<s>
Structural	B-General_Concept
risk	I-General_Concept
minimization	I-General_Concept
(	O
SRM	O
)	O
is	O
an	O
inductive	O
principle	O
of	O
use	O
in	O
machine	O
learning	O
.	O
</s>
<s>
Commonly	O
in	O
machine	O
learning	O
,	O
a	O
generalized	O
model	O
must	O
be	O
selected	O
from	O
a	O
finite	O
data	O
set	O
,	O
with	O
the	O
consequent	O
problem	O
of	O
overfitting	B-Error_Name
the	O
model	O
becoming	O
too	O
strongly	O
tailored	O
to	O
the	O
particularities	O
of	O
the	O
training	O
set	O
and	O
generalizing	O
poorly	O
to	O
new	O
data	O
.	O
</s>
<s>
In	O
practical	O
terms	O
,	O
Structural	B-General_Concept
Risk	I-General_Concept
Minimization	I-General_Concept
is	O
implemented	O
by	O
minimizing	O
,	O
where	O
is	O
the	O
train	O
error	O
,	O
the	O
function	O
is	O
called	O
a	O
regularization	O
function	O
,	O
and	O
is	O
a	O
constant	O
.	O
</s>
<s>
Note	O
that	O
as	O
the	O
weights	O
become	O
zero	O
,	O
and	O
as	O
,	O
the	O
model	O
typically	O
suffers	O
from	O
overfitting	B-Error_Name
.	O
</s>
