<s>
In	O
machine	O
learning	O
,	O
pattern	O
recognition	O
,	O
and	O
image	B-Algorithm
processing	I-Algorithm
,	O
feature	B-Algorithm
extraction	I-Algorithm
starts	O
from	O
an	O
initial	O
set	O
of	O
measured	O
data	O
and	O
builds	O
derived	O
values	O
(	O
features	B-Algorithm
)	O
intended	O
to	O
be	O
informative	O
and	O
non-redundant	O
,	O
facilitating	O
the	O
subsequent	O
learning	O
and	O
generalization	O
steps	O
,	O
and	O
in	O
some	O
cases	O
leading	O
to	O
better	O
human	O
interpretations	O
.	O
</s>
<s>
Feature	B-Algorithm
extraction	I-Algorithm
is	O
related	O
to	O
dimensionality	B-Algorithm
reduction	I-Algorithm
.	O
</s>
<s>
the	O
same	O
measurement	O
in	O
both	O
feet	O
and	O
meters	O
,	O
or	O
the	O
repetitiveness	O
of	O
images	O
presented	O
as	O
pixels	O
)	O
,	O
then	O
it	O
can	O
be	O
transformed	O
into	O
a	O
reduced	O
set	O
of	O
features	B-Algorithm
(	O
also	O
named	O
a	O
feature	B-Algorithm
vector	I-Algorithm
)	O
.	O
</s>
<s>
Determining	O
a	O
subset	O
of	O
the	O
initial	O
features	B-Algorithm
is	O
called	O
feature	B-General_Concept
selection	I-General_Concept
.	O
</s>
<s>
The	O
selected	O
features	B-Algorithm
are	O
expected	O
to	O
contain	O
the	O
relevant	O
information	O
from	O
the	O
input	O
data	O
,	O
so	O
that	O
the	O
desired	O
task	O
can	O
be	O
performed	O
by	O
using	O
this	O
reduced	O
representation	O
instead	O
of	O
the	O
complete	O
initial	O
data	O
.	O
</s>
<s>
Feature	B-Algorithm
extraction	I-Algorithm
involves	O
reducing	O
the	O
number	O
of	O
resources	O
required	O
to	O
describe	O
a	O
large	O
set	O
of	O
data	O
.	O
</s>
<s>
Analysis	O
with	O
a	O
large	O
number	O
of	O
variables	O
generally	O
requires	O
a	O
large	O
amount	O
of	O
memory	O
and	O
computation	O
power	O
,	O
also	O
it	O
may	O
cause	O
a	O
classification	B-General_Concept
algorithm	O
to	O
overfit	B-Error_Name
to	O
training	O
samples	O
and	O
generalize	O
poorly	O
to	O
new	O
samples	O
.	O
</s>
<s>
Feature	B-Algorithm
extraction	I-Algorithm
is	O
a	O
general	O
term	O
for	O
methods	O
of	O
constructing	O
combinations	O
of	O
the	O
variables	O
to	O
get	O
around	O
these	O
problems	O
while	O
still	O
describing	O
the	O
data	O
with	O
sufficient	O
accuracy	O
.	O
</s>
<s>
Many	O
machine	O
learning	O
practitioners	O
believe	O
that	O
properly	O
optimized	O
feature	B-Algorithm
extraction	I-Algorithm
is	O
the	O
key	O
to	O
effective	O
model	O
construction	O
.	O
</s>
<s>
Results	O
can	O
be	O
improved	O
using	O
constructed	O
sets	O
of	O
application-dependent	O
features	B-Algorithm
,	O
typically	O
built	O
by	O
an	O
expert	O
.	O
</s>
<s>
One	O
such	O
process	O
is	O
called	O
feature	B-General_Concept
engineering	I-General_Concept
.	O
</s>
<s>
Alternatively	O
,	O
general	O
dimensionality	B-Algorithm
reduction	I-Algorithm
techniques	O
are	O
used	O
such	O
as	O
:	O
</s>
<s>
One	O
very	O
important	O
area	O
of	O
application	O
is	O
image	B-Algorithm
processing	I-Algorithm
,	O
in	O
which	O
algorithms	O
are	O
used	O
to	O
detect	O
and	O
isolate	O
various	O
desired	O
portions	O
or	O
shapes	O
(	O
features	B-Algorithm
)	O
of	O
a	O
digitized	B-Algorithm
image	I-Algorithm
or	O
video	B-Application
stream	I-Application
.	O
</s>
<s>
It	O
is	O
particularly	O
important	O
in	O
the	O
area	O
of	O
optical	B-Application
character	I-Application
recognition	I-Application
.	O
</s>
<s>
Many	O
data	B-Algorithm
analysis	I-Algorithm
software	I-Algorithm
packages	O
provide	O
for	O
feature	B-Algorithm
extraction	I-Algorithm
and	O
dimension	B-Algorithm
reduction	I-Algorithm
.	O
</s>
<s>
Common	O
numerical	O
programming	O
environments	O
such	O
as	O
MATLAB	B-Language
,	O
SciLab	B-Application
,	O
NumPy	B-Application
,	O
scikit-learn	B-Application
and	O
the	B-Language
R	I-Language
language	I-Language
provide	O
some	O
of	O
the	O
simpler	O
feature	B-Algorithm
extraction	I-Algorithm
techniques	O
(	O
e.g.	O
</s>
<s>
principal	B-Application
component	I-Application
analysis	I-Application
)	O
via	O
built-in	O
commands	O
.	O
</s>
<s>
There	O
are	O
also	O
software	O
packages	O
targeting	O
specific	O
software	O
machine	O
learning	O
applications	O
that	O
specialize	O
in	O
feature	B-Algorithm
extraction	I-Algorithm
.	O
</s>
