<s>
In	O
machine	O
learning	O
,	O
one-class	B-General_Concept
classification	I-General_Concept
(	O
OCC	O
)	O
,	O
also	O
known	O
as	O
unary	B-General_Concept
classification	I-General_Concept
or	O
class-modelling	O
,	O
tries	O
to	O
identify	O
objects	O
of	O
a	O
specific	O
class	O
amongst	O
all	O
objects	O
,	O
by	O
primarily	O
learning	O
from	O
a	O
training	O
set	O
containing	O
only	O
the	O
objects	O
of	O
that	O
class	O
,	O
although	O
there	O
exist	O
variants	O
of	O
one-class	O
classifiers	B-General_Concept
where	O
counter-examples	O
are	O
used	O
to	O
further	O
refine	O
the	O
classification	B-General_Concept
boundary	O
.	O
</s>
<s>
This	O
is	O
different	O
from	O
and	O
more	O
difficult	O
than	O
the	O
traditional	O
classification	B-General_Concept
problem	O
,	O
which	O
tries	O
to	O
distinguish	O
between	O
two	O
or	O
more	O
classes	O
with	O
the	O
training	O
set	O
containing	O
objects	O
from	O
all	O
the	O
classes	O
.	O
</s>
<s>
While	O
many	O
of	O
the	O
above	O
approaches	O
focus	O
on	O
the	O
case	O
of	O
removing	O
a	O
small	O
number	O
of	O
outliers	O
or	O
anomalies	O
,	O
one	O
can	O
also	O
learn	O
the	O
other	O
extreme	O
,	O
where	O
the	O
single	O
class	O
covers	O
a	O
small	O
coherent	O
subset	O
of	O
the	O
data	O
,	O
using	O
an	O
information	B-General_Concept
bottleneck	I-General_Concept
approach	O
.	O
</s>
<s>
The	O
term	O
one-class	B-General_Concept
classification	I-General_Concept
(	O
OCC	O
)	O
was	O
coined	O
by	O
Moya	O
&	O
Hush	O
(	O
1996	O
)	O
and	O
many	O
applications	O
can	O
be	O
found	O
in	O
scientific	O
literature	O
,	O
for	O
example	O
outlier	B-Algorithm
detection	I-Algorithm
,	O
anomaly	B-Algorithm
detection	I-Algorithm
,	O
novelty	B-Algorithm
detection	I-Algorithm
.	O
</s>
<s>
SVM	B-Algorithm
based	O
one-class	B-General_Concept
classification	I-General_Concept
(	O
OCC	O
)	O
relies	O
on	O
identifying	O
the	O
smallest	O
hypersphere	O
(	O
with	O
radius	O
r	O
,	O
and	O
center	O
c	O
)	O
consisting	O
of	O
all	O
the	O
data	O
points	O
.	O
</s>
<s>
The	O
introduction	O
of	O
kernel	O
function	O
provide	O
additional	O
flexibility	O
to	O
the	O
One-class	O
SVM	B-Algorithm
(	O
OSVM	O
)	O
algorithm	O
.	O
</s>
<s>
A	O
similar	O
problem	O
is	O
PU	B-General_Concept
learning	I-General_Concept
,	O
in	O
which	O
a	O
binary	B-General_Concept
classifier	I-General_Concept
is	O
learned	O
in	O
a	O
semi-supervised	B-General_Concept
way	O
from	O
only	O
positive	O
and	O
unlabeled	O
sample	O
points	O
.	O
</s>
<s>
In	O
PU	B-General_Concept
learning	I-General_Concept
,	O
two	O
sets	O
of	O
examples	O
are	O
assumed	O
to	O
be	O
available	O
for	O
training	O
:	O
the	O
positive	O
set	O
and	O
a	O
mixed	O
set	O
,	O
which	O
is	O
assumed	O
to	O
contain	O
both	O
positive	O
and	O
negative	O
samples	O
,	O
but	O
without	O
these	O
being	O
labeled	O
as	O
such	O
.	O
</s>
<s>
A	O
variety	O
of	O
techniques	O
exist	O
to	O
adapt	O
supervised	B-General_Concept
classifiers	B-General_Concept
to	O
the	O
PU	B-General_Concept
learning	I-General_Concept
setting	O
,	O
including	O
variants	O
of	O
the	O
EM	B-Algorithm
algorithm	I-Algorithm
.	O
</s>
<s>
PU	B-General_Concept
learning	I-General_Concept
has	O
been	O
successfully	O
applied	O
to	O
text	B-Algorithm
,	O
time	O
series	O
,	O
bioinformatics	O
tasks	O
,	O
and	O
Remote-Sensing	O
Data	O
.	O
</s>
<s>
Several	O
approaches	O
have	O
been	O
proposed	O
to	O
solve	O
one-class	B-General_Concept
classification	I-General_Concept
(	O
OCC	O
)	O
.	O
</s>
<s>
Gaussian	O
model	O
is	O
one	O
of	O
the	O
simplest	O
methods	O
to	O
create	O
one-class	O
classifiers	B-General_Concept
.	O
</s>
<s>
The	O
basic	O
Support	B-Algorithm
Vector	I-Algorithm
Machine	I-Algorithm
(	O
SVM	B-Algorithm
)	O
paradigm	O
is	O
trained	O
using	O
both	O
positive	O
and	O
negative	O
examples	O
,	O
however	O
studies	O
have	O
shown	O
there	O
are	O
many	O
valid	O
reasons	O
for	O
using	O
only	O
positive	O
examples	O
.	O
</s>
<s>
When	O
the	O
SVM	B-Algorithm
algorithm	O
is	O
modified	O
to	O
only	O
use	O
positive	O
examples	O
,	O
the	O
process	O
is	O
considered	O
one-class	B-General_Concept
classification	I-General_Concept
.	O
</s>
<s>
One	O
situation	O
where	O
this	O
type	O
of	O
classification	B-General_Concept
might	O
prove	O
useful	O
to	O
the	O
SVM	B-Algorithm
paradigm	O
is	O
in	O
trying	O
to	O
identify	O
a	O
web	O
browser	O
’s	O
sites	O
of	O
interest	O
based	O
only	O
off	O
of	O
the	O
user	O
’s	O
browsing	O
history	O
.	O
</s>
<s>
One-class	B-General_Concept
classification	I-General_Concept
can	O
be	O
particularly	O
useful	O
in	O
biomedical	O
studies	O
where	O
often	O
data	O
from	O
other	O
classes	O
can	O
be	O
difficult	O
or	O
impossible	O
to	O
obtain	O
.	O
</s>
<s>
In	O
studying	O
biomedical	O
data	O
it	O
can	O
be	O
difficult	O
and/or	O
expensive	O
to	O
obtain	O
the	O
set	O
of	O
labeled	O
data	O
from	O
the	O
second	O
class	O
that	O
would	O
be	O
necessary	O
to	O
perform	O
a	O
two-class	O
classification	B-General_Concept
.	O
</s>
<s>
To	O
apply	O
typicality	O
to	O
one-class	B-General_Concept
classification	I-General_Concept
for	O
biomedical	O
studies	O
,	O
each	O
new	O
observation	O
,	O
,	O
is	O
compared	O
to	O
the	O
target	O
class	O
,	O
,	O
and	O
identified	O
as	O
an	O
outlier	O
or	O
a	O
member	O
of	O
the	O
target	O
class	O
.	O
</s>
<s>
One-class	B-General_Concept
classification	I-General_Concept
has	O
similarities	O
with	O
unsupervised	O
concept	O
drift	O
detection	O
,	O
where	O
both	O
aim	O
to	O
identify	O
whether	O
the	O
unseen	O
data	O
share	O
similar	O
characteristics	O
to	O
the	O
initial	O
data	O
.	O
</s>
<s>
In	O
one-class	B-General_Concept
classification	I-General_Concept
,	O
the	O
flow	O
of	O
data	O
is	O
not	O
important	O
.	O
</s>
<s>
Unsupervised	O
concept	O
drift	O
detection	O
can	O
be	O
identified	O
as	O
the	O
continuous	O
form	O
of	O
one-class	B-General_Concept
classification	I-General_Concept
.	O
</s>
<s>
One-class	O
classifiers	B-General_Concept
are	O
used	O
for	O
detecting	O
concept	O
drifts	O
.	O
</s>
