<s>
A	O
liquid	B-Algorithm
state	I-Algorithm
machine	I-Algorithm
(	O
LSM	O
)	O
is	O
a	O
type	O
of	O
reservoir	B-Algorithm
computer	I-Algorithm
that	O
uses	O
a	O
spiking	B-Algorithm
neural	I-Algorithm
network	I-Algorithm
.	O
</s>
<s>
The	O
recurrent	B-Algorithm
nature	O
of	O
the	O
connections	O
turns	O
the	O
time	O
varying	O
input	O
into	O
a	O
spatio-temporal	O
pattern	O
of	O
activations	O
in	O
the	O
network	O
nodes	O
.	O
</s>
<s>
The	O
spatio-temporal	O
patterns	O
of	O
activation	O
are	O
read	O
out	O
by	O
linear	B-General_Concept
discriminant	I-General_Concept
units	O
.	O
</s>
<s>
Given	O
a	O
large	O
enough	O
variety	O
of	O
such	O
nonlinear	O
functions	O
,	O
it	O
is	O
theoretically	O
possible	O
to	O
obtain	O
linear	O
combinations	O
(	O
using	O
the	O
read	O
out	O
units	O
)	O
to	O
perform	O
whatever	O
mathematical	O
operation	O
is	O
needed	O
to	O
perform	O
a	O
certain	O
task	O
,	O
such	O
as	O
speech	B-Application
recognition	I-Application
or	O
computer	B-Application
vision	I-Application
.	O
</s>
<s>
If	O
a	O
reservoir	O
has	O
fading	O
memory	O
and	O
input	O
separability	O
,	O
with	O
help	O
of	O
a	O
readout	O
,	O
it	O
can	O
be	O
proven	O
the	O
liquid	B-Algorithm
state	I-Algorithm
machine	I-Algorithm
is	O
a	O
universal	O
function	O
approximator	O
using	O
Stone	O
–	O
Weierstrass	O
theorem	O
.	O
</s>
