Professional Documents
Culture Documents
1-Introduction To Machine Learning
1-Introduction To Machine Learning
to
Machine
Learning:
CS
436/580L
Introduc)on
2
Grading
• Five
homeworks
(30%)
– 6%
each
– Due
one-‐two
weeks
later
– Some
programming,
some
exercises
– Assigned
via
myCourses.
• One
Midterm
(20%),
One
Final
(25%)
– Exams
are
closed
book.
You
will
be
allowed
a
cheat
sheet,
a
double-‐sided
8.5
x
11
page.
• Quizzes
(5%)
3
Grading
• Project
(15%)
– Groups
of
three
• Class
Par)cipa)on
(5%)
– based
on
abendance
and
how
ac)vely
a
student
par)cipates
in
class
discussions.
• Abendance
and
par)cipa)on
is
mandatory
– Grade
reduced
by
a
leber
grade
for
lack
of
abendance
(e.g.,
A-‐
becomes
B-‐;
B-‐
becomes
a
C-‐;
etc)
– Miss
three
classes
without
prior
approval
will
receive
an
F
in
the
class
4
Project
• Class
project
info
–
Get
into
teams
of
size
3
as
soon
as
you
can
– Group
forming
deadline:
Feb
5th,
2017
– Would
prefer
size
3
groups,
but
I
might
allow
some
groups
of
size
2
if
there
is
a
strong
reason
– individual
projects
are
strongly
discouraged
– If
you
are
unable
to
form
a
group,
email
me/TAs
and
we
will
assign
you
to
a
group.
5
Project
Pre-‐proposal
• Pre-‐proposal
due
on
Feb
5th
– One
person
from
each
team
email
me
and
the
TAs
the
team
members
along
with
an
abstract
– Should
include
topic,
possible
datasets/algorithms
you
plan
to
explore
6
Source
Materials
• T.
Mitchell,
Machine
Learning,
McGraw-‐Hill
(required)
• C.
Bishop,
PaRern
Recogni)on
and
Machine
Learning,
Springer
(required)
• Kevin
Murphy,
Machine
Learning:
A
probabilis)c
perspec)ve
(recommended)
• Class
Notes/Slides
7
Class
Ac)vity
• Discuss
with
your
neighbor
• What
do
you
think
is
machine
learning?
Define
in
your
own
words.
• Think
of
one
task
you
think
that
is
already
using/
could
greatly
benefit
from
machine
learning.
E.g.,
self-‐driving
cars!
Be
crea)ve!!!
I
would
like
as
many
different
answers
as
possible!!!
8
Introducing
Machine
Learning
Play
video:
hbps://www.ted.com/talks/
jeremy_howard_the_wonderful_and_terrifying_i
mplica)ons_of_computers_that_can_learn#t-‐484
56
9
Glassdoor:
Data
Scien)st
Jobs
Best Jobs
Want a new job? Glassdoor is here to help, identifying the 25 Best Jobs in America for 2016. The jobs that make
this list have the highest overall Glassdoor Job Score, determined by combining three key factors – number of job
openings, salary and career opportunities rating. These jobs stand out across all three categories.
United States 2016
1 Data Scientist
Job Openings 1,736
Median Base Salary $116,840
Career Opportunity 4.1
Job Score 4.7
2 Tax Manager
Job Openings 1,574
Median Base Salary $108,000
10
http://www.bloomberg.com/news/articles/
2015-06-04/help-wanted-black-belts-in-data
11
What
is
Data
Science?
12
Most
Important
Technical
Skills
for
Data
Science
13
Most
Important
Technical
Skills
for
Data
Science
14
Why
Study
Machine
Learning:
A
Few
Quotes
• “A
breakthrough
in
machine
learning
would
be
worth
ten
Microsons”
(Bill
Gates,
Microson)
• “Machine
learning
is
the
next
Internet”
(Tony
Tether,
Former
Director,
DARPA)
• Machine
learning
is
the
hot
new
thing”
(John
Hennessy,
President,
Stanford)
• “Web
rankings
today
are
mostly
a
maber
of
machine
learning”
(Prabhakar
Raghavan,
Former
Dir.
Research,
Yahoo)
• “Machine
learning
is
going
to
result
in
a
real
revolu)on”
(Greg
Papadopoulos,
CTO,
Sun)
15
• Geqng
computers
to
program
themselves
• Wri)ng
sonware
is
the
bobleneck,
let
data
do
the
work
Tradi)onal
Programming
Requirements
Program
Output
Data
Human
Computer
Input
Requirements
and
data
change
o^en
Machine
Learning
Requirements
Output
Machine
Program
Data
Computer
Learning
Input
16
Training
Data
Training
Example
Two
Classes:
{Yes,No}
17
• Geqng
computers
to
program
themselves
• Wri)ng
sonware
is
the
bobleneck,
let
data
do
the
work
Tradi)onal
Programming
Data
Computer
Output
Program
Machine
Learning
Data
Computer
Program
Output
18
Magic?
No,
more
like
gardening
• Seeds
=
Algorithms
• Nutrients
=
Data
• Gardener
=
You
• Plants
=
Programs
19
Defini)on:
Machine
Learning!
• T.
Mitchell:
Well
posed
machine
learning
– Improving
performance
via
experience
– Formally,
A
computer
program
is
said
to
learn
from
experience
E
with
respect
to
some
class
of
tasks
T
and
performance
measure
P,
it
its
performance
at
tasks
in
T
as
measured
by
P,
improves
with
experience.
• H.
Simon
– Learning
denotes
changes
in
the
system
that
are
adap)ve
in
the
sense
that
they
enable
the
system
to
do
the
task
or
tasks
drawn
from
the
same
popula)on
more
efficiently
and
more
effec)vely
the
next
)me.
The
ability
to
perform
a
task
in
a
situa)on
which
has
never
been
encountered
before
20
Example
1:
A
Chess
learning
problem
• Task
T:
playing
chess
• Performance
measure
P:
percent
of
games
won
against
opponents
• Training
Experience
E:
playing
prac)ce
games
against
itself
21
Example
2:
Autonomous
Vehicle
Problem
• Task
T:
driving
on
a
public
highway/roads
using
vision
sensors
• Performance
Measure
P:
percentage
of
)me
the
vehicle
is
involved
in
an
accident
• Training
Experience
E:
a
sequence
of
images
and
steering
commands
recorded
while
observing
a
human
driver
22
When
to
use
Machine
Learning?
• Human
exper)se
is
absent
– Example:
naviga)ng
on
mars
• Humans
are
unable
to
explain
their
exper)se
– Example:
vision,
speech,
language
• Requirements
and
data
change
over
)me
– Example:
Tracking,
Biometrics,
Personalized
fingerprint
recogni)on
• The
problem
or
the
data
size
is
just
too
large
– Example:
Web
Search
23
Types
of
Learning
• Supervised
(induc)ve)
learning
– Training
data
includes
desired
outputs
• Unsupervised
learning
– Training
data
does
not
include
desired
outputs
– Find
hidden/interes)ng
structure
in
data
• Semi-‐supervised
learning
– Training
data
includes
a
few
desired
outputs
• Reinforcement
learning
– the
learner
interacts
with
the
world
via
“ac)ons”
and
tries
to
find
an
op)mal
policy
of
behavior
with
respect
to
“rewards”
it
receives
from
the
environment
24
Examples/Types
of
Machine
Learning
Tasks
• Forecas)ng
or
Predic)on
– Stock
price
of
Google
tomorrow?
• Classifica)on
and
Regression
– Is
Ana
credit-‐worthy?
– What
is
Ana’s
credit
score?
• Ranking
– How
to
rank
images
that
contain
“An
awesome
machine
learning
model”?
• Outlier/Anomaly/Fraud
detec)on
– Is
it
Ana”
using
the
credit
card
in
Mexico
or
is
it
someone
else?
• Finding
paberns
– Almost
60%
of
shoppers
buy
Diapers
and
Milk
together!
25
Machine
Learning:
Applica)ons
• Examples
of
real-‐world
scenarios
where
machine
learning
performs
its
magic!
26
Classifica)on
Example:
Spam
Filtering
27
Classifica)on
Example:
Weather
Predic)on
28
Regression
example:
Predic)ng
Gold/
Stock
prices
30
Collabora)ve
Filtering
• The
problem
of
collabora/ve
filtering
is
to
predict
how
well
a
user
will
like
an
item
that
he
has
not
rated
given
a
set
of
historical
preference
judgments
for
a
community
of
users.
31
Collabora)ve
Filtering
32
Collabora)ve
Filtering
33
Clustering:
Discover
Structure
in
data
34
Clustering
35
ML
in
a
Nutshell
• Tens
of
thousands
of
machine
learning
algorithms
• Hundreds
new
every
year
• Every
machine
learning
algorithm
has
three
components:
– Representa)on
– Evalua)on
– Op)miza)on
36
Representa)on
• Decision
trees
• Sets
of
rules
/
Logic
programs
• Instances
• Graphical
models
(Bayes/Markov
nets)
• Neural
networks
• Support
vector
machines
• Model
ensembles
• Etc.
37
Evalua)on
• Accuracy
• Precision
and
recall
• Squared
error
• Likelihood
• Posterior
probability
• Cost
/
U)lity
• Margin
• Entropy
• K-‐L
divergence
• Etc.
38
Op)miza)on
• Combinatorial
op)miza)on
– E.g.:
Greedy
search
• Convex
op)miza)on
– E.g.:
Gradient
descent
• Constrained
op)miza)on
– E.g.:
Linear
programming
39
Machine
learning
has
grown
in
leaps
and
bounds
• The
main
approach
for
– Speech
Recogni)on
– Robo)cs
– Natural
Language
Processing
– Computa)onal
Biology
– Sensor
networks
– Computer
Vision
– Web
– And
so
on
40
What
We’ll
Cover
• Supervised
learning:
Decision
tree
induc)on,
Rule
induc)on,
Instance-‐based
learning,
Bayesian
learning,
Neural
networks,
Support
vector
machines,
Linear
Regression,
Model
ensembles
• Unsupervised
learning:
Clustering,
Dimensionality
reduc)on
• General
machine
learning
concepts
and
techniques:
Feature
selec)on,
cross-‐valida)on,
maximum
likelihood
es)ma)on,
gradient
descent,
expecta)on-‐maximiza)on
• And
some
special
topics
(if
)me
permits):
probabilis)c
graphical
models,
topic
models
• Your
responsibility:
– Brush
up
on
some
important
background
– Linear
algebra,
Sta)s)cs
101,
Vectors,
Probability
theory
41