<s>
An	O
adaptive	B-Algorithm
neuro-fuzzy	I-Algorithm
inference	I-Algorithm
system	I-Algorithm
or	O
adaptive	B-Algorithm
network-based	I-Algorithm
fuzzy	I-Algorithm
inference	I-Algorithm
system	I-Algorithm
(	O
ANFIS	B-Algorithm
)	O
is	O
a	O
kind	O
of	O
artificial	B-Architecture
neural	I-Architecture
network	I-Architecture
that	O
is	O
based	O
on	O
Takagi	O
–	O
Sugeno	O
fuzzy	O
inference	O
system	O
.	O
</s>
<s>
Since	O
it	O
integrates	O
both	O
neural	B-Architecture
networks	I-Architecture
and	O
fuzzy	O
logic	O
principles	O
,	O
it	O
has	O
potential	O
to	O
capture	O
the	O
benefits	O
of	O
both	O
in	O
a	O
single	O
framework	O
.	O
</s>
<s>
Its	O
inference	O
system	O
corresponds	O
to	O
a	O
set	O
of	O
fuzzy	O
IF	B-Language
–	I-Language
THEN	I-Language
rules	I-Language
that	O
have	O
learning	O
capability	O
to	O
approximate	O
nonlinear	O
functions	O
.	O
</s>
<s>
Hence	O
,	O
ANFIS	B-Algorithm
is	O
considered	O
to	O
be	O
a	O
universal	O
estimator	O
.	O
</s>
<s>
For	O
using	O
the	O
ANFIS	B-Algorithm
in	O
a	O
more	O
efficient	O
and	O
optimal	O
way	O
,	O
one	O
can	O
use	O
the	O
best	O
parameters	O
obtained	O
by	O
genetic	B-Algorithm
algorithm	I-Algorithm
.	O
</s>
<s>
The	O
first	O
layer	O
takes	O
the	O
input	O
values	O
and	O
determines	O
the	O
membership	B-General_Concept
functions	I-General_Concept
belonging	O
to	O
them	O
.	O
</s>
<s>
The	O
first	O
layer	O
of	O
an	O
ANFIS	B-Algorithm
network	O
describes	O
the	O
difference	O
to	O
a	O
vanilla	O
neural	B-Architecture
network	I-Architecture
.	O
</s>
<s>
Neural	B-Architecture
networks	I-Architecture
in	O
general	O
are	O
operating	O
with	O
a	O
data	B-General_Concept
pre-processing	I-General_Concept
step	O
,	O
in	O
which	O
the	O
features	B-Algorithm
are	O
converted	O
into	O
normalized	O
values	O
between	O
0	O
and	O
1	O
.	O
</s>
<s>
An	O
ANFIS	B-Algorithm
neural	B-Architecture
network	I-Architecture
does	O
n't	O
need	O
a	O
sigmoid	B-Algorithm
function	I-Algorithm
,	O
but	O
it	O
's	O
doing	O
the	O
preprocessing	O
step	O
by	O
converting	O
numeric	O
values	O
into	O
fuzzy	O
values	O
.	O
</s>
<s>
Converting	O
the	O
numerical	O
values	O
into	O
Fuzzy	O
numbers	O
is	O
done	O
with	O
the	O
membership	B-General_Concept
function	I-General_Concept
which	O
consists	O
of	O
semantic	B-General_Concept
descriptions	I-General_Concept
like	O
near	O
,	O
middle	O
and	O
far	O
.	O
</s>
<s>
Each	O
possible	O
linguistic	O
value	O
is	O
given	O
by	O
an	O
individual	O
neuron	B-Algorithm
.	O
</s>
<s>
The	O
neuron	B-Algorithm
“	O
near	O
”	O
fires	O
with	O
a	O
value	O
from	O
0	O
until	O
1	O
,	O
if	O
the	O
distance	O
is	O
located	O
within	O
the	O
category	O
"	O
near	O
"	O
.	O
</s>
<s>
While	O
the	O
neuron	B-Algorithm
“	O
middle	O
”	O
fires	O
,	O
if	O
the	O
distance	O
in	O
that	O
category	O
.	O
</s>
