I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Ro
bo
t
ics a
nd
Aut
o
m
a
t
io
n
(
I
J
RA)
Vo
l.
13
,
No
.
4
,
Dec
em
b
er
20
24
,
p
p
.
4
9
5
~
5
0
5
I
SS
N:
2722
-
2
5
8
6
,
DOI
:
1
0
.
1
1
5
9
1
/i
jr
a
.
v
13
i
4
.
pp
4
9
5
-
5
05
495
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
r
a
.
ia
esco
r
e.
co
m
Ana
ly
sis
of
sing
le
la
y
er artif
icia
l ne
ura
l net
wo
rk
neu
ro
mo
rphic
ha
rdwa
re c
hip
Aruna
P
a
nt,
Adesh
K
um
a
r,
P
iy
us
h K
uch
ha
l
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
i
c
a
l
a
n
d
El
e
c
t
r
o
n
i
c
s
En
g
i
n
e
e
r
i
n
g
,
S
c
h
o
o
l
o
f
A
d
v
a
n
c
e
d
En
g
i
n
e
e
r
i
n
g
,
U
P
ES,
D
e
h
r
a
d
u
n
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
1
5
,
2
0
2
4
R
ev
is
ed
J
u
l 3
0
,
2
0
2
4
Acc
ep
ted
Au
g
1
2
,
2
0
2
4
Th
e
n
e
u
ro
m
o
r
p
h
ic
a
rc
h
it
e
c
tu
re
s
a
re
h
a
r
d
wa
re
n
e
two
r
k
s
y
ste
m
s
d
e
sig
n
e
d
with
n
e
u
ra
l
f
u
n
c
ti
o
n
s.
Ne
u
ra
l
n
e
two
r
k
s
se
e
n
i
n
b
io
l
o
g
y
se
rv
e
a
s
a
n
in
sp
irati
o
n
f
o
r
n
e
two
r
k
sy
ste
m
s.
A sy
n
a
p
se
c
o
n
n
e
c
ts ev
e
ry
n
o
d
e
o
r
n
e
u
ro
n
in
a
n
a
rti
ficia
l
n
e
u
ra
l
n
e
two
r
k
(AN
N)
to
e
v
e
ry
o
t
h
e
r
n
o
d
e
.
As
in
b
io
l
o
g
ica
l
b
ra
in
s,
th
e
a
m
p
li
tu
d
e
o
f
th
e
li
n
k
in
g
b
e
twe
e
n
n
o
d
e
s
re
fe
rre
d
to
a
s
sy
n
a
p
ti
c
we
ig
h
ts
will
re
g
u
late
th
e
c
o
n
n
e
c
ti
o
n
.
I
n
c
o
n
t
ra
st
to
c
o
n
v
e
n
ti
o
n
a
l
d
e
sig
n
,
AN
N
u
se
s
m
a
n
y
h
ig
h
ly
o
rg
a
n
ize
d
d
e
a
li
n
g
p
iec
e
s
th
a
t
wo
rk
to
g
e
th
e
r
to
so
lv
e
re
a
l
-
wo
rld
issu
e
s.
Th
e
d
e
sig
n
o
f
th
e
n
e
u
r
o
m
o
rp
h
ic
h
a
rd
wa
re
c
h
ip
is
d
isc
u
ss
e
d
in
th
e
p
a
p
e
r.
T
h
e
targ
e
t
d
e
v
ice
u
se
d
is
a
Virte
x
-
5
F
ield
P
ro
g
ra
m
m
a
b
le
G
a
te
Arr
a
y
(F
P
GA
)
a
n
d
th
e
sim
u
lati
o
n
is
tak
e
n
o
n
Xili
n
x
M
o
d
e
lS
im
so
ftwa
re
.
Th
is
c
h
ip
is
d
e
sig
n
e
d
f
o
r
2
0
n
e
u
ro
n
in
p
u
ts,
e
a
c
h
o
f
th
e
n
e
u
ro
n
in
p
u
ts
is
8
-
b
it
.
Eac
h
2
0
-
n
e
u
r
o
n
i
n
p
u
t
is
m
u
lt
ip
li
e
d
b
y
2
0
i
n
p
u
t
we
ig
h
ts
a
n
d
e
a
c
h
we
ig
h
t
is
8
-
b
it
so
wh
e
n
th
e
se
2
0
in
p
u
t
w
e
ig
h
ts
a
re
m
u
lt
ip
li
e
d
b
y
2
0
n
e
u
r
o
n
i
n
p
u
ts
in
t
h
e
m
u
lt
ip
li
e
r
it
g
iv
e
s
1
6
-
b
it
o
u
t
p
u
t
.
A
c
o
n
tro
l
lo
g
ic
is
u
se
d
i
n
th
is
n
e
u
ro
m
o
rp
h
ic
h
a
rd
wa
re
c
h
ip
d
e
sig
n
wh
ich
is
u
se
d
t
o
fe
e
d
m
u
lt
ip
li
e
r
o
u
t
p
u
t
t
o
e
a
c
h
in
p
u
t
o
f
t
h
e
h
id
d
e
n
lay
e
r.
T
h
e
sy
ste
m
-
lev
e
l
o
u
tco
m
e
o
f
t
h
e
h
id
d
e
n
lay
e
r
is
t
h
e
n
g
i
v
e
n
to
th
e
m
u
lt
i
p
lex
e
r
wh
ich
h
a
s
2
0
i
n
p
u
ts
a
n
d
o
n
e
sin
g
le
o
u
t
p
u
t
.
Th
e
m
u
lt
ip
lex
e
r
is
u
se
d
to
se
lec
t
a
n
y
o
f
t
h
e
2
0
o
u
t
p
u
ts
o
f
th
e
c
o
n
tr
o
l
l
o
g
ic.
F
in
a
ll
y
,
to
g
a
in
a
n
u
n
d
e
rsta
n
d
i
n
g
o
f
t
h
e
p
e
rfo
rm
a
n
c
e
o
f
th
is
n
e
u
r
o
m
o
rp
h
ic
h
a
rd
wa
re
c
h
i
p
,
we
h
a
v
e
c
o
m
p
u
ted
th
e
h
a
rd
wa
re
u
ti
l
iza
ti
o
n
p
a
ra
m
e
ters
.
Th
e
se
p
a
ra
m
e
ters
in
c
lu
d
e
slice
s,
in
p
u
t/
o
u
t
p
u
t
b
l
o
c
k
s
(I
O
Bs),
re
g
isters
u
se
d
,
m
e
m
o
ry
,
a
n
d
t
h
e
o
v
e
ra
ll
p
ro
p
a
g
a
ti
o
n
d
e
la
y
u
se
d
b
y
th
e
h
a
r
d
wa
re
c
h
ip
.
K
ey
w
o
r
d
s
:
C
o
n
tr
o
l lo
g
ic
Field
p
r
o
g
r
a
m
m
ab
le
g
ate
a
r
r
a
y
Sin
g
le
lay
er
ar
tific
ial
n
eu
r
al
n
etwo
r
k
Neu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
c
h
ip
Sy
s
tem
p
er
f
o
r
m
a
n
ce
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ad
esh
Ku
m
ar
Dep
ar
tm
en
t o
f
E
lectr
ical
an
d
E
lectr
o
n
ics E
n
g
in
ee
r
i
n
g
,
Sch
o
o
l o
f
Ad
v
an
ce
d
E
n
g
in
ee
r
in
g
,
UPES
Deh
r
ad
u
n
I
n
d
ia
E
m
ail: a
d
esh
m
an
av
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Neu
r
o
m
o
r
p
h
ic
co
m
p
u
tin
g
s
ee
k
s
to
m
im
ic
th
e
f
u
n
ctio
n
ality
o
f
th
e
h
u
m
an
b
r
ai
n
.
I
t
is
p
r
e
d
icate
d
o
n
th
e
n
o
tio
n
th
at
th
e
b
r
ain
is
a
s
u
p
er
co
m
p
u
ter
with
g
r
ea
t
ef
f
ici
en
cy
an
d
p
o
wer
th
at
ca
n
p
r
o
ce
s
s
in
f
o
r
m
atio
n
in
a
non
-
lin
ea
r
f
ash
io
n
.
T
o
ac
co
m
p
lis
h
th
is
,
n
eu
r
o
m
o
r
p
h
ic
co
m
p
u
tin
g
m
ak
es
u
s
e
o
f
a
n
etwo
r
k
o
f
in
ter
co
n
n
ec
ted
n
eu
r
o
n
s
,
wh
ich
ar
e
b
asic
p
r
o
ce
s
s
in
g
u
n
its
th
at
p
r
o
ce
s
s
d
at
a.
T
h
e
elec
tr
ical
an
d
ch
em
ica
l
ch
ar
ac
ter
is
tics
o
f
b
io
lo
g
ical
n
eu
r
o
n
s
ar
e
m
im
ic
k
ed
in
th
e
d
esig
n
o
f
th
e
ar
ti
f
icial
n
eu
r
o
n
s
.
B
ec
au
s
e
o
f
its
ab
ilit
y
to
h
an
d
le
in
f
o
r
m
atio
n
in
a
m
an
n
er
clo
s
e
to
th
at
o
f
th
e
h
u
m
an
b
r
ain
,
n
eu
r
o
m
o
r
p
h
ic
co
m
p
u
tin
g
is
a
p
er
f
ec
t
f
it
f
o
r
th
e
im
p
r
o
v
em
e
n
t
o
f
b
r
ai
n
-
co
m
p
u
ter
in
ter
f
ac
es
(
B
C
I
s
)
.
T
h
e
c
ap
ac
ity
o
f
n
eu
r
o
m
o
r
p
h
ic
c
o
m
p
u
tin
g
to
p
r
o
ce
s
s
in
f
o
r
m
atio
n
in
a
m
an
n
er
lik
e
th
e
h
u
m
an
b
r
ain
is
o
n
e
o
f
its
p
r
im
ar
y
ad
v
an
tag
es.
T
h
is
im
p
lies
th
at
it
ca
n
b
e
ap
p
lied
to
d
ev
elo
p
B
C
I
s
th
at
m
ak
e
s
en
s
e
to
u
s
er
s
m
o
r
e
n
atu
r
ally
.
Fo
r
in
s
tan
ce
,
a
n
eu
r
o
m
o
r
p
h
ic
B
C
I
m
ig
h
t
m
ak
e
u
s
e
o
f
th
e
s
am
e
s
ig
n
als
s
u
ch
as
elec
tr
o
en
ce
p
h
alo
g
r
ap
h
y
(
E
E
G)
s
ig
n
als
t
h
at
th
e
b
r
a
in
u
s
es
to
r
eg
u
late
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
2
5
8
6
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
,
Vo
l
.
13
,
No
.
4
,
Dec
em
b
er
20
24
:
4
9
5
-
5
0
5
496
m
o
v
em
en
t.
T
h
e
ca
p
ac
ity
o
f
n
eu
r
o
m
o
r
p
h
ic
co
m
p
u
tin
g
t
o
p
r
o
ce
s
s
m
ass
iv
e
am
o
u
n
ts
o
f
d
at
a
q
u
ick
ly
is
an
o
th
e
r
ad
v
an
tag
e.
Ad
d
itio
n
ally
,
b
y
lo
wer
in
g
th
e
am
o
u
n
t
o
f
p
o
wer
n
ee
d
ed
to
o
p
er
ate
B
C
I
s
,
n
eu
r
o
m
o
r
p
h
ic
co
m
p
u
tin
g
m
ay
en
h
an
ce
t
h
eir
o
v
er
all
p
er
f
o
r
m
a
n
ce
.
T
h
is
m
ay
r
esu
lt
in
th
e
cr
ea
tio
n
o
f
B
C
I
s
th
at
ar
e
lo
n
g
er
-
last
in
g
,
lig
h
ter
,
an
d
m
o
r
e
p
o
r
tab
le.
A
d
ev
ice
k
n
o
wn
as
b
r
ain
-
c
o
m
p
u
ter
in
ter
ac
tio
n
en
a
b
les
d
ir
ec
t
co
n
n
ec
tio
n
b
etwe
en
th
e
b
r
ain
an
d
e
x
ter
n
al
d
ev
ice
s
with
o
u
t
r
ely
in
g
o
n
p
e
r
ip
h
er
al
n
er
v
es.
No
wad
ay
s
,
B
C
I
is
m
o
r
e
o
f
a
s
cien
ce
-
in
v
en
ted
to
p
ic
th
an
it
was
i
n
th
e
p
r
ec
e
d
en
t
b
ec
a
u
s
e
o
f
s
ig
n
if
ican
t
b
r
ea
k
th
r
o
u
g
h
s
in
n
eu
r
o
l
o
g
y
,
s
ig
n
al
p
r
o
ce
s
s
in
g
,
co
m
p
u
tatio
n
al
n
eu
r
o
s
cien
ce
,
s
en
s
o
r
s
,
an
d
o
th
e
r
r
elate
d
f
ield
s
.
B
C
I
is
an
ex
p
an
d
in
g
f
ield
o
f
s
tu
d
y
.
T
h
r
o
u
g
h
th
e
u
s
e
o
f
th
e
E
E
G
o
r
o
th
er
b
r
ain
s
ig
n
als,
p
eo
p
le
with
s
ev
er
e
m
o
b
ilit
y
d
is
ab
ilit
ies
ca
n
co
m
m
u
n
icate
with
ex
ter
n
al
ass
is
tiv
e
d
ev
ices
th
r
o
u
g
h
th
e
u
s
e
o
f
B
C
I
tec
h
n
o
lo
g
y
[
1
]
.
T
h
e
in
ter
c
o
n
n
ec
t
ed
m
icr
o
p
r
o
ce
s
s
in
g
u
n
its
ar
e
ca
lled
n
e
u
r
o
n
s
[
2
]
F
o
u
r
g
e
n
er
al
f
u
n
ctio
n
s
o
f
n
e
u
r
o
n
s
ar
e
in
p
u
t,
tr
ig
g
er
,
c
o
n
d
u
ctio
n
,
an
d
o
u
t
p
u
t
[
3
]
.
T
h
e
elec
tr
ical
ac
tiv
ity
p
r
o
d
u
ce
d
b
y
b
r
ain
s
tr
u
ctu
r
es
an
d
r
ec
o
r
d
ed
f
r
o
m
th
e
s
u
r
f
ac
e
o
f
th
e
s
ca
lp
u
s
in
g
elec
tr
o
d
es
is
k
n
o
wn
as
a
n
elec
tr
o
en
ce
p
h
al
o
g
r
am
,
o
r
E
E
G.
E
E
G
is
th
e
p
r
im
ar
y
m
eth
o
d
u
s
ed
b
y
r
esear
ch
e
r
s
to
d
ef
in
e
b
r
ain
ac
tiv
ity
s
in
ce
it
ca
n
b
e
r
ec
o
r
d
e
d
n
o
n
-
in
v
asiv
ely
with
th
e
u
s
e
o
f
d
is
p
o
s
ab
le
eq
u
ip
m
en
t.
T
h
r
o
u
g
h
a
f
u
ll
B
C
I
s
y
s
tem
,
th
e
E
E
G
o
r
b
r
ain
ac
tiv
ity
ca
n
b
e
ap
p
li
ed
in
r
ea
l
-
tim
e
p
r
o
ce
s
s
in
g
t
o
co
n
tr
o
l
ex
ter
n
al
eq
u
ip
m
en
t.
A
d
ata
co
llectin
g
s
y
s
tem
,
p
r
e
-
p
r
o
ce
s
s
in
g
o
f
t
h
e
o
b
tain
ed
s
ig
n
als,
f
ea
tu
r
e
ex
tr
ac
tio
n
,
f
ea
tu
r
e
class
if
icatio
n
,
p
o
s
t
-
p
r
o
ce
s
s
in
g
o
f
th
e
class
if
ier
o
u
tp
u
t,
a
n
d
last
ly
th
e
co
n
t
r
o
l
u
s
er
in
ter
f
ac
e
an
d
d
e
v
ice
co
n
tr
o
ller
ar
e
th
e
s
tan
d
ar
d
co
m
p
o
n
e
n
ts
o
f
a
B
C
I
s
ch
em
e.
I
n
r
ec
en
t
y
ea
r
s
,
th
e
f
ield
o
f
B
C
I
h
as
witn
es
s
ed
s
ig
n
if
ican
t
b
r
ea
k
th
r
o
u
g
h
s
in
b
o
th
th
e
m
ed
ical
an
d
in
f
o
r
m
atio
n
tech
n
o
lo
g
y
f
ield
s
.
T
h
e
n
o
n
-
in
v
asiv
e
v
er
s
io
n
o
f
B
C
I
s
,
wh
ich
is
b
ased
o
n
E
E
G,
is
p
ar
ticu
lar
ly
p
o
p
u
lar
[
4
]
.
B
C
I
is
b
ased
o
n
an
E
E
G
an
d
p
r
esen
ts
n
ew
av
en
u
es
o
f
co
m
m
u
n
icatio
n
c
o
n
n
ec
tin
g
th
e
h
u
m
an
b
r
ain
an
d
a
c
o
m
p
u
ter
[
5
]
.
Am
o
n
g
th
e
n
u
m
er
o
u
s
d
if
f
er
en
t
m
eth
o
d
s
th
at
ar
e
ac
ce
s
s
ib
le,
ar
tific
ial
n
eu
r
al
n
etwo
r
k
s
(
ANNs
)
ar
e
a
well
-
estab
lis
h
ed
m
eth
o
d
th
at
h
as
s
ev
er
al
s
u
cc
ess
f
u
l
ap
p
licatio
n
s
in
th
e
f
ield
o
f
B
C
I
r
esear
ch
[
6
]
.
I
t
is
p
o
s
s
ib
le
to
ap
p
ly
n
eu
r
al
n
etwo
r
k
s
u
s
in
g
eith
er
an
al
o
g
o
r
d
ig
ital
s
y
s
tem
s
.
B
ec
au
s
e
o
f
its
b
en
ef
its
p
r
ec
is
io
n
,
m
o
r
e
r
e
p
ea
tab
ilit
y
,
r
ed
u
ce
d
s
en
s
itiv
ity
to
n
o
is
e,
im
p
r
o
v
ed
test
ab
ilit
y
,
in
cr
ea
s
ed
f
lex
ib
ilit
y
,
an
d
co
m
p
atib
ilit
y
with
v
ar
i
o
u
s
p
r
ep
r
o
ce
s
s
o
r
ty
p
es
-
th
e
d
ig
ital
ap
p
r
o
ac
h
is
m
o
r
e
o
f
ten
u
s
ed
.
T
h
e
h
ar
d
war
e
im
p
lem
en
tatio
n
s
o
f
d
ig
ital
n
e
u
r
al
n
etwo
r
k
s
ar
e
f
u
r
t
h
er
d
iv
i
d
ed
in
to
th
r
ee
ca
teg
o
r
ies:
f
ield
p
r
o
g
r
am
m
ab
le
g
ate
ar
r
ay
(
FP
GA)
b
ased
,
d
ig
ital
s
ig
n
al
p
r
o
ce
s
s
in
g
(
DSP)
b
a
s
ed
,
an
d
ap
p
licatio
n
-
s
p
ec
if
ic
in
teg
r
ated
cir
cu
its
(
ASI
C
)
b
ased
[
7
]
.
Sin
ce
DSP
im
p
lem
en
tatio
n
is
s
eq
u
en
tial,
th
e
p
ar
allel
ar
ch
itectu
r
e
o
f
t
h
e
n
eu
r
o
n
s
in
a
lay
er
is
lo
s
t
[
8
]
.
T
h
e
ASI
C
ap
p
licatio
n
s
d
o
n
o
t
allo
w
f
o
r
u
s
er
r
ec
o
n
f
ig
u
r
atio
n
.
B
ec
au
s
e
it
allo
ws
f
o
r
f
lex
ib
le
r
ec
o
n
f
ig
u
r
atio
n
an
d
m
ai
n
tain
s
th
e
p
a
r
allel
ar
ch
itectu
r
e
o
f
th
e
n
er
v
e
ce
ll
n
eu
r
o
n
s
in
a
lay
er
,
FP
GA
is
an
ap
p
r
o
p
r
iate
p
iece
o
f
h
ar
d
war
e
f
o
r
n
e
u
r
al
n
etwo
r
k
im
p
lem
e
n
t
atio
n
.
2.
RE
L
AT
E
D
WO
RK
T
h
r
ee
c
o
m
p
u
tatio
n
al
tr
aits
c
o
m
m
o
n
l
y
lin
k
e
d
with
ANNs
in
clu
d
e
p
ar
allelis
m
,
m
o
d
u
l
ar
ity
,
an
d
d
y
n
am
ic
ad
a
p
tatio
n
.
B
ec
au
s
e
FP
G
A
-
b
ased
r
ec
o
n
f
ig
u
r
a
b
le
co
m
p
u
tin
g
ar
ch
itectu
r
es
allo
w
f
o
r
r
ap
id
r
ec
o
n
f
ig
u
r
atio
n
an
d
co
n
cu
r
r
en
cy
to
ad
ju
s
t
an
ANN
'
s
weig
h
t
s
an
d
to
p
o
lo
g
y
,
th
e
y
ar
e
an
ex
ce
llen
t
ch
o
ice
f
o
r
im
p
lem
en
tin
g
ANNs.
L
ar
g
e
-
s
ca
le
n
eu
r
al
n
etwo
r
k
FP
GA
r
ea
lizatio
n
r
em
ain
s
a
d
if
f
ic
u
lt
task
d
u
e
to
t
h
e
‘
m
u
ltip
licatio
n
-
r
ich
’
n
atu
r
e
o
f
ANN
alg
o
r
ith
m
s
an
d
th
eir
co
m
p
ar
ativ
el
y
h
ig
h
im
p
lem
en
tatio
n
co
s
ts
.
Nu
m
er
o
u
s
s
tu
d
ies
h
av
e
b
ee
n
p
u
b
lis
h
ed
i
n
th
is
f
ield
,
in
cl
u
d
in
g
m
u
lti
-
ch
ip
r
ea
lizatio
n
[
9
]
,
NNs
with
ce
r
tain
lim
itatio
n
s
to
o
b
tain
f
aster
s
p
ee
d
o
f
o
p
er
atio
n
at
a
ch
ea
p
er
co
s
t
[
1
0
]
,
an
d
n
ew
NN
m
u
ltip
licatio
n
alg
o
r
ith
m
s
[
1
1
]
.
T
h
e
ass
ess
m
en
t
o
f
E
E
G
s
ig
n
als
b
y
in
co
r
p
o
r
atin
g
th
e
f
ee
d
-
f
o
r
war
d
m
u
ltil
ay
er
ANN
an
d
FP
GA
u
tili
zin
g
VHDL
lan
g
u
ag
e
in
th
e
tim
e
-
f
r
eq
u
e
n
cy
r
ep
r
esen
tatio
n
s
f
o
r
th
e
d
o
m
ain
an
aly
s
is
h
as
b
ee
n
u
s
ed
to
d
iag
n
o
s
e
ep
ilep
s
y
,
a
n
eu
r
o
lo
g
ical
c
o
n
d
itio
n
[
1
2
]
.
T
h
e
cu
r
r
e
n
t
r
eq
u
ir
em
en
t
f
o
r
n
eu
r
o
m
o
r
p
h
ic
ch
ip
s
in
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
th
at
r
eq
u
ir
e
th
e
b
est
h
ar
d
war
e
an
d
m
em
o
r
y
is
wh
at
d
r
o
v
e
th
e
d
ev
elo
p
m
en
t
o
f
ANN
h
ar
d
war
e
ch
ip
s
.
T
h
e
ar
ch
itectu
r
e
o
f
th
e
d
ee
p
lear
n
in
g
-
b
ased
ANN
is
in
ten
d
e
d
to
d
eliv
er
th
e
b
est
p
er
f
o
r
m
an
ce
[
1
3
]
.
Ar
ch
itectu
r
e
f
o
r
u
s
in
g
ANNs
o
n
FP
GAs.
T
h
e
f
r
am
ewo
r
k
f
r
ee
s
th
e
u
s
er
f
r
o
m
h
a
n
d
lin
g
lo
w
-
lev
el
h
ar
d
war
e
d
etails,
allo
win
g
f
o
r
th
e
q
u
ick
d
ep
lo
y
m
e
n
t o
f
alr
ea
d
y
tr
ain
e
d
ANNs o
n
FP
GA
s
y
s
tem
s
[
1
4
]
.
T
h
er
e
is
o
n
ly
o
n
e
ac
tiv
atio
n
o
p
er
atio
n
al
b
lo
ck
i
n
co
r
p
o
r
ated
in
th
e
n
ew
h
ar
d
w
ar
e
ar
ch
itectu
r
e
f
o
r
f
ee
d
-
f
o
r
w
ar
d
n
e
u
r
al
n
etwo
r
k
(
FF
NN)
,
th
at
is
ex
ter
n
al
to
th
e
wh
o
le
n
etwo
r
k
.
T
o
m
in
im
iz
e
co
m
p
lex
ity
an
d
ar
ea
u
s
ag
e.
I
n
ter
m
ed
iate
r
esu
lts
ar
e
ea
s
ily
ac
ce
s
s
ib
le
b
ec
au
s
e
th
e
s
in
g
le
ac
tiv
atio
n
b
lo
ck
is
n
o
t
p
ar
t
o
f
th
e
n
etwo
r
k
[
1
5
]
.
Ad
d
itio
n
ally
,
o
n
ly
o
n
e
h
id
d
en
lay
er
is
u
s
ed
in
t
h
is
in
s
tan
ce
;
all
o
th
er
lay
er
s
o
u
g
h
t
to
m
ak
e
u
s
e
o
f
t
h
e
s
am
e
h
ar
d
war
e.
T
h
e
la
y
er
th
at
is
p
u
t
in
to
p
r
ac
tice
will
b
e
th
e
b
ig
g
est
lay
er
with
th
e
m
o
s
t
n
eu
r
o
n
s
[
1
6
]
.
T
h
e
ar
c
h
itectu
r
e
is
o
f
ten
a
s
y
s
to
lic
ar
r
ay
.
I
t
is
a
u
n
if
o
r
m
n
etwo
r
k
o
f
lin
k
e
d
d
ata
p
r
o
ce
s
s
in
g
u
n
its
(
DPU)
k
n
o
wn
as
ce
lls
o
r
n
o
d
es
[
1
7
]
.
T
h
e
id
ea
b
eh
in
d
ANNs
o
r
ig
in
ated
with
b
r
ain
c
o
n
ce
p
ts
th
a
t
wer
e
m
o
d
if
ied
f
o
r
u
s
e
with
d
ig
ital
co
m
p
u
ter
s
.
ANNs
wer
e
f
ir
s
t
d
ev
elo
p
ed
a
s
m
ath
em
atica
lly
m
o
d
eled
b
r
ain
n
eu
r
o
n
s
[
1
8
]
.
T
h
ese
s
tu
d
i
es
d
em
o
n
s
tr
ate
t
h
at
ev
er
y
n
e
u
r
o
n
in
ANNs
r
ec
ei
v
es
in
f
o
r
m
atio
n
as
in
p
u
t
f
r
o
m
eith
er
ex
ter
n
al
o
r
o
th
er
n
e
u
r
o
n
s
.
T
h
is
d
ata
is
d
is
tr
ib
u
ted
as
an
o
u
t
p
u
t
th
at
i
s
ap
p
lied
as
a
n
o
n
-
lin
ea
r
f
u
n
ctio
n
an
d
ca
lcu
late
d
as
th
e
weig
h
ted
to
tal
o
f
th
e
in
p
u
ts
[
1
9
]
.
T
h
e
b
est
o
p
tio
n
f
o
r
FF
NN
r
ea
lizatio
n
is
h
ar
d
war
e
f
o
r
FP
GA
im
p
lem
en
tati
o
n
s
,
b
u
t
th
er
e
ar
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
I
SS
N:
2722
-
2
5
8
6
A
n
a
lysi
s
o
f sin
g
le
la
ye
r
a
r
tifi
c
ia
l n
eu
r
a
l n
etw
o
r
k
n
eu
r
o
mo
r
p
h
ic
h
a
r
d
w
a
r
e
ch
ip
(
A
r
u
n
a
P
a
n
t
)
497
n
u
m
er
o
u
s
o
b
s
tacle
s
to
o
v
er
co
m
e
[
2
0
]
.
Activ
atio
n
f
u
n
ctio
n
im
p
lem
en
tatio
n
,
m
atr
ix
m
u
ltip
licatio
n
im
p
lem
en
tatio
n
,
weig
h
t
m
atr
ix
s
to
r
ag
e,
p
ar
tial
r
esu
lts
,
c
o
ef
f
icien
ts
,
etc.
ar
e
im
p
o
r
ta
n
t
f
o
r
s
elec
tin
g
an
ap
p
r
o
p
r
iate
n
u
m
b
e
r
f
o
r
m
at
to
o
p
tim
ize
ar
ea
,
p
o
wer
,
an
d
s
p
e
ed
[
2
1
]
.
Neu
r
al
n
etwo
r
k
s
h
a
v
e
m
an
y
ap
p
licatio
n
s
in
v
er
y
lar
g
e
-
s
ca
le
in
teg
r
atio
n
(
VL
SI)
d
esig
n
,
o
n
e
o
f
wh
i
ch
is
lear
n
in
g
h
o
w
to
ca
teg
o
r
ize
b
r
ain
b
eh
a
v
io
r
p
atter
n
s
.
Fu
r
th
er
m
o
r
e,
it
is
in
s
itu
atio
n
s
wh
er
e
th
er
e
is
a
s
t
r
o
n
g
c
o
r
r
elatio
n
[
2
2
]
.
I
t
is
illu
s
tr
ated
with
VL
SI
n
etwo
r
k
s
o
f
s
p
ik
e
n
eu
r
o
n
s
an
d
a
m
u
ltid
im
en
s
io
n
al
p
atter
n
o
f
v
ar
y
in
g
s
ac
k
in
g
r
ates
[
2
3
]
.
T
h
e
r
o
b
u
s
t
s
p
ik
e
-
d
r
iv
en
f
lex
i
b
ilit
y
ap
p
r
o
ac
h
is
u
s
ed
to
im
p
lem
en
t
d
y
n
am
ic
s
y
n
ap
s
es
[
2
4
]
.
Mo
r
eo
v
e
r
,
th
e
i
n
p
u
t
p
atter
n
to
th
e
o
u
tp
u
t
n
eu
r
o
n
,
t
h
e
in
p
u
t
s
ig
n
a
l
will
o
f
f
er
an
ex
tr
a
co
n
tr
ib
u
tio
n
s
p
ik
e
tr
ai
n
d
u
r
in
g
tr
ain
in
g
.
I
t
is
em
p
lo
y
e
d
to
r
esp
o
n
d
to
th
e
i
n
p
u
t
p
r
o
to
ty
p
e
at
a
m
o
d
est
r
ate
o
r
with
a
h
ig
h
ch
ar
g
e
[
2
5
]
.
Sil
ico
n
n
eu
r
o
n
s
wer
e
im
p
lem
en
ted
b
ased
o
n
th
e
ap
p
licatio
n
n
e
ed
s
o
f
a
co
llectio
n
o
f
s
o
lu
t
io
n
s
.
Dif
f
er
en
t
c
o
m
p
u
tatio
n
a
l
m
o
d
els
ca
n
b
e
im
p
lem
en
ted
u
s
in
g
th
e
wid
e
r
an
g
e
o
f
n
e
u
r
o
m
o
r
p
h
ic
s
ilico
n
cir
cu
its
av
ailab
le
[
2
6
]
.
T
h
e
p
r
ed
ictio
n
o
f
th
e
o
v
er
all
n
u
m
b
er
o
f
to
u
r
is
ts
ar
r
iv
in
g
in
Sp
ain
was m
ad
e
u
s
in
g
n
eu
r
al
s
in
g
le
-
in
p
u
t sin
g
le
-
o
u
tp
u
t n
eu
r
al
n
etwo
r
k
s
[
2
7
]
an
d
m
u
ltip
le
-
in
p
u
t
m
u
ltip
le
-
o
u
tp
u
t
n
eu
r
al
n
etwo
r
k
s
.
A
n
ANN
r
ec
eiv
es
a
lar
g
e
n
u
m
b
er
o
f
s
ig
n
al
in
p
u
ts
f
r
o
m
th
e
d
is
tin
ct
d
ata
c
o
llectio
n
o
r
o
u
tp
u
t
o
f
p
r
ev
io
u
s
ly
ass
o
ciate
d
n
e
u
r
o
n
s
.
E
v
er
y
i
n
p
u
t
e
n
ter
s
th
r
o
u
g
h
a
lin
k
k
n
o
wn
as
a
s
y
n
ap
s
e,
wh
ich
h
as
a
m
as
s
[
2
8
]
.
A
s
ca
lab
le
A
NN
ch
ip
ca
n
b
e
cr
ea
ted
with
f
ea
tu
r
es
lik
e
q
u
ick
r
esp
o
n
s
e
tim
es,
lo
w
co
s
ts
,
lo
w
p
o
wer
co
n
s
u
m
p
tio
n
,
a
n
d
t
h
e
ab
ilit
y
to
in
teg
r
ate
with
F
PGA
an
d
wo
r
k
with
em
b
ed
d
e
d
cir
cu
its
.
T
o
im
p
le
m
en
t
th
e
ANN,
it
is
n
ec
ess
a
r
y
to
en
s
u
r
e
t
h
at
th
er
e
ar
e
e
n
o
u
g
h
FP
GA
lo
g
ic
elem
en
ts
(
L
E
s
)
o
r
lo
o
k
u
p
ta
b
les
(
L
UT
s
)
av
ailab
le.
T
h
is
is
cr
u
cial
f
o
r
h
an
d
lin
g
th
e
r
eq
u
ir
ed
n
u
m
b
er
o
f
n
eu
r
o
n
s
an
d
th
eir
c
o
n
n
ec
tio
n
s
.
E
f
f
icien
t
u
tili
za
tio
n
o
f
r
e
g
is
ter
s
o
r
f
lip
-
f
lo
p
s
is
cr
u
cial
f
o
r
th
e
s
to
r
ag
e
o
f
weig
h
ts
an
d
in
ter
m
ed
iate
v
alu
es.
I
t
is
r
eq
u
ir
e
d
to
ca
lc
u
late
th
e
n
ec
ess
ar
y
th
r
o
u
g
h
p
u
t
t
o
d
eter
m
in
e
th
e
clo
c
k
f
r
eq
u
e
n
cy
f
o
r
y
o
u
r
d
esig
n
.
F
o
r
y
o
u
r
wh
o
le
d
esig
n
,
m
a
k
e
s
u
r
e
th
at
th
e
s
etu
p
an
d
h
o
ld
t
im
es
ar
e
with
in
th
e
allo
wed
r
an
g
e.
3.
M
E
T
H
O
DS
Neu
r
o
m
o
r
p
h
ic
s
y
s
tem
s
n
ec
ess
itate
u
n
iq
u
e
n
etwo
r
k
to
p
o
lo
g
ies.
T
h
e
m
o
s
t
co
m
m
o
n
k
in
d
o
f
n
etwo
r
k
m
o
d
el
im
p
lem
en
tatio
n
is
th
e
f
ee
d
-
f
o
r
war
d
n
e
u
r
al
n
etwo
r
k
,
wh
ich
is
em
p
lo
y
ed
in
m
u
ltil
ay
er
s
en
s
in
g
.
I
n
ad
d
itio
n
to
t
h
ese,
th
er
e
ar
e
m
a
n
y
o
th
er
ty
p
es
o
f
n
eu
r
al
n
etw
o
r
k
to
p
o
l
o
g
ies,
s
u
ch
as
a
r
tific
ial
n
eu
r
al
n
etwo
r
k
s
,
ce
llu
lar
n
eu
r
al
n
etwo
r
k
s
,
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
,
ce
llu
lar
au
to
m
ata,
s
p
ik
i
n
g
n
eu
r
al
n
etwo
r
k
s
,
ar
tific
ial
n
eu
r
al
n
etwo
r
k
s
,
an
d
n
e
u
r
al
n
etwo
r
k
s
th
at
ar
e
p
u
ls
e
-
co
u
p
led
.
I
n
k
ee
p
in
g
with
th
e
cu
r
r
en
t
tr
en
d
in
n
eu
r
al
n
etwo
r
k
r
esear
ch
,
th
e
h
o
p
-
f
iel
d
n
etwo
r
k
,
a
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
(
R
NN)
n
etwo
r
k
d
esi
g
n
,
was
p
ar
ticu
lar
ly
p
o
p
u
lar
in
ea
r
ly
n
e
u
r
al
m
o
r
p
h
o
lo
g
y
im
p
lem
en
tatio
n
s
.
Mo
r
e
m
o
d
er
n
im
p
lem
e
n
tatio
n
s
o
f
th
is
ar
ch
itectu
r
e
ex
is
t.
Fo
r
ex
am
p
le,
d
ata
ca
teg
o
r
izatio
n
,
f
au
lt
d
etec
tio
n
,
g
r
ap
h
p
ar
titi
o
n
,
etc.
wo
r
k
to
g
eth
er
to
s
o
lv
e
r
ea
l
-
wo
r
ld
is
s
u
es.
Ar
tific
ial
n
eu
r
al
n
etw
o
r
k
s
,
o
r
ANNs,
ar
e
co
m
p
u
ter
p
r
o
g
r
am
s
th
at
d
r
aw
s
tim
u
latio
n
f
r
o
m
th
e
h
u
m
an
b
r
ain
[
2
9
]
.
T
h
ey
ca
n
r
eso
lv
e
e
x
tr
em
ely
co
m
p
licated
is
s
u
es
t
h
at
th
e
tr
ad
itio
n
al
p
r
o
ce
s
s
in
g
s
y
s
tem
is
u
n
ab
le
to
h
an
d
le.
B
ec
au
s
e
o
f
its
q
u
ick
p
r
o
ce
s
s
in
g
s
p
ee
d
an
d
ac
cu
r
at
e
o
u
tp
u
t,
ANN
is
u
s
ed
in
m
an
y
d
if
f
er
e
n
t
f
ield
s
.
T
h
er
e
ar
e
alr
ea
d
y
s
ev
er
al
v
a
r
ieties
o
f
ANN
s
.
T
h
e
m
o
s
t
p
o
p
u
lar
an
d
s
tr
aig
h
tf
o
r
war
d
o
n
e
is
f
ee
d
-
f
o
r
war
d
n
eu
r
al
n
etwo
r
k
s
(
FF
NN)
.
Neu
r
al
n
etwo
r
k
s
an
d
n
eu
r
o
m
o
r
p
h
i
c
n
etwo
r
k
s
ar
e
th
e
two
way
s
to
d
o
ANN.
Neu
r
al
n
etwo
r
k
s
ar
e
s
im
p
ly
a
s
o
f
twar
e
ap
p
r
o
ac
h
th
at
r
u
n
s
o
n
PC
-
b
ased
p
latf
o
r
m
s
.
Du
e
to
its
in
ab
ilit
y
to
m
an
ag
e
th
e
r
eq
u
ir
ed
th
r
o
u
g
h
p
u
t,
a
PC
-
b
ased
s
y
s
tem
is
u
n
f
it f
o
r
h
ig
h
-
s
p
ee
d
p
r
o
ce
s
s
in
g
,
p
r
ed
ictio
n
,
d
ec
is
io
n
-
m
ak
in
g
,
o
r
ca
te
g
o
r
iz
atio
n
.
T
h
er
ef
o
r
e,
n
eu
r
o
m
o
r
p
h
ic
d
esig
n
o
r
a
n
o
th
e
r
h
a
r
d
war
e
ap
p
r
o
ac
h
will
b
e
th
e
way
to
g
o
f
o
r
ANN
im
p
le
m
en
tatio
n
[
3
0
]
.
C
en
tr
al
p
r
o
ce
s
s
in
g
u
n
its
(
C
P
Us)
an
d
g
r
ap
h
i
cs
p
r
o
ce
s
s
in
g
u
n
its
(
GPUs
)
ar
e
id
ea
l
f
o
r
th
is
task
d
u
e
t
o
th
eir
lig
h
tn
in
g
-
f
ast
d
a
ta
p
r
o
ce
s
s
in
g
ca
p
ab
ilit
ies;
n
ev
er
th
eless
,
th
ey
a
r
e
r
elativ
ely
p
o
wer
f
u
l
wh
en
c
o
m
p
ar
ed
t
o
FP
GAs
an
d
o
th
er
h
ar
d
war
e
p
latf
o
r
m
s
.
Fig
u
r
e
1
p
r
esen
ts
th
e
ar
ch
itectu
r
e
o
f
th
e
s
in
g
le
-
lay
e
r
f
ee
d
-
f
o
r
war
d
n
eu
r
al
n
etwo
r
k
.
Fig
u
r
e
2
p
r
esen
ts
th
e
f
u
n
ct
io
n
al
b
lo
ck
s
o
f
t
h
e
s
in
g
le
-
s
tag
e
n
eu
r
o
n
n
etwo
r
k
.
W
e
d
esig
n
ed
a
s
y
s
tem
u
s
in
g
VHDL
m
o
d
elin
g
,
wh
ich
in
clu
d
es
a
s
in
g
le
-
s
tag
e
n
eu
r
o
n
ch
ip
.
W
e
d
esig
n
e
d
th
r
ee
p
r
im
a
r
y
u
n
its
:
th
e
m
u
ltip
licatio
n
u
n
it,
t
h
e
co
n
tr
o
l
u
n
it,
an
d
a
m
u
ltip
lex
e
r
[
3
1
]
.
T
h
e
n
e
u
r
o
m
o
r
p
h
ic
ch
ip
wh
ich
h
ad
b
ee
n
d
esig
n
ed
ca
n
b
e
d
iv
id
e
d
m
ain
ly
i
n
to
th
r
ee
p
ar
ts
.
−
Mu
ltip
lier
u
n
it:
T
h
e
m
u
ltip
lier
th
at
was
u
s
ed
to
co
n
f
ig
u
r
e
m
u
ltip
lies
2
0
n
eu
r
o
n
in
p
u
ts
with
2
0
n
eu
r
o
n
weig
h
ts
ea
ch
h
av
in
g
8
-
b
it
in
f
o
r
m
atio
n
a
n
d
8
-
b
it
weig
h
ts
a
n
d
g
iv
es
o
n
e
s
in
g
le
o
u
tp
u
t
w
h
ich
is
1
6
b
its
.
So
,
af
ter
t
h
e
m
u
ltip
licatio
n
o
f
n
eu
r
o
n
weig
h
t,
we
g
et
th
e
v
al
u
es,
M
0
,
M
1
,
M
2
,
M
2
,
M
3
,
M
4
,
M
5
,
M
6
,
M
7
,
…
M
19
.
−
C
o
n
tr
o
l
u
n
it:
I
n
th
e
co
n
tr
o
l
l
o
g
ic
we
h
av
e
th
e
in
p
u
ts
f
r
o
m
M
0
,
M
1
,
M
2
,
M
2
,
M
3
,
M
4
,
M
5
,
M
6
,
M
7
,
.
.
.
M
19
,
o
n
e
s
o
u
r
ce
a
d
d
r
ess
o
f
5
b
it
s
,
a
n
d
th
e
o
u
t
p
u
ts
ar
e
N
0
,
N
1
,
N
2
,
N
3
,
N
4
,
N
5
,
N
6
,
N
7
,
…
,
N
19
.
T
h
e
d
esti
n
atio
n
ad
d
r
ess
is
also
5
b
its
.
T
h
er
ef
o
r
e,
b
ased
o
n
th
e
s
o
u
r
ce
ad
d
r
ess
an
d
d
esti
n
atio
n
ad
d
r
ess
th
e
o
u
tp
u
t
o
f
th
e
m
u
ltip
lier
is
f
ed
t
o
ea
ch
n
o
d
e
o
f
th
e
h
id
d
en
lay
e
r
.
−
Mu
ltip
lex
er
u
n
it:
Fro
m
th
e
co
n
tr
o
l
u
n
it
we
g
et
2
0
o
u
tp
u
ts
wh
ich
ar
e
f
r
o
m
N
0
,
N
1
,
N
2
,
N
3
,
N
4
,
…
,
N
19
.
I
n
th
e
m
u
ltip
lex
er
u
n
it,
we
h
av
e
2
0
in
p
u
ts
.
E
ac
h
in
p
u
t
is
1
6
b
its
an
d
o
n
e
s
elec
tio
n
lin
e
wh
ich
is
5
b
its
an
d
b
ased
u
p
o
n
th
e
s
elec
tio
n
lin
e
t
h
e
o
u
tp
u
t o
f
t
h
e
m
u
ltip
lex
e
r
u
n
it we
g
et.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
2
5
8
6
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
,
Vo
l
.
13
,
No
.
4
,
Dec
em
b
er
20
24
:
4
9
5
-
5
0
5
498
T
o
ac
h
iev
e
p
ar
allelis
m
,
it
is
n
ec
es
s
ar
y
to
ex
ec
u
te
th
e
co
m
p
u
tatio
n
s
o
f
s
ev
er
a
l
n
eu
r
o
n
s
s
im
u
ltan
eo
u
s
ly
,
if
th
e
av
ailab
le
FP
GA
r
e
s
o
u
r
ce
s
allo
w
f
o
r
it.
Pip
elin
in
g
im
p
r
o
v
es
th
e
n
e
two
r
k
'
s
th
r
o
u
g
h
p
u
t
b
y
b
r
ea
k
in
g
d
o
w
n
th
e
p
r
o
ce
s
s
in
g
in
to
s
m
aller
p
ar
ts
.
T
o
p
r
o
v
id
e
p
r
o
p
er
m
an
a
g
em
en
t
a
n
d
s
y
n
ch
r
o
n
izatio
n
o
f
th
e
in
p
u
t/o
u
t
p
u
t
(
I
/O)
p
i
n
s
,
p
ar
ticu
lar
ly
wh
en
d
ea
lin
g
with
m
u
ltip
le
clo
ck
d
o
m
ai
n
s
,
it
is
ess
en
tial
to
en
s
u
r
e
th
at
all
co
m
p
o
n
e
n
ts
o
f
th
e
d
e
s
ig
n
ar
e
s
y
n
ch
r
o
n
ized
with
ea
ch
o
th
e
r
.
T
o
m
in
im
ize
p
o
wer
co
n
s
u
m
p
tio
n
,
it
is
n
ec
ess
ar
y
to
o
p
tim
ize
th
e
d
esig
n
b
y
r
ed
u
cin
g
s
witch
in
g
ac
tiv
ity
an
d
im
p
lem
e
n
tin
g
lo
w
-
p
o
wer
d
esig
n
s
o
lu
tio
n
s
.
B
ased
o
n
th
e
s
im
u
la
tio
n
,
th
e
d
esig
n
is
co
n
f
ir
m
e
d
t
o
h
av
e
d
ata
r
ates
f
o
r
t
h
e
ANN
in
p
u
ts
an
d
o
u
tp
u
ts
th
at
ar
e
co
n
s
is
ten
t
with
th
e
in
p
u
t/o
u
tp
u
t
b
an
d
wid
th
o
f
th
e
FP
GA.
T
h
e
co
n
s
tr
ain
ts
o
f
y
o
u
r
FP
GA
b
o
ar
d
will
d
eter
m
in
e
th
e
allo
ca
tio
n
o
f
p
i
n
s
f
o
r
d
ata
i
n
p
u
ts
,
o
u
tp
u
ts
,
a
n
d
co
n
t
r
o
l
s
ig
n
als.
T
h
e
d
esig
n
is
v
er
if
ied
o
n
th
e
FP
GA,
u
tili
zin
g
in
-
cir
cu
it
te
s
tin
g
an
d
d
eb
u
g
g
in
g
tech
n
i
q
u
es,
s
u
ch
as
test
b
en
c
h
es
a
n
d
in
ter
n
al
s
ig
n
al
m
o
n
ito
r
in
g
.
T
h
e
ANN
is
s
tr
u
ctu
r
ed
i
n
a
m
o
d
u
lar
f
ash
io
n
,
en
ab
lin
g
ea
s
y
e
x
p
an
s
io
n
an
d
m
o
d
if
icatio
n
t
o
ac
co
m
m
o
d
ate
m
o
r
e
in
t
r
icate
n
etwo
r
k
s
o
r
d
iv
er
s
e
u
s
es
in
th
e
f
u
t
u
r
e.
T
h
is
f
u
n
ctio
n
a
lity
allo
ws
y
o
u
t
o
cu
s
to
m
ize
th
e
ANN
to
s
u
it
d
if
f
er
en
t
d
atasets
an
d
s
ce
n
ar
io
s
b
y
m
o
d
if
y
in
g
p
ar
am
eter
s
s
u
c
h
as
th
e
n
u
m
b
e
r
o
f
n
eu
r
o
n
s
o
r
lay
er
s
.
Fig
u
r
e
1
.
Ar
c
h
itectu
r
e
o
f
s
in
g
l
e
-
lay
er
f
ee
d
-
f
o
r
war
d
n
e
u
r
al
n
e
two
r
k
Fig
u
r
e
2
.
Fu
n
ctio
n
al
b
lo
c
k
s
o
f
s
in
g
le
-
s
tag
e
n
eu
r
o
n
n
etwo
r
k
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
h
a
r
d
war
e
n
eu
r
o
m
o
r
p
h
ic
ch
ip
f
o
r
2
0
n
e
u
r
o
n
s
in
p
u
t
is
d
esig
n
ed
a
n
d
ea
ch
n
eu
r
o
n
is
8
-
b
it
in
p
u
t
wh
ich
is
g
iv
en
f
ir
s
t to
th
e
m
u
l
tip
lier
.
T
h
e
n
eu
r
o
n
in
p
u
t w
h
ic
h
we
h
av
e
g
iv
en
is
f
r
o
m
X
0
, X
1
, X
2
to
X
19
an
d
th
e
weig
h
ts
wh
ich
we
h
av
e
u
s
ed
h
er
e
ar
e
W
0
,
W
1
,
W
2
,
W
3
to
W
19
.
T
h
e
weig
h
ts
with
th
e
in
p
u
t
ar
e
f
ir
s
t
m
u
ltip
lied
b
y
th
e
m
u
ltip
lier
.
I
t
is
im
p
o
r
tan
t
to
o
p
tim
ize
tim
in
g
an
d
d
ata
p
ath
way
s
,
p
r
o
v
id
e
ad
eq
u
a
te
s
y
n
ch
r
o
n
izatio
n
,
m
an
ag
e
FP
GA
r
eso
u
r
ce
s
ef
f
icien
tly
,
co
n
s
id
er
p
o
wer
co
n
s
u
m
p
tio
n
an
d
I
/O
lim
its
,
an
d
d
e
s
ig
n
a
s
in
g
le
-
lay
er
ANN
in
HDL
f
o
r
FP
GA.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
I
SS
N:
2722
-
2
5
8
6
A
n
a
lysi
s
o
f sin
g
le
la
ye
r
a
r
tifi
c
ia
l n
eu
r
a
l n
etw
o
r
k
n
eu
r
o
mo
r
p
h
ic
h
a
r
d
w
a
r
e
ch
ip
(
A
r
u
n
a
P
a
n
t
)
499
Fig
u
r
e
3
.
Sin
g
le
s
tag
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
c
h
ip
r
e
g
is
ter
tr
an
s
is
to
r
lo
g
ic
(
R
T
L
)
T
h
e
p
r
o
d
u
cts
we
g
et
ar
e
M
0
,
M
1
,
M
2
to
M
19
.
wh
ich
a
r
e
th
e
i
n
p
u
ts
o
f
th
e
co
n
tr
o
l
l
o
g
ic,
a
n
d
th
e
o
u
t
p
u
ts
o
f
th
e
co
n
tr
o
l
l
o
g
ic
a
r
e
No
,
N
1
,
N
2
to
N
19
.
T
h
e
co
n
tr
o
l
lo
g
i
c
is
u
s
ed
t
o
f
ee
d
o
n
e
n
eu
r
o
n
i
n
p
u
t
t
o
ea
c
h
o
f
th
e
in
p
u
ts
in
th
e
h
id
d
e
n
lay
er
.
T
h
e
o
u
tp
u
t
o
f
t
h
e
co
n
t
r
o
l
lo
g
ic
is
g
iv
en
to
M
3
th
e
m
u
ltip
lex
e
r
wh
ich
m
u
x
all
th
e
in
p
u
ts
an
d
g
iv
es
a
s
in
g
le
o
u
tp
u
t
Z
0
.
T
h
e
m
u
ltip
lex
er
in
p
u
ts
ar
e
f
r
o
m
D
0
to
D
19
,
an
d
it
h
as
o
n
e
s
elec
tio
n
lin
e
wh
ich
is
o
f
5
b
its
,
an
d
o
n
e
s
in
g
le
o
u
tp
u
t
Z
0
w
h
ich
is
o
f
1
6
b
i
ts
.
T
h
e
in
p
u
ts
ar
e
p
r
o
ce
s
s
ed
as
Neu
r
o
n
_
X0
_
in
p
u
t
(
7
:0
)
to
Ne
u
r
o
n
_
X1
9
_
in
p
u
t
(
7
:0
)
p
r
esen
tin
g
th
e
2
0
n
eu
r
o
n
s
i
n
p
u
ts
,
t
h
e
c
o
r
r
esp
o
n
d
in
g
weig
h
ts
ar
e
Neu
r
o
n
_
weig
h
t_
w0
(
7
:0
)
to
N
eu
r
o
n
_
weig
h
t_
w1
9
(
7
:0
)
.
T
h
e
p
o
r
t
a
d
d
r
ess
(
4
:
0
)
d
en
o
tes
lay
er
ad
d
r
ess
wh
ich
is
lim
ited
to
s
in
g
le
lay
er
h
e
r
e
.
T
h
e
n
eu
r
o
n
_
o
u
tp
u
t(
1
5
:
0
)
is
th
e
1
6
-
b
it
o
u
tp
u
t
o
f
t
h
e
n
eu
r
a
l
n
etwo
r
k
.
Fig
u
r
e
3
p
r
esen
ts
th
e
R
T
L
o
f
a
s
tag
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
c
h
ip
.
Fig
u
r
e
4
r
e
p
r
esen
ts
th
e
s
ch
em
atic
o
f
s
in
g
le
-
s
tag
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
c
h
ip
s
,
r
esp
ec
tiv
ely
.
T
h
e
u
s
e
o
f
R
T
L
in
a
VHDL
p
r
o
g
r
am
m
in
g
lan
g
u
ag
e
is
cr
u
cial
to
d
ev
elo
p
in
g
a
s
in
g
le
-
lay
er
AN
N
th
at
o
p
tim
izes
s
p
ee
d
,
r
eso
u
r
ce
co
n
s
u
m
p
tio
n
,
an
d
p
o
wer
ef
f
icien
cy
,
wh
ile
also
en
s
u
r
in
g
d
eter
m
i
n
is
tic
p
er
f
o
r
m
an
ce
,
s
ca
lab
ilit
y
,
an
d
f
lex
ib
i
lity
.
T
o
estab
lis
h
a
s
ca
lab
le
A
NN
,
it
i
s
im
p
er
ativ
e
to
co
n
s
tr
u
ct
an
ar
ch
itectu
r
e
th
at
ca
n
e
f
f
ec
tiv
ely
h
an
d
le
h
ig
h
-
d
im
en
s
io
n
al,
n
o
is
y
,
an
d
tem
p
o
r
al
d
ata.
T
o
cr
ea
t
e
a
v
er
s
atile
n
etwo
r
k
f
o
r
E
E
G
ap
p
licatio
n
s
,
f
o
cu
s
o
n
in
c
o
r
p
o
r
ati
n
g
m
o
d
u
lar
ity
,
o
p
tim
izin
g
m
em
o
r
y
u
s
e,
im
p
lem
en
tin
g
ef
f
ec
tiv
e
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es,
an
d
em
p
lo
y
in
g
r
o
b
u
s
t
tr
ain
in
g
m
eth
o
d
s
.
T
o
en
h
a
n
ce
th
e
s
y
s
tem
'
s
p
er
f
o
r
m
an
ce
a
n
d
u
s
er
-
f
r
ien
d
lin
ess
,
it
is
ad
v
i
s
ab
le
to
co
n
s
id
er
th
e
u
tili
za
tio
n
o
f
h
ar
d
war
e
ac
ce
ler
atio
n
,
as
well
as
th
e
as
p
ec
ts
o
f
in
te
r
p
r
etab
ilit
y
an
d
r
ea
l
-
wo
r
ld
v
alid
atio
n
.
ANNs
in
E
E
G
a
p
p
licatio
n
s
ca
n
r
ec
eiv
e
a
d
iv
er
s
e
r
an
g
e
o
f
in
p
u
ts
,
wh
ich
ca
n
b
e
co
m
p
lex
an
d
v
ar
ied
.
T
h
ese
in
p
u
ts
in
clu
d
e
r
aw
tim
e
-
s
er
ies
s
ig
n
als,
p
r
e
-
p
r
o
ce
s
s
ed
an
d
f
ilter
ed
d
ata,
an
d
ex
tr
ac
ted
f
ea
tu
r
es.
T
o
p
r
o
p
er
ly
u
tili
ze
E
E
G
d
ata
in
ANN,
it
is
cr
u
cial
to
p
er
f
o
r
m
ap
p
r
o
p
r
iate
p
r
e
p
r
o
ce
s
s
in
g
,
f
ea
tu
r
e
ex
tr
ac
tio
n
,
an
d
d
ata
s
tr
u
ctu
r
e.
T
o
ac
h
iev
e
th
e
b
est
p
er
f
o
r
m
an
ce
an
d
p
r
ec
is
e
o
u
tco
m
es,
it
is
ad
v
is
ab
le
to
ad
ju
s
t
th
e
in
p
u
t
r
ep
r
esen
tatio
n
b
ased
o
n
th
e
s
p
ec
if
ic
p
r
o
b
lem
a
n
d
n
eu
r
al
n
etwo
r
k
s
t
r
u
ctu
r
e.
Fig
u
r
es
5
an
d
6
r
ep
r
esen
t
th
e
s
im
u
latio
n
wav
ef
o
r
m
f
o
r
s
in
g
l
e
s
tag
e
2
0
in
p
u
ts
an
d
o
u
t
p
u
ts
o
f
ANN
in
b
in
ar
y
u
s
in
g
Mo
d
elSim
.
Fig
u
r
es
7
a
n
d
8
r
ep
r
esen
t
th
e
s
im
u
latio
n
wav
e
f
o
r
m
f
o
r
s
in
g
le
s
tag
e
2
0
in
p
u
ts
an
d
o
u
tp
u
ts
o
f
ANN
in
an
in
teg
er
u
s
in
g
Mo
d
elSim
.
T
ab
les
1
an
d
2
s
h
o
w
th
e
am
o
u
n
t
o
f
h
ar
d
war
e
p
ar
am
eter
u
s
ed
an
d
its
p
er
ce
n
tag
e
u
s
ed
.
T
h
e
d
ev
ice
u
tili
za
tio
n
r
e
p
o
r
t
f
o
r
th
is
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
is
tak
en
f
r
o
m
Xilin
x
s
o
f
twar
e.
T
h
e
n
u
m
b
er
o
f
s
lices
u
s
ed
i
s
3
2
0
in
th
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
wh
ich
is
o
u
t
o
f
a
to
tal
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
2
5
8
6
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
,
Vo
l
.
13
,
No
.
4
,
Dec
em
b
er
20
24
:
4
9
5
-
5
0
5
500
o
f
1
2
4
8
0
.
T
h
e
n
u
m
b
er
o
f
L
UT
s
u
s
ed
is
2
3
2
o
u
t
o
f
a
to
tal
o
f
1
2
4
8
0
.
I
n
p
u
t
-
o
u
t
p
u
t b
lo
ck
s
(
I
OB
)
u
s
ed
is
3
5
1
o
u
t
o
f
a
to
tal
o
f
1
7
2
.
T
h
e
to
tal
d
e
lay
u
s
ed
in
th
is
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
is
6
.
3
5
9
n
s
.
T
h
e
m
em
o
r
y
u
tili
ze
d
f
o
r
t
h
is
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
wa
r
e
ch
ip
is
4
6
6
5
3
0
4
k
ilo
b
y
tes.
T
h
e
tar
g
et
d
ev
ice
u
s
ed
is
Vir
tex
-
5
FP
GA
with
t
h
e
d
ev
ice
(
x
c5
v
lx
2
0
t
-
2
-
f
f
3
2
3
)
f
o
r
s
im
u
latio
n
a
n
d
s
y
n
th
esis
.
T
h
e
d
esig
n
u
tili
ze
d
a
p
o
wer
o
f
1
2
.
9
0
m
illi
watts
an
d
ex
p
er
ien
ce
d
a
d
elay
o
f
9
.
2
0
n
a
n
o
s
ec
o
n
d
s
.
Fig
u
r
e
9
d
ep
icts
th
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
i
p
u
tili
za
tio
n
ch
ar
t.
Fig
u
r
e
4
.
Sch
em
atic
o
f
s
in
g
le
-
s
tag
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
Evaluation Warning : The document was created with Spire.PDF for Python.
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
I
SS
N:
2722
-
2
5
8
6
A
n
a
lysi
s
o
f sin
g
le
la
ye
r
a
r
tifi
c
ia
l n
eu
r
a
l n
etw
o
r
k
n
eu
r
o
mo
r
p
h
ic
h
a
r
d
w
a
r
e
ch
ip
(
A
r
u
n
a
P
a
n
t
)
501
Fig
u
r
e
5
.
Mo
d
elSim
s
im
u
latio
n
o
f
s
in
g
le
s
tag
e
2
0
in
p
u
ts
ANN
in
b
in
ar
y
Fig
u
r
e
6
.
Mo
d
elSim
s
im
u
latio
n
o
f
th
e
s
in
g
le
-
s
tag
e
o
u
t
p
u
ts
o
f
ANN
in
b
in
ar
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
2
5
8
6
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
,
Vo
l
.
13
,
No
.
4
,
Dec
em
b
er
20
24
:
4
9
5
-
5
0
5
502
Fig
u
r
e
7
.
Mo
d
elSim
s
im
u
latio
n
o
f
s
in
g
le
s
tag
e
2
0
in
p
u
t A
N
N
in
in
teg
er
Fig
u
r
e
8
.
Mo
d
elSim
s
im
u
latio
n
o
f
s
in
g
le
-
s
tag
e
o
u
tp
u
ts
o
f
A
NN
in
in
teg
er
Evaluation Warning : The document was created with Spire.PDF for Python.
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
I
SS
N:
2722
-
2
5
8
6
A
n
a
lysi
s
o
f sin
g
le
la
ye
r
a
r
tifi
c
ia
l n
eu
r
a
l n
etw
o
r
k
n
eu
r
o
mo
r
p
h
ic
h
a
r
d
w
a
r
e
ch
ip
(
A
r
u
n
a
P
a
n
t
)
503
T
ab
le
1
.
Har
d
war
e
p
ar
a
m
eter
s
u
tili
ze
d
b
y
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
N
e
u
r
o
mo
r
p
h
i
c
C
h
i
p
s
i
z
e
M
u
l
t
i
p
l
i
e
r
u
se
d
Ta
r
g
e
t
d
e
v
i
c
e
(
F
P
G
A
)
LU
Ts
S
l
i
c
e
s
D
e
l
a
y
(
n
s)
I
O
B
s
M
e
m
o
r
y
(
K
B
)
ANN
-
20
20
V
i
r
t
e
x
-
5
(
x
c
5
v
l
x
2
0
t
-
2
-
f
f
3
2
3
)
2
3
2
3
2
0
6
.
3
5
9
3
5
1
4
6
6
5
3
0
4
T
ab
le
2
.
Har
d
war
e
p
ar
a
m
eter
s
u
tili
ze
d
b
y
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
H
a
r
d
w
a
r
e
p
a
r
a
me
t
e
r
s
P
a
r
a
me
t
e
r
u
se
d
A
v
a
i
l
a
b
l
e
P
e
r
c
e
n
t
a
g
e
(
%)
o
f
p
a
r
a
m
e
t
e
r
s
u
se
d
S
l
i
c
e
s
3
2
0
1
2
4
8
0
2
R
e
g
i
s
t
e
r
s
2
3
2
1
2
4
8
0
1
I
O
B
s
3
5
1
1
7
2
2
0
4
Fig
u
r
e
9
.
Neu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
u
tili
za
tio
n
c
h
ar
t
5.
CO
NCLU
SI
O
N
T
h
e
ar
c
h
itectu
r
e
an
d
o
p
e
r
atio
n
o
f
th
e
h
u
m
an
b
r
ain
s
er
v
e
as
th
e
in
s
p
ir
atio
n
f
o
r
th
e
d
esig
n
o
f
a
ch
ip
k
n
o
wn
as
a
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
i
p
.
T
h
e
n
eu
r
o
m
o
r
p
h
i
c
h
ar
d
war
e
c
h
ip
f
o
r
20
-
n
eu
r
o
n
in
p
u
t
was
d
esig
n
e
d
s
u
cc
ess
f
u
lly
in
Vir
tex
-
5
FP
G
A
Xilin
x
I
SE
1
4
.
7
.
T
h
e
s
ch
e
m
atic
o
f
th
e
s
in
g
le
-
s
tag
e
n
eu
r
o
m
o
r
p
h
ic
ch
ip
is
ex
tr
ac
ted
s
u
cc
ess
f
u
lly
an
d
th
e
Mo
d
elSim
s
im
u
latio
n
f
o
r
d
i
f
f
er
en
t
test
in
p
u
ts
,
r
esp
ec
tiv
el
y
.
T
h
e
R
T
L
o
f
th
e
s
in
g
le
-
s
tag
e
n
eu
r
o
m
o
r
p
h
ic
ch
i
p
is
d
o
n
e
u
s
in
g
t
h
e
d
etailed
h
ar
d
war
e
p
ar
a
m
eter
s
u
tili
ze
d
b
y
th
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
i
p
.
T
h
e
h
a
r
d
war
e
u
tili
za
tio
n
ch
ar
t
f
o
r
t
h
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
in
cl
u
d
es
th
e
n
u
m
b
er
o
f
L
UT
s
u
tili
ze
d
in
th
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
c
h
ip
,
wh
ic
h
is
2
3
2
,
th
e
Sli
ce
s
u
tili
ze
d
ar
e
3
2
0
,
a
n
d
th
e
I
OB
s
u
s
ed
ar
e
3
5
1
.
T
h
e
t
o
tal
p
ath
d
elay
is
6
.
3
5
9
n
s
,
an
d
th
e
m
e
m
o
r
y
u
tili
ze
d
in
th
e
d
esig
n
o
f
th
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
i
p
is
4
6
,
6
5
3
0
4
KB
.
Har
d
war
e
p
ar
am
eter
s
p
er
ce
n
tag
e
u
tili
ze
d
in
th
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
wa
r
e
ch
i
p
d
esig
n
,
n
u
m
b
er
o
f
L
UT
u
tili
z
ed
in
th
e
n
eu
r
o
m
o
r
p
h
ic
h
ar
d
war
e
ch
ip
wh
ich
is
2
3
2
an
d
a
v
a
ilab
le
ar
e
1
2
4
8
0
s
o
p
er
ce
n
tag
e
o
f
p
ar
am
eter
u
tili
ze
d
in
o
n
ly
1
%.
an
d
th
e
Sli
ce
s
u
tili
ze
d
ar
e
3
2
0
an
d
av
ail
ab
le
ar
e
1
2
4
8
0
,
s
o
p
er
ce
n
tag
e
o
f
p
ar
am
eter
u
tili
z
ed
is
o
n
ly
2
%,
a
n
d
th
e
in
p
u
t
-
o
u
tp
u
t
o
r
I
OB
s
u
s
ed
a
r
e
3
5
1
a
n
d
av
ailab
le
a
r
e
1
7
2
wh
ich
is
2
0
4
% o
f
th
e
wh
o
le.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
an
k
s
to
th
e
UPES
VL
SI
d
esig
n
an
d
DSP L
ab
to
p
r
o
v
id
e
th
e
p
latf
o
r
m
f
o
r
ca
r
r
y
in
g
o
u
t th
e
wo
r
k
.
RE
F
E
R
E
NC
E
S
[
1
]
A
.
G
o
e
l
,
A
.
K
.
G
o
e
l
,
a
n
d
A
.
K
u
m
a
r
,
“
P
e
r
f
o
r
ma
n
c
e
a
n
a
l
y
s
i
s
o
f
m
u
l
t
i
p
l
e
i
n
p
u
t
si
n
g
l
e
l
a
y
e
r
n
e
u
r
a
l
n
e
t
w
o
r
k
h
a
r
d
w
a
r
e
c
h
i
p
,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
2
,
n
o
.
1
8
,
p
p
.
2
8
2
1
3
–
2
8
2
3
4
,
Ju
l
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
0
4
2
-
0
2
3
-
1
4
6
2
7
-
3.
[
2
]
J.
M
i
sr
a
a
n
d
I
.
S
a
h
a
,
“
A
r
t
i
f
i
c
i
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s i
n
h
a
r
d
w
a
r
e
:
A
su
r
v
e
y
o
f
t
w
o
d
e
c
a
d
e
s
o
f
p
r
o
g
r
e
ss,
”
N
e
u
r
o
c
o
m
p
u
t
i
n
g
,
v
o
l
.
7
4
,
n
o
.
1
–
3
,
p
p
.
2
3
9
–
2
5
5
,
D
e
c
.
2
0
1
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m.
2
0
1
0
.
0
3
.
0
2
1
.
[
3
]
M
.
M
a
r
c
h
e
si
,
G
.
O
r
l
a
n
d
i
,
F
.
P
i
a
z
z
a
,
a
n
d
A
.
U
n
c
i
n
i
,
“
F
a
st
N
e
u
r
a
l
N
e
t
w
o
r
k
s
W
i
t
h
o
u
t
M
u
l
t
i
p
l
i
e
r
s,
”
I
EE
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
N
e
u
r
a
l
N
e
t
w
o
rks
,
v
o
l
.
4
,
n
o
.
1
,
p
p
.
5
3
–
6
2
,
1
9
9
3
,
d
o
i
:
1
0
.
1
1
0
9
/
7
2
.
1
8
2
6
9
5
.
[
4
]
A
.
G
o
e
l
,
A
.
K
.
G
o
e
l
,
a
n
d
A
.
K
u
mar
,
“
Th
e
r
o
l
e
o
f
a
r
t
i
f
i
c
i
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
a
n
d
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
i
n
u
t
i
l
i
z
i
n
g
s
p
a
t
i
a
l
i
n
f
o
r
ma
t
i
o
n
,
”
S
p
a
t
i
a
l
I
n
f
o
rm
a
t
i
o
n
R
e
se
a
rc
h
,
v
o
l
.
3
1
,
n
o
.
3
,
p
p
.
2
7
5
–
2
8
5
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
1
3
2
4
-
0
2
2
-
0
0
4
9
4
-
x.
[
5
]
A
.
S
.
R
a
w
a
t
,
A
.
R
a
n
a
,
A
.
K
u
m
a
r
,
a
n
d
A
.
B
a
g
w
a
r
i
,
“
A
p
p
l
i
c
a
t
i
o
n
o
f
mu
l
t
i
l
a
y
e
r
a
r
t
i
f
i
c
i
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
i
n
t
h
e
d
i
a
g
n
o
s
i
s
s
y
st
e
m
:
a
sy
st
e
ma
t
i
c
r
e
v
i
e
w
,
”
I
AE
S
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
I
J
-
AI
)
,
v
o
l
.
7
,
n
o
.
3
,
p
p
.
1
3
8
–
1
4
2
,
A
u
g
.
2
0
1
8
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
a
i
.
v
7
.
i
3
.
p
p
1
3
8
-
1
4
2
.
[
6
]
U
.
P
a
n
d
e
y
,
A
.
P
a
t
h
a
k
,
A
.
K
u
mar,
a
n
d
S
.
M
o
n
d
a
l
,
“
A
p
p
l
i
c
a
t
i
o
n
s
o
f
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
i
n
p
o
w
e
r
s
y
s
t
e
m
o
p
e
r
a
t
i
o
n
,
c
o
n
t
r
o
l
a
n
d
p
l
a
n
n
i
n
g
:
a
r
e
v
i
e
w
,
”
C
l
e
a
n
E
n
e
r
g
y
,
v
o
l
.
7
,
n
o
.
6
,
p
p
.
1
1
9
9
–
1
2
1
8
,
N
o
v
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
9
3
/
c
e
/
z
k
a
d
0
6
1
.
[
7
]
A
.
K
u
mar
,
P
.
C
h
a
u
d
a
,
a
n
d
A
.
D
e
v
r
a
r
i
,
“
M
a
c
h
i
n
e
Le
a
r
n
i
n
g
A
p
p
r
o
a
c
h
f
o
r
B
r
a
i
n
T
u
m
o
r
D
e
t
e
c
t
i
o
n
a
n
d
S
e
g
me
n
t
a
t
i
o
n
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
232
320
351
A
N
N
-
20
UT
IL
IZA
T
IO
N
Neu
ro
m
o
rph
ic
Ch
ip
H
a
rdw
a
re
Utiliza
t
io
n
L
UT
S
l
i
c
e
s
IO
B
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
7
2
2
-
2
5
8
6
I
AE
S
I
n
t
J
R
o
b
&
A
u
to
m
,
Vo
l
.
13
,
No
.
4
,
Dec
em
b
er
20
24
:
4
9
5
-
5
0
5
504
J
o
u
rn
a
l
o
f
O
r
g
a
n
i
za
t
i
o
n
a
l
a
n
d
C
o
l
l
e
c
t
i
v
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
1
,
n
o
.
3
,
p
p
.
6
8
–
8
4
,
Ju
l
.
2
0
2
1
,
d
o
i
:
1
0
.
4
0
1
8
/
i
j
o
c
i
.
2
0
2
1
0
7
0
1
0
5
.
[
8
]
S
.
D
h
y
a
n
i
,
A
.
K
u
mar,
a
n
d
S
.
C
h
o
u
d
h
u
r
y
,
“
A
n
a
l
y
si
s
o
f
EC
G
-
b
a
s
e
d
a
r
r
h
y
t
h
m
i
a
d
e
t
e
c
t
i
o
n
s
y
st
e
m
u
s
i
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
Me
t
h
o
d
s
X
,
v
o
l
.
1
0
,
p
.
1
0
2
1
9
5
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
me
x
.
2
0
2
3
.
1
0
2
1
9
5
.
[
9
]
S
.
D
h
y
a
n
i
,
A
.
K
u
m
a
r
,
a
n
d
S
.
C
h
o
u
d
h
u
r
y
,
“
R
e
v
i
e
w
o
f
a
n
a
l
y
si
s
o
f
EC
G
b
a
se
d
a
r
r
h
y
t
h
m
i
a
d
e
t
e
c
t
i
o
n
s
y
s
t
e
m
u
si
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
2
0
2
3
,
p
.
2
0
0
2
9
,
d
o
i
:
1
0
.
1
0
6
3
/
5
.
0
1
7
8
0
6
2
.
[
1
0
]
R
.
H
.
Tu
r
n
e
r
a
n
d
R
.
F
.
W
o
o
d
s
,
“
H
i
g
h
l
y
e
f
f
i
c
i
e
n
t
,
l
i
m
i
t
e
d
r
a
n
g
e
mu
l
t
i
p
l
i
e
r
s
f
o
r
LU
T
-
b
a
se
d
F
P
G
A
a
r
c
h
i
t
e
c
t
u
r
e
s,”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
V
e
ry
L
a
r
g
e
S
c
a
l
e
I
n
t
e
g
r
a
t
i
o
n
(
VL
S
I
)
S
y
st
e
m
s
,
v
o
l
.
1
2
,
n
o
.
1
0
,
p
p
.
1
1
1
3
–
1
1
1
7
,
O
c
t
.
2
0
0
4
,
d
o
i
:
1
0
.
1
1
0
9
/
TV
LSI
.
2
0
0
4
.
8
3
3
3
9
9
.
[
1
1
]
R
.
S
a
r
i
ć
,
D
.
J
o
k
i
ć
,
N
.
B
e
g
a
n
o
v
i
ć
,
L.
G
.
P
o
k
v
i
ć
,
a
n
d
A
.
B
a
d
n
j
e
v
i
ć
,
“
F
P
G
A
-
b
a
se
d
r
e
a
l
-
t
i
me
e
p
i
l
e
p
t
i
c
se
i
z
u
r
e
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
A
r
t
i
f
i
c
i
a
l
N
e
u
r
a
l
N
e
t
w
o
r
k
,
”
B
i
o
m
e
d
i
c
a
l
S
i
g
n
a
l
Pro
c
e
ss
i
n
g
a
n
d
C
o
n
t
ro
l
,
v
o
l
.
6
2
,
p
.
1
0
2
1
0
6
,
S
e
p
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
b
s
p
c
.
2
0
2
0
.
1
0
2
1
0
6
.
[
1
2
]
V
.
S
.
D
h
a
k
a
e
t
a
l
.
,
“
A
s
u
r
v
e
y
o
f
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
a
p
p
l
i
e
d
f
o
r
p
r
e
d
i
c
t
i
o
n
o
f
p
l
a
n
t
l
e
a
f
d
i
se
a
s
e
s,”
S
e
n
s
o
rs
,
v
o
l
.
2
1
,
n
o
.
1
4
,
p
.
4
7
4
9
,
Ju
l
.
2
0
2
1
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
1
1
4
4
7
4
9
.
[
1
3
]
P
.
Š
k
o
d
a
,
T.
L
i
p
i
ć
,
A
.
S
r
p
,
B
.
M
.
R
o
g
i
n
a
,
K
.
S
k
a
l
a
,
a
n
d
F
.
V
a
j
d
a
,
“
I
mp
l
e
men
t
a
t
i
o
n
f
r
a
m
e
w
o
r
k
f
o
r
A
r
t
i
f
i
c
i
a
l
N
e
u
r
a
l
N
e
t
w
o
r
k
s
o
n
F
P
G
A
,
”
i
n
MIPRO
2
0
1
1
-
3
4
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
v
e
n
t
i
o
n
o
n
I
n
f
o
rm
a
t
i
o
n
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
El
e
c
t
r
o
n
i
c
s
a
n
d
Mi
c
r
o
e
l
e
c
t
r
o
n
i
c
s
-
Pr
o
c
e
e
d
i
n
g
s
,
2
0
1
1
,
p
p
.
2
7
4
–
2
7
8
.
[
1
4
]
J.
S
c
h
mi
d
h
u
b
e
r
,
“
D
e
e
p
L
e
a
r
n
i
n
g
i
n
n
e
u
r
a
l
n
e
t
w
o
r
k
s:
A
n
o
v
e
r
v
i
e
w
,
”
N
e
u
r
a
l
N
e
t
w
o
r
k
s
,
v
o
l
.
6
1
,
p
p
.
8
5
–
1
1
7
,
J
a
n
.
2
0
1
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
n
e
t
.
2
0
1
4
.
0
9
.
0
0
3
.
[
1
5
]
J.
K
i
m,
K
.
H
w
a
n
g
,
a
n
d
W
.
S
u
n
g
,
“
X
1
0
0
0
r
e
a
l
-
t
i
m
e
p
h
o
n
e
me
r
e
c
o
g
n
i
t
i
o
n
V
LSI
u
si
n
g
f
e
e
d
-
f
o
r
w
a
r
d
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
i
n
2
0
1
4
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ac
o
u
st
i
c
s
,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
P
ro
c
e
ssi
n
g
(
I
C
A
S
S
P)
,
M
a
y
2
0
1
4
,
p
p
.
7
5
1
0
–
7
5
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
4
.
6
8
5
5
0
6
0
.
[
1
6
]
L.
D
.
M
e
d
u
s,
T
.
I
a
k
y
mc
h
u
k
,
J.
V
.
F
r
a
n
c
e
s
-
V
i
l
l
o
r
a
,
M
.
B
a
t
a
l
l
e
r
-
M
o
m
p
e
a
n
,
a
n
d
A
.
R
o
sa
d
o
-
M
u
n
o
z
,
“
A
N
o
v
e
l
S
y
s
t
o
l
i
c
P
a
r
a
l
l
e
l
H
a
r
d
w
a
r
e
A
r
c
h
i
t
e
c
t
u
r
e
f
o
r
t
h
e
F
P
G
A
A
c
c
e
l
e
r
a
t
i
o
n
o
f
F
e
e
d
f
o
r
w
a
r
d
N
e
u
r
a
l
N
e
t
w
o
r
k
s,”
I
EEE
A
c
c
e
ss
,
v
o
l
.
7
,
p
p
.
7
6
0
8
4
–
7
6
1
0
3
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
9
.
2
9
2
0
8
8
5
.
[
1
7
]
T.
V
.
H
u
y
n
h
,
“
D
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
a
c
c
e
l
e
r
a
t
o
r
b
a
se
d
o
n
F
P
G
A
,
”
i
n
2
0
1
7
4
t
h
N
AFO
S
T
ED
C
o
n
f
e
re
n
c
e
o
n
I
n
f
o
rm
a
t
i
o
n
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
N
I
C
S
2
0
1
7
-
P
ro
c
e
e
d
i
n
g
s
,
N
o
v
.
2
0
1
7
,
v
o
l
.
2
0
1
7
-
Jan
u
a
r
y
,
p
p
.
2
5
4
–
2
5
7
,
d
o
i
:
1
0
.
1
1
0
9
/
N
A
F
O
S
TED
.
2
0
1
7
.
8
1
0
8
0
7
3
.
[
1
8
]
J.
P
a
r
k
a
n
d
W
.
S
u
n
g
,
“
F
P
G
A
b
a
s
e
d
i
m
p
l
e
me
n
t
a
t
i
o
n
o
f
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
u
si
n
g
o
n
-
c
h
i
p
mem
o
r
y
o
n
l
y
,
”
i
n
2
0
1
6
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Ac
o
u
st
i
c
s
,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pro
c
e
ss
i
n
g
(
I
C
A
S
S
P)
,
M
a
r
.
2
0
1
6
,
p
p
.
1
0
1
1
–
1
0
1
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
6
.
7
4
7
1
8
2
8
.
[
1
9
]
K
.
G
u
o
,
S
.
Ze
n
g
,
J
.
Y
u
,
Y
.
W
a
n
g
,
a
n
d
H
.
Y
a
n
g
,
“
[
D
L]
A
s
u
r
v
e
y
o
f
F
P
G
A
-
b
a
se
d
n
e
u
r
a
l
n
e
t
w
o
r
k
i
n
f
e
r
e
n
c
e
a
c
c
e
l
e
r
a
t
o
r
s,
”
A
C
M
T
ra
n
s
a
c
t
i
o
n
s
o
n
Re
c
o
n
f
i
g
u
ra
b
l
e
T
e
c
h
n
o
l
o
g
y
a
n
d
S
y
s
t
e
m
s
,
v
o
l
.
1
2
,
n
o
.
1
,
p
p
.
1
–
2
6
,
M
a
r
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
4
5
/
3
2
8
9
1
8
5
.
[
2
0
]
T.
C
h
o
u
d
h
a
r
y
,
V
.
M
i
sh
r
a
,
A
.
G
o
sw
a
mi
,
a
n
d
J.
S
a
r
a
n
g
a
p
a
n
i
,
“
A
c
o
m
p
r
e
h
e
n
si
v
e
s
u
r
v
e
y
o
n
mo
d
e
l
c
o
m
p
r
e
ss
i
o
n
a
n
d
a
c
c
e
l
e
r
a
t
i
o
n
,
”
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
Re
v
i
e
w
,
v
o
l
.
5
3
,
n
o
.
7
,
p
p
.
5
1
1
3
–
5
1
5
5
,
F
e
b
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
4
6
2
-
0
2
0
-
0
9
8
1
6
-
7.
[
2
1
]
A
.
P
.
D
.
A
.
F
e
r
r
e
i
r
a
a
n
d
E.
N
.
D
.
S
.
B
a
r
r
o
s
,
“
A
h
i
g
h
p
e
r
f
o
r
ma
n
c
e
f
u
l
l
p
i
p
e
l
i
n
e
d
a
r
q
u
i
t
e
c
t
u
r
e
o
f
M
LP
n
e
u
r
a
l
n
e
t
w
o
r
k
s
i
n
F
P
G
A
,
”
i
n
2
0
1
0
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
E
l
e
c
t
r
o
n
i
c
s
,
C
i
rc
u
i
t
s,
a
n
d
S
y
s
t
e
m
s,
I
C
E
C
S
2
0
1
0
-
Pr
o
c
e
e
d
i
n
g
s
,
D
e
c
.
2
0
1
0
,
p
p
.
7
4
2
–
7
4
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
EC
S
.
2
0
1
0
.
5
7
2
4
6
1
9
.
[
2
2
]
I
.
G
i
a
c
o
m
o
e
t
a
l
.
,
“
N
e
u
r
o
mo
r
p
h
i
c
si
l
i
c
o
n
n
e
u
r
o
n
c
i
r
c
u
i
t
s,
”
Fr
o
n
t
i
e
rs
i
n
N
e
u
r
o
s
c
i
e
n
c
e
,
v
o
l
.
5
,
2
0
1
1
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
w
w
w
.
f
r
o
n
t
i
e
r
si
n
.
o
r
g
/
J
o
u
r
n
a
l
/
A
b
st
r
a
c
t
.
a
s
p
x
?
s
=
7
5
5
&n
a
me
=
n
e
u
r
o
m
o
r
p
h
i
c
_
e
n
g
i
n
e
e
r
i
n
g
&A
R
T
_
D
O
I
=
1
0
.
3
3
8
9
/
f
n
i
n
s.
2
0
1
1
.
0
0
0
7
3
.
[
2
3
]
M
.
K
.
S
i
n
g
h
,
N
.
S
i
n
g
h
,
a
n
d
A
.
K
.
S
i
n
g
h
,
“
S
p
e
a
k
e
r
’
s
V
o
i
c
e
C
h
a
r
a
c
t
e
r
i
st
i
c
s
a
n
d
S
i
m
i
l
a
r
i
t
y
M
e
a
s
u
r
e
m
e
n
t
u
si
n
g
E
u
c
l
i
d
e
a
n
D
i
st
a
n
c
e
s,
”
i
n
2
0
1
9
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
,
I
C
S
C
2
0
1
9
,
M
a
r
.
2
0
1
9
,
p
p
.
3
1
7
–
3
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
C
4
5
6
2
2
.
2
0
1
9
.
8
9
3
8
3
6
6
.
[
2
4
]
F
.
C
o
r
r
a
d
i
,
D
.
B
o
n
t
r
a
g
e
r
,
a
n
d
G
.
I
n
d
i
v
e
r
i
,
“
To
w
a
r
d
n
e
u
r
o
m
o
r
p
h
i
c
i
n
t
e
l
l
i
g
e
n
t
b
r
a
i
n
-
m
a
c
h
i
n
e
i
n
t
e
r
f
a
c
e
s
:
A
n
e
v
e
n
t
-
b
a
s
e
d
n
e
u
r
a
l
r
e
c
o
r
d
i
n
g
a
n
d
p
r
o
c
e
ss
i
n
g
s
y
st
e
m,
”
i
n
I
EEE
2
0
1
4
Bi
o
m
e
d
i
c
a
l
C
i
rc
u
i
t
s
a
n
d
S
y
s
t
e
m
s
C
o
n
f
e
re
n
c
e
,
B
i
o
C
A
S
2
0
1
4
-
Pro
c
e
e
d
i
n
g
s
,
O
c
t
.
2
0
1
4
,
p
p
.
5
8
4
–
5
8
7
,
d
o
i
:
1
0
.
1
1
0
9
/
B
i
o
C
A
S
.
2
0
1
4
.
6
9
8
1
7
9
3
.
[
2
5
]
T.
P
o
s
e
w
sk
y
a
n
d
D
.
Z
i
e
n
e
r
,
“
Th
r
o
u
g
h
p
u
t
o
p
t
i
mi
z
a
t
i
o
n
s
f
o
r
F
P
G
A
-
b
a
s
e
d
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
i
n
f
e
r
e
n
c
e
,
”
M
i
c
ro
p
r
o
c
e
ss
o
rs
a
n
d
Mi
c
r
o
s
y
st
e
m
s
,
v
o
l
.
6
0
,
p
p
.
1
5
1
–
1
6
1
,
J
u
l
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
m
i
c
p
r
o
.
2
0
1
8
.
0
4
.
0
0
4
.
[
2
6
]
A
.
K
u
mar,
G
.
V
e
r
ma
,
M
.
K
.
G
u
p
t
a
,
M
.
S
a
l
a
u
d
d
i
n
,
B
.
K
.
R
e
h
ma
n
,
a
n
d
D
.
K
u
mar,
“
3
D
mu
l
t
i
l
a
y
e
r
mes
h
N
o
C
c
o
mm
u
n
i
c
a
t
i
o
n
a
n
d
F
P
G
A
sy
n
t
h
e
s
i
s,
”
Wi
r
e
l
e
ss Pe
rs
o
n
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
s
,
v
o
l
.
1
0
6
,
n
o
.
4
,
p
p
.
1
8
5
5
–
1
8
7
3
,
J
u
n
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
2
7
7
-
0
1
8
-
5
7
2
4
-
3.
[
2
7
]
S
.
A
n
w
a
r
,
K
.
H
w
a
n
g
,
a
n
d
W
.
S
u
n
g
,
“
F
i
x
e
d
p
o
i
n
t
o
p
t
i
mi
z
a
t
i
o
n
o
f
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
o
b
j
e
c
t
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
2
0
1
5
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ac
o
u
st
i
c
s,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
P
ro
c
e
ssi
n
g
(
I
C
A
S
S
P)
,
A
p
r
.
2
0
1
5
,
p
p
.
1
1
3
1
–
1
1
3
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
5
.
7
1
7
8
1
4
6
.
[
2
8
]
P
.
D
.
D
e
o
t
a
l
e
a
n
d
L.
D
o
l
e
,
“
D
e
si
g
n
o
f
F
P
G
A
b
a
s
e
d
g
e
n
e
r
a
l
p
u
r
p
o
se
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
F
e
b
.
2
0
1
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
C
ES.2
0
1
4
.
7
0
3
3
8
4
3
.
[
2
9
]
S
.
H
i
m
a
v
a
t
h
i
,
D
.
A
n
i
t
h
a
,
a
n
d
A
.
M
u
t
h
u
r
a
mal
i
n
g
a
m,
“
F
e
e
d
f
o
r
w
a
r
d
n
e
u
r
a
l
n
e
t
w
o
r
k
i
m
p
l
e
me
n
t
a
t
i
o
n
i
n
F
P
G
A
u
s
i
n
g
l
a
y
e
r
mu
l
t
i
p
l
e
x
i
n
g
f
o
r
e
f
f
e
c
t
i
v
e
r
e
s
o
u
r
c
e
u
t
i
l
i
z
a
t
i
o
n
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
N
e
u
ra
l
N
e
t
w
o
rk
s
,
v
o
l
.
1
8
,
n
o
.
3
,
p
p
.
8
8
0
–
8
8
8
,
M
a
y
2
0
0
7
,
d
o
i
:
1
0
.
1
1
0
9
/
TN
N
.
2
0
0
7
.
8
9
1
6
2
6
.
[
3
0
]
Y
.
Zh
a
n
g
a
n
d
S
.
Z
h
a
n
g
,
“
O
p
t
i
m
i
z
e
d
d
e
e
p
l
e
a
r
n
i
n
g
a
r
c
h
i
t
e
c
t
u
r
e
s
w
i
t
h
f
a
st
mat
r
i
x
o
p
e
r
a
t
i
o
n
k
e
r
n
e
l
s
o
n
p
a
r
a
l
l
e
l
p
l
a
t
f
o
r
m,
”
i
n
2
0
1
3
I
EEE
2
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
T
o
o
l
s
w
i
t
h
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
N
o
v
.
2
0
1
3
,
p
p
.
7
1
–
7
8
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
TA
I
.
2
0
1
3
.
2
1
.
[
3
1
]
Z.
Y
u
,
A
.
M
.
A
b
d
u
l
g
h
a
n
i
,
A
.
Z
a
h
i
d
,
H
.
H
e
i
d
a
r
i
,
M
.
A
.
I
mr
a
n
,
a
n
d
Q
.
H
.
A
b
b
a
si
,
“
A
n
o
v
e
r
v
i
e
w
o
f
n
e
u
r
o
mo
r
p
h
i
c
c
o
m
p
u
t
i
n
g
f
o
r
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
e
n
a
b
l
e
d
h
a
r
d
w
a
r
e
-
b
a
se
d
h
o
p
f
i
e
l
d
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
8
,
p
p
.
6
7
0
8
5
–
6
7
0
9
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
2
9
8
5
8
3
9
.
[
2
9
]
S
.
H
i
m
a
v
a
t
h
i
,
D
.
A
n
i
t
h
a
a
n
d
A
.
M
u
t
h
u
r
a
mal
i
n
g
a
m,
“
F
e
e
d
f
o
r
w
a
r
d
n
e
u
r
a
l
n
e
t
w
o
r
k
i
mp
l
e
m
e
n
t
a
t
i
o
n
i
n
F
P
G
A
u
si
n
g
l
a
y
e
r
mu
l
t
i
p
l
e
x
i
n
g
f
o
r
e
f
f
e
c
t
i
v
e
r
e
so
u
r
c
e
u
t
i
l
i
z
a
t
i
o
n
,
”
in
I
E
EE
T
r
a
n
sa
c
t
i
o
n
s
o
n
N
e
u
ra
l
N
e
t
w
o
r
k
s
,
v
o
l
.
1
8
,
n
o
.
3
,
p
p
.
8
8
0
-
8
8
8
,
M
a
y
2
0
0
7
,
d
o
i
:
1
0
.
1
1
0
9
/
TN
N
.
2
0
0
7
.
8
9
1
6
2
6
.
[
3
0
]
Y
.
Zh
a
n
g
a
n
d
S
.
Z
h
a
n
g
,
“
O
p
t
i
m
i
z
e
d
d
e
e
p
l
e
a
r
n
i
n
g
a
r
c
h
i
t
e
c
t
u
r
e
s
w
i
t
h
f
a
st
mat
r
i
x
o
p
e
r
a
t
i
o
n
k
e
r
n
e
l
s
o
n
p
a
r
a
l
l
e
l
p
l
a
t
f
o
r
m
,
”
i
n
2
0
1
3
I
EEE
2
5
th
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
T
o
o
l
s
w
i
t
h
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
H
e
r
n
d
o
n
,
V
A
,
U
S
A
,
2
0
1
3
,
p
p
.
7
1
-
7
8
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
TA
I
.
2
0
1
3
.
2
1
.
[
3
1
]
Z.
Y
u
,
A
.
M
.
A
b
d
u
l
g
h
a
n
i
,
A
.
Za
h
i
d
,
H
.
H
e
i
d
a
r
i
,
M
.
A
.
I
mr
a
n
a
n
d
Q
.
H
.
A
b
b
a
si
,
“
A
n
o
v
e
r
v
i
e
w
o
f
n
e
u
r
o
m
o
r
p
h
i
c
c
o
m
p
u
t
i
n
g
f
o
r
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
e
n
a
b
l
e
d
h
a
r
d
w
a
r
e
-
b
a
se
d
h
o
p
f
i
e
l
d
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
in
I
EEE
A
c
c
e
ss
,
v
o
l
.
8
,
p
p
.
6
7
0
8
5
-
6
7
0
9
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
2
9
8
5
8
3
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.