0% found this document useful (0 votes)
40 views17 pagesUnit 1 ML
Uploaded by
Satyam PandeyCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
40 views17 pagesUnit 1 ML
Uploaded by
Satyam PandeyCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 17
| sianlseanlaialininieiay
i
easel
(Cee a
nit=4 —_
Macnine Lenening
Jeseringitiii) isthe U branchrulet
RPC doe aia
ESA couctt aNd either
*& 2 te =r the may homan learn. Is
Heaeakh ats aranama ie
“The perma pills is te allow lhe comp
to learn Auto matically witbest human
iniervestions
JSP es of MW ArqokitTHM
Previseo
T4 iwi mchelinalls Whembathe pmedeli is] gettin
Apeina) slepidallel sballed gidialaeedt Ciel
hal cles dala or predict
wvutcomes accuratel
x= Cass ifyng spam
Ti, jean.) be. Sebaracted=sinite ers Apes, ot
problem Pee eee reg reason
Un suPERNMISED g
It is us hen wikthelillinte ilemmenal a
class’ Pied | sor illlals elle LiaThisiiie ies Ries res
ML in which the al enrihmas sree mst
aa ned labels
vided with ar so rns
"GL the ‘veining oy
@ scanned with OKEN Scanner ‘ hidden patterns
jal Aajcovers Miles p Me
: deta! grouping without Harte) of
human d in RA becios
Tis ability te disever sick GF
differences bas made Han V ideal sein
fer data ene) is
3 | Sem- S vpeevised
Tis Ba, appheach vehi) ML «shat cont bing
a__sra\\ ameunt sf a i) abelved
data wih a lars ameun\ ©
unlabelled data.
Ex Speech mt pete
Internet content classification
Geegle in 2016, Vaunched a new 4
Serot= supervised S”\¥el_ called Geepie
expander
4)| Keineorcemeny — Lengning 4
TH ig the area of MUL itt js about
ae suitable actions. te maximize |
[ine particular)» the aoe of a
guitcome,
| App ucerions 4
Usedid ins Srebetits (ri nductrial svfemationll
Used in creating draini Sse
{ reapiveme(ht
ace te the
@ scanned with OKEN Scanner ADVANTAGES OF ML
Avtomation of everythin
Wide a ¢ q apt
Mi hay itd goal» en) i medical business,
eo Mech nd | \
halling et dlats
identries ! deendaly OA
sge ot cations
Effi bil
Easily d
Disnox enrages
ML requires massive data, sets: te train
{thee fahould be unbiased st goed quality
ML needs enovg Aime to Net the Ch celine
ja | ps Fatt) | thei
sireorilnen saieaty
develop __enev
purpase toth a considerable
ip error acceptability
1Z EARNING ii
Ti iis pls pybcess alert vndey standing Fnowlady
b eh aviour kills, values A VY atdyibe tes
The ability -Tamitisyniaae wisi processes 4
ho mane | animals lade Sroaghin ells
cetgied ain’
“ba, wleoenng Soh eed lees
ie a
@ scanned with OKEN Scanner ran}
x
Wer Derines Le PRNING P\c 9 RYT HMg)
PRobLems Ay
computer proaram is said te learn From
experiences in cabtext lo seme task (to
X performance measure (P)
Seme example that Gefine LP are .
Te filter emails Spam a
Task Classi fyi emails as spam, or not
Eiathe ipeeens of emails Sears clais ty
aS Spam or net
os ar yeor label as of email
meee meres
Hanclas siting iaten piel ler a
Task — oRisaliaies ait sah fen words Tw
= percerslage at V Words joccurartely ese
e= A directory ef handwritten | words a
with | (eet classification
Rebet Drivin Preblern |
+3 Dea en Hpsiic hs ha bs in syht Sscannets
po Aivelye dittance es Belweos 2
fall (wu
en oes f
lostr ction
limage B & steer
noted he phils homay
ae
@ scanned with OKEN Scanner __ARFAT5.
| From Preptenow — PRoscey
t Fer casting differest fruit for reesnition
po Able to prddict maximum varie of q Freie
£2 —lralnitg machine wlth ‘dats sel of
0
Fort ee
(wo Free Recognitién — PRogeem
to Fer casting different face of homan for
Qniteetae nities
ASR ee
of
aining machine with largest data sel
differest Vo Faces of human
Bw | Automation TTRanstation of Documents
a A> Te jvandate hemaiene lag oaye -lo ether
yay
= NV
j po Able fo convert one language to anether
Ts lita a
q ¢
anners i e> Dab sets of differed ype of |
wid A cnecver LERkning Proavems
Playing cheeker's game
7 tt rae ae
ad of ining —oaiat.
q | ee Pla implereration ie
@ scanned with OKEN Scanner Met
Hig ORY of
Turtry «Test
Gel of checkers :
> Per ception
> Neighbour ‘ala orcthro T-
> _ Explana Viten based. Jearnin ai
> (Net talk roprarm) Cerny t y
Clearn “te Pronoorce coord
Data driven approach
Deep Jearnin
Micresot +
tow Pees
= + human features
2e3 Goagle bral develapsd ae
7 eo
DESIGNING “THE Leprning © YSTEM
tte used “fo design learning eastern lare —
j Ea cthe jeeuel cask i t - y
Choose
tthe suitable! cet of tratnina data
Divide he [raining data inte Sreops or
lectin a
Determine the Knowledge represedt-tion
chasses then
Croore 2 learning " classifies that can a
fe hy pethdhis fiem the Be Baas
Ppp the Vearner clasifies to teat data
Conca Seaham etn, of atte ete
The fl |
| rpet system
@ scanned with OKEN Scanner & Scanned with OKEN Scanner task for 4
uy] Marka ea P
iy Statist
