Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
13 views

DL Assignment 5

Random forest notes sppu B.E entc notes

Uploaded by

Madhura Peshave
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
13 views

DL Assignment 5

Random forest notes sppu B.E entc notes

Uploaded by

Madhura Peshave
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

PAGE Na.

DATE

ASSIGNMENT-5
QExplain Ridivechiünal RNN Tn Nahural lanquagr
Pocutlng (NLP)
A Bidirechbnal Recurrent Aeural Nehooyk
(BiRNN)Qre a aniabon of ctandard RNN
that Imp roe seaunemadalling by, prcasin
hput data in both fonward and backunard
diechiona In tradition al RNNE nfoxmaton flaag
only ona direohbn fom_past to futuse
whtch Itmits the modell abilty to undertand
contert, apeciallyIn Nateal Language
Processin NL)Houwcver, in many task,
Auture contat j as mpoxt ant as the_patt
In BiRNNg, tu RNNe YUn gimultaneoul
One tom the start fo the cnd of the sequnu
and thL other fom the end o the start Thic
Lstructu enabl the model to capture both
past and tuhut dep endencies in the seauence,
akingit parhulary ued for take like
machin tranlationpeech coqnltionand
named entiby cOqn thon. BiRNNS are more
effective n handling task ohere cundere tandia
he entire staunca conkut casental such
lanquage modelhg and queston
aniloenn here sunounding oonda inftluenc
the mening
E parbalar ord
Exarspla?
Kackpropsaation Thuah Tirn (APIT) it an
alaorithan ued to tranina Recunsnt Neual
NehooxkeNNe)buunfoldina th hetuok
applymabatkpzopogatian.
Tna standard nural nctntk, backprapegata
it azingle forund
straigit since there
pas snda contApondin backusad pa
Houatverin eNN; thenehaotkes biddn gtatk
depcndDn prewioutA itrneepshtch
otroduCe temperaldepeodundea
BPTTthe RNN. unzald for
LStxtral ime steprand gradient art Lompitd
at each tine tep andacurnlatd to
update the cight..
hoz ekampla, ina siaúna task ltke
prcditing
8PT allos thenehwotk to leaco fom
depeo derndea betscen words acroL
målhplk time teps Achallenge aith RPTT
that it Can lead fo thevoniching
plading gradint proklnaher qradent
learn long
famdependsrica
O'a Explain in Dekail Any Toc Aduand Ea
1 LsTM (ong Shart -Tean Memany)

LCTM Lang ShortTesns Menany):


tSTM an aduanad ty pe of ÂNN
deaiqned to oVPCOme tha vanizhing
4radientDzohlum ntndihonalRNNe LCTHa
have pedal struchirts Called anaxi cel
ohichcah atorei nfomation for long
Lpeioc fachlSIM cell containsthree
-gaty the input qate haxct gate and
outputqateThesc aotee zealat th flau
of information decidtngahich nfamoatan
to keepdisearooc paleon to thie bext
tme itep Tht shruchu maku LaTMa
highly eflechve incaptuning long teron deperde.
nciea,making thenipopular ntadki dik
language modellingptech ecognitibn and
machin tranalation

2.RU(rattd Recurent Uoit)


GRU _anathr advancad RNN Vaiont
similar to LIM but coth asaplee
anchitcue GRUs comb+n th forget and
nput gata Tato a pibale updte Qate
and A l oalio
thiy hav a eet aate. This
tloplied struchurt tduc computatianal
complaiky ohile ttaininq th abity t
captue lang-tm depndundeGeU otte
pertorm cimilary to L TME but ar
PAGE M.

DATE

computationallyfarter,makingteno
ters suitable
for tauk loithV mitedcOUrce, cuch aA
real - ime speech RCOgnihon ahd tet
qenération
dhat bensrakile Adersdial Netsoks
CGAN)2
Generative Adversariau NehuDOcki GANs) an
clas of dapleaning modele canistng
of two neLral hetwnti aachenator and
discrioinat - that compete aqainatf cach
o ther in a game -theoretic mannir The
qntrator creaka anthetie idata that
rambesthe rtal data,wbile the
discriminator attarsp la to diinquish betwcen
2caland foke data The gencrator tap vea
hy teaoimg o produc data that can foo
the disriminator uhilethe disrimnin ator gth.
Lbettor at tdentiytag fake data Dver timu the
qcheratorhecome4 LBrDficient at qcntratinq
datathat Anearly indithnquchable frora
talidata
GAN havea cotde range of applicotba
fnceding qenerating lichiemages, ideo,
and -audioas csclas fn takà ikimage-to
translation and shyle tranfer Thuy
Lhave aluo been ueed n crcatie
frelde
Such as qenerating artuork, mste . ahd sD
modelsHoweuer training GANs s challenginq
and the
Ldue to nababi lity and th potenial fo
mode
Itmited
collapse, oher
anatiors
thegcncrato
ofdlata
poo duc

You might also like