sherO Each Inpat thls sigal to al
hensinit tais
his
Suny its
Shep (3) Eoch hdden Lnit Z; J=1 top
input signat
buction 2'= P(2int) ehd çedy signal
ativ ation to
to

Step( Each bctpd nit yr (k=| tom) Suny is weigited

Jiny = Wox
J=1
acta vlon Fwhun
 Error
Backprgogalion )
9, (k=lton recievey
Step ) Each out put
tuget pattesn
unit
chm inpat
patt on
Prformaion 8v=t-e) FC yink)
Croy
Sune is A input
unit 2: (Itoh)
Ste Each hleelen
from its In the laye cubore

Siny

erorr n elziy)
ABias

(k=l kom) updete its bis

9k
Ste9) Eech odpcd Cuit k oeignt
cton teim giver
CorrectÜn
Cmol weigt Gj= to p)
Dy

bicy Correction tem

wjul he lolol ) -+ Awjk
Theneyoe WbK( new) wolold) + AWoK

unit 2 ( ltop) updde is biay 4

each h aden Currve hun
loeigd (i= o to n) weigkt

blcy cosstealion tem AVo = d &

AVi
V Cod) +

4est sto pping cbncition

Con Te sto png Caciton a
he te
 rag'an

HE BB NETWORk
'uhen an Cuiun f cel A is heay Cnough to eencitecelE
Cmd repectedly Or metchol'e
pema dhankes firiy iy
g'hooth poceg Uy
Some both cells Such thut
in
ne cell o B0s ntraayd

’ ) 1f 2 neurn

dare
 lUeigkt updete

vetu
redor

Ftowchart
Stet

ittize welgts

for ozah No
S:t

Bjcs Upbte oneuW0r tA w

b(nej bla)*y
bhew boid t A

Stop
 hebb hatwoe mplensd
) Sesign a to

AND

1 1

1
1

(n)
W

initial w=0
b

Inpts b
toeiylt chgey
Ab
New
blwew)

1 1 1 1 1 1
1
-1 2
1 -1 1 -1
-1
1 1
-1 -1 1 1 -1 3
-1 -1 1 -1
-1 2 2 -2

w2
W2
b= -2

testng 2

=-2+ L(2) +()2)

 2) Deaign hebb hetwod weigte roquirey
to. peufo clysificiti u 4 gie inpd ptHen
Paten 0=-1)

+ + +

-1
Tauget
Paten
1 -J -1 1 -1 -1 1
1 1 1

1 -1 1 1 -1 1

Wy = W W= W
Wz
stepl

1 -1-1 1-J-1|1 1 11 -1 -]-1 1 1 1

1-1 1 1 t1|11 1 1 -1L1-1 1 1

Neuw welqhts
wy Ws wg
-1 -1 1-1 - 1 1 1 1

-2
-2
o

wt w
40

0 +o+ 2 r0 to t2 +0,a0
positabe
 poplatgn
Cindividealy
analogoy
to osone)
ot chzo to
ndirideuc Eeah
Cnponcut (odeda S
alogo clo ve fre sepMesent indiideed Each
Popdion
(om pob guen forSolutiuh indivicua
is,
Spuce
speace sean within mountanod
ch spaee’ Seaych
niiome Geulted hore
nt bele ate Heele pe
plo
tho
n i fhat
eeet Soati
e OHhey. Cteele to
put Aitest Prom Cenes " ndiideal Theose
ce who Indvdael
ete Hhen (Pteyt ) for umpte populetn Founlationn
mate. eyoee
4
sonechoo cunologo
chauatte to
steiny Tu bits /PbatI /itee ble poust
vidual
is Endi Eeeh jolection
in
point poplehon a eseht
d Hepy of cunict
genenaion,nd Each
n fsolution
or ahquatry qenencte phoble To
ielaa ’Baed
on
lutiory beleng
to
evolutior pott
hat algorttn Secoch heaistc cudqplive CMe
Algosithm Genentic
 Fness Score
to Cach Sndividual olih
A ney suore & givon to
os Om Yndlvideal to npete

nied inchividewd alon uith teiy Pitney ore

1he Pali ridea wth bHee hithey
who hete bet e tsey Populet n ize is
Prcue bete
cretec - Leut
So stattie So be
will de.
Oee tbsping peclecd haviy o sguteied il«.
rom preuoy population
the opulectin Condeyed
Opeotust aenatic Aaanithm lnctividualy wit
prefetene
give thes to allouo Hae to pey
:) detection poratas ’ to
ore A
sLLesive qeuatOL.
heir enes
betwen intiidinl,
2-) C40sso veH Opeato ’ His reprejeut matsaton oplutor
Too individals seloctod wty eniangeel
Geney at thee HUSSOve
Than
ereatig cmpletelg na individual (ofPs priy)

fnsestt randem geney in of

3:) Mutation Opeato to-To metntn in poplatier
to dnojd prematue
Veonynqpe
Before Mutaton PaH|B|D|E)A
After Muttlion paMBC |D|E]N|
hoo to represet a solcton to Cwit
u) En codng U
Frame wbork of QA

Steut

etiaize
Populton

Evalee Fitne
Conver 1 Seledion
(selea mte
Yes
S tup Reprodulion
Crossoven
Mctaton
hvegjon

+Hninaion Giteion
Conyergehe how to ceide hant coleLion
-
hous to evalaate solta un
to
fitney evcusn
loe the the so latiy
Cross orel ho to eup One optina to
muttiun move
Inyosibr To
 (Start)
Crecte tnitiaf
Populeton(N)
vidcs
Evaluete each
Îniidual
Select Np indi

halny poel
| Crestto
Klonvergis No

Ciss0vee
Reu ndividual
with be,t fitne
vcle
outdton
Reproludie

Stop Replace altihdiiduey

9nectn ith
in lasl Creee
he oPspriny
population sZa -0
Paametes Ihitid
tuveslur m,
Gonveug & hucteon
mputatnonally epeueive
feaues oveulappirg gneration
bette hhoe Citted indivjduass
is bícyel
towcud iteLathon
Selecn
mud «ppee
beyt fit
good when
 Stead Stale Genetit Algont
Stat)

|Cenote initil

Sdte each idiviele

tepeuhun
Sale et two nithout

ossov
Matation
inveser

Reyet te offspr
is deplicuud

No
|Evauie offspri
hephae e
Gsost heliviceu t n
with
worst

Yeg
|Retun solction

Stop
chonosome of Loyalengh
Tes eap yie
init prencde
unogima
 buu on
ucu on ohich taley
Rrness funition 8- lt Simply olefined cuy a
cevndidote Joliion to
Cy
fnpat eund prpclcey hespet to pro blem
t Solction is

stid euty lo cmpute.

fest to heayce
how itagienstis
chact eistrc q u t i vely
0- khaps acle
A s(100)
B- Lso)
C-15 (20:)
MS

AB = 116 15 Rso
20 30e
|25 6 s31 Ac- 701
Pathl -01j QS
3s

peth AB - 111
124 s G31
path3
1 25 6 341
puthy
 two peuLLt
Cross Over : ploey q9 faling a chitd
Soution Lprodrcçy fom
fron them pockae Sug
segeion
iceedl in ovde fo
Tlot shigs ae

thoee step randsim eir col ihdlie leeal

dauiion opeator Sel ecy tuo calony
alory'sylengt
repro remdom
1) us ste K Selectec at Swapged betwe
two
2) position elee are
3*) finaly qositin
Se

CHOSSOv inty Chd sestrong ayte Cul

set
pornty
Cet Covvevpovatigpoird
Crossovel
is
CAtheyed

cho Yene 10410

chhoni some 1 0 01! 111

1o 11o 211

10 10I 0 L0

pojd Croscove I1611 00100 1010

pnent 10101 11000 01111°

11000 110110
11011
childhen: 011110
10 101 001 00

Un'foom Geoss ore ando ml

Seleclocl
Gch bit
reundoud qensatid tor each
eonaing.
Co rrep hoy k
Cecs6Lpet
palr
panent L01 1 O01 1
10
Paneud2 0 001 10 110
Mae 1 104 o
ehild
cla 2
 Arithmete Recmhinalion
Mohole

6|4|2)4 |sl6 ihlsl91

puocle tlegal aftspoiy.

have high chauee to
Sonple. c0sovel
(HOSOVel
4 ome atiy
In TSp
Simple
bing
hease tha Oce
/legal
wil be thee for ths phoblom
nodified ryn
Unafom cHOSOV er Cou be peret
citiey jn Urde f
ohee hayk is 1, bpy Hemeui nin citiey
whee eyk is o, cheose
othe ptent'

Mtation ’ it atro clce neuo utic strucee in popltn

randomy
blocke
’ muttion haps escope zom DCal neiimus trap4
baeiuteny dirnsity pepulation
clso keaps qmenuo] Stock ed,thy engwning
eHODciity ^ socinch spues is Sald to be
engedic soti on

Fliprg oy
)flipping! hcuyed hits frvlyey changyiy oto g cmd s to o
n nlalion chuoone qnated
A panet is
randomdy gnenatad
for mtaion chones sone arrepo ndi Ped

Paxend 1011 010 1

Mutaion dhou 100 1
chel d 001 1 1 1 00
 lutechang Two randM posiiog
coydiponting to thoge
tenchwyl

1 1 11 00 01L

postuon is cheseh d bit Aent to

Ruesing hanalom
positiong
's pceue
eersec nd ahld chweot

ft l01 1o( 401

childl Jo1 10110
hetili con whole chonosloe
Poobahilky ko chengo
chauge)
Mutatiun Should hot ves syten
becaye theh A wl change tu random Securch

delection SleTion is a methcl that andonly picke

ChoOSone ot r populetion tory to tLeir

Tuo type selecton schame

Pbppilionate bayed c electen icleg od indivdul besed on
4ho fnes velue teletire to athedg,
ainal heused seladti on Solaet ndiricuel,ttnhot upon thein
bnes bct upon 4hoih nane poplation

individeay Popalation y
po

Tournarne nt Selec lion

0unanent Selec on providey seleite preco by haleiny
tndividduely
toonamont

fadvidu
’ The bet Touthaneit ron to whaet ingelea into hotin
Cuupetrt on repe alhd
pod
 oplimal
is is
&oletion

To cwinancut

Roulete wheel Seleztion

Ftnes |s Selacion itney

2

D
lo/

Finecd Poiter
Momun wea than possibi ty koqet selded
ftney Area a seleeton

Qheel is diidd ito n piy Each indiriduS aets

pasti on crcle whi ch is Pio potiU nal to Aitnes
Vcue
poi t s choen on
wheel Cirlm feQnce ay
AAed rotaled
Tho
Slown tn ront is cho(en ay pawtevt
Coymey
repeoteel
þoocey
cle thet Citey indi vidual hey gseot en p'e
hete a gsodo hoce in
bodey

is eay in populain
Slep Calclete Selme
heneele remdnm
Popelion kesp eclding
rom top af
Stautig
bitneyey to Sem , H <

Tla inclivdud for which p eneecu 3 is

chusen
Stochaslie Unveal Samplng
Sihileu to Roulete wheal howeven tugfeecct
SUs qute we hae hultiple Picedpoi
CLe chosen in ject One spin
this setp
be choscn at leayt Once
hdividualy to

fincd
Port

hegalire aleo Hus wathod cloit

whee Fitnes cow tulee

Rank Bcyed Selection

coith, negative tithes, ale 4 0s hastty cuee
Ayo woyk
the ndividual in population have veLy loe
ohen lmosl equal Shau or p°e
bitey ale
each hdiidud o matt e
hene
eah otte Sene prohablity
Tis leedy to a wy in sel eoticn prese to weud fter

Seh situat'on. nauny A poor pet selectiony ih

n Huis ind'idual is ranked ao rdiy to theln

biney o
eyeuy
The seloalion of paeny eoch
( Not bitnens)
Fins
JoJA
4 26 6J.

l3- 33 )
 Selastiuh random Set pet Fom
Randum populauo No presw toe

Boltzman Selectiuh baceel on etopq hd inepurte

Sanmpting hathsde in chte caulo
Sinleeon t hturelly leacy to
aclaplive itey in ohic tuzlien doey d
valey oith uoiomed
stag bined nt

In boltz man selezlion a Conlincus elcale . The

Bte a seleeti on LLLOBi
Staut at hih which Oswwedmeny' Çeleti on prayue
which haealng Seletin
is Ww A to hawsoo
Ihehca
les the bet
appoplae dsqee
Fitneg stf cusety Cuailehle bejl c
cêu
Let fmau - ha-t such, f(x;) zfman
Stig has fitney
thewe it is seletled
thed

hon hey
with boltzmon
Picbabity

Cmputdon fiistac
Few hest
Eihsm - first hagt ehnomUSune Or
Popletio'
lopied

Muttion Operator Bitfip mutation

Rauncu reseriy
Scrcmple mudaaer
Invenian uttelon
 3:) 2)ho
u.) Seach
Aprlic)whehationg
ein
onal
thmat'
Gnalyis
cuspac
dy is
Seuch cug
uenetc
ALyo(n
is Cnplen

sceee
e poorty
eup
et
knouol
edsye