Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Introduction to Machine Learning
1. Introduc)on
to
Machine
Learning
NHM
Tanveer
Hossain
Khan
(Hasan)
2. About
Me
• I
“work
for
fun”
and
mostly
work
with
Ruby.
• Love
programming
and
learning.
• Skilled
on
Ruby,
Java,
PHP,
Nodejs
and
Go.
• Love
to
take
challenge
• I
am
working
with
Tweek.tv
(one
of
the
Berlin
startups)
3. What’s
in
?
• What
is
Machine
learning
?
• GeQng
rid
of
fear
• Where
to
use
it
?
• Who
is
using
?
• Discussion
on
few
Machine
learning
algorithms.
• Few
books
and
references.
• Q/A
5. Defini)on
?
“Field of study that gives computers the
ability to learn without being explicitly
programmed”
By
Arthur
Samuel
(Collected
from
wiki)
6. What
is
Machine
Learning?
1. Train
machine
with
examples
2. Algorithm
stores
the
trained
data
into
a
internal
mathema)cal
model.
3. Predict
new
data
based
on
the
trained
model.
8. Where
to
use
it?
• Automa)cally
categoriza)on
• Preparing
recommenda)on
• Analyzing
sen)ment
and
behaviors
• Recognizing
pa]erns
• Grouping
unrecognized
pa]erns
• OCR,
Voice
recogni)on,
Image
recogni)on
• Discovering
likelihood
and
many
more.
9. Who
is
using
?
• Facebook
(Image
tagging,
Newsfeed)
• Gmail
(Spam
detec)on,
Important
email
detec)on)
• YouTube
(Video
recommenda)on,
What
to
watch)
• Google
search
(Preparing
search
result)
• Amazon
(Sugges)ng
similar
product)
• Many
more…
11. ML
in
Ac)on
• Supervised
learning
– Classifica)on
– Regression
• Unsupervised
learning
– Clustering
• Recommenda)on
– Content
based
– Collabora)ve
filtering
12. Supervised
Learning
• Machine
doesn’t
own
any
cogni)ve
system
like
human
does
hence
they
need
human
intervened
feature
extrac)on!
• Classifica)on
&
Regression
– Naïve
Bayes
– Decision
Tree
• ID3
Algorithm
– k-‐NN
(k
nearest
neighbors)
– SVM
(Support
Vector
Machine)
– Many
more…
13. Naïve
Bayes
• Mul)
class
classifica)on
• Base
on
bayes
theorem
• Text
categoriza)on
• Works
with
small
training
data
14. Support
Vector
Machine
(SVM)
• Binary
classifica)on
• None
probabilis)c
binary
linear
classifica)on
• Represents
examples
as
points
in
space
• Linear
classifier
• Text
categoriza)on
• Uses
loss
func)on
15. ID3
• Decision
tree
• Predic)ve
model
• Itera)ve
• Uses
in
Informa)on
Retrieval
(IR)
technologies
17. k-‐means
• Signal
processing
• Data
mining
• Itera)ve
• Feature
learning
• Cluster
analysis
• Color
quan)za)on
(Reduce
number
of
dis)nct
colors
from
an
image)
18. Recommenda)ons
• Content
based
– Natural
language
processing
– Named
En)ty
Recogni)on
– Disambigua)on
(VW
Golf
or
Sports
Golf)
• Collabora)ve
Filtering
– Using
SVM,
Naïve
bayes
– Implicit
or
explicit
feedback
– Distance
calcula)on
&
k-‐nn
based
filtering
– User
or
item
based
19. Few
pointers
• h]p://guidetodatamining.com/
– Very
easy
learning
and
programmer
focused
• Introduc)on
to
Machine
Learning
–
Ethem
Alpaydin
(The
MIT
Press)
• Mahout
in
Ac)on
• Mlbase
documenta)on
21. You
can
use
in
produc)on
(without
coding)
• h]p://predic)on.io/
-‐
For
Collabora)ve
filtering
based
recommenda)on
engine.
• Google
Predic)on
API
-‐
h]ps://developers.google.com/predic)on/
• Algorithm.io
-‐
h]p://www.algorithms.io/
(Not
sure
about
it)