<s>
In	O
the	O
context	O
of	O
artificial	B-Architecture
neural	I-Architecture
network	I-Architecture
,	O
pruning	B-Algorithm
is	O
the	O
practice	O
of	O
removing	O
parameters	O
(	O
which	O
may	O
entail	O
removing	O
individual	O
parameters	O
,	O
or	O
parameters	O
in	O
groups	O
such	O
as	O
by	O
neurons	B-Algorithm
)	O
from	O
an	O
existing	O
network	O
.	O
</s>
<s>
This	O
can	O
be	O
done	O
to	O
reduce	O
the	O
computational	O
resources	O
required	O
to	O
run	O
the	O
neural	B-Architecture
network	I-Architecture
.	O
</s>
<s>
A	O
basic	O
algorithm	O
for	O
pruning	B-Algorithm
is	O
as	O
follows	O
:	O
</s>
<s>
Rank	O
the	O
neurons	B-Algorithm
according	O
to	O
their	O
importance	O
(	O
assuming	O
there	O
is	O
a	O
clearly	O
defined	O
measure	O
for	O
"	O
importance	O
"	O
)	O
.	O
</s>
<s>
Check	O
a	O
termination	O
condition	O
(	O
to	O
be	O
determined	O
by	O
the	O
user	O
)	O
to	O
see	O
whether	O
to	O
continue	O
pruning	B-Algorithm
.	O
</s>
<s>
Recently	O
a	O
highly	O
pruning	B-Algorithm
three	O
layer	O
tree	O
architecture	O
,	O
has	O
achieved	O
a	O
similar	O
success	O
rate	O
to	O
that	O
of	O
LeNet-5	O
on	O
the	O
CIFAR-10	O
dataset	O
with	O
a	O
lesser	O
computational	O
complexity	O
.	O
</s>
