Scribd
Upload
16 views13 pages

On Machine Learning

Machine learning enables computers to learn from experience, improving their performance on specific tasks. It has various applications, including spam detection, facial recognition, and speech recognition. The machine learning process involves steps such as preprocessing, feature extraction/selection, and learning.

Uploaded by

Thimothy Basava
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
16 views13 pages

On Machine Learning

Machine learning enables computers to learn from experience, improving their performance on specific tasks. It has various applications, including spam detection, facial recognition, and speech recognition. The machine learning process involves steps such as preprocessing, feature extraction/selection, and learning.

Uploaded by

Thimothy Basava
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 13

Methods, a p plica tions, etc.

Ove rview

- oW h a t is mac hine lea rning?


oA p p lications of mac hine
lea rning o Ty pes of
a lg o r ithms
o Basics of statistical pattern
recognition
c Preprocessing
c Feature Extraction/ Selection
c Learning
o Fina l remarks
W ha t is Ma c hine
- Le a rning ?
o " Fie l d of study tha t g ive s computers the a b ility
to le a rn w ithout b e ing ex p licitly p r o g r ammed"
D Arthur Samuel (1959)
o "A comp uter p r o g r am is said to lea rn f rom
ex p e r ience E w ith respect to some class of tasks T
and performance measure P, if its p e r formance
a t tasks in T, as measured by P, improves w ith
ex p e r ience E"
Ex a mp le: Spam Email

- Detection
o "A comp uter p r o g r a m is said to lea r n f rom
ex p e r ience E w ith respect to some class of tasks
T a n d p e r fo r m a nce measur e P, if its p e r f ormance
a t tasks in T, as measured by P, improves w ith
ex p e r ience E"

o In our p r o ject ,
c T: classif y emails as spam or not spam
c E: watch the user l a bel ema ils as spam or not
spam
c P: percentage of correctly -identif ie d emails
A p p lica tions of M ac hine Learning

- o Facia l recognition
- o Speec h
recognition
A pplications of M achine Learning
- o Othe r a p p lications:
c Spam detection
c Intern et search
c Da t a m inin g m e dical
recor ds
c M any more!
Supe rvised
Lea rning

- oG iven la beled t ra ining e x a m p les


oFind corr ect p r e d iction f or an unla beled ex ample
o E.g . Handw r iting recognition

·- ..
""'•• f inger Dar>•

N'f .. .:;J;'" ""'


"
""'
..
• • •
t
I "''
·-
i;z Lt<
Types of Supe rvised Learning
Pro b lems

- Classif ica tion Regression

o Predict whether a o G iven the input,


g iven inp ut belongs to p r e d ict a continuous
one of multi ple output
classes o O ut put is
o O ut put is d iscrete continuous '·.
15

•
. .
. .. .... .....
.. •
'
. . :· . •
10 ... , · •
I •.··:0-"t:llis. ··..:. • •: • ••
• .... . ..
...,.,.
• • ·!" • •' •• •. •

_
• .. . •
• .. • •
.· . ·.. ... ..
. ... .
,
-
• • • •

'
.
.. •

· -l -l 0 ·2 ·10 10 20 30 40 so 60
• •
. ' .
.
-}
Unsup er vised Lea rning

- oG iven unla b e l e d t r a ining samples


oDiscover low d imensional patterns, structur e
o E.g. C luster new s a rticles b y t o p ic

Sports " 1!19 •

Lakera even ser ies with Rockets after chippy Game


2 v eto!)'
USA T y • 2 ho ur s

""°
By O .nny
ANGELES
1 • The e
. !Wulars By DiMd Leon Mooro, USA TODAY
. . . . , . " * ' Y 111.ngt ll'<l l.ol ArQeles Lakers
LOS
oo\Jd doChey
tnoughl In f Q IO the• l uLim A:•gt.Ga.
p .'-tong • ... I - IO l!te & ! t.
B!ltl.M . .11"<1\!ii Hol.dlcn
e•
l<obe, . l . . . , Jhc>N !04gh t o ld' ·0 .'lme 2 w.r Ho.i....., CJ
ro. . c l • QlD!! P p S • • s . . . t ; Poi! 1.-.;t!I Q I ON K • $!
• Y 1. f ' l f ! !! ! I ••rhcl<os
cgm
•I
Reinforcement Lea rning

- o Lea rn f rom d e l a y e d f eedback


o E.g . M ac hine learns how to p l a y
chess
1b 2b

r-- - ---- - - -- - --- Predtction

1
-
2a
1 4
a
I
I[ 3,
J
Production
o. . .

I
Summa ry

o M achine le a rning: computers lea r n f rom


ex p e r ience to improve performance o f c e r t a in
tasks
oM any usef ul a p p lications that we use t o d a y
o Dif f er ent steps to machine lea r ning
process
c Preprocessing
c Feature Extraction/ Selection
c Learning

You might also like