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.
14
,
No
.
4
,
Dec
em
b
er
20
25
,
p
p
.
3
6
6
~
3
7
5
I
SS
N:
2722
-
2
5
8
6
,
DOI
:
1
0
.
1
1
5
9
1
/i
jr
a
.
v
14
i
4
.
pp
3
6
6
-
375
366
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
r
a
.
ia
esco
r
e.
co
m
A
surv
ey
o
n
co
nv
o
lutiona
l neural n
etwork
ha
rd
wa
re
a
cceleratio
n t
h
ro
ug
h app
ro
x
ima
te
co
mputing
mul
tip
le and
a
ccumula
tes
uni
t
Su
v
it
ha
P
a
t
hiy
a
da
n Sud
ha
k
a
ra
n
,
Aa
t
hm
a
nes
a
n T
ha
ng
a
k
a
la
i
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
mm
u
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
V
e
l
T
e
c
h
R
a
n
g
a
r
a
j
a
n
D
r
.
S
a
g
u
n
t
h
a
l
a
R
&D
I
n
st
i
t
u
t
e
o
f
S
c
i
e
n
c
e
a
n
d
Te
c
h
n
o
l
o
g
y
,
C
h
e
n
n
a
i
,
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
an
6
,
2
0
2
5
R
ev
is
ed
Au
g
6
,
2
0
2
5
Acc
ep
ted
Au
g
2
6
,
2
0
2
5
Co
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
s
(
CNN
s)
a
re
a
p
p
li
e
d
to
a
d
iffere
n
t
ra
n
g
e
o
f
re
a
l
-
wo
rld
c
o
m
p
le
x
tas
k
s
t
o
p
r
o
v
id
e
e
ffe
c
ti
v
e
so
l
u
ti
o
n
s
wit
h
h
i
g
h
a
c
c
u
ra
c
y
.
Ba
se
d
o
n
t
h
e
a
p
p
li
c
a
ti
o
n
'
s
c
o
m
p
lex
it
y
,
CNN
d
e
m
a
n
d
s
a
l
o
t
o
f
p
ro
c
e
ss
in
g
u
n
it
s
a
n
d
m
e
m
o
ry
sp
a
c
e
s
fo
r
it
s
e
ffe
c
ti
v
e
imp
lem
e
n
tatio
n
.
Bri
n
g
i
n
g
th
is
c
o
m
p
u
tati
o
n
a
l
tas
k
to
h
a
rd
wa
re
fo
r
p
r
o
c
e
ss
in
g
t
h
e
d
a
ta
t
o
e
n
h
a
n
c
e
th
e
a
c
c
e
ler
a
ti
o
n
h
e
lp
s
i
n
a
c
h
iev
i
n
g
r
e
a
l
-
ti
m
e
p
e
rfo
rm
a
n
c
e
imp
ro
v
e
m
e
n
t.
Re
c
e
n
t
stu
d
ies
f
o
c
u
se
d
o
n
a
p
p
ro
x
ima
ti
o
n
m
e
th
o
d
o
lo
g
y
t
o
o
v
e
rc
o
m
e
t
h
is
p
r
o
b
lem
.
Th
is
p
r
o
p
o
se
d
su
r
v
e
y
a
n
a
l
y
z
e
s
v
a
rio
u
s
re
c
e
n
t
m
e
th
o
d
s
in
v
o
lv
e
d
in
imp
lem
e
n
ti
n
g
a
p
p
r
o
x
ima
ti
n
g
c
o
m
p
u
ti
n
g
-
b
a
se
d
p
r
o
c
e
ss
in
g
e
le
m
e
n
ts
a
n
d
th
e
ir
u
sa
g
e
in
CNN
s.
P
rima
ril
y
,
th
e
su
r
v
e
y
fo
c
u
se
s
o
n
m
u
l
ti
p
le
a
n
d
a
c
c
u
m
u
late
s
(M
AC)
u
n
it
a
n
d
th
e
ir
v
a
rio
u
s
a
p
p
ro
x
ima
ti
o
n
m
e
th
o
d
s,
wh
ich
a
c
ts
a
s
a
fu
n
d
a
m
e
n
tal
b
l
o
c
k
a
s
a
p
ro
c
e
ss
in
g
e
lem
e
n
t
i
n
t
h
e
CN
N
lay
e
rs.
S
e
c
o
n
d
l
y
,
it
f
o
c
u
se
s
o
n
v
a
ri
o
u
s
CNN
h
a
rd
wa
re
a
c
c
e
lera
ti
o
n
a
rc
h
it
e
c
tu
re
s
a
n
d
th
e
ir
lay
e
rs
d
e
sig
n
e
d
u
si
n
g
d
iffere
n
t
m
e
th
o
d
s
a
n
d
th
e
ir
wi
d
e
ra
n
g
e
o
f
a
p
p
li
c
a
ti
o
n
s.
S
o
m
e
o
f
t
h
e
re
c
e
n
t
d
e
sig
n
m
e
th
o
d
s
a
p
p
li
e
d
to
v
a
rio
u
s
ra
n
g
e
s
o
f
a
p
p
li
c
a
ti
o
n
s
a
re
a
lso
a
n
a
ly
z
e
d
i
n
th
e
p
r
o
p
o
se
d
su
rv
e
y
.
T
h
is
d
e
taile
d
a
n
a
ly
sis
g
iv
e
s
a
n
o
u
tl
o
o
k
o
n
e
ffe
c
ti
v
e
a
p
p
ro
x
ima
ti
o
n
b
l
o
c
k
s
a
n
d
th
e
CNN
a
rc
h
it
e
c
tu
re
to
b
e
e
ffe
c
ti
v
e
ly
u
se
d
in
v
a
ri
o
u
s
d
e
sig
n
s
,
wit
h
a
sc
o
p
e
o
f
a
re
a
in
wh
ich
fu
tu
re
imp
r
o
v
e
m
e
n
t
c
a
n
b
e
m
a
d
e
.
K
ey
w
o
r
d
s
:
Ap
p
r
o
x
im
ate
m
u
ltip
le
an
d
ac
cu
m
u
lates
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
Dee
p
lear
n
in
g
E
r
r
o
r
t
o
ler
an
t d
esig
n
Har
d
war
e
ac
ce
ler
atio
n
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
:
Su
v
ith
a
Path
iy
ad
an
Su
d
h
ak
a
r
a
n
Dep
ar
tm
en
t o
f
E
lectr
o
n
ics an
d
C
o
m
m
u
n
icatio
n
E
n
g
in
ee
r
i
n
g
,
Vel
T
ec
h
R
an
g
ar
ajan
Dr
.
Sag
u
n
th
ala
R
&
D
I
n
s
titu
te
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
Av
ad
i,
C
h
en
n
ai,
T
a
m
il Na
d
u
,
6
0
0
0
6
2
,
I
n
d
ia
E
m
ail:
s
u
v
ith
ap
s
v
td
1
2
5
2
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
W
ith
cu
r
r
en
t
ad
v
an
ce
m
e
n
ts
in
tech
n
o
lo
g
y
,
ar
tific
ial
in
tellig
en
ce
(
AI
)
h
as
its
p
r
esen
ce
in
alm
o
s
t
all
f
ield
s
,
r
an
g
in
g
f
r
o
m
b
io
m
ed
ic
al
to
ag
r
icu
ltu
r
e
an
d
au
to
m
ati
o
n
to
co
m
m
u
n
icatio
n
[
1
]
.
AI
c
o
n
tin
u
o
u
s
ly
ev
o
lv
es
with
th
e
h
elp
o
f
v
ar
io
u
s
tr
ain
i
n
g
m
o
d
els
o
r
d
atasets
av
ailab
le
an
d
f
r
eq
u
en
tly
u
p
d
ated
d
at
a
f
o
r
im
p
r
o
v
in
g
its
ac
cu
r
ac
y
an
d
ac
ce
ler
atio
n
th
r
o
u
g
h
in
-
b
u
ilt
m
em
o
r
y
[
2
]
.
So
m
e
co
n
v
e
n
tio
n
al
d
i
g
ital
s
ig
n
al
p
r
o
ce
s
s
in
g
(
DSP)
b
lo
ck
s
lack
f
lex
ib
ilit
y
f
o
r
p
r
o
ce
s
s
in
g
v
ar
io
u
s
b
it
len
g
t
h
d
ata;
to
o
v
er
co
m
e
th
is
d
is
ad
v
an
t
ag
e
h
ar
d
co
r
e
DSP
b
lo
ck
in
f
ield
p
r
o
g
r
am
m
a
b
le
g
ated
ar
r
ay
(
FP
GA)
is
u
t
ilized
.
FP
GA,
with
it
s
h
ig
h
p
er
f
o
r
m
a
n
ce
an
d
r
ec
o
n
f
ig
u
r
ab
ilit
y
,
h
elp
s
in
en
h
an
cin
g
t
h
e
ac
ce
ler
atio
n
o
f
n
eu
r
al
n
etwo
r
k
p
r
o
ce
s
s
es
th
r
o
u
g
h
m
o
d
e
s
elec
tio
n
in
an
d
r
u
n
-
tim
e
r
ec
o
n
f
ig
u
r
atio
n
o
f
ac
ce
ler
ato
r
b
lo
ck
s
[
3
]
.
Ap
p
r
o
x
im
ate
c
o
m
p
u
tin
g
ac
ts
as
an
ef
f
ec
tiv
e
m
eth
o
d
o
lo
g
y
in
r
ed
u
cin
g
r
eso
u
r
ce
u
tili
za
tio
n
,
p
o
wer
co
n
s
u
m
p
t
io
n
,
al
o
n
g
wi
th
p
e
r
f
o
r
m
an
ce
i
m
p
r
o
v
e
m
e
n
t
i
n
v
a
r
i
o
u
s
d
ee
p
le
ar
n
i
n
g
m
et
h
o
d
s
,
wh
i
ch
a
cc
ep
ts
m
i
n
i
m
al
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
s
u
r
ve
y
o
n
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
h
a
r
d
w
a
r
e
a
cc
elera
tio
n
…
(
S
u
vith
a
P
a
t
h
iya
d
a
n
S
u
d
h
a
ka
r
a
n
)
367
to
ler
ab
le
v
a
r
iatio
n
s
in
th
e
ar
i
th
m
etic
o
u
tp
u
t
th
r
o
u
g
h
p
ar
all
el
co
m
p
u
tatio
n
s
an
d
b
y
u
s
in
g
co
m
p
r
ess
o
r
s
[
4
]
.
B
ec
au
s
e
m
ac
h
in
e
lear
n
in
g
in
v
o
lv
es
a
lar
g
e
n
u
m
b
er
o
f
M
AC
b
lo
ck
s
[
5
]
,
wh
ich
ar
e
u
s
ed
in
co
n
v
o
lu
tio
n
al
lay
er
s
in
v
o
lv
in
g
k
er
n
el
m
at
r
ix
p
r
o
ce
s
s
in
g
.
T
h
e
h
y
b
r
id
s
y
s
to
lic
d
esig
n
with
f
ac
to
r
ed
p
r
o
p
ag
ate
ad
d
er
[
6
]
,
MA
C
with
in
ter
n
al
co
m
p
r
es
s
o
r
f
o
r
p
a
r
tial
p
r
o
d
u
ct
r
ed
u
c
tio
n
,
im
p
r
ec
is
e
ad
d
er
an
d
m
u
ltip
lier
co
m
b
i
n
ed
s
tr
u
ctu
r
e
[
7
]
,
an
d
p
a
r
allel
MA
C
f
o
r
ef
f
icien
t
DNN
in
ter
f
ac
e
ar
e
v
ar
i
o
u
s
MA
C
d
esig
n
s
[
8
]
.
MA
C
u
n
it
in
C
NN
lay
er
in
v
o
lv
es
m
u
ltip
lier
an
d
ad
d
er
u
n
its
wh
ich
ac
co
u
n
t
to
5
5
to
6
5
%
o
f
th
e
o
v
er
all
ar
ea
.
T
h
e
ap
p
r
o
x
im
atio
n
ca
n
b
e
d
o
n
e
in
v
ar
io
u
s
wa
y
s
in
ad
d
e
r
an
d
m
u
ltip
lier
cir
cu
its
with
o
u
t
co
m
p
r
o
m
is
in
g
th
e
d
e
s
ir
ab
le
ac
cu
r
ac
y
b
y
in
p
u
t
o
p
e
r
an
d
awa
r
e
a
p
p
r
o
x
im
atio
n
,
a
p
p
r
o
x
im
atio
n
with
s
elf
-
er
r
o
r
ad
ju
s
tm
en
t,
lo
g
ar
ith
m
ic
b
ase
d
ap
p
r
o
x
im
atio
n
m
u
ltip
licatio
n
,
an
d
th
r
o
u
g
h
p
a
r
tial
p
r
o
d
u
ct
s
eg
m
en
tatio
n
an
d
p
ar
tial
ad
d
itio
n
.
Sen
s
itiv
ity
-
b
ased
er
r
o
r
to
ler
an
t
d
esig
n
u
s
e
s
v
o
ltag
e
s
ca
lin
g
an
d
s
ca
lin
g
b
y
elim
in
atin
g
u
n
u
s
ed
b
l
o
ck
s
[
5
]
.
T
h
e
in
cr
ea
s
e
in
th
e
u
s
ag
e
o
f
C
NN
n
ee
d
s
ef
f
ec
tiv
e
ac
ce
ler
a
tio
n
ar
ch
itectu
r
e
f
o
r
d
esig
n
[
9
]
with
lar
g
e
weig
h
t
b
o
u
n
d
an
d
m
atr
ix
o
p
er
atio
n
s
lik
e
weig
h
t
-
o
r
ien
ted
a
p
p
r
o
x
im
atio
n
,
C
NN
with
er
r
o
r
co
r
r
ec
tio
n
m
o
d
u
le,
d
u
al
p
r
ec
is
io
n
with
r
eu
s
in
g
o
u
tp
u
t
p
a
r
tially
,
s
h
if
tin
g
b
ased
C
NN
to
r
e
d
u
ce
th
e
m
u
ltip
lier
c
o
m
p
lex
ity
[
1
0
]
a
n
d
C
NN
wi
th
er
r
o
r
r
ate
an
aly
s
is
.
C
er
tain
m
eth
o
d
s
th
at
ar
e
ca
p
ab
le
o
f
ad
ju
s
tin
g
th
e
p
r
o
ce
s
s
in
g
elem
en
ts
d
u
r
in
g
r
u
n
tim
e
im
p
r
o
v
e
th
e
r
ec
o
n
f
ig
u
r
atio
n
[
1
1
]
.
T
h
e
ap
p
licati
o
n
-
s
p
ec
if
ic
ac
ce
ler
ato
r
s
u
ch
as
Sq
u
ee
ze
Net
an
d
Mo
b
ileNet,
also
m
in
i
m
izes
th
e
ab
o
v
e
-
m
e
n
tio
n
ed
p
r
o
b
lem
b
y
o
p
tim
izin
g
t
h
e
in
te
r
n
al
lay
er
s
in
v
o
l
v
ed
i
n
co
m
p
u
tatio
n
co
m
p
a
r
ed
to
c
o
n
v
en
tio
n
al
a
r
ch
itectu
r
es.
Var
io
u
s
o
th
er
m
eth
o
d
o
l
o
g
ies
lik
e
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
ar
e
ef
f
ec
tiv
e
in
ac
h
ie
v
in
g
h
ig
h
p
r
ec
is
io
n
in
p
r
o
ce
s
s
in
g
E
C
G
s
ig
n
als,
b
u
t
i
t
n
ee
d
s
o
th
er
alg
o
r
ith
m
s
to
class
if
y
th
e
o
u
tp
u
t sig
n
al
[
1
0
]
.
C
u
r
r
en
t
r
esear
ch
r
elate
d
t
o
th
e
ap
p
r
o
x
im
ate
co
m
p
u
tin
g
o
f
t
h
e
MA
C
u
n
it
f
o
cu
s
es
o
n
th
e
m
u
ltip
lier
s
an
d
ad
d
er
d
esig
n
.
I
n
2
0
2
1
,
FP
GAs
p
r
o
ce
s
s
two
m
u
ltip
ly
-
an
d
-
ac
cu
m
u
late
(
MA
C
)
o
p
er
atio
n
s
at
o
n
ce
in
s
id
e
o
n
e
DSP
b
lo
ck
.
An
ap
p
r
o
x
im
ate
co
m
p
u
tin
g
ap
p
r
o
ac
h
p
r
o
v
i
d
es
en
er
g
y
-
ef
f
icien
t
m
u
ltip
ly
-
ac
cu
m
u
late
(
MA
C
)
p
r
o
ce
s
s
in
g
[
1
2
]
.
I
n
o
r
d
er
to
f
u
r
th
er
m
i
n
im
ize
p
o
wer
c
o
n
s
u
m
p
tio
n
,
a
l
o
w
-
p
o
wer
C
I
M
tec
h
n
iq
u
e
m
o
d
if
ies
th
e
ca
n
o
n
ical
s
ig
n
ed
d
i
g
it
f
o
r
th
e
n
etwo
r
k
weig
h
ts
an
d
a
m
o
d
if
i
ed
r
ad
ix
-
4
B
o
o
th
alg
o
r
ith
m
at
th
e
in
p
u
t.
T
h
e
d
ata
p
ath
en
ab
les th
e
in
ter
n
al
ex
p
lo
itatio
n
o
f
s
elf
-
h
ea
lin
g
in
p
a
r
allel
d
esig
n
[
1
3
]
.
A
n
en
er
g
y
-
ef
f
i
cien
t d
ee
p
lear
n
in
g
ac
ce
ler
ato
r
with
cu
s
to
m
izab
le
r
u
n
tim
e
ac
cu
r
ac
y
is
p
r
e
d
icate
d
o
n
t
h
e
v
o
ltag
e
o
v
e
r
s
ca
lin
g
(
VOS)
ap
p
r
o
ac
h
.
T
h
e
r
ed
u
ctio
n
m
eth
o
d
lo
wer
s
en
er
g
y
a
n
d
p
o
wer
,
a
n
d
d
ela
y
wh
ile
m
ain
tain
in
g
a
to
le
r
ab
le
l
ev
el
o
f
q
u
ality
.
Fro
m
r
ec
en
t
r
esear
ch
,
ex
is
tin
g
tech
n
iq
u
es
co
n
s
u
m
e
m
o
r
e
ar
ea
,
p
o
we
r
ef
f
icien
c
y
a
n
d
d
elay
.
T
h
e
ex
is
tin
g
tech
n
iq
u
e
h
ad
p
er
f
o
r
m
ed
h
i
g
h
c
o
m
p
u
tatio
n
al
c
o
m
p
lex
ity
in
th
e
h
ar
d
war
e
.
Ap
p
r
o
x
im
ate
co
m
p
u
tin
g
in
cr
ea
s
es
th
e
ar
ea
an
d
also
af
f
ec
ts
o
v
er
all
e
f
f
icien
cy
.
T
o
o
v
er
co
m
e
th
ese
is
s
u
es,
ap
p
r
o
x
i
m
ate
co
m
p
u
tin
g
is
u
s
ed
in
th
e
Mu
ltip
ly
an
d
Acc
u
m
u
late
u
n
it,
in
C
NN
h
ar
d
war
e
ac
ce
ler
atio
n
h
as b
ee
n
p
r
o
p
o
s
e
d
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
u
tili
ze
s
ap
p
r
o
x
im
ate
MA
C
in
C
NN
f
o
r
h
ar
d
war
e
ac
ce
ler
ato
r
s
.
T
h
e
s
u
r
v
ey
f
o
cu
s
es
o
n
n
o
v
el
ap
p
r
o
ac
h
es
f
o
r
d
ev
elo
p
i
n
g
th
e
MA
C
u
n
it
in
th
e
p
r
o
ce
s
s
in
g
elem
en
t
o
f
C
NN
ac
ce
ler
ato
r
s
.
T
h
is
s
tu
d
y
s
h
o
ws
th
at
u
tili
zin
g
d
y
n
a
m
ic
v
o
ltag
e
s
ca
lin
g
,
ap
p
r
o
x
im
atio
n
MA
C
-
b
ased
C
NN
ac
ce
ler
ato
r
s
m
ay
m
in
im
ize
p
o
wer
u
s
ag
e
a
n
d
m
ain
tain
ac
ce
p
tab
le
ac
cu
r
ac
y
lev
els.
T
h
ese
ap
p
r
o
ac
h
es
o
f
f
er
f
r
am
ewo
r
k
s
f
o
r
m
in
im
izin
g
en
e
r
g
y
u
s
ag
e
a
n
d
r
eso
u
r
ce
r
ed
u
ctio
n
with
ac
c
ep
tab
le
er
r
o
r
.
T
h
is
s
tu
d
y
a
n
aly
s
es
th
e
tr
ad
e
-
o
f
f
s
b
etwe
en
ac
cu
r
ac
y
lo
s
s
an
d
en
e
r
g
y
s
av
in
g
s
ac
r
o
s
s
d
if
f
er
en
t a
p
p
r
o
x
im
atio
n
lev
els.
T
h
e
r
esear
ch
q
u
esti
o
n
s
o
f
th
e
s
u
g
g
ested
m
eth
o
d
o
lo
g
y
ar
e
a.
Ho
w
ca
n
ap
p
r
o
x
im
ate
MA
C
o
p
er
atio
n
s
b
e
ef
f
ec
tiv
ely
u
tili
ze
d
in
C
NN
ac
ce
ler
ato
r
s
?
b.
Ho
w
ca
n
an
o
p
tim
al
b
alan
ce
b
etwe
en
en
er
g
y
ef
f
icien
cy
an
d
i
n
f
er
en
ce
ac
c
u
r
ac
y
b
e
m
ain
tain
ed
?
T
h
e
s
u
r
v
ey
p
a
p
e
r
is
p
r
es
en
te
d
as
f
o
l
lo
ws.
T
h
e
a
p
p
r
o
x
im
at
e
c
o
m
p
u
t
in
g
o
f
t
h
e
MA
C
u
n
i
t
b
ase
d
o
n
v
a
r
i
o
u
s
r
ec
e
n
t
m
u
lti
p
l
ie
r
s
a
n
d
a
d
d
e
r
d
esi
g
n
s
is
lis
t
e
d
in
S
e
cti
o
n
2
.
An
ex
p
l
a
n
at
io
n
o
f
C
NN
la
y
e
r
s
a
n
d
th
ei
r
r
e
ce
n
t
m
et
h
o
d
o
l
o
g
ies
f
o
r
h
ar
d
wa
r
e
a
cc
ele
r
ati
o
n
is
d
es
cr
i
b
ed
i
n
S
ec
ti
o
n
3
.
T
h
e
r
es
u
lts
o
f
v
a
r
i
o
u
s
m
u
lti
p
l
ie
r
d
esi
g
n
s
a
r
e
d
is
c
u
s
s
e
d
i
n
Se
cti
o
n
4
.
F
u
t
u
r
e
r
e
co
m
m
e
n
d
ati
o
n
s
a
r
e
s
u
g
g
este
d
i
n
Se
cti
o
n
5
.
Secti
o
n
6
co
n
cl
u
d
es
th
e
s
u
r
v
ey
.
2.
P
RO
P
O
SE
D
M
E
T
H
O
D
T
h
is
s
e
cti
o
n
d
es
cr
i
b
es
t
h
e
a
p
p
r
o
x
i
m
at
e
co
m
p
u
ti
n
g
-
b
as
ed
MA
C
u
n
i
ts
a
n
d
t
h
ei
r
i
n
t
eg
r
a
ti
o
n
i
n
t
o
t
h
e
C
NN
ac
c
ele
r
at
o
r
t
o
e
n
h
a
n
c
e
h
ar
d
wa
r
e
ef
f
i
ci
en
cy
w
h
i
le
m
ai
n
tai
n
i
n
g
ac
ce
p
ta
b
l
e
ac
c
u
r
ac
y
le
v
els
.
B
y
em
p
l
o
y
i
n
g
ap
p
r
o
x
i
m
a
te
c
o
m
p
u
t
in
g
,
t
h
e
d
esi
g
n
d
y
n
a
m
i
ca
l
ly
a
d
j
u
s
ts
MA
C
p
r
e
cisi
o
n
ac
co
r
d
i
n
g
t
o
t
h
e
co
m
p
u
tat
io
n
al
d
e
m
a
n
d
o
f
e
ac
h
C
NN
la
y
e
r
.
Ov
e
r
a
ll
,
t
h
e
p
r
o
p
o
s
e
d
m
e
th
o
d
ac
h
i
e
v
es
a
n
e
f
f
ic
ie
n
t
t
r
a
d
e
-
o
f
f
b
e
twe
en
a
cc
u
r
ac
y
an
d
p
er
f
o
r
m
a
n
ce
,
p
r
o
v
i
d
i
n
g
a
s
ca
l
ab
le
a
n
d
p
o
we
r
-
e
f
f
ici
en
t C
NN
a
cc
ele
r
at
o
r
s
u
ita
b
l
e
f
o
r
r
e
a
l
-
ti
m
e
a
p
p
l
ica
ti
o
n
s
.
2
.
1
.
Ana
ly
s
is
o
f
v
a
rio
us
des
i
g
ns
inv
o
lv
ing
a
pp
ro
x
im
a
t
e
c
o
m
pu
t
ing
M
AC
I
n
C
NN
a
cc
ele
r
at
o
r
s
,
a
d
a
p
ti
v
e
a
p
p
r
o
x
i
m
at
io
n
d
y
n
a
m
i
ca
l
ly
m
o
d
if
ies
t
h
e
a
cc
u
r
ac
y
o
f
MA
C
o
p
e
r
at
io
n
s
ac
c
o
r
d
in
g
to
t
h
e
d
em
a
n
d
s
a
n
d
in
tr
ica
c
y
o
f
A
I
t
ask
s
.
B
y
ass
i
g
n
i
n
g
g
r
e
at
er
p
r
e
cisi
o
n
t
o
c
r
u
c
ial
ca
l
c
u
la
ti
o
n
s
a
n
d
p
e
r
m
itti
n
g
ap
p
r
o
x
im
ati
o
n
i
n
le
s
s
s
en
s
iti
v
e
p
r
o
c
ed
u
r
es
,
t
h
is
m
eth
o
d
m
a
x
i
m
i
ze
s
ac
cu
r
a
c
y
a
n
d
en
e
r
g
y
ef
f
i
ci
en
cy
.
L
o
w
-
p
r
i
o
r
it
y
p
o
s
it
io
n
s
,
li
k
e
v
o
ice
r
e
co
g
n
it
io
n
o
r
lo
w
-
r
es
o
l
u
t
i
o
n
v
i
d
eo
p
r
o
ce
s
s
in
g
,
f
o
r
i
n
s
ta
n
c
e,
m
a
y
wi
th
s
ta
n
d
g
r
ea
te
r
a
p
p
r
o
x
im
ati
o
n
le
v
els
,
wh
ic
h
r
es
u
lts
i
n
s
u
b
s
ta
n
ti
al
e
n
e
r
g
y
s
av
in
g
s
.
O
n
th
e
o
t
h
e
r
h
a
n
d
,
l
o
w
er
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
.
14
,
No
.
4
,
Dec
em
b
er
20
25
:
3
6
6
-
3
7
5
368
ap
p
r
o
x
i
m
a
ti
o
n
l
ev
els
a
r
e
r
e
q
u
ir
e
d
t
o
p
r
ese
r
v
e
ac
c
u
r
ac
y
in
h
i
g
h
-
p
r
ec
is
i
o
n
a
cti
v
i
ties
li
k
e
au
t
o
n
o
m
o
u
s
d
r
i
v
i
n
g
an
d
m
e
d
ic
al
i
m
a
g
i
n
g
.
C
NN
ac
c
ele
r
a
to
r
s
m
a
y
m
o
d
i
f
y
M
AC
p
r
ec
is
i
o
n
in
r
ea
l
-
ti
m
e
b
y
i
n
c
lu
d
i
n
g
a
d
ap
ti
v
e
ap
p
r
o
x
i
m
a
ti
o
n
t
ec
h
n
i
q
u
es,
g
u
ar
an
tee
in
g
a
n
id
ea
l
b
al
a
n
ce
b
etwe
e
n
a
cc
u
r
a
cy
a
n
d
p
o
we
r
u
s
e.
C
NN
h
ar
d
w
ar
e
m
a
y
b
e
m
a
d
e
m
o
r
e
v
e
r
s
atil
e
b
y
i
m
p
le
m
e
n
ti
n
g
r
u
n
ti
m
e
-
co
n
f
i
g
u
r
ab
le
M
AC
u
n
its
an
d
p
r
ec
is
e
tas
k
al
lo
ca
ti
o
n
.
A
b
asi
c
d
ia
g
r
a
m
o
f
t
h
e
s
t
r
u
c
tu
r
e
o
f
t
h
e
MA
C
u
n
it
is
s
h
o
w
n
i
n
Fig
u
r
e
1
.
T
h
e
i
n
p
u
ts
o
f
M
A
C
ar
e
t
h
e
ac
t
iv
ati
o
n
in
p
u
t(
X)
a
n
d
th
e
w
ei
g
h
t
(
Y)
.
Mu
l
ti
p
li
er
o
u
t
p
u
t
(
Z
)
an
d
r
eg
is
ter
v
al
u
e
(
R
)
ar
e
ad
d
ed
t
o
p
r
o
d
u
c
e
t
h
e
ac
c
u
m
u
lat
e
d
o
u
t
p
u
t
(
A
)
in
e
ac
h
cl
o
c
k
c
y
cl
e
.
T
h
e
e
q
u
at
io
n
o
f
t
h
e
g
e
n
e
r
a
liz
e
d
MA
C
is
g
i
v
e
n
i
n
(
1
)
.
=
(
∗
)
+
=
+
(
1
)
T
h
e
h
ar
d
war
e
C
NN
ac
ce
ler
ato
r
s
ar
e
g
en
e
r
ally
d
esig
n
e
d
ar
o
u
n
d
MA
C
b
lo
ck
s
.
E
f
f
icien
t
M
AC
b
ased
o
n
er
r
o
r
-
t
o
ler
ab
le
a
p
p
r
o
x
im
at
e
co
m
p
u
tin
g
r
esu
lts
in
ar
ea
an
d
en
e
r
g
y
-
ef
f
icien
t
h
ar
d
war
e
ac
ce
ler
atio
n
.
T
o
im
p
r
o
v
e
th
e
th
r
o
u
g
h
p
u
t,
th
e
c
o
m
m
o
n
weig
h
t
f
o
r
two
ac
tiv
a
tio
n
in
p
u
ts
is
m
u
ltip
lied
u
s
in
g
a
p
ar
allel
d
o
u
b
le
MA
C
o
p
er
atio
n
.
An
o
th
e
r
lo
w
co
m
p
lex
ity
d
esig
n
f
o
r
p
r
o
c
ess
in
g
ad
d
er
s
an
d
m
u
ltip
lier
s
is
u
s
in
g
s
to
ch
asti
c
co
m
p
u
tin
g
,
in
wh
ich
m
u
ltip
lie
r
s
ar
e
d
esig
n
ed
u
s
in
g
an
AND
g
ate
ar
r
ay
.
I
t
r
eq
u
i
r
es
a
lo
w
-
co
m
p
lex
ity
r
an
d
o
m
b
it
g
en
er
ato
r
f
o
r
g
e
n
er
atin
g
t
h
e
p
r
o
b
ab
ilit
y
.
Fu
s
ed
a
d
d
er
-
m
u
ltip
lier
d
esig
n
also
elim
in
ates
a
co
n
s
id
er
ab
le
am
o
u
n
t o
f
r
eso
u
r
ce
s
th
r
o
u
g
h
r
eu
s
in
g
th
e
ad
d
er
in
t
h
e
m
u
ltip
l
ier
p
o
r
tio
n
f
o
r
th
e
ac
c
u
m
u
latio
n
o
p
er
atio
n
.
So
m
e
p
h
y
s
ical
lev
el
ap
p
r
o
ac
h
es u
s
e
a
v
o
ltag
e
s
ca
lin
g
m
eth
o
d
to
i
n
ject
ap
p
r
o
x
im
atio
n
in
th
e
cir
cu
it to
lev
el
u
p
to
t
h
e
p
er
f
o
r
m
an
ce
is
n
o
t
af
f
ec
te
d
.
Fig
u
r
e
2
s
h
o
ws
th
e
v
a
r
io
u
s
ap
p
r
o
x
im
ati
o
n
m
eth
o
d
s
f
o
r
d
esig
n
in
g
an
ap
p
r
o
x
im
ate
m
u
ltip
lier
an
d
an
ap
p
r
o
x
im
ate
ad
d
e
r
-
b
ased
MA
C
u
n
it.
T
h
e
m
ajo
r
ity
o
f
r
ec
e
n
t
ap
p
r
o
x
im
ate
m
u
ltip
lier
u
s
es
ap
p
r
o
x
im
ate
co
m
p
r
ess
o
r
s
f
r
o
m
4
:
2
to
8
:2
f
o
r
p
ar
tial
p
r
o
d
u
ct
r
e
d
u
ctio
n
.
I
n
4
:2
co
m
p
r
ess
o
r
-
b
ased
MA
C
with
to
ler
ab
le
er
r
o
r
in
v
o
lv
es
th
e
in
ter
leav
in
g
o
f
p
o
s
itiv
e
an
d
n
eg
ativ
e
co
m
p
r
ess
o
r
s
.
A
lo
w
co
m
p
lex
ity
m
u
ltip
lier
d
esig
n
in
v
o
lv
in
g
a
f
r
ee
g
ate
-
b
ased
co
m
p
r
ess
o
r
u
s
in
g
p
r
o
b
ab
ilit
y
an
aly
s
is
is
d
e
s
cr
ib
ed
an
d
ap
p
lied
to
v
ar
io
u
s
er
r
o
r
-
to
ler
an
t
ap
p
licatio
n
s
.
Hig
h
-
p
er
f
o
r
m
an
ce
m
u
lti
-
lev
el
co
m
p
r
ess
o
r
s
in
v
o
lv
in
g
4
:2
,
5
:2
,
an
d
7
:2
with
a
m
o
d
if
ied
s
tr
u
ctu
r
e
in
wh
ich
ca
r
r
y
o
u
t
d
o
es
n
o
t
d
ep
e
n
d
o
n
ca
r
r
y
in
ar
e
d
escr
ib
ed
.
A
n
o
v
el
cu
s
to
m
ized
co
m
p
r
ess
o
r
is
u
s
ed
f
o
r
r
ed
u
ci
n
g
t
h
e
p
o
wer
co
n
s
u
m
p
tio
n
.
T
h
e
a
p
p
r
o
x
im
at
io
n
is
ap
p
lied
to
v
ar
io
u
s
tr
ee
-
b
ased
m
u
ltip
lier
d
esig
n
s
,
s
u
ch
as
th
e
W
allac
e
tr
ee
m
u
ltip
lier
an
d
Dad
d
a
m
u
ltip
li
er
.
Fig
u
r
e
1
.
Gen
e
r
al
MA
C
ar
ch
itectu
r
e
Fig
u
r
e
2
.
Ap
p
r
o
x
im
atio
n
m
et
h
o
d
in
v
o
lv
ed
i
n
MA
C
,
m
u
ltip
lier
,
an
d
a
d
d
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
s
u
r
ve
y
o
n
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
h
a
r
d
w
a
r
e
a
cc
elera
tio
n
…
(
S
u
vith
a
P
a
t
h
iya
d
a
n
S
u
d
h
a
ka
r
a
n
)
369
A
m
o
d
i
f
ie
d
b
o
o
t
h
m
u
l
ti
p
li
er
m
i
n
im
iz
es
t
h
e
p
ar
tia
l
p
r
o
d
u
ct
b
y
h
al
f
an
d
is
a
ls
o
a
p
p
r
o
x
i
m
at
ed
b
y
u
s
i
n
g
an
a
p
p
r
o
x
i
m
at
e
b
o
o
th
e
n
c
o
d
e
r
.
A
n
ap
p
r
o
x
i
m
a
te
B
o
o
t
h
m
u
l
tip
lie
r
b
ase
d
o
n
t
r
u
n
ca
ti
o
n
a
n
d
ca
r
r
y
-
b
ase
d
e
r
r
o
r
co
m
p
e
n
s
at
io
n
is
u
til
ize
d
,
a
n
d
a
n
a
p
p
r
o
x
i
m
a
te
r
a
d
i
x
-
4
B
o
o
t
h
e
n
c
o
d
e
r
s
u
it
ab
le
f
o
r
i
m
a
g
e
p
r
o
c
ess
in
g
a
p
p
lic
ati
o
n
s
is
d
esc
r
i
b
e
d
.
I
n
p
u
t
o
p
e
r
a
n
d
-
b
a
s
ed
s
e
g
m
e
n
ta
ti
o
n
is
als
o
a
n
o
t
h
er
m
et
h
o
d
o
l
o
g
y
t
h
a
t
m
i
n
i
m
iz
e
s
t
h
e
c
o
m
p
u
ta
ti
o
n
al
co
m
p
le
x
it
y
o
f
t
h
e
m
u
l
ti
p
li
er
b
y
o
p
e
r
a
ti
n
g
wi
th
o
n
l
y
a
n
o
n
-
z
e
r
o
e
le
m
e
n
t
p
o
r
ti
o
n
o
f
th
e
p
ar
ti
al
p
r
o
d
u
c
t.
Ad
d
er
is
u
s
ed
in
th
e
last
s
tag
e,
ad
d
itio
n
an
d
ac
cu
m
u
latin
g
o
f
th
e
d
ata
s
to
r
ed
in
th
e
MA
C
u
n
it.
I
n
th
e
L
SB
-
b
ased
ap
p
r
o
x
im
atio
n
lo
wer
p
ar
t
o
f
th
e
a
d
d
er
is
r
e
p
lace
d
with
a
co
n
s
tan
t
OR
-
g
ate,
wh
ich
m
in
im
izes
o
v
er
all
e
r
r
o
r
co
m
p
ar
ed
to
d
ir
e
ct
tr
u
n
ca
tio
n
.
E
r
r
o
r
m
et
r
ics
o
f
th
e
L
SB
ad
d
e
r
ar
e
ca
lcu
lated
an
d
c
o
m
p
ar
e
d
.
T
h
e
ca
r
r
y
s
p
ec
u
lativ
e
lo
g
ic
is
also
u
s
ed
in
v
ar
io
u
s
liter
atu
r
e
wo
r
k
s
f
o
r
m
in
im
izin
g
th
e
cr
itical
p
ath
d
elay
b
y
p
r
e
-
g
en
er
atin
g
th
e
ca
r
r
y
.
An
er
r
o
r
co
m
p
en
s
atio
n
cir
cu
it
is
in
co
r
p
o
r
ate
d
in
to
b
lo
ck
-
b
ased
s
p
ec
u
latio
n
to
r
ec
tify
s
p
ec
u
lativ
e
er
r
o
r
s
.
T
o
m
ak
e
th
e
ad
d
er
a
p
p
r
o
p
r
iate
f
o
r
h
ig
h
-
s
p
ee
d
ap
p
licatio
n
s
,
h
ig
h
-
s
p
ee
d
p
ar
allel
p
r
ef
i
x
ad
d
er
s
lik
e
B
r
en
t
-
Ku
n
g
,
Han
ca
r
ls
o
n
,
Sk
lan
s
k
y
,
a
n
d
K
o
g
g
e
Sto
n
e
a
r
e
also
ap
p
r
o
x
im
ated
b
y
elim
in
atin
g
th
e
in
ter
m
ed
iate
p
r
o
p
ag
ate
-
g
e
n
er
ate
s
tag
e
to
r
ed
u
ce
th
e
r
eso
u
r
ce
u
tili
za
tio
n
.
T
ab
le
1
lis
ts
ap
p
r
o
x
im
ate
MA
C
d
esig
n
s
in
v
o
lv
in
g
v
ar
i
o
u
s
m
u
l
tip
lier
s
an
d
ad
d
er
s
in
th
e
r
ec
en
t liter
atu
r
e.
T
ab
le
1
.
Ap
p
r
o
x
i
m
ate
MA
C
d
esig
n
in
v
o
lv
in
g
v
a
r
io
u
s
m
u
ltip
lier
s
an
d
ad
d
er
s
M
e
t
h
o
d
o
l
o
g
y
D
e
v
i
c
e
B
i
t
w
i
d
t
h
A
p
p
l
i
c
a
t
i
o
n
Ty
p
e
R
e
s
u
l
t
o
u
t
c
o
me
D
o
u
b
l
e
M
A
C
[
1
3
]
X
i
l
i
n
x
V
i
r
t
e
x
7
4
8
5
T
F
P
G
A
8
-
b
i
t
A
l
e
x
N
e
t
a
n
d
V
G
G
P
e
r
f
o
r
ma
n
c
e
i
m
p
r
o
v
e
me
n
t
f
r
o
m 1
4
%
to
8
0
%
M
A
C
u
s
i
n
g
a
p
p
r
o
x
i
ma
t
e
c
o
m
p
r
e
ss
o
r
w
i
t
h
b
a
l
a
n
c
e
d
e
r
r
o
r
a
c
c
u
mu
l
a
t
i
o
n
[
1
4
]
S
y
n
o
p
sy
s
d
e
si
g
n
c
o
m
p
i
l
e
r
i
n
a
2
8
n
m
C
M
O
S
16
-
b
i
t
M
o
b
i
l
e
N
e
t
V
2
a
n
d
S
q
u
e
e
z
e
N
e
t
En
e
r
g
y
o
p
t
i
m
i
z
a
t
i
o
n
b
y
m
o
r
e
t
h
a
n
3
5
%
R
e
c
u
r
si
v
e
m
u
l
t
i
p
l
i
c
a
t
i
o
n
[
1
5
]
4
5
n
m
t
e
c
h
n
o
l
o
g
y
n
o
d
e
w
i
t
h
TS
M
C
l
i
b
r
a
r
y
8
-
b
i
t
G
a
u
ss
i
a
n
sm
o
o
t
h
i
n
g
6
0
%
e
n
e
r
g
y
s
a
v
i
n
g
s
R
a
d
i
x
-
4
a
l
g
o
r
i
t
h
m
a
l
o
n
g
w
i
t
h
c
a
n
o
n
i
c
a
l
r
e
p
r
e
se
n
t
a
t
i
o
n
[
1
6
]
45
n
m
a
n
a
l
o
g
d
e
si
g
n
u
si
n
g
c
a
d
e
n
c
e
8
-
b
i
t
A
l
e
x
N
e
t
En
e
r
g
y
e
f
f
i
c
i
e
n
c
y
a
c
h
i
e
v
e
d
b
y
4
1
.
6
%
B
i
n
a
r
y
w
a
r
e
h
a
r
d
w
a
r
e
a
c
c
e
l
e
r
a
t
o
r
[
1
7
]
2
8
n
m
C
M
O
S
t
e
c
h
n
o
l
o
g
y
1
0
2
4
-
b
i
t
A
l
e
x
N
e
t
a
n
d
V
G
G
N
e
t
s
H
i
g
h
t
h
r
o
u
g
h
p
u
t
b
y
1
.
5
–
1
3
.
3
×
P
a
r
a
l
l
e
l
m
u
l
t
i
p
l
i
c
a
t
i
o
n
u
n
i
t
s
[
1
8
]
X
i
l
i
n
x
Z
y
n
q
-
7
0
0
0
S
o
C
–
F
P
G
A
d
e
v
i
c
e
16
-
b
i
t
C
N
N
c
o
n
v
o
l
u
t
i
o
n
a
l
a
n
d
f
u
l
l
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
A
r
e
a
r
e
d
u
c
t
i
o
n
s
o
f
2
8
.
1
9
%
−
5
6
.
0
9
%
R
u
n
t
i
me
a
c
c
u
r
a
c
y
c
o
n
f
i
g
u
r
a
b
l
e
b
y
v
o
l
t
a
g
e
o
v
e
r
s
c
a
l
i
n
g
[
1
9
]
1
5
n
m F
i
n
F
ET
t
e
c
h
n
o
l
o
g
y
8
-
b
i
t
N
V
D
LA
P
o
w
e
r
e
f
f
i
c
i
e
n
c
y
a
c
h
i
e
v
e
d
t
h
r
o
u
g
h
F
i
n
F
ET
t
e
c
h
n
o
l
o
g
y
A
p
p
r
o
x
i
ma
t
e
c
o
mp
u
t
i
n
g
M
A
C
u
si
n
g
I
n
t
e
r
n
a
l
-
S
e
l
f
-
H
e
a
l
i
n
g
[
2
0
]
TSM
C
4
0
n
m
t
e
c
h
n
o
l
o
g
y
u
si
n
g
sy
n
o
p
si
s
d
e
si
g
n
c
o
m
p
i
l
e
r
16
-
b
i
t
R
a
d
i
o
a
st
r
o
n
o
m
y
c
a
l
i
b
r
a
t
i
o
n
p
r
o
c
e
ss
i
n
g
1
8
%
a
r
e
a
a
n
d
1
4
%
p
o
w
e
r
r
e
d
u
c
t
i
o
n
I
n
p
u
t
-
C
o
n
sc
i
o
u
s
A
p
p
r
o
x
i
m
a
t
e
M
A
C
[
2
1
]
V
i
r
t
e
x
-
6
X
C
6
V
LX
7
5
T
F
P
G
A
8
-
b
i
t
I
mag
e
b
l
e
n
d
i
n
g
a
n
d
G
a
u
ss
i
a
n
sm
o
o
t
h
i
n
g
6
5
%
e
n
e
r
g
y
e
f
f
i
c
i
e
n
c
y
a
n
d
5
%
i
n
c
r
e
a
s
e
i
n
r
e
so
u
r
c
e
u
t
i
l
i
z
a
t
i
o
n
3.
M
E
T
H
O
D
T
h
is
s
ec
tio
n
d
escr
ib
es
t
h
e
C
NN
ac
ce
ler
atio
n
f
o
r
f
ea
tu
r
e
e
x
tr
ac
tio
n
a
n
d
class
if
icatio
n
p
e
r
f
o
r
m
a
n
ce
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
em
p
h
asizes
o
p
tim
izin
g
C
NN
h
ar
d
war
e
ac
ce
ler
atio
n
th
r
o
u
g
h
a
p
p
r
o
x
im
ate
MA
C
-
b
ased
ar
ch
itectu
r
es
th
at
en
h
an
ce
co
m
p
u
tatio
n
al
ef
f
icien
cy
with
o
u
t
s
ig
n
if
ican
tly
co
m
p
r
o
m
is
in
g
ac
cu
r
ac
y
.
T
h
e
C
NN
s
tr
u
ctu
r
e
in
teg
r
ates
co
n
v
o
lu
tio
n
al,
R
eL
U,
p
o
o
lin
g
,
an
d
f
u
lly
co
n
n
ec
ted
lay
er
s
,
wh
er
e
t
h
e
c
o
n
v
o
lu
ti
o
n
al
lay
e
r
s
p
er
f
o
r
m
th
e
co
r
e
f
ea
tu
r
e
ex
tr
a
ctio
n
u
s
in
g
o
p
tim
ized
ap
p
r
o
x
im
ate
m
u
ltip
lier
s
an
d
a
d
d
er
s
.
T
h
e
p
r
o
p
o
s
ed
C
NN
ac
ce
ler
ato
r
,
im
p
lem
en
ted
o
n
FP
GA
p
latf
o
r
m
s
,
ac
h
iev
es
s
u
b
s
tan
tial
r
ed
u
ctio
n
s
in
p
o
wer
co
n
s
u
m
p
tio
n
an
d
laten
cy
wh
ile
m
ain
tain
in
g
h
ig
h
class
if
icatio
n
ac
cu
r
ac
y
,
m
ak
in
g
it
s
u
itab
le
f
o
r
r
ea
l
-
tim
e,
r
e
s
o
u
r
ce
-
co
n
s
tr
ain
e
d
,
an
d
en
er
g
y
-
s
en
s
itiv
e
ap
p
licatio
n
s
.
3
.
1
.
Ana
ly
s
is
o
f
v
a
rio
us
des
i
g
ns
inv
o
lv
ing
CNN
ha
rdwa
r
e
a
cc
eler
a
t
io
n
Fig
u
r
e
3
d
e
p
i
cts
t
h
e
d
i
f
f
er
e
n
t
la
y
er
s
o
f
th
e
C
NN
a
n
d
th
ei
r
i
n
te
r
co
n
n
e
cti
o
n
.
C
o
n
v
o
l
u
ti
o
n
al,
R
e
L
U
,
Ma
x
p
o
o
li
n
g
,
an
d
f
u
ll
y
c
o
n
n
e
cte
d
la
y
er
s
m
a
k
e
u
p
t
h
e
C
N
N
ar
ch
ite
ct
u
r
e.
C
o
n
v
o
l
u
t
io
n
al
l
ay
er
f
o
r
m
s
a
co
r
e
o
p
e
r
at
io
n
i
n
v
o
l
v
i
n
g
k
e
r
n
e
l
-
b
as
ed
m
at
r
i
x
co
m
p
u
tat
io
n
s
t
o
ex
t
r
a
ct
i
m
p
o
r
ta
n
t
f
ea
t
u
r
es
f
r
o
m
t
h
e
ac
t
iv
ati
n
g
i
n
p
u
t.
I
t
u
s
es
a
wi
n
d
o
w
in
g
m
e
th
o
d
to
p
r
o
ce
s
s
e
ac
h
s
m
al
l
s
e
g
m
en
t
o
f
t
h
e
i
n
p
u
t
d
at
a.
T
h
e
r
e
ar
e
v
a
r
i
o
u
s
ac
ti
v
at
in
g
f
u
n
cti
o
n
s
a
v
ai
la
b
le
,
b
u
t
R
eL
U
is
co
m
m
o
n
l
y
u
s
e
d
t
o
eli
m
i
n
at
e
t
h
e
n
e
g
at
iv
e
te
r
m
s
to
p
e
r
f
ec
t
th
e
f
ea
t
u
r
e
ex
t
r
ac
ti
o
n
p
r
o
ce
s
s
wit
h
o
u
t
h
ar
d
wa
r
e
c
o
m
p
le
x
it
y
.
P
o
o
li
n
g
l
ay
er
s
,
w
h
i
ch
a
r
e
cl
ass
i
f
ie
d
as
g
l
o
b
al
a
v
e
r
ag
e
p
o
o
li
n
g
an
d
m
a
x
im
u
m
p
o
o
li
n
g
,
a
r
e
p
l
ac
e
d
b
et
wee
n
v
a
r
i
o
u
s
c
o
n
v
o
l
u
ti
o
n
al
la
y
e
r
s
t
o
m
in
im
i
ze
t
h
e
d
at
a
v
o
l
u
m
e
t
o
b
e
p
r
o
ce
s
s
e
d
b
y
s
el
ec
ti
n
g
th
e
m
a
x
i
m
u
m
o
r
a
v
e
r
a
g
e
o
f
t
h
e
c
o
n
v
o
l
u
ti
o
n
al
r
es
u
l
t.
F
u
l
ly
c
o
n
n
e
cte
d
l
ay
er
s
f
o
r
m
an
in
t
er
f
a
ce
b
e
twe
en
in
p
u
t
a
n
d
o
u
tp
u
t
l
a
y
e
r
s
,
l
ik
e
a
n
e
u
r
o
n
,
an
d
r
e
q
u
ir
e
a
l
o
t
o
f
s
to
r
ag
e
f
o
r
s
to
r
i
n
g
t
h
e
in
t
er
m
e
d
i
at
e
d
a
ta
.
Sy
s
to
lic
ar
r
ay
is
o
n
e
o
f
th
e
ef
f
ec
tiv
e
ar
ch
itectu
r
es
en
h
an
cin
g
th
e
d
ata
p
ath
in
th
e
m
ac
h
in
e
lear
n
in
g
ac
ce
ler
ato
r
.
I
n
a
h
y
b
r
id
f
ac
t
o
r
ed
ad
d
er
a
n
d
ac
c
u
m
u
latio
n
b
ased
o
n
a
s
y
s
to
lic
ar
ch
itect
u
r
e
f
o
r
g
e
n
er
alize
d
m
atr
ix
m
u
ltip
licatio
n
in
ter
f
ac
e
is
u
tili
ze
d
,
wh
ich
ca
n
b
e
ap
p
lied
to
v
ar
io
u
s
d
ee
p
lear
n
in
g
ap
p
licatio
n
s
.
T
h
e
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
.
14
,
No
.
4
,
Dec
em
b
er
20
25
:
3
6
6
-
3
7
5
370
h
ar
d
war
e
ac
ce
ler
atio
n
in
C
NN
is
d
o
n
e
th
r
o
u
g
h
v
ar
io
u
s
co
m
p
r
ess
io
n
an
d
a
r
ch
itectu
r
al
m
o
d
if
icatio
n
s
.
Ver
y
r
ec
en
t
r
esear
ch
f
o
c
u
s
ed
o
n
co
m
p
r
ess
io
n
tech
n
iq
u
es
lik
e
th
e
d
ictio
n
ar
y
m
eth
o
d
,
wh
ich
is
u
s
ed
to
m
in
im
ize
th
e
s
to
r
ag
e
s
p
ac
e
wh
e
n
s
im
ilar
-
weig
h
t
d
ata
ar
e
p
r
esen
t.
I
n
a
th
r
ee
-
lay
e
r
ed
a
p
p
r
o
ac
h
th
r
o
u
g
h
weig
h
t
p
r
u
n
in
g
,
co
m
p
r
ess
io
n
,
a
n
d
d
ec
o
m
p
r
e
s
s
io
n
o
f
p
r
o
ce
s
s
in
g
elem
en
t
s
th
r
o
u
g
h
th
e
d
ictio
n
a
r
y
m
eth
o
d
a
r
e
u
tili
ze
d
.
R
ed
u
n
d
an
cy
r
e
m
o
v
al
th
r
o
u
g
h
in
tr
a
-
an
d
in
ter
-
ch
a
n
n
el
co
m
p
r
ess
io
n
is
ad
o
p
ted
f
o
r
r
es
o
u
r
ce
m
in
im
izatio
n
.
Op
tim
ized
weig
h
t
allo
ca
tio
n
b
ased
o
n
f
in
e
-
g
r
ain
ed
ar
ch
itect
u
r
e
is
u
tili
ze
d
to
im
p
r
o
v
e
co
m
p
r
ess
io
n
ef
f
icien
c
y
.
I
n
C
NN
ac
ce
ler
ato
r
s
,
ap
p
r
o
x
im
atio
n
in
tr
o
d
u
ce
s
co
m
p
u
tatio
n
al
er
r
o
r
th
at
m
ig
h
t
ca
u
s
e
v
ar
ian
ce
s
in
o
u
tp
u
t
p
r
ed
ictio
n
s
,
h
en
ce
af
f
ec
tin
g
i
n
f
er
en
ce
ac
cu
r
ac
y
.
Stro
n
g
ap
p
r
o
x
im
atio
n
s
ca
n
s
ig
n
if
ican
tly
r
ed
u
ce
ac
cu
r
ac
y
,
wh
er
ea
s
g
en
tle
ap
p
r
o
x
i
m
atio
n
s
m
ay
h
av
e
m
in
im
al
co
n
s
eq
u
e
n
ce
s
.
Par
ticu
lar
ly
in
d
ee
p
n
etwo
r
k
s
,
ap
p
r
o
x
im
ate
MA
C
u
n
its
,
wh
ich
d
ec
r
ea
s
e
ac
cu
r
ac
y
to
i
n
cr
ea
s
e
ef
f
icien
c
y
,
h
av
e
t
h
e
p
o
te
n
tial
to
m
ag
n
if
y
s
m
all
er
r
o
r
s
i
n
ac
tiv
atio
n
s
an
d
weig
h
ts
,
r
es
u
ltin
g
in
in
co
r
r
ec
t
class
if
icatio
n
s
.
Ap
p
r
o
x
im
ate
co
m
p
u
ta
tio
n
s
ca
n
ass
is
t
i
n
p
r
eser
v
in
g
a
lev
el
o
f
ac
cu
r
ac
y
wh
ile
g
ain
in
g
ad
v
a
n
tag
es
in
t
er
m
s
o
f
r
e
d
u
ce
d
p
o
wer
u
s
ag
e
an
d
d
elay
.
A
s
p
ar
s
e
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
is
also
ad
o
p
ted
in
v
ar
io
u
s
d
esig
n
s
alo
n
g
with
s
elf
s
elf
-
r
ec
o
v
er
y
s
o
lu
tio
n
d
u
r
in
g
a
m
alf
u
n
ctio
n
in
th
e
p
r
o
ce
s
s
in
g
elem
en
t.
Sp
ec
u
lativ
e
ap
p
r
o
a
ch
es
ar
e
also
u
s
ed
in
er
r
o
r
-
to
l
er
an
t
d
esig
n
,
wh
er
e
th
e
weig
h
t
v
alu
e
d
o
esn
’
t
h
av
e
m
u
ch
im
p
ac
t
in
th
e
o
u
tp
u
t.
Secu
r
ity
is
s
u
es
s
u
ch
as
s
id
e
-
ch
an
n
el
attac
k
s
ar
e
ex
p
o
s
ed
b
y
p
o
wer
f
lu
ctu
at
io
n
s
.
T
o
p
r
o
tect
ag
ain
s
t
s
id
e
-
ch
an
n
el
attac
k
s
,
r
a
n
d
o
m
ized
ap
p
r
o
x
im
ate
co
m
p
u
tatio
n
a
d
d
s
co
n
tr
o
lled
r
an
d
o
m
izatio
n
.
C
NN
o
u
tp
u
t
s
ar
e
p
r
o
tecte
d
f
r
o
m
s
m
all
er
r
o
r
s
b
y
th
e
e
r
r
o
r
-
r
esil
ien
t
ap
p
r
o
x
im
atio
n
.
T
h
e
m
o
d
el
e
x
tr
ac
tio
n
is
p
r
ev
e
n
ted
b
y
h
ar
d
war
e
-
lev
el
d
is
g
u
is
e
m
eth
o
d
s
,
in
cl
u
d
in
g
d
y
n
am
ic
r
ec
o
n
f
ig
u
r
atio
n
an
d
n
o
is
e
in
jectio
n
.
H
y
b
r
id
ap
p
r
o
x
im
ate
-
ex
ac
t
co
m
p
u
tin
g
an
d
s
ec
u
r
e
f
a
u
lt
d
etec
tio
n
also
im
p
r
o
v
e
s
ec
u
r
ity
with
o
u
t
s
ac
r
if
icin
g
e
f
f
ec
tiv
en
ess
.
T
h
e
C
NN
-
b
ased
h
a
r
d
war
e
ac
c
eler
atio
n
ap
p
lied
t
o
ab
n
o
r
m
al
h
ea
r
tb
ea
t
d
etec
tio
n
f
o
r
class
if
y
in
g
v
ar
i
o
u
s
ab
n
o
r
m
alities
was
a
ls
o
s
tu
d
ied
in
th
is
s
u
r
v
ey
wo
r
k
.
I
n
T
ab
le
2
,
r
elate
d
wo
r
k
s
in
v
o
lv
i
n
g
C
NN
h
ar
d
war
e
ac
ce
ler
atio
n
in
th
e
s
u
r
v
ey
ar
e
lis
ted
.
Fig
u
r
e
3
.
C
NN
lay
er
s
an
d
t
h
eir
in
ter
co
n
n
ec
tio
n
T
ab
le
2
.
R
elate
d
wo
r
k
s
in
v
o
lv
in
g
C
NN
h
ar
d
war
e
ac
ce
ler
atio
n
M
e
t
h
o
d
o
l
o
g
y
D
e
v
i
c
e
B
i
t
w
i
d
t
h
A
p
p
l
i
c
a
t
i
o
n
Ty
p
e
R
e
s
u
l
t
o
u
t
c
o
me
H
y
b
r
i
d
a
c
c
u
m
u
l
a
t
o
r
f
a
c
t
o
r
e
d
sy
st
o
l
i
c
a
r
r
a
y
[
2
2
]
32
-
n
m
t
e
c
h
n
o
l
o
g
y
u
s
i
n
g
S
y
n
o
p
sy
s
D
e
si
g
n
C
o
m
p
i
l
e
r
32
-
b
i
t
G
e
n
e
r
a
l
m
a
t
r
i
x
mu
l
t
i
p
l
i
c
a
t
i
o
n
i
n
t
e
r
f
a
c
e
A
r
e
a
e
f
f
i
c
i
e
n
c
y
i
n
t
h
e
r
a
n
g
e
1
2
.
8
%
−
5
0
.
2
%
a
n
d
i
n
p
o
w
e
r
e
f
f
i
c
i
e
n
c
y
b
e
t
w
e
e
n
1
8
.
6
%
−
4
1
%
S
p
a
r
se
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
a
c
c
e
l
e
r
a
t
o
r
w
i
t
h
p
r
e
-
e
n
c
o
d
i
n
g
[
2
3
]
2
8
n
m
u
s
i
n
g
S
y
n
o
p
s
y
s Des
i
g
n
C
o
m
p
i
l
e
r
8
-
b
i
t
V
G
G
1
6
,
A
l
e
x
N
e
t
,
M
o
b
i
l
e
N
e
t
V
3
En
e
r
g
y
e
f
f
i
c
i
e
n
c
y
o
f
A
c
c
e
l
e
r
a
t
o
r
a
c
h
i
e
v
e
d
b
y
9
0
.
0
3
%
F
i
r
e
F
l
y
:
H
i
g
h
-
p
e
r
f
o
r
m
a
n
c
e
n
e
u
r
a
l
n
e
t
w
o
r
k
u
si
n
g
D
S
P
p
r
o
c
e
sso
r
[
2
4
]
V
a
r
i
o
u
s X
i
l
i
n
x
F
P
G
A
f
a
m
i
l
i
e
s
u
se
d
a
r
e
x
c
7
k
3
2
5
t
,
a
n
d
x
c
v
u
4
4
0
64
-
b
i
t
N
V
I
D
I
A
p
e
a
k
p
e
r
f
o
r
ma
n
c
e
o
f
5
.
5
3
TO
P
/
s(
Tr
i
l
l
i
o
n
s
o
f
O
p
e
r
a
t
i
o
n
s
P
e
r
S
e
c
o
n
d
)
a
t
3
0
0
M
H
z
A
c
c
e
l
e
r
a
t
o
r
b
a
se
d
o
n
C
o
m
p
r
e
ssi
o
n
[
2
5
]
28
n
m
TS
M
C
u
s
i
n
g
S
y
n
o
p
s
y
s
D
e
si
g
n
C
o
mp
i
l
e
r
(
D
C
)
16
-
b
i
t
VGG
-
1
6
,
R
e
sN
e
t
-
5
0
a
n
d
I
n
c
e
p
t
i
o
n
-
v3
C
o
m
p
r
e
ss
i
o
n
r
a
t
i
o
s
9
.
8
%
∼
1
9
.
3
%
S
p
e
c
u
l
a
t
i
v
e
p
r
o
c
e
ss
i
n
g
f
o
r
e
n
e
r
g
y
e
f
f
i
c
i
e
n
t
a
c
c
e
l
e
r
a
t
o
r
[
2
6
]
40
-
n
m
C
M
O
S
p
r
o
c
e
ss
d
e
s
i
g
n
c
o
m
p
i
l
e
r
f
r
o
m S
y
n
o
p
sy
s
8
-
b
i
t
A
l
e
x
N
e
t
M
o
r
e
t
h
a
n
2
0
%
e
n
e
r
g
y
e
f
f
i
c
i
e
n
c
y
w
i
t
h
mi
n
i
mal
a
c
c
u
r
a
c
y
d
e
g
r
a
d
a
t
i
o
n
a
n
d
l
o
w
e
r
i
mp
l
e
m
e
n
t
a
t
i
o
n
o
v
e
r
h
e
a
d
Th
e
t
h
r
e
e
-
s
t
e
p
a
p
p
r
o
a
c
h
t
r
i
mm
i
n
g
o
f
w
e
i
g
h
t
s
,
d
i
c
t
i
o
n
a
r
y
-
c
o
mp
r
e
ssi
o
n
a
n
d
d
e
c
o
m
p
r
e
ss
i
o
n
[
2
7
]
S
y
n
o
p
sy
s
t
o
o
l
w
i
t
h
9
0
n
m
t
e
c
h
n
o
l
o
g
y
8
-
b
i
t
R
e
sN
e
t
-
5
0
H
i
g
h
e
r
c
o
m
p
r
e
ss
i
o
n
e
f
f
i
c
i
e
n
c
y
w
i
t
h
o
p
t
i
m
i
z
e
d
mem
o
r
y
u
s
a
g
e
H
i
g
h
-
p
e
r
f
o
r
ma
n
c
e
a
d
d
i
t
i
o
n
f
o
r
h
a
r
d
w
a
r
e
a
c
c
e
l
e
r
a
t
o
r
o
f
n
e
u
r
a
l
n
e
t
w
o
r
k
s
[
2
8
]
C
a
d
e
n
c
e
V
i
r
t
u
o
so
t
o
o
l
w
i
t
h
45
-
n
m
p
r
o
c
e
ss
8
-
b
i
t
R
ESN
ET
-
18
D
o
u
b
l
e
t
h
r
o
u
g
h
p
u
t
a
l
o
n
g
w
i
t
h
15
-
2
0
%
a
r
e
a
a
n
d
p
o
w
e
r
e
f
f
i
c
i
e
n
c
y
C
o
m
p
r
e
ss
i
n
g
p
r
o
c
e
ss
i
n
g
e
l
e
me
n
t
b
a
s
e
d
o
n
w
e
i
g
h
t
s
p
a
r
s
i
t
y
[
2
9
]
S
y
n
o
p
sy
s
d
e
si
g
n
c
o
m
p
i
l
e
r
w
i
t
h
N
a
n
g
a
t
e
4
5
-
n
m O
p
e
n
C
e
l
l
L
i
b
r
a
r
y
1
2
8
-
b
i
t
C
I
F
A
R
-
1
0
0
a
n
d
I
mag
e
N
e
t
d
a
t
a
s
e
t
H
i
g
h
t
h
r
o
u
g
h
p
u
t
a
n
d
p
o
w
e
r
o
p
t
i
m
i
z
a
t
i
o
n
a
c
h
i
e
v
e
d
b
y
3
0
%
H
a
r
d
w
a
r
e
d
e
si
g
n
f
o
r
C
N
N
f
o
r
EC
G
a
b
n
o
r
m
a
l
i
t
y
[
3
0
]
X
i
l
i
n
x
Z
y
n
q
ZC
7
0
6
16
-
b
i
t
1
-
D
C
N
N
A
c
c
u
r
a
c
y
f
o
r
v
a
r
i
o
u
s
EC
G
a
b
n
o
r
m
a
l
i
t
i
e
s
i
s
9
9
.
1
0
%
A
b
n
o
r
ma
l
E
C
G
d
e
t
e
c
t
i
o
n
u
si
n
g
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
[
3
1
]
TSM
C
0
.
1
8
µ
m
C
M
O
S
u
si
n
g
S
y
n
o
p
sy
s
t
o
o
l
8
-
b
i
t
M
I
T
-
B
I
H
9
6
.
3
%
a
c
c
u
r
a
c
y
f
o
r
d
i
f
f
e
r
e
n
t
EC
G
sa
mp
l
e
s
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
s
u
r
ve
y
o
n
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
h
a
r
d
w
a
r
e
a
cc
elera
tio
n
…
(
S
u
vith
a
P
a
t
h
iya
d
a
n
S
u
d
h
a
ka
r
a
n
)
371
4.
RE
SU
L
T
AND
DI
SCUS
SI
O
NS
T
h
is
s
tu
d
y
im
p
lem
e
n
ts
a
p
r
o
t
o
ty
p
e
u
s
in
g
Xilin
x
Z
y
n
q
FP
GAs
to
v
er
if
y
p
o
wer
s
av
in
g
s
.
T
h
is
s
tu
d
y
co
m
p
ar
es
ex
ac
t
an
d
ap
p
r
o
x
i
m
ate
MA
C
im
p
lem
en
tatio
n
s
to
ass
e
s
s
C
NN
ac
cu
r
ac
y
r
ed
u
ctio
n
ca
u
s
ed
b
y
ap
p
r
o
x
im
ate
MA
C
o
p
er
atio
n
s
.
C
NN
m
is
cla
s
s
if
icatio
n
r
ates
o
u
g
h
t
to
b
e
co
n
n
ec
ted
with
er
r
o
r
m
ea
s
u
r
es
s
u
ch
as
MRED
an
d
NM
E
D.
T
h
e
s
e
f
in
d
in
g
s
will
im
p
r
o
v
e
th
e
s
tu
d
y
b
y
o
f
f
e
r
in
g
a
th
o
r
o
u
g
h
ev
alu
atio
n
o
f
ap
p
r
o
x
im
atio
n
ef
f
ec
ts
.
Fo
r
an
ap
p
r
o
x
im
ate
MA
C
-
b
ased
C
NN,
th
e
v
ar
iatio
n
in
th
e
ex
ac
t
s
u
m
o
u
tp
u
t
an
d
th
e
an
ticip
ated
s
u
m
o
u
tp
u
t
is
k
n
o
wn
as
th
e
er
r
o
r
d
is
tan
c
e.
T
h
e
Me
an
R
E
D
(
MRED)
f
o
r
ev
e
r
y
N
-
b
it
ap
p
r
o
x
im
atio
n
ad
d
er
is
g
iv
e
n
i
n
(
2
)
,
=
1
2
2
∑
2
2
=
1
(
2
)
wh
er
e
m
ea
n
er
r
o
r
d
is
tan
ce
(
ME
D)
is
th
e
av
er
ag
e
er
r
o
r
d
is
tan
ce
an
d
Mi
is
p
r
ec
is
e
m
u
ltip
licatio
n
.
ME
D
n
o
r
m
alize
d
is
g
iv
e
n
in
(
3
)
,
=
1
2
2
(
2
−
1
)
2
∑
2
2
=
1
(
3
)
wh
er
e
E
D
is
th
e
er
r
o
r
r
ate
o
f
t
h
e
m
etr
ic.
As
m
en
tio
n
ed
in
Fig
u
r
e
2
,
v
ar
io
u
s
ap
p
r
o
x
im
ate
a
d
d
er
an
d
m
u
ltip
lier
d
esig
n
s
ar
e
u
tili
ze
d
in
d
esig
n
in
g
a
p
p
r
o
x
im
ate
m
u
ltip
le
an
d
ac
cu
m
u
late
u
n
its
.
T
h
e
m
u
ltip
lier
u
s
es
v
ar
io
u
s
co
m
p
r
ess
io
n
-
b
ased
d
esig
n
s
,
tr
ee
-
b
ased
r
ed
u
ctio
n
s
tr
u
ctu
r
es,
in
p
u
t
o
p
er
an
d
-
b
a
s
ed
an
aly
s
is
,
r
ad
ix
-
4
en
c
o
d
er
d
esig
n
,
an
d
e
r
r
o
r
co
m
p
en
s
atio
n
ar
ch
itectu
r
es.
T
h
ese
d
esig
n
s
in
v
o
lv
in
g
r
ec
en
t
liter
atu
r
e
ar
e
lis
ted
in
T
ab
le
3
.
T
h
e
co
m
p
r
ess
io
n
-
b
ased
ar
ch
itectu
r
e
o
cc
u
p
ies
m
in
im
al
ar
ea
an
d
p
o
wer
co
n
s
u
m
p
tio
n
.
Fre
e
g
ate
co
m
p
r
es
s
o
r
-
b
ased
m
u
ltip
lier
d
esig
n
s
T
E
PM1
a
n
d
T
E
PM2
r
esu
lt
in
m
in
im
al
d
elay
co
m
p
ar
ed
to
o
th
er
tr
ee
-
b
ased
a
n
d
b
o
th
m
u
ltip
lier
ar
ch
itectu
r
es.
T
h
e
er
r
o
r
co
m
p
en
s
atio
n
m
u
ltip
lier
AB
ME
C
d
esig
n
h
as
an
ar
ea
o
v
e
r
h
ea
d
f
o
r
th
e
er
r
o
r
co
r
r
ec
tio
n
cir
c
u
it
co
m
p
a
r
ed
t
o
o
th
er
d
esig
n
s
in
T
ab
le
3
.
A
n
in
p
u
t
o
p
er
an
d
-
b
ased
s
eg
m
e
n
ted
AM
SS
d
esig
n
co
n
s
u
m
es m
in
im
al
p
o
wer
b
y
d
is
ca
r
d
in
g
ce
r
tain
ir
r
ele
v
an
t p
o
r
tio
n
s
o
f
th
e
m
u
ltip
lier
.
T
ab
le
3
.
C
o
m
p
a
r
ativ
e
an
aly
s
is
f
o
r
ap
p
r
o
x
im
ate
m
u
ltip
lier
in
ex
is
tin
g
an
d
p
r
o
p
o
s
ed
tech
n
iq
u
es
D
e
si
g
n
A
r
e
a
(
u
m
2
)
D
e
l
a
y
(
n
s)
P
o
w
e
r
(
u
W
)
P
D
P
(
f
J)
ADP
N
M
ED
TEPM
1
[
3
2
]
3
3
1
.
4
0
.
0
9
7
6
.
1
6
.
8
2
9
.
8
0
.
0
1
8
TEPM
2
[
3
2
]
3
1
8
.
5
0
.
1
0
7
3
.
4
7
.
3
3
4
.
4
0
.
0
1
7
A
B
M
E
C
[
3
3
]
1
0
6
2
.
6
1
.
4
8
7
1
.
2
1
0
5
.
3
8
1
5
7
2
.
6
5
0
.
8
5
A
B
M
C
B
E
C
[
3
4
]
3
0
7
0
.
9
5
1
6
.
3
1
1
5
.
4
9
2
9
.
6
5
0
.
5
6
7
A
D
M
1
[
3
5
]
4
3
8
0
.
9
4
3
7
.
9
3
3
5
.
6
5
4
1
1
.
7
2
0
.
7
5
4
A
M
S
S
[
3
6
]
6
6
.
5
0
.
1
6
5
4
9
.
9
8
.
2
3
1
0
.
9
7
0
.
0
4
P
r
o
p
o
se
d
5
4
.
2
0
.
0
8
1
4
.
1
1
5
.
3
2
2
.
1
3
0
.
0
3
T
ab
le
4
co
m
p
ar
es
p
r
o
p
o
s
ed
ap
p
r
o
x
im
ate
C
NN
ac
ce
ler
ato
r
s
with
ex
is
tin
g
C
NNs.
T
h
e
f
in
d
in
g
s
d
em
o
n
s
tr
ate
th
at
ap
p
r
o
x
im
ati
o
n
ac
ce
ler
ato
r
s
u
s
e
less
p
o
wer
an
d
p
er
f
o
r
m
q
u
ick
e
r
.
T
h
e
r
e
s
u
lts
s
h
o
w
th
at
th
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
ac
h
iev
es
a
s
ig
n
if
ican
t
im
p
r
o
v
em
en
t
in
b
o
th
p
o
we
r
ef
f
icien
c
y
an
d
co
m
p
u
tatio
n
a
l
p
er
f
o
r
m
an
ce
.
T
h
e
p
r
o
p
o
s
ed
a
cc
eler
ato
r
co
n
s
u
m
es
o
n
l
y
3
5
W
,
wh
ich
is
n
o
tab
ly
lo
wer
th
an
th
e
300
–
7
0
0
W
co
n
s
u
m
ed
b
y
NVI
DI
A
T
en
s
o
r
C
o
r
es,
in
d
icatin
g
a
d
r
asti
c
r
ed
u
ctio
n
in
p
o
wer
u
s
ag
e.
T
h
e
s
e
r
esu
lts
h
ig
h
lig
h
t
th
at
th
e
ap
p
r
o
x
im
ate
MA
C
-
b
a
s
ed
C
NN
ac
ce
ler
ato
r
o
f
f
er
s
s
u
p
er
io
r
s
p
ee
d
an
d
e
n
er
g
y
ef
f
ici
en
cy
.
T
ab
le
4
.
Nu
m
e
r
ical
co
m
p
a
r
is
o
n
with
p
r
o
p
o
s
ed
ap
p
r
o
x
im
ate
co
m
p
u
tin
g
-
b
ased
C
NN
ac
ce
ler
ato
r
an
d
ex
is
tin
g
C
NN
ac
ce
ler
ato
r
s
P
r
o
p
o
se
d
A
p
p
r
o
x
i
ma
t
e
c
o
m
p
u
t
i
n
g
-
b
a
sed
C
N
N
a
c
c
e
l
e
r
a
t
o
r
Ex
i
s
t
i
n
g
C
N
N
a
c
c
e
l
e
r
a
t
o
r
N
V
I
D
I
A
Te
n
so
r
C
o
r
e
s
35W
3
0
0
-
7
0
0
W
G
o
o
g
l
e
E
d
g
e
TPU
0
.
5
ms
>
1
0
ms
I
n
t
e
l
M
o
v
i
d
i
u
s
M
y
r
i
a
d
X
1
2
m
s
>
2
0
ms
T
ab
le
5
s
h
o
ws
th
e
t
r
ad
e
-
o
f
f
s
i
n
ac
cu
r
ac
y
,
th
r
o
u
g
h
p
u
t,
an
d
p
o
wer
co
n
s
u
m
p
tio
n
b
etwe
en
co
n
v
en
tio
n
al
C
NN
im
p
lem
en
tatio
n
s
an
d
p
r
o
p
o
s
ed
a
p
p
r
o
x
im
atio
n
MA
C
-
b
ased
C
NN
ac
ce
ler
ato
r
s
.
Ap
p
r
o
x
im
ate
MA
C
-
b
ased
m
eth
o
d
s
ar
e
ap
p
r
o
p
r
iate
f
o
r
p
o
wer
c
o
n
s
u
m
p
tio
n
,
th
r
o
u
g
h
p
u
t
with
less
lo
s
s
o
f
ac
c
u
r
ac
y
.
Sp
ec
if
ically
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
e
x
h
ib
its
o
n
ly
a
1
.
7
5
%
ac
cu
r
ac
y
lo
s
s
,
wh
ich
is
2
3
.
9
%
lo
wer
t
h
an
GPU
-
b
ased
C
NNs
(
2
.
3
%)
a
n
d
3
5
.
2
%
lo
wer
th
an
T
PU
-
b
ased
C
NNs
(
2
.
7
%).
Ov
er
all,
th
ese
co
m
p
a
r
is
o
n
s
clea
r
ly
d
em
o
n
s
tr
ate
th
at
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
.
14
,
No
.
4
,
Dec
em
b
er
20
25
:
3
6
6
-
3
7
5
372
th
e
p
r
o
p
o
s
ed
ap
p
r
o
x
im
ate
M
AC
-
b
ased
C
NN
ac
ce
ler
ato
r
p
r
o
v
id
es
s
u
p
er
io
r
en
er
g
y
ef
f
ic
ien
cy
an
d
r
e
d
u
ce
d
p
o
wer
u
s
ag
e
wh
ile
m
ain
tain
in
g
h
ig
h
i
n
f
er
e
n
ce
ac
cu
r
ac
y
.
T
ab
le
6
s
h
o
ws
th
e
s
im
u
latio
n
r
esu
lts
th
at
d
em
o
n
s
tr
ate
d
if
f
er
en
t
ap
p
r
o
x
im
atio
n
lev
e
ls
f
o
r
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
As
th
e
ap
p
r
o
x
im
atio
n
lev
el
in
cr
ea
s
es,
ac
cu
r
ac
y
lo
s
s
r
is
es
f
r
o
m
0
.
6
5
%
to
3
.
2
5
%,
wh
ile
en
er
g
y
s
av
in
g
s
im
p
r
o
v
e
f
r
o
m
1
2
.
3
%
to
3
5
.
1
%.
T
h
is
h
ig
h
li
g
h
ts
th
e
b
alan
ce
b
etwe
en
p
o
wer
ef
f
icien
cy
an
d
ac
cu
r
ac
y
in
a
p
p
r
o
x
im
ate
MA
C
-
b
ased
C
NN
ac
ce
ler
ato
r
s
.
T
ab
le
5
.
C
o
m
p
a
r
is
o
n
tab
le
f
o
r
p
r
o
p
o
s
ed
ap
p
r
o
x
im
ate
MA
C
-
b
ased
C
NN
ac
ce
ler
ato
r
s
v
s
.
tr
ad
itio
n
al
C
NN
im
p
lem
en
tatio
n
s
M
e
t
r
i
c
P
r
o
p
o
se
d
A
p
p
r
o
x
i
ma
t
e
M
A
C
-
b
a
se
d
C
N
N
Tr
a
d
i
t
i
o
n
a
l
C
N
N
G
P
U
TPU
A
c
c
u
r
a
c
y
l
o
ss
1
.
7
5
%
2
.
3
%
2
.
7
%
En
e
r
g
y
sa
v
i
n
g
s
3
0
.
5
%
2
3
.
8
%
2
7
.
6
%
P
o
w
e
r
c
o
n
su
mp
t
i
o
n
4
.
5
W
9
.
5
W
7
.
8
W
T
ab
le
6
.
C
o
m
p
a
r
is
o
n
tab
le
f
o
r
v
ar
io
u
s
ap
p
r
o
x
im
atio
n
lev
els
with
ac
cu
r
ac
y
lo
s
s
an
d
e
n
er
g
y
s
av
in
g
s
in
th
e
p
r
o
p
o
s
ed
m
eth
o
d
A
p
p
r
o
x
i
ma
t
i
o
n
l
e
v
e
l
A
c
c
u
r
a
c
y
l
o
ss
En
e
r
g
y
sa
v
i
n
g
s
5
0
.
6
5
1
2
.
3
10
1
.
7
5
2
3
.
8
15
3
.
2
5
3
5
.
1
5.
DIS
CU
SS
I
O
N
Ad
ap
tiv
e
p
r
ec
is
io
n
s
ca
lin
g
h
elp
s
o
p
tim
ize
C
NN
ac
ce
ler
ato
r
s
b
y
b
alan
cin
g
p
o
wer
co
n
s
u
m
p
tio
n
,
p
er
f
o
r
m
an
ce
,
an
d
ac
cu
r
ac
y
.
Dy
n
am
ic
v
o
ltag
e
s
ca
lin
g
a
d
ju
s
ts
th
e
o
p
er
atin
g
v
o
ltag
e
a
n
d
f
r
eq
u
e
n
cy
o
f
MA
C
u
n
its
b
ased
o
n
wo
r
k
lo
a
d
r
e
q
u
i
r
em
en
ts
.
B
y
lo
wer
in
g
v
o
ltag
e
d
u
r
in
g
ap
p
r
o
x
im
ate
c
o
m
p
u
tatio
n
s
in
n
o
n
-
cr
itical
C
NN
lay
er
s
an
d
m
ain
tain
in
g
h
ig
h
er
p
r
ec
is
io
n
f
o
r
s
en
s
itiv
e
lay
er
s
,
DVS
ca
n
ac
h
iev
e
s
ig
n
if
ican
t p
o
wer
s
av
in
g
s
with
m
in
im
al
ac
cu
r
ac
y
lo
s
s
.
FP
GA
-
b
ased
ac
ce
ler
ato
r
s
,
s
u
ch
as Xilin
x
Z
y
n
q
,
s
u
p
p
o
r
t a
d
ap
ti
v
e
v
o
ltag
e
s
ca
lin
g
,
allo
win
g
r
ea
l
-
tim
e
ad
ju
s
tm
en
t
s
to
o
p
tim
ize
en
er
g
y
ef
f
icien
c
y
.
6.
RE
CO
M
M
E
NDA
T
I
O
N
O
N
F
URTHER
WO
RK
T
h
is
s
tu
d
y
an
aly
ze
d
v
ar
i
o
u
s
r
ec
en
t
d
ev
el
o
p
m
en
ts
o
n
ap
p
r
o
x
im
ate
MA
C
u
n
it
an
d
th
eir
u
s
ag
e
in
h
ar
d
war
e
ac
ce
ler
ato
r
s
f
o
r
C
NN.
Ap
p
r
o
x
im
atio
n
is
a
co
m
m
o
n
f
ac
to
r
in
all
th
ese
wo
r
k
s
an
d
is
d
o
n
e
th
r
o
u
g
h
v
ar
io
u
s
s
im
p
lific
atio
n
p
r
o
ce
s
s
es
f
o
r
ar
ea
-
p
o
we
r
-
d
elay
r
ed
u
c
tio
n
.
Fu
tu
r
e
r
esear
ch
wo
r
k
ca
n
f
o
cu
s
o
n
s
ev
er
al
h
y
b
r
id
d
esig
n
s
o
f
ef
f
ec
tiv
e
m
e
th
o
d
s
to
r
ed
u
ce
t
h
e
co
m
p
le
x
ity
an
d
en
e
r
g
y
u
s
ag
e
o
f
th
e
MA
C
u
n
it.
I
n
p
u
t
-
awa
r
e
s
eg
m
en
ted
MA
C
d
esig
n
an
d
ca
r
r
y
p
r
o
p
a
g
atio
n
f
r
ee
,
f
r
ee
g
ate
d
esig
n
b
ased
e
x
ten
d
e
d
ac
cu
m
u
latio
n
u
n
it
ca
n
also
b
e
in
v
esti
g
ated
in
f
u
tu
r
e
r
esear
ch
.
I
n
ter
m
s
o
f
C
NN,
v
ar
io
u
s
co
m
p
r
ess
io
n
ap
p
r
o
ac
h
es
f
o
r
r
eso
u
r
ce
o
p
tim
izatio
n
an
d
p
o
we
r
g
atin
g
m
eth
o
d
o
lo
g
ies
to
m
in
im
ize
th
e
s
witch
in
g
ac
tiv
ity
in
th
e
p
r
o
ce
s
s
in
g
u
n
it
ca
n
b
e
f
o
cu
s
ed
o
n
.
T
h
e
o
p
tim
ize
d
C
NN
h
ar
d
war
e
ac
ce
ler
ato
r
w
ith
ap
p
r
o
x
im
atio
n
ca
n
also
b
e
v
er
if
ied
in
v
ar
io
u
s
b
io
m
ed
ical
s
ig
n
al
p
r
o
ce
s
s
in
g
,
lik
e
elec
tr
o
ca
r
d
io
g
r
am
(
E
C
G)
,
elec
tr
o
en
ce
p
h
alo
g
r
am
(
E
E
G)
,
a
n
d
elec
tr
o
m
y
o
g
r
am
(
E
MG
)
,
with
its
m
etr
ics lik
e
s
en
s
it
iv
ity
,
p
r
e
d
ictio
n
,
an
d
ac
cu
r
ac
y
.
7.
CO
NCLU
SI
O
N
T
h
is
s
u
r
v
ey
wo
r
k
e
x
p
lo
r
e
d
v
ar
io
u
s
ap
p
r
o
x
im
ate
MA
C
u
n
i
ts
an
d
h
ar
d
war
e
ac
ce
ler
ato
r
s
f
o
r
C
NN.
Firstl
y
,
th
e
s
u
r
v
ey
co
n
s
is
ts
o
f
n
ew
m
eth
o
d
o
lo
g
ies
in
v
o
lv
e
d
in
d
esig
n
in
g
th
e
MA
C
u
n
i
t,
wh
ich
ac
ts
as
a
p
r
o
ce
s
s
in
g
elem
en
t
in
th
e
C
NN
ac
ce
ler
ato
r
.
T
h
e
s
tu
d
y
d
em
o
n
s
tr
ates
ap
p
r
o
x
im
ate
MA
C
-
b
ased
C
NN
ac
ce
ler
ato
r
s
an
d
is
s
ca
led
u
s
in
g
d
y
n
am
ic
v
o
ltag
e
s
ca
lin
g
t
o
s
ig
n
if
ican
tly
r
ed
u
ce
p
o
wer
co
n
s
u
m
p
tio
n
w
h
ile
m
ain
tain
in
g
ac
ce
p
tab
le
ac
cu
r
ac
y
lev
els.
T
h
ese
m
eth
o
d
o
lo
g
ies
p
r
o
v
id
ed
d
if
f
er
en
t
f
r
am
e
wo
r
k
s
f
o
r
r
ed
u
cin
g
en
er
g
y
co
n
s
u
m
p
tio
n
an
d
r
eso
u
r
ce
m
in
im
izatio
n
with
a
to
ler
ab
le
am
o
u
n
t
o
f
er
r
o
r
.
A
wid
e
r
esear
ch
o
n
r
ec
en
t
wo
r
k
in
clu
d
es
r
ec
u
r
s
iv
e
MA
C
,
v
o
ltag
e
s
ca
lin
g
,
in
ter
n
al
s
elf
-
h
ea
lin
g
,
s
to
ch
asti
c
co
m
p
u
tin
g
,
p
ar
allel
m
u
ltip
licatio
n
,
an
d
MA
C
b
ase
d
o
n
ap
p
r
o
x
i
m
ate
co
m
p
r
ess
o
r
s
.
T
h
is
s
u
r
v
ey
in
clu
d
es
a
d
etai
led
ex
am
in
atio
n
o
f
v
ar
io
u
s
a
p
p
r
o
x
im
atio
n
lev
els.
I
t
c
o
m
p
ar
es
s
im
u
latio
n
r
esu
lt
s
with
r
ea
l
h
ar
d
war
e
p
er
f
o
r
m
an
ce
b
y
m
ea
s
u
r
in
g
in
f
er
en
ce
s
p
ee
d
,
en
er
g
y
ef
f
ici
en
cy
,
an
d
class
if
icatio
n
ac
cu
r
ac
y
.
I
n
th
is
wo
r
k
en
er
g
y
en
e
r
g
y
-
ef
f
icien
t
C
NN
ac
ce
ler
ato
r
s
u
s
in
g
ap
p
r
o
x
im
at
e
MA
C
u
n
it
b
alan
ce
ap
p
r
o
x
i
m
ate
p
r
ec
is
io
n
f
o
r
r
ed
u
cin
g
p
o
wer
co
n
s
u
m
p
tio
n
an
d
r
e
d
u
cin
g
er
r
o
r
.
T
h
is
s
tu
d
y
p
r
o
v
id
es
a
f
o
u
n
d
atio
n
f
o
r
d
esig
n
in
g
en
e
r
g
y
-
ef
f
icien
t
C
N
N
ac
ce
ler
ato
r
s
u
s
in
g
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
s
u
r
ve
y
o
n
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
h
a
r
d
w
a
r
e
a
cc
elera
tio
n
…
(
S
u
vith
a
P
a
t
h
iya
d
a
n
S
u
d
h
a
ka
r
a
n
)
373
ap
p
r
o
x
im
ate
MA
C
u
n
its
.
B
y
b
alan
cin
g
ap
p
r
o
x
im
atio
n
a
n
d
p
r
ec
is
io
n
,
we
o
f
f
er
a
v
iab
le
ap
p
r
o
ac
h
f
o
r
r
ed
u
cin
g
p
o
wer
co
n
s
u
m
p
tio
n
wh
ile
m
a
in
tain
in
g
in
f
e
r
en
ce
ac
c
u
r
ac
y
.
Ap
p
r
o
x
im
ate
MA
C
o
p
er
atio
n
s
ca
n
b
e
u
tili
ze
d
b
y
ap
p
ly
in
g
s
elec
tiv
e
ap
p
r
o
x
im
a
tio
n
to
n
o
n
-
cr
itical
lay
er
s
,
u
s
in
g
h
y
b
r
id
p
r
ec
is
io
n
,
a
n
d
e
m
p
lo
y
in
g
r
u
n
tim
e
-
co
n
f
ig
u
r
ab
le
MA
C
u
n
its
.
B
al
an
cin
g
e
n
er
g
y
ef
f
icien
c
y
a
n
d
ac
cu
r
ac
y
is
ac
h
iev
ed
t
h
r
o
u
g
h
ad
ap
tiv
e
p
r
ec
is
io
n
s
ca
lin
g
,
d
y
n
a
m
ic
v
o
ltag
e
s
ca
lin
g
,
an
d
s
elec
tiv
e
r
e
-
c
o
m
p
u
tatio
n
b
ased
o
n
a
p
p
licatio
n
-
s
p
ec
if
ic
to
ler
an
ce
s
.
L
ater
,
th
e
s
u
r
v
ey
r
elate
d
to
th
e
r
ec
en
t
C
NN
ac
ce
ler
ato
r
is
lis
ted
,
an
d
its
v
ar
io
u
s
r
esu
lt
o
u
tco
m
es
in
ter
m
s
o
f
th
r
o
u
g
h
p
u
t,
e
n
er
g
y
,
an
d
r
eso
u
r
ce
u
tili
za
tio
n
ar
e
co
m
p
a
r
ed
.
T
h
e
s
y
s
to
lic
ar
r
ay
o
p
tim
izatio
n
,
th
e
s
p
ec
u
lativ
e
ap
p
r
o
ac
h
f
o
r
C
NN,
v
ar
io
u
s
co
m
p
r
ess
io
n
m
et
h
o
d
o
lo
g
ies
o
f
C
NN,
s
p
ar
s
e
C
NN,
an
d
C
NN
f
o
r
ab
n
o
r
m
al
h
ea
r
tb
ea
t
d
etec
tio
n
ar
e
also
an
aly
ze
d
in
th
is
ar
ticle.
E
f
f
i
cien
t
MA
C
an
d
C
NN
d
esig
n
s
ca
n
b
e
u
s
ed
in
ap
p
licatio
n
s
lik
e
m
ed
ical
u
s
ag
e,
au
to
m
o
b
ile,
co
m
m
u
n
icatio
n
,
an
d
s
ec
u
r
ity
ap
p
licatio
n
s
.
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
t
h
o
r
wo
u
ld
lik
e
to
ex
p
r
ess
h
is
h
ea
r
tf
elt
g
r
atitu
d
e
t
o
th
e
s
u
p
er
v
is
o
r
f
o
r
h
is
g
u
i
d
an
ce
a
n
d
u
n
wav
er
in
g
s
u
p
p
o
r
t d
u
r
in
g
t
h
is
r
esear
ch
f
o
r
h
is
g
u
id
an
ce
an
d
s
u
p
p
o
r
t.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Su
v
ith
a
Path
iy
ad
an
Su
d
h
ak
ar
a
n
✓
✓
✓
✓
✓
✓
✓
Aath
m
an
esan
T
h
an
g
ak
alai
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
e
au
t
h
o
r
s
d
ec
lar
e
th
at
th
e
y
h
av
e
n
o
k
n
o
wn
c
o
m
p
etin
g
f
in
an
cial
in
ter
ests
o
r
p
er
s
o
n
al
r
el
atio
n
s
h
ip
s
th
at
co
u
ld
h
av
e
ap
p
ea
r
ed
t
o
in
f
lu
en
ce
th
e
wo
r
k
r
e
p
o
r
te
d
in
t
h
is
p
ap
er
.
I
NF
O
RM
E
D
CO
NS
E
N
T
W
e
ce
r
tify
th
at
we
h
av
e
ex
p
lain
ed
th
e
n
atu
r
e
an
d
p
u
r
p
o
s
e
o
f
th
is
s
tu
d
y
to
th
e
ab
o
v
e
-
n
am
e
d
in
d
iv
id
u
al
a
n
d
h
av
e
d
is
cu
s
s
ed
th
e
p
o
ten
tial
b
e
n
ef
its
o
f
p
ar
ticip
atio
n
.
All
q
u
esti
o
n
s
r
aise
d
b
y
th
e
in
d
iv
id
u
al
h
av
e
b
ee
n
an
s
wer
ed
,
a
n
d
we
will r
em
ain
av
ailab
le
to
ad
d
r
e
s
s
an
y
f
u
tu
r
e
in
q
u
ir
ies.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
e
r
esear
ch
g
u
id
es r
ev
iewe
d
an
d
eth
ically
ap
p
r
o
v
ed
th
is
m
a
n
u
s
cr
ip
t f
o
r
p
u
b
lis
h
in
g
in
th
is
jo
u
r
n
al.
DATA AV
AI
L
AB
I
L
I
T
Y
Data
s
h
ar
in
g
is
n
o
t
ap
p
licab
l
e
to
th
is
ar
ticle
as
n
o
d
atasets
we
r
eg
en
er
ated
o
r
an
aly
ze
d
d
u
r
in
g
th
e
cu
r
r
en
t stu
d
y
.
RE
F
E
R
E
NC
E
S
[
1
]
A
.
D
u
a
,
Y
.
L
i
,
a
n
d
F
.
R
e
n
,
“
S
y
s
t
o
l
i
c
-
C
N
N
:
A
n
O
p
e
n
C
L
-
d
e
f
i
n
e
d
s
c
a
l
a
b
l
e
r
u
n
-
t
i
m
e
-
f
l
e
x
i
b
l
e
F
P
G
A
a
c
c
e
l
e
r
a
t
o
r
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
a
c
c
e
l
e
r
a
t
i
n
g
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
i
n
f
e
r
e
n
c
e
i
n
c
l
o
u
d
/
e
d
g
e
c
o
m
p
u
t
i
n
g
,
”
2
0
2
0
I
E
EE
2
8
t
h
A
n
n
u
a
l
I
n
t
e
r
n
a
t
i
o
n
a
l
S
y
m
p
o
si
u
m
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
.
14
,
No
.
4
,
Dec
em
b
er
20
25
:
3
6
6
-
3
7
5
374
o
n
Fi
e
l
d
-
Pr
o
g
r
a
m
m
a
b
l
e
C
u
st
o
m
C
o
m
p
u
t
i
n
g
M
a
c
h
i
n
e
s (F
C
C
M)
.
I
EEE,
p
.
2
3
1
,
2
0
2
0
.
d
o
i
:
1
0
.
1
1
0
9
/
f
c
c
m
4
8
2
8
0
.
2
0
2
0
.
0
0
0
6
4
.
[
2
]
C
.
R
a
g
h
u
r
a
m,
V
.
R
.
D
a
n
d
u
,
a
n
d
B
.
J
a
i
s
o
n
,
“
H
y
b
r
i
d
i
z
a
t
i
o
n
o
f
d
i
l
a
t
e
d
c
n
n
w
i
t
h
a
t
t
e
n
t
i
o
n
l
i
n
k
n
e
t
f
o
r
b
r
a
i
n
c
a
n
c
e
r
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
D
a
t
a
S
c
i
e
n
c
e
a
n
d
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
2
,
n
o
.
0
2
,
p
p
.
3
5
–
4
2
,
2
0
2
4
.
[
3
]
H
.
Li
u
,
L.
S
o
n
g
,
R
.
S
u
n
d
a
r
a
se
k
a
r
,
a
n
d
A
.
J.
G
.
M
a
l
a
r
,
“
C
o
mp
u
t
e
r
n
e
t
w
o
r
k
d
a
t
a
ma
n
a
g
e
m
e
n
t
mo
d
e
l
b
a
s
e
d
o
n
e
d
g
e
c
o
mp
u
t
i
n
g
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
R
e
l
i
a
b
i
l
i
t
y
,
Q
u
a
l
i
t
y
a
n
d
S
a
f
e
t
y
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
3
1
,
n
o
.
0
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
4
2
/
s
0
2
1
8
5
3
9
3
2
3
5
0
0
3
0
4
.
[
4
]
A
.
A
h
i
l
a
n
,
A
.
A
l
b
e
r
t
R
a
j
,
A
.
G
o
r
a
n
t
l
a
,
R
.
J
o
t
h
i
n
,
M
.
S
h
u
n
m
u
g
a
t
h
a
m
mal
,
a
n
d
G
.
A
.
S
a
f
d
a
r
,
“
D
e
si
g
n
o
f
e
n
e
r
g
y
-
e
f
f
i
c
i
e
n
t
a
p
p
r
o
x
i
m
a
t
e
a
r
i
t
h
m
e
t
i
c
c
i
r
c
u
i
t
s
f
o
r
e
r
r
o
r
t
o
l
e
r
a
n
t
me
d
i
c
a
l
i
m
a
g
e
p
r
o
c
e
ssi
n
g
a
p
p
l
i
c
a
t
i
o
n
s
,
”
L
e
c
t
u
re
N
o
t
e
s
i
n
E
l
e
c
t
r
i
c
a
l
En
g
i
n
e
e
ri
n
g
.
S
p
r
i
n
g
e
r
N
a
t
u
r
e
S
i
n
g
a
p
o
r
e
,
p
p
.
6
7
9
–
6
9
2
,
2
0
2
4
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
981
-
99
-
8
6
4
6
-
0
_
5
3
.
[
5
]
X
.
X
i
e
,
J.
L
i
n
,
Z.
W
a
n
g
,
a
n
d
J.
W
e
i
,
“
A
n
e
f
f
i
c
i
e
n
t
a
n
d
f
l
e
x
i
b
l
e
a
c
c
e
l
e
r
a
t
o
r
d
e
si
g
n
f
o
r
sp
a
r
s
e
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,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
st
e
m
s
I
:
Re
g
u
l
a
r
Pa
p
e
rs
,
v
o
l
.
6
8
,
n
o
.
7
,
p
p
.
2
9
3
6
–
2
9
4
9
,
J
u
l
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
S
I
.
2
0
2
1
.
3
0
7
4
3
0
0
.
[
6
]
M
.
S
.
K
i
m,
A
.
A
.
D
e
l
B
a
r
r
i
o
,
H
.
K
i
m,
a
n
d
N
.
B
a
g
h
e
r
z
a
d
e
h
,
“
Th
e
e
f
f
e
c
t
s
o
f
a
p
p
r
o
x
i
ma
t
e
m
u
l
t
i
p
l
i
c
a
t
i
o
n
o
n
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,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
Em
e
rg
i
n
g
T
o
p
i
c
s
i
n
C
o
m
p
u
t
i
n
g
,
v
o
l
.
1
0
,
n
o
.
2
,
p
p
.
9
0
4
–
9
1
6
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
t
e
t
c
.
2
0
2
1
.
3
0
5
0
9
8
9
.
[
7
]
A
.
D
e
m
i
d
o
v
sk
i
j
a
n
d
E
.
S
m
i
r
n
o
v
,
“
E
f
f
e
c
t
i
v
e
p
o
s
t
-
t
r
a
i
n
i
n
g
q
u
a
n
t
i
z
a
t
i
o
n
o
f
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
i
n
f
e
r
e
n
c
e
o
n
l
o
w
p
o
w
e
r
n
e
u
r
a
l
a
c
c
e
l
e
r
a
t
o
r
,
”
2
0
2
0
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
re
n
c
e
o
n
N
e
u
r
a
l
N
e
t
w
o
r
k
s
(
I
J
C
N
N
)
.
I
EEE,
p
p
.
1
–
7
,
2
0
2
0
.
d
o
i
:
1
0
.
1
1
0
9
/
i
j
c
n
n
4
8
6
0
5
.
2
0
2
0
.
9
2
0
7
2
8
1
.
[
8
]
Q
.
S
o
n
g
,
W
.
C
u
i
,
L.
S
u
n
,
a
n
d
G
.
Ji
n
,
“
D
e
si
g
n
a
n
d
i
m
p
l
e
me
n
t
a
t
i
o
n
o
f
a
u
n
i
v
e
r
sa
l
s
h
i
f
t
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
a
c
c
e
l
e
r
a
t
o
r
,
”
I
EEE
Em
b
e
d
d
e
d
S
y
st
e
m
s Le
t
t
e
rs
,
v
o
l
.
1
6
,
n
o
.
1
,
p
p
.
1
7
–
2
0
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
l
e
s.
2
0
2
2
.
3
2
3
3
7
9
6
.
[
9
]
J.
L
e
e
,
G
.
K
i
m,
J.
P
a
r
k
,
a
n
d
H
.
-
M
.
B
a
e
,
“
Li
n
k
b
i
t
-
e
r
r
o
r
-
r
a
t
e
r
e
q
u
i
r
e
me
n
t
a
n
a
l
y
s
i
s
f
o
r
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
s,
”
2
0
2
1
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
S
y
m
p
o
si
u
m
o
n
C
i
r
c
u
i
t
s
a
n
d
S
y
st
e
m
s (I
S
C
A
S
)
.
I
EEE,
p
p
.
1
–
5
,
2
0
2
1
.
d
o
i
:
1
0
.
1
1
0
9
/
i
sc
a
s5
1
5
5
6
.
2
0
2
1
.
9
4
0
1
1
1
2
.
[
1
0
]
Q
.
Z
h
o
u
,
H
.
X
u
,
J.
Li
,
T.
M
a
,
a
n
d
C
.
C
h
e
n
,
“
O
P
A
S
C
A
:
O
u
t
e
r
p
r
o
d
u
c
t
-
b
a
s
e
d
a
c
c
e
l
e
r
a
t
o
r
w
i
t
h
u
n
i
f
i
e
d
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
s
p
a
r
se
c
o
n
v
o
l
u
t
i
o
n
a
n
d
a
t
t
e
n
t
i
o
n
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
r
-
Ai
d
e
d
D
e
s
i
g
n
o
f
I
n
t
e
g
r
a
t
e
d
C
i
r
c
u
i
t
s
a
n
d
S
y
st
e
m
s
,
v
o
l
.
4
4
,
n
o
.
4
,
p
p
.
1
2
9
0
–
1
3
0
3
,
A
p
r
.
2
0
2
5
,
d
o
i
:
1
0
.
1
1
0
9
/
TC
A
D
.
2
0
2
4
.
3
4
8
3
0
9
2
.
[
1
1
]
C.
-
X
.
X
u
e
e
t
a
l
.
,
“
2
4
.
1
A
1
M
b
m
u
l
t
i
b
i
t
R
e
R
A
M
c
o
m
p
u
t
i
n
g
-
in
-
mem
o
r
y
mac
r
o
w
i
t
h
1
4
.
6
n
s
p
a
r
a
l
l
e
l
M
A
C
c
o
mp
u
t
i
n
g
t
i
m
e
f
o
r
C
N
N
b
a
s
e
d
A
I
e
d
g
e
p
r
o
c
e
ss
o
r
s,
”
2
0
1
9
I
EEE
I
n
t
e
rn
a
t
i
o
n
a
l
S
o
l
i
d
-
S
t
a
t
e
C
i
r
c
u
i
t
s
C
o
n
f
e
re
n
c
e
-
(
I
S
S
C
C
)
.
I
EEE,
2
0
1
9
.
d
o
i
:
1
0
.
1
1
0
9
/
i
ss
c
c
.
2
0
1
9
.
8
6
6
2
3
9
5
.
[
1
2
]
S
.
M
.
W
a
q
a
s
,
M
.
Z
a
k
w
a
n
,
M
.
A
sh
r
a
f
,
G
.
N
a
i
f
A
l
W
a
k
i
d
,
a
n
d
M
.
H
u
m
a
y
u
n
,
“
A
su
r
v
e
y
o
n
a
p
p
r
o
x
i
ma
t
e
h
a
r
d
w
a
r
e
a
c
c
e
l
e
r
a
t
o
r
f
o
r
e
r
r
o
r
-
t
o
l
e
r
a
n
t
a
p
p
l
i
c
a
t
i
o
n
s,
”
S
e
c
u
r
i
n
g
t
h
e
D
i
g
i
t
a
l
Re
a
l
m
.
C
R
C
P
r
e
ss,
p
p
.
1
1
5
–
1
2
5
,
2
0
2
5
.
d
o
i
:
1
0
.
1
2
0
1
/
9
7
8
1
0
0
3
4
9
7
8
5
1
-
1
1
.
[
1
3
]
S
.
Le
e
,
D
.
K
i
m
,
D
.
N
g
u
y
e
n
,
a
n
d
J.
Le
e
,
“
D
o
u
b
l
e
M
A
C
o
n
a
D
S
P
:
B
o
o
st
i
n
g
t
h
e
p
e
r
f
o
r
ma
n
c
e
o
f
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
o
n
F
P
G
A
s,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
r
-
A
i
d
e
d
D
e
s
i
g
n
o
f
I
n
t
e
g
ra
t
e
d
C
i
r
c
u
i
t
s
a
n
d
S
y
st
e
m
s
,
v
o
l
.
3
8
,
n
o
.
5
,
p
p
.
8
8
8
–
8
9
7
,
M
a
y
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
A
D
.
2
0
1
8
.
2
8
2
4
2
8
0
.
[
1
4
]
G
.
P
a
r
k
,
J.
K
u
n
g
,
a
n
d
Y
.
Le
e
,
“
D
e
si
g
n
a
n
d
a
n
a
l
y
si
s
o
f
a
p
p
r
o
x
i
m
a
t
e
c
o
mp
r
e
ss
o
r
s
f
o
r
b
a
l
a
n
c
e
d
e
r
r
o
r
a
c
c
u
m
u
l
a
t
i
o
n
i
n
M
A
C
o
p
e
r
a
t
o
r
,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
s
t
e
m
s
I
:
R
e
g
u
l
a
r
P
a
p
e
rs
,
v
o
l
.
6
8
,
n
o
.
7
,
p
p
.
2
9
5
0
–
2
9
6
1
,
Ju
l
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
S
I
.
2
0
2
1
.
3
0
7
3
1
7
7
.
[
1
5
]
S
.
D
.
S
.
,
T
.
K
a
r
t
h
i
k
e
y
a
n
,
a
n
d
N
.
M
.
S
k
.
,
“
En
e
r
g
y
e
f
f
i
c
i
e
n
t
m
u
l
t
i
p
l
y
-
a
c
c
u
m
u
l
a
t
e
u
n
i
t
u
s
i
n
g
n
o
v
e
l
r
e
c
u
r
si
v
e
m
u
l
t
i
p
l
i
c
a
t
i
o
n
f
o
r
e
r
r
o
r
-
t
o
l
e
r
a
n
t
a
p
p
l
i
c
a
t
i
o
n
s,
”
I
n
t
e
g
ra
t
i
o
n
,
v
o
l
.
9
2
,
p
p
.
2
4
–
3
4
,
S
e
p
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
v
l
si
.
2
0
2
3
.
0
4
.
0
0
6
.
[
1
6
]
R
.
X
i
a
o
e
t
a
l
.
,
“
A
l
o
w
-
p
o
w
e
r
i
n
-
mem
o
r
y
m
u
l
t
i
p
l
i
c
a
t
i
o
n
a
n
d
a
c
c
u
m
u
l
a
t
i
o
n
a
r
r
a
y
w
i
t
h
mo
d
i
f
i
e
d
R
a
d
i
x
-
4
i
n
p
u
t
a
n
d
c
a
n
o
n
i
c
a
l
s
i
g
n
e
d
d
i
g
i
t
w
e
i
g
h
t
s
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
Ve
ry
L
a
r
g
e
S
c
a
l
e
I
n
t
e
g
r
a
t
i
o
n
(
VLS
I
)
S
y
st
e
m
s
,
v
o
l
.
3
1
,
n
o
.
1
1
,
p
p
.
1
7
0
0
–
1
7
1
2
,
N
o
v
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TV
LSI
.
2
0
2
3
.
3
3
0
6
3
7
6
.
[
1
7
]
S
.
R
y
u
,
Y
.
O
h
,
a
n
d
J.
-
J.
K
i
m,
“
B
i
n
a
r
y
w
a
r
e
:
A
h
i
g
h
-
p
e
r
f
o
r
m
a
n
c
e
d
i
g
i
t
a
l
h
a
r
d
w
a
r
e
a
c
c
e
l
e
r
a
t
o
r
f
o
r
b
i
n
a
r
y
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
Ve
r
y
L
a
r
g
e
S
c
a
l
e
I
n
t
e
g
r
a
t
i
o
n
(
VLS
I
)
S
y
st
e
m
s
,
v
o
l
.
3
1
,
n
o
.
1
2
,
p
p
.
2
1
3
7
–
2
1
4
1
,
D
e
c
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TV
LSI
.
2
0
2
3
.
3
3
2
4
8
3
4
.
[
1
8
]
S.
-
N
.
Ta
n
g
,
“
A
r
e
a
-
e
f
f
i
c
i
e
n
t
p
a
r
a
l
l
e
l
mu
l
t
i
p
l
i
c
a
t
i
o
n
u
n
i
t
s
f
o
r
C
N
N
a
c
c
e
l
e
r
a
t
o
r
s
w
i
t
h
o
u
t
p
u
t
c
h
a
n
n
e
l
p
a
r
a
l
l
e
l
i
z
a
t
i
o
n
,
”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
Ve
r
y
L
a
r
g
e
S
c
a
l
e
I
n
t
e
g
ra
t
i
o
n
(
VL
S
I
)
S
y
s
t
e
m
s
,
v
o
l
.
3
1
,
n
o
.
3
,
p
p
.
4
0
6
–
4
1
0
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TV
LSI
.
2
0
2
3
.
3
2
3
5
7
7
6
.
[
1
9
]
H
.
A
f
z
a
l
i
-
K
u
sh
a
a
n
d
M
.
P
e
d
r
a
m,
“
X
-
N
V
D
LA
:
R
u
n
t
i
me
a
c
c
u
r
a
c
y
c
o
n
f
i
g
u
r
a
b
l
e
N
V
D
LA
b
a
s
e
d
o
n
a
p
p
l
y
i
n
g
v
o
l
t
a
g
e
o
v
e
r
sca
l
i
n
g
t
o
c
o
m
p
u
t
i
n
g
a
n
d
m
e
mo
r
y
u
n
i
t
s,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
r
c
u
i
t
s
a
n
d
S
y
s
t
e
m
s
I
:
R
e
g
u
l
a
r
P
a
p
e
rs
,
v
o
l
.
7
0
,
n
o
.
5
,
p
p
.
1
9
8
9
–
2
0
0
2
,
M
a
y
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
S
I
.
2
0
2
3
.
3
2
4
7
7
4
3
.
[
2
0
]
G
.
A
.
G
i
l
l
a
n
i
,
M
.
A
.
H
a
n
i
f
,
B
.
V
e
r
s
t
o
e
p
,
S
.
H
.
G
e
r
e
z
,
M
.
S
h
a
f
i
q
u
e
,
a
n
d
A
.
B
.
J
.
K
o
k
k
e
l
e
r
,
“
M
A
C
I
S
H
:
D
e
si
g
n
i
n
g
a
p
p
r
o
x
i
m
a
t
e
M
A
C
a
c
c
e
l
e
r
a
t
o
r
s
w
i
t
h
i
n
t
e
r
n
a
l
-
sel
f
-
h
e
a
l
i
n
g
,
”
I
EE
E
A
c
c
e
ss
,
v
o
l
.
7
,
p
p
.
7
7
1
4
2
–
7
7
1
6
0
,
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
3
3
5
.
[
2
1
]
M
.
M
a
sa
d
e
h
,
O
.
H
a
s
a
n
,
a
n
d
S
.
Ta
h
a
r
,
“
I
n
p
u
t
-
c
o
n
s
c
i
o
u
s
a
p
p
r
o
x
i
m
a
t
e
m
u
l
t
i
p
l
y
-
a
c
c
u
m
u
l
a
t
e
(
M
A
C
)
u
n
i
t
f
o
r
e
n
e
r
g
y
-
e
f
f
i
c
i
e
n
c
y
,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
7
,
p
p
.
1
4
7
1
2
9
–
1
4
7
1
4
2
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
9
.
2
9
4
6
5
1
3
.
[
2
2
]
K
.
I
n
a
y
a
t
a
n
d
J
.
C
h
u
n
g
,
“
H
y
b
r
i
d
a
c
c
u
mu
l
a
t
o
r
f
a
c
t
o
r
e
d
sy
s
t
o
l
i
c
a
r
r
a
y
f
o
r
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
c
c
e
l
e
r
a
t
i
o
n
,
”
I
E
E
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
Ve
ry
L
a
r
g
e
S
c
a
l
e
I
n
t
e
g
ra
t
i
o
n
(
VLS
I
)
S
y
s
t
e
m
s
,
v
o
l
.
3
0
,
n
o
.
7
,
p
p
.
8
8
1
–
8
9
2
,
J
u
l
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
TV
LSI
.
2
0
2
2
.
3
1
7
0
2
3
3
.
[
2
3
]
Q
.
C
h
e
n
g
e
t
a
l
.
,
“
A
l
o
w
-
p
o
w
e
r
s
p
a
r
s
e
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
a
c
c
e
l
e
r
a
t
o
r
w
i
t
h
p
r
e
-
e
n
c
o
d
i
n
g
r
a
d
i
x
-
4
b
o
o
t
h
mu
l
t
i
p
l
i
e
r
,
”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
st
e
m
s II
:
Ex
p
r
e
ss
Bri
e
f
s
,
v
o
l
.
7
0
,
n
o
.
6
,
p
p
.
2
2
4
6
–
2
2
5
0
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
t
c
s
i
i
.
2
0
2
2
.
3
2
3
1
3
6
1
.
[
2
4
]
J.
L
i
,
G
.
S
h
e
n
,
D
.
Z
h
a
o
,
Q
.
Z
h
a
n
g
,
a
n
d
Y
.
Ze
n
g
,
“
F
i
r
e
f
l
y
:
A
h
i
g
h
-
t
h
r
o
u
g
h
p
u
t
h
a
r
d
w
a
r
e
a
c
c
e
l
e
r
a
t
o
r
f
o
r
s
p
i
k
i
n
g
n
e
u
r
a
l
n
e
t
w
o
r
k
s
w
i
t
h
e
f
f
i
c
i
e
n
t
D
S
P
a
n
d
mem
o
r
y
o
p
t
i
m
i
z
a
t
i
o
n
,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s o
n
Ve
ry
L
a
rg
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
.
3
1
,
n
o
.
8
,
p
p
.
1
1
7
8
–
1
1
9
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
t
v
l
si
.
2
0
2
3
.
3
2
7
9
3
4
9
.
[
2
5
]
C
.
X
i
e
,
Z
.
S
h
a
o
,
N
.
Zh
a
o
,
Y
.
D
u
,
a
n
d
L.
D
u
,
“
A
n
e
f
f
i
c
i
e
n
t
C
N
N
i
n
f
e
r
e
n
c
e
a
c
c
e
l
e
r
a
t
o
r
b
a
s
e
d
o
n
i
n
t
r
a
-
a
n
d
i
n
t
e
r
-
c
h
a
n
n
e
l
f
e
a
t
u
r
e
map
c
o
mp
r
e
ssi
o
n
,
”
I
E
EE
T
r
a
n
sa
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
s
t
e
m
s
I
:
R
e
g
u
l
a
r
P
a
p
e
rs
,
v
o
l
.
7
0
,
n
o
.
9
,
p
p
.
3
6
2
5
–
3
6
3
8
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
t
c
si
.
2
0
2
3
.
3
2
8
7
6
0
2
.
[
2
6
]
R.
-
X
.
Z
h
e
n
g
,
Y
.
-
C
.
K
o
,
a
n
d
T
.
-
T.
L
i
u
,
“
A
sp
e
c
u
l
a
t
i
v
e
c
o
m
p
u
t
a
t
i
o
n
a
p
p
r
o
a
c
h
f
o
r
e
n
e
r
g
y
-
e
f
f
i
c
i
e
n
t
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
E
E
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
r
-
A
i
d
e
d
D
e
si
g
n
o
f
I
n
t
e
g
r
a
t
e
d
C
i
r
c
u
i
t
s
a
n
d
S
y
st
e
m
s
,
v
o
l
.
4
2
,
n
o
.
3
,
p
p
.
7
9
5
–
8
0
6
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
A
D
.
2
0
2
2
.
3
1
8
3
5
6
1
.
[
2
7
]
A
.
A
r
u
n
a
c
h
a
l
a
m,
S
.
K
u
n
d
u
,
A
.
R
a
h
a
,
S
.
B
a
n
e
r
j
e
e
,
S
.
N
a
t
a
r
a
j
a
n
,
a
n
d
K
.
B
a
s
u
,
“
A
n
o
v
e
l
l
o
w
-
p
o
w
e
r
c
o
mp
r
e
ssi
o
n
s
c
h
e
me
f
o
r
sy
st
o
l
i
c
a
r
r
a
y
-
b
a
se
d
d
e
e
p
l
e
a
r
n
i
n
g
a
c
c
e
l
e
r
a
t
o
r
s
,
”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
r
-
Ai
d
e
d
D
e
si
g
n
o
f
I
n
t
e
g
r
a
t
e
d
C
i
rc
u
i
t
s
a
n
d
S
y
s
t
e
m
s
,
v
o
l
.
4
2
,
n
o
.
4
,
p
p
.
1
0
8
5
–
1
0
9
8
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
A
D
.
2
0
2
2
.
3
1
9
8
0
3
6
.
[
2
8
]
S
.
Z
h
u
,
L.
H
.
K
.
D
u
o
n
g
,
H
.
C
h
e
n
,
D
.
L
i
u
,
a
n
d
W
.
Li
u
,
“
F
A
T
:
A
n
i
n
-
me
mo
r
y
a
c
c
e
l
e
r
a
t
o
r
w
i
t
h
f
a
s
t
a
d
d
i
t
i
o
n
f
o
r
t
e
r
n
a
r
y
w
e
i
g
h
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
r
-
A
i
d
e
d
D
e
si
g
n
o
f
I
n
t
e
g
r
a
t
e
d
C
i
r
c
u
i
t
s
a
n
d
S
y
st
e
m
s
,
v
o
l
.
4
2
,
n
o
.
3
,
p
p
.
7
8
1
–
7
9
4
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TC
A
D
.
2
0
2
2
.
3
1
8
4
2
7
6
.
[
2
9
]
X
.
C
h
e
n
,
J.
Zh
u
,
J
.
Ji
a
n
g
,
a
n
d
C
.
-
Y
.
Tsu
i
,
“
Ti
g
h
t
c
o
m
p
r
e
ss
i
o
n
:
C
o
m
p
r
e
ssi
n
g
C
N
N
t
h
r
o
u
g
h
f
i
n
e
-
g
r
a
i
n
e
d
p
r
u
n
i
n
g
a
n
d
w
e
i
g
h
t
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
s
u
r
ve
y
o
n
co
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
k
h
a
r
d
w
a
r
e
a
cc
elera
tio
n
…
(
S
u
vith
a
P
a
t
h
iya
d
a
n
S
u
d
h
a
ka
r
a
n
)
375
p
e
r
m
u
t
a
t
i
o
n
f
o
r
e
f
f
i
c
i
e
n
t
i
mp
l
e
m
e
n
t
a
t
i
o
n
,
”
I
E
EE
T
r
a
n
sa
c
t
i
o
n
s
o
n
C
o
m
p
u
t
e
r
-
Ai
d
e
d
D
e
s
i
g
n
o
f
I
n
t
e
g
r
a
t
e
d
C
i
rc
u
i
t
s
a
n
d
S
y
st
e
m
s
,
v
o
l
.
4
2
,
n
o
.
2
,
p
p
.
6
4
4
–
6
5
7
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
A
D
.
2
0
2
2
.
3
1
7
8
0
4
7
.
[
3
0
]
J.
L
u
e
t
a
l
.
,
“
Ef
f
i
c
i
e
n
t
h
a
r
d
w
a
r
e
a
r
c
h
i
t
e
c
t
u
r
e
o
f
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
f
o
r
E
C
G
c
l
a
ssi
f
i
c
a
t
i
o
n
i
n
w
e
a
r
a
b
l
e
h
e
a
l
t
h
c
a
r
e
d
e
v
i
c
e
,
”
I
EE
E
T
r
a
n
s
a
c
t
i
o
n
s
o
n
C
i
rcu
i
t
s
a
n
d
S
y
st
e
m
s
I
:
R
e
g
u
l
a
r
P
a
p
e
rs
,
v
o
l
.
6
8
,
n
o
.
7
,
p
p
.
2
9
7
6
–
2
9
8
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
t
c
si
.
2
0
2
1
.
3
0
7
2
6
2
2
.
[
3
1
]
Y.
-
H
.
C
h
e
n
,
S
.
-
W
.
C
h
e
n
,
P
.
-
J.
C
h
a
n
g
,
H
.
-
T.
H
u
a
,
S
.
-
Y
.
Li
n
,
a
n
d
R
.
-
S
.
C
h
e
n
,
“
A
V
LSI
c
h
i
p
f
o
r
t
h
e
a
b
n
o
r
m
a
l
h
e
a
r
t
b
e
a
t
d
e
t
e
c
t
i
o
n
u
si
n
g
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
e
n
so
rs
,
v
o
l
.
2
2
,
n
o
.
3
,
p
.
7
9
6
,
Ja
n
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
s
2
2
0
3
0
7
9
6
.
[
3
2
]
L.
S
a
y
a
d
i
,
S
.
T
i
marc
h
i
,
a
n
d
A
.
S
h
e
i
k
h
-
A
k
b
a
r
i
,
“
Tw
o
e
f
f
i
c
i
e
n
t
a
p
p
r
o
x
i
m
a
t
e
u
n
s
i
g
n
e
d
m
u
l
t
i
p
l
i
e
r
s
b
y
d
e
v
e
l
o
p
i
n
g
n
e
w
c
o
n
f
i
g
u
r
a
t
i
o
n
f
o
r
a
p
p
r
o
x
i
ma
t
e
4
:
2
c
o
m
p
r
e
ss
o
r
s,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
s
t
e
m
s
I
:
Re
g
u
l
a
r
Pa
p
e
r
s
,
v
o
l
.
7
0
,
n
o
.
4
,
p
p
.
1
6
4
9
–
1
6
5
9
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
S
I
.
2
0
2
3
.
3
2
4
2
5
5
8
.
[
3
3
]
S
.
Y
o
n
g
x
i
a
e
t
a
l
.
,
“
D
e
si
g
n
o
f
a
p
p
r
o
x
i
ma
t
e
B
o
o
t
h
m
u
l
t
i
p
l
i
e
r
s
b
a
se
d
o
n
e
r
r
o
r
c
o
mp
e
n
s
a
t
i
o
n
,
”
I
n
t
e
g
r
a
t
i
o
n
,
v
o
l
.
9
0
,
p
p
.
1
8
3
–
1
8
9
,
M
a
y
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
v
l
si
.
2
0
2
3
.
0
2
.
0
0
1
.
[
3
4
]
Z.
A
i
z
a
z
a
n
d
K
.
K
h
a
r
e
,
“
A
r
e
a
a
n
d
p
o
w
e
r
e
f
f
i
c
i
e
n
t
t
r
u
n
c
a
t
e
d
b
o
o
t
h
m
u
l
t
i
p
l
i
e
r
s
u
si
n
g
a
p
p
r
o
x
i
ma
t
e
c
a
r
r
y
-
b
a
se
d
e
r
r
o
r
c
o
m
p
e
n
sa
t
i
o
n
,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
r
c
u
i
t
s
a
n
d
S
y
s
t
e
m
s
I
I
:
Ex
p
r
e
ss
Bri
e
f
s
,
v
o
l
.
6
9
,
n
o
.
2
,
p
p
.
5
7
9
–
5
8
3
,
F
e
b
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
S
I
I
.
2
0
2
1
.
3
0
9
4
9
1
0
.
[
3
5
]
G
.
A
n
u
s
h
a
a
n
d
P
.
D
e
e
p
a
,
“
D
e
s
i
g
n
o
f
a
p
p
r
o
x
i
m
a
t
e
a
d
d
e
r
s
a
n
d
mu
l
t
i
p
l
i
e
r
s
f
o
r
e
r
r
o
r
t
o
l
e
r
a
n
t
i
ma
g
e
p
r
o
c
e
ss
i
n
g
,
”
Mi
c
r
o
p
ro
c
e
ss
o
rs
a
n
d
Mi
c
ro
s
y
st
e
m
s
,
v
o
l
.
7
2
,
p
.
1
0
2
9
4
0
,
F
e
b
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
mi
c
p
r
o
.
2
0
1
9
.
1
0
2
9
4
0
.
[
3
6
]
A
.
G
.
M
.
S
t
r
o
l
l
o
,
E
.
N
a
p
o
l
i
,
D
.
D
e
C
a
r
o
,
N
.
P
e
t
r
a
,
G
.
S
a
g
g
e
se
,
a
n
d
G
.
D
i
M
e
o
,
“
A
p
p
r
o
x
i
ma
t
e
m
u
l
t
i
p
l
i
e
r
s
u
si
n
g
st
a
t
i
c
seg
m
e
n
t
a
t
i
o
n
:
e
r
r
o
r
a
n
a
l
y
s
i
s
a
n
d
i
m
p
r
o
v
e
me
n
t
s
,
”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
st
e
m
s
I
:
Re
g
u
l
a
r
Pa
p
e
r
s
,
v
o
l
.
6
9
,
n
o
.
6
,
p
p
.
2
4
4
9
–
2
4
6
2
,
Ju
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
T
C
S
I
.
2
0
2
2
.
3
1
5
2
9
2
1
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
S
u
v
it
h
a
P
a
th
i
y
a
d
a
n
S
u
d
h
a
k
a
r
a
n
is
a
P
h
D
S
c
h
o
lar
i
n
t
h
e
De
p
a
rtme
n
t
o
f
El
e
c
tro
n
ics
a
n
d
Co
m
m
u
n
ica
ti
o
n
En
g
i
n
e
e
rin
g
a
t
Ve
l
Tec
h
Ra
n
g
a
r
a
jan
Dr.
S
a
g
u
n
t
h
a
la
R
&
D
In
stit
u
te
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
l
o
g
y
,
Ch
e
n
n
a
i,
In
d
ia
a
n
d
c
u
rre
n
tl
y
wo
rk
i
n
g
a
s
a
n
As
sista
n
t
P
ro
fe
ss
o
r
i
n
th
e
d
e
p
a
rtme
n
t
o
f
El
e
c
tro
n
ics
a
n
d
C
o
m
m
u
n
ica
ti
o
n
En
g
in
e
e
rin
g
a
t
IES
C
o
ll
e
g
e
o
f
En
g
i
n
e
e
rin
g
,
In
d
ia.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
lo
w
p
o
we
r
VL
S
I
d
e
sig
n
a
n
d
d
e
e
p
lea
rn
i
n
g
a
c
c
e
ler
a
to
rs.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
su
v
it
h
a
p
sv
t
d
1
2
5
2
@g
m
a
il
.
c
o
m
.
Aa
th
m
a
n
e
sa
n
T
h
a
n
g
a
k
a
l
a
i
is
a
n
As
sista
n
t
P
r
o
fe
ss
o
r
in
t
h
e
De
p
a
rtme
n
t
o
f
El
e
c
tro
n
ics
a
n
d
Co
m
m
u
n
ica
ti
o
n
En
g
in
e
e
rin
g
a
t
Ve
l
Tec
h
Un
i
v
e
rsity
,
Ch
e
n
n
a
i,
In
d
ia.
He
re
c
e
iv
e
d
h
is
P
h
.
D.
i
n
El
e
c
tro
n
ic
s
a
n
d
Co
m
m
u
n
ica
ti
o
n
En
g
in
e
e
ri
n
g
wit
h
re
se
a
rc
h
fo
c
u
s
o
n
m
e
tam
a
t
e
rial
a
n
d
tera
h
e
rtz
a
n
te
n
n
a
s
fo
r
b
io
m
e
d
ica
l
a
p
p
li
c
a
ti
o
n
s.
His
a
re
a
s
o
f
in
tere
st
i
n
c
lu
d
e
a
n
ten
n
a
d
e
sig
n
,
m
e
tam
a
teria
ls,
we
a
ra
b
le
a
n
ten
n
a
s,
a
n
d
THz
sy
ste
m
s
fo
r
m
e
d
ica
l
d
iag
n
o
st
ics
.
He
h
a
s
p
u
b
li
sh
e
d
se
v
e
ra
l
S
CI
-
i
n
d
e
x
e
d
re
se
a
rc
h
p
a
p
e
rs
a
n
d
is
c
u
rre
n
tl
y
e
x
p
lo
ri
n
g
a
d
v
a
n
c
e
d
a
n
ten
n
a
tec
h
n
o
lo
g
ies
f
o
r
wire
le
ss
a
n
d
b
i
o
m
e
d
ica
l
a
p
p
li
c
a
ti
o
n
s.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
c
e
g
n
e
sa
n
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.