<s>
In	O
machine	O
learning	O
,	O
instance-based	B-General_Concept
learning	I-General_Concept
(	O
sometimes	O
called	O
memory-based	B-General_Concept
learning	I-General_Concept
)	O
is	O
a	O
family	O
of	O
learning	O
algorithms	O
that	O
,	O
instead	O
of	O
performing	O
explicit	O
generalization	O
,	O
compare	O
new	O
problem	O
instances	O
with	O
instances	O
seen	O
in	O
training	O
,	O
which	O
have	O
been	O
stored	O
in	O
memory	O
.	O
</s>
<s>
This	O
means	O
that	O
the	O
hypothesis	O
complexity	O
can	O
grow	O
with	O
the	O
data	O
:	O
in	O
the	O
worst	O
case	O
,	O
a	O
hypothesis	O
is	O
a	O
list	O
of	O
n	O
training	O
items	O
and	O
the	O
computational	O
complexity	O
of	O
classifying	B-General_Concept
a	O
single	O
new	O
instance	O
is	O
O(n )	O
.	O
</s>
<s>
One	O
advantage	O
that	O
instance-based	B-General_Concept
learning	I-General_Concept
has	O
over	O
other	O
methods	O
of	O
machine	O
learning	O
is	O
its	O
ability	O
to	O
adapt	O
its	O
model	O
to	O
previously	O
unseen	O
data	O
.	O
</s>
<s>
Examples	O
of	O
instance-based	B-General_Concept
learning	I-General_Concept
algorithms	O
are	O
the	O
k-nearest	B-General_Concept
neighbors	I-General_Concept
algorithm	I-General_Concept
,	O
kernel	B-Algorithm
machines	I-Algorithm
and	O
RBF	B-Algorithm
networks	I-Algorithm
.	O
</s>
<s>
To	O
battle	O
the	O
memory	O
complexity	O
of	O
storing	O
all	O
training	O
instances	O
,	O
as	O
well	O
as	O
the	O
risk	O
of	O
overfitting	B-Error_Name
to	O
noise	O
in	O
the	O
training	O
set	O
,	O
instance	O
reduction	O
algorithms	O
have	O
been	O
proposed	O
.	O
</s>
