<s>
Instantaneously	B-Algorithm
trained	I-Algorithm
neural	I-Algorithm
networks	I-Algorithm
are	O
feedforward	B-Algorithm
artificial	I-Algorithm
neural	I-Algorithm
networks	I-Algorithm
that	O
create	O
a	O
new	O
hidden	O
neuron	O
node	O
for	O
each	O
novel	O
training	O
sample	O
.	O
</s>
<s>
These	O
networks	O
use	O
unary	B-Algorithm
coding	I-Algorithm
for	O
an	O
effective	O
representation	O
of	O
the	O
data	O
sets	O
.	O
</s>
<s>
Since	O
then	O
,	O
instantaneously	B-Algorithm
trained	I-Algorithm
neural	I-Algorithm
networks	I-Algorithm
have	O
been	O
proposed	O
as	O
models	O
of	O
short	O
term	O
learning	O
and	O
used	O
in	O
web	B-Application
search	I-Application
,	O
and	O
financial	O
time	O
series	O
prediction	O
applications	O
.	O
</s>
<s>
They	O
have	O
also	O
been	O
used	O
in	O
instant	O
classification	B-Algorithm
of	I-Algorithm
documents	I-Algorithm
and	O
for	O
deep	B-Algorithm
learning	I-Algorithm
and	O
data	B-Application
mining	I-Application
.	O
</s>
<s>
As	O
in	O
other	O
neural	O
networks	O
,	O
their	O
normal	O
use	O
is	O
as	O
software	O
,	O
but	O
they	O
have	O
also	O
been	O
implemented	O
in	O
hardware	O
using	O
FPGAs	O
and	O
by	O
optical	B-Algorithm
implementation	I-Algorithm
.	O
</s>
<s>
In	O
feedback	O
networks	O
the	O
Willshaw	O
network	O
as	O
well	O
as	O
the	O
Hopfield	B-Algorithm
network	I-Algorithm
are	O
able	O
to	O
learn	O
instantaneously	O
.	O
</s>
