I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
13
,
No
.
4
,
Dec
em
b
er
2
0
2
4
,
p
p
.
4
6
9
5
~
4
7
0
2
I
SS
N:
2
2
5
2
-
8938
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ai.
v
13
.i
4
.
p
p
4
6
9
5
-
4
7
0
2
4695
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Sa
rcas
m
d
etec
tion
on
so
cia
l
da
ta:
heuristic
sea
rch
a
nd
deep
lea
rning
Aru
m
ug
ha
m
P
a
la
nia
mm
a
l,
P
urus
ho
t
ha
m
a
n
A
na
nd
a
ba
b
u
D
e
p
a
r
t
me
n
t
of
C
o
mp
u
t
e
r
a
n
d
I
n
f
o
r
ma
t
i
o
n
S
c
i
e
n
c
e
,
A
n
n
a
mal
a
i
U
n
i
v
e
r
si
t
y
,
C
h
i
d
a
m
b
a
r
a
m
,
I
n
d
i
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
6
,
2024
R
ev
i
s
ed
A
p
r
9
,
2
0
2
4
A
cc
ep
ted
A
p
r
17
,
2
0
2
4
Du
e
to
t
h
e
sig
n
if
ica
n
t
s
u
rg
e
in
o
n
li
n
e
a
c
ti
v
it
y
,
sa
rc
a
s
m
d
e
tec
ti
o
n
(S
D)
h
a
s
a
tt
ra
c
ted
m
a
jo
r
a
tt
e
n
ti
o
n
in
so
c
ial
m
e
d
ia
n
e
t
w
o
rk
s.
S
a
rc
a
s
m
is
a
le
x
ica
l
it
e
m
of
n
e
g
a
ti
v
e
se
n
ti
m
e
n
ts
or
d
isl
ik
e
s
by
u
ti
li
z
in
g
e
x
a
g
g
e
r
a
ted
lan
g
u
a
g
e
c
o
n
stru
c
ts.
SD
h
a
s
c
re
a
ted
a
n
a
t
u
ra
l
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
(
NL
P
)
p
ro
c
e
d
u
re
f
o
c
u
se
d
on
t
h
e
in
t
rica
te
a
n
d
u
n
c
lea
r
a
sp
e
c
ts
of
sa
r
c
a
s
m
,
p
ri
m
a
ril
y
u
se
d
in
se
n
ti
m
e
n
t
a
n
a
ly
sis
(S
A
),
h
u
m
a
n
-
c
o
m
p
u
ter
i
n
tera
c
ti
o
n
,
a
n
d
v
a
rio
u
s
NL
P
a
p
p
li
c
a
ti
o
n
s.
C
o
n
c
u
rre
n
tl
y
,
a
d
v
a
n
c
e
m
e
n
ts
in
m
a
c
h
in
e
lea
rn
in
g
(M
L
)
a
p
p
ro
a
c
h
e
s
f
a
c
il
it
a
te
th
e
c
re
a
ti
o
n
of
e
ff
e
c
ti
v
e
SD
s
y
st
e
m
s
.
T
h
is
m
a
n
u
sc
rip
t
p
re
se
n
ts
th
e
f
u
tu
re
se
a
rc
h
a
lg
o
r
it
h
m
w
it
h
d
e
e
p
lea
rn
in
g
a
ss
isted
sa
rc
a
s
m
d
e
tec
ti
o
n
a
n
d
c
las
sif
ica
ti
o
n
on
s
o
c
ial
n
e
tw
o
rk
in
g
d
a
ta
(
F
S
A
DL
-
S
DCSND)
a
p
p
ro
a
c
h
.
T
h
e
m
a
jo
r
in
ten
t
of
th
e
F
S
A
D
L
-
S
DCSND
a
p
p
ro
a
c
h
is
in
th
e
e
ffe
c
ti
v
e
a
n
d
a
u
t
o
m
a
ted
re
c
o
g
n
it
io
n
of
sa
rc
a
stic
tex
t.
In
th
e
p
re
se
n
ted
F
S
A
D
L
-
S
DCSND
tec
h
n
iq
u
e
,
se
v
e
ra
l
d
a
ta
p
re
-
p
ro
c
e
ss
in
g
sta
g
e
s
a
r
e
a
c
h
iev
e
d
to
tran
sf
o
rm
th
e
d
a
ta
i
n
to
a
c
o
m
p
a
ti
b
le
f
o
rm
a
t.
Be
sid
e
s,
th
e
F
S
A
DL
-
S
DCSND
a
p
p
ro
a
c
h
a
p
p
li
e
s
a
b
i
d
irec
ti
o
n
a
l
se
rial
-
p
a
ra
ll
e
l
lo
n
g
sh
o
rt
-
term
m
e
m
o
r
y
(BS
-
P
L
S
T
M
)
a
p
p
ro
a
c
h
f
o
r
SD
a
n
d
c
las
si
f
ica
ti
o
n
.
T
h
e
h
y
p
e
rp
a
ra
m
e
t
e
r
tu
n
in
g
p
ro
c
e
ss
is
a
c
c
o
m
p
li
sh
e
d
by
e
m
p
lo
y
in
g
th
e
f
u
tu
re
se
a
rc
h
a
lg
o
rit
h
m
f
o
r
im
p
ro
v
in
g
th
e
re
c
o
g
n
it
io
n
of
th
e
BS
-
P
L
S
T
M
m
o
d
e
l.
F
o
r
su
p
e
rio
r
o
u
t
p
u
t
of
th
e
F
S
A
D
L
-
S
DCSND
m
o
d
e
l,
a
s
e
q
u
e
n
c
e
of
sim
u
latio
n
s
can
be
a
p
p
li
e
d
.
T
h
e
in
v
e
stig
a
ti
o
n
a
l
o
u
tp
u
ts
h
ig
h
li
g
h
t
e
d
th
e
im
p
ro
v
e
d
so
l
u
ti
o
n
s
of
th
e
F
S
A
D
L
-
S
DCSND
m
o
d
e
l
w
it
h
o
th
e
r
a
p
p
ro
a
c
h
e
s
u
n
d
e
r
d
iv
e
rse
p
e
rf
o
r
m
a
n
c
e
m
e
a
su
re
s
.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
F
u
t
u
r
e
s
ea
r
ch
al
g
o
r
ith
m
Natu
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
Sar
ca
s
m
d
etec
tio
n
So
cial
m
ed
ia
T
h
is
is
an
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r
th
e
CC
BY
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
A
r
u
m
u
g
h
a
m
P
alan
ia
m
m
al
Dep
ar
t
m
en
t
of
C
o
m
p
u
ter
an
d
I
n
f
o
r
m
a
tio
n
Scie
n
ce
,
An
n
a
m
a
lai
Un
i
v
er
s
it
y
C
h
id
a
m
b
ar
a
m
,
I
n
d
ia
E
m
ail:
r
ev
a
th
y
2
0
k
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Mic
r
o
b
lo
g
g
in
g
p
lat
f
o
r
m
s
ar
e
th
e
m
ain
m
ed
i
u
m
s
f
o
r
ex
p
r
es
s
in
g
t
h
e
i
n
d
iv
id
u
al
’
s
o
p
in
io
n
s
,
th
o
u
g
h
ts
,
an
d
v
ie
w
s
on
d
i
f
f
er
en
t
d
o
m
ai
n
s
an
d
ac
tio
n
s
[
1
]
.
Sar
ca
s
m
is
a
co
m
p
lica
ted
p
r
o
ce
s
s
of
ir
o
n
y
th
at
o
f
te
n
o
r
ig
i
n
ates
in
s
o
cial
n
et
w
o
r
k
i
n
g
s
ite
s
an
d
m
icr
o
b
lo
g
g
i
n
g
p
latf
o
r
m
s,
as
t
h
ese
p
lat
f
o
r
m
s
f
r
eq
u
e
n
tl
y
s
u
p
p
o
r
t
co
n
d
em
n
atio
n
an
d
tr
o
llin
g
of
o
th
er
s
[
2
]
.
In
tex
t
clas
s
if
icatio
n
,
s
ar
ca
s
m
d
ete
ctio
n
(
SD
)
can
be
a
cr
u
cial
to
o
l,
w
h
ich
h
as
s
e
v
er
al
ef
f
ec
ts
on
n
u
m
er
o
u
s
f
ield
s
i
n
c
lu
d
in
g
s
ale
s
,
h
ea
lt
h,
an
d
s
ec
u
r
it
y
[
3
]
.
By
u
s
i
n
g
SD
m
e
th
o
d
s
,
co
m
p
a
n
ies
a
n
al
y
s
e
clien
t
s
’
e
m
o
tio
n
s
ab
o
u
t
t
h
eir
p
r
o
d
u
cts.
It
o
f
f
er
s
an
i
m
p
o
r
ta
n
t
b
en
ef
it
of
in
cr
ea
s
in
g
t
h
e
q
u
al
it
y
of
p
r
o
d
u
cts
[
4
]
.
In
s
en
ti
m
e
n
t
an
al
y
s
is
(
SA
)
,
th
e
s
ar
ca
s
m
cla
s
s
i
f
icat
io
n
h
a
s
a
cr
itical
s
u
b
-
ac
ti
v
it
y
,
p
ar
ticu
l
ar
l
y
f
o
r
class
i
f
y
in
g
t
w
ee
t
s
,
an
d
f
o
r
s
h
ar
i
n
g
i
m
p
li
ed
d
etails
w
it
h
i
n
t
h
e
in
f
o
r
m
a
t
io
n
th
at
an
i
n
d
iv
id
u
al
ex
p
r
es
s
es
or
co
n
v
e
y
s
w
it
h
o
th
er
s
[
5
]
.
A
d
d
itio
n
all
y
,
t
h
e
f
r
a
m
e
w
o
r
k
of
th
e
t
w
ee
t
ca
n
be
also
u
til
iz
ed
f
o
r
p
r
ed
ictin
g
s
ar
ca
s
m
(
f
o
r
ex
a
m
p
le,
ch
an
g
i
n
g
t
h
e
p
o
lar
it
y
of
n
e
g
at
iv
e
or
p
o
s
itiv
e
s
tate
m
e
n
ts
i
n
to
th
eir
o
p
p
o
s
ite
ty
p
es).
On
T
w
i
tter
,
th
e
n
u
m
er
o
u
s
p
r
o
b
lem
s
cr
ea
te
SD
as
a
c
h
all
en
g
i
n
g
tas
k
[
6]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
13
,
No
.
4
,
Dec
em
b
er
2024
:
4
6
9
5
-
4
7
0
2
4696
W
h
ile
u
ti
lizi
n
g
t
h
e
tr
ad
itio
n
al
m
ac
h
in
e
lear
n
i
n
g
(
ML
)
a
p
p
r
o
ac
h
,
f
ea
tu
r
e
ex
tr
ac
tio
n
p
er
f
o
r
m
s
a
s
ig
n
i
f
ica
n
t
p
ar
t
[
7
]
.
A
s
et
of
f
e
atu
r
es
is
d
e
s
cr
ib
ed
f
o
r
ex
tr
ac
ti
n
g
m
a
n
y
co
n
tex
ts
f
r
o
m
t
h
e
in
f
o
r
m
at
io
n
.
In
a
s
t
u
d
y
,
a
s
et
of
f
ea
t
u
r
es
to
f
in
d
w
h
eth
er
an
a
n
al
y
s
is
is
s
ar
ca
s
tic
or
n
o
t
f
o
r
m
ed
.
T
h
ese
f
ea
t
u
r
es
ar
e
i
m
b
alan
ce
,
p
u
n
ct
u
atio
n
,
a
n
d
h
y
p
er
b
o
le.
It
d
ec
id
es
a
co
m
m
e
n
t
i
m
b
ala
n
ce
w
h
en
th
e
p
r
o
v
id
ed
s
tar
r
atin
g
is
i
n
cr
ea
s
ed
h
o
w
ev
er
,
th
e
m
aj
o
r
it
y
of
w
o
r
d
s
in
th
e
te
x
t
h
a
v
e
n
e
g
ati
v
e
o
p
in
io
n
s
,
an
d
alter
n
atel
y
[
8
]
.
T
h
e
co
m
m
en
ts
ar
e
lab
eled
h
y
p
er
b
o
licall
y
w
h
en
m
o
r
e
th
a
n
th
r
ee
n
e
g
ati
v
e
or
p
o
s
iti
v
e
w
o
r
d
s
.
T
h
e
p
u
n
ctu
atio
n
m
ar
k
s
th
e
ex
i
s
ten
ce
of
ex
cla
m
atio
n
m
ar
ks
or
m
a
n
y
q
u
esti
o
n
s
[
9
]
.
Deep
lear
n
i
n
g
(
DL
)
tec
h
n
iq
u
es
ar
e
also
e
m
p
lo
y
ed
f
o
r
SD
b
ec
a
u
s
e
of
t
h
eir
ca
p
ab
ilit
y
to
au
to
m
ati
ca
ll
y
lear
n
d
i
f
f
icu
l
t
r
ep
r
esen
tat
io
n
s
a
n
d
p
atter
n
s
in
r
a
w
te
x
t
d
ata.
C
o
n
ce
r
n
in
g
th
e
DL
ap
p
r
o
ac
h
,
tech
n
iq
u
e
s
e
lect
io
n
is
t
h
e
ch
al
len
g
i
n
g
p
ar
t
[
1
0
]
.
T
r
an
s
f
o
r
m
er
m
et
h
o
d
s
ar
e
a
cu
r
r
en
t
d
ev
elo
p
m
en
t
in
DL
m
e
th
o
d
s
an
d
ar
e
ex
tr
e
m
el
y
ef
f
icie
n
t
f
o
r
an
ex
te
n
s
iv
e
r
an
g
e
of
n
at
u
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
i
n
g
ta
s
k
s
.
T
h
is
m
an
u
s
cr
ip
t
o
f
f
er
s
th
e
d
es
ig
n
of
th
e
f
u
t
u
r
e
s
ea
r
ch
alg
o
r
it
h
m
w
i
th
d
ee
p
lear
n
in
g
ass
i
s
te
d
s
ar
ca
s
m
d
etec
tio
n
an
d
cla
s
s
i
f
icatio
n
o
n
s
o
cial
n
et
w
o
r
k
in
g
d
ata
(
F
S
ADL
-
SD
C
S
ND)
ap
p
r
o
ac
h
.
T
h
e
m
aj
o
r
in
ten
t
of
t
h
e
FS
A
D
L
-
SD
C
SN
D
ap
p
r
o
ac
h
is
in
t
h
e
e
f
f
ec
ti
v
e
a
n
d
au
to
m
at
ed
r
ec
o
g
n
itio
n
of
s
ar
ca
s
t
ic
te
x
t.
In
t
h
e
p
r
esen
ted
FS
A
D
L
-
SD
C
SN
D
tech
n
iq
u
e,
s
ev
er
al
s
tag
e
s
of
d
ata
p
r
ep
r
o
c
ess
i
n
g
ar
e
ac
h
ie
v
ed
to
tr
an
s
f
o
r
m
th
e
d
ata
i
n
to
a
co
m
p
atib
le
f
o
r
m
at.
B
esid
e
s
,
th
e
F
S
A
D
L
-
SD
C
SND
ap
p
r
o
ac
h
ap
p
lies
a
b
id
ir
ec
tio
n
al
s
er
ial
-
p
ar
allel
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
B
S
-
P
L
S
T
M)
a
p
p
r
o
ac
h
f
or
r
ec
o
g
n
izi
n
g
an
d
class
i
f
y
in
g
t
h
e
s
ar
ca
s
m
.
T
h
e
h
y
p
er
p
ar
am
e
ter
tu
n
in
g
p
r
o
ce
s
s
is
ac
co
m
p
lis
h
e
d
by
e
m
p
lo
y
i
n
g
th
e
f
u
t
u
r
e
s
ea
r
ch
alg
o
r
ith
m
(
FS
A
)
f
o
r
i
m
p
r
o
v
in
g
t
h
e
r
ec
o
g
n
itio
n
ac
h
iev
e
m
e
n
t
of
t
h
e
BS
-
P
L
ST
M
m
o
d
el.
To
ex
h
ib
it
th
e
g
r
ea
ter
s
o
lu
tio
n
of
t
h
e
FS
AD
L
-
SDC
S
ND
m
o
d
el,
a
s
eq
u
en
ce
of
s
i
m
u
latio
n
s
can
be
ap
p
lied
.
T
h
e
s
u
b
s
eq
u
e
n
t
s
ec
tio
n
s
of
th
e
ar
ticle
ar
e
ar
r
an
g
ed
a
s
:
s
ec
tio
n
2
s
h
o
w
s
t
h
e
li
ter
atu
r
e
r
ev
ie
w
.
Sectio
n
3
o
u
tli
n
es
t
h
e
p
r
o
p
o
s
e
d
m
o
d
el.
Sectio
n
4
elab
o
r
ates
on
th
e
p
er
f
o
r
m
a
n
ce
e
v
alu
a
tio
n
.
Sectio
n
5
co
m
p
letes
th
e
w
o
r
k
.
2.
RE
L
AT
E
D
WO
RK
S
A
cc
o
r
in
g
to
Siv
a
lin
g
a
m
et
a
l.
[
1
1
]
,
an
u
n
s
u
p
er
v
i
s
ed
lear
n
in
g
tech
n
iq
u
e,
th
e
co
n
d
itio
n
a
l
r
an
d
o
m
f
ield
s
tr
u
ct
u
r
e
m
o
d
i
f
ied
ex
p
ec
tatio
n
m
ax
i
m
iza
tio
n
(
C
R
F
-
MEM)
tech
n
iq
u
e
is
d
ev
elo
p
ed
to
id
en
tify
s
ar
ca
s
m
in
t
w
ee
t
s
.
T
h
is
d
ev
elo
p
ed
alg
o
r
ith
m
tar
g
et
s
to
ad
d
r
ess
th
e
d
r
a
w
b
ac
k
ex
i
s
ti
n
g
in
th
e
s
ta
n
d
ar
d
ex
p
ec
tatio
n
m
ax
i
m
izatio
n
(
EM
)
m
et
h
o
d
,
th
e
r
an
d
o
m
allo
ca
tio
n
f
ac
to
r
s
w
i
th
th
e
p
r
ese
n
ted
f
ea
tu
r
e
c
o
r
r
elatio
n
v
al
u
es.
P
an
d
ey
an
d
Si
n
g
h
[
1
2
]
pr
o
p
o
s
ed
a
tech
n
iq
u
e
i
n
cl
u
d
in
g
b
id
ir
e
ctio
n
al
en
co
d
er
r
ep
r
esen
tatio
n
s
f
r
o
m
tr
an
s
f
o
r
m
er
s
(
B
E
R
T
)
s
tack
ed
w
it
h
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
L
ST
M)
.
T
h
e
e
m
b
ed
d
ed
v
ec
to
r
s
ar
e
u
t
il
ized
by
t
h
e
L
ST
M
n
et
w
o
r
k
co
m
p
r
i
s
i
n
g
a
s
i
n
g
le
la
y
er
.
K
u
m
ar
et
a
l.
[
1
3
]
i
n
tr
o
d
u
ce
d
a
m
u
lti
-
h
ea
d
atte
n
tio
n
(
MH
A
)
-
b
ased
b
id
ir
ec
tio
n
al
-
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(B
i
-
L
ST
M)
m
o
d
el
f
o
r
d
etec
tin
g
s
ar
ca
s
tic
co
m
m
e
n
t
ar
ies
in
a
s
p
ec
i
f
ied
q
u
an
tit
y
.
P
alan
ia
m
m
al
an
d
A
n
a
n
d
ab
ab
u
[
1
4
]
in
tr
o
d
u
ce
d
a
n
o
v
el
ch
a
o
s
s
p
ec
if
icatio
n
c
u
r
v
e
an
a
l
y
s
i
s
(
C
SC
A
)
w
it
h
a
g
r
ap
h
co
n
v
o
lu
t
io
n
al
n
et
w
o
r
k
f
o
r
s
ar
ca
s
m
d
etec
tio
n
(
GC
NSD)
ap
p
r
o
ac
h
.
Af
ter
p
r
e
-
p
r
o
ce
s
s
in
g
,
th
e
m
et
h
o
d
i
m
p
le
m
en
ts
t
h
e
g
r
ap
h
co
n
v
o
lu
tio
n
al
n
et
w
o
r
k
(
GC
N
)
f
r
a
m
e
w
o
r
k
f
o
r
id
en
ti
f
y
i
n
g
a
n
d
class
i
f
y
in
g
d
i
v
er
s
e
t
y
p
es
of
s
ar
ca
s
m
.
L
astl
y
,
t
h
e
C
SC
A
m
et
h
o
d
w
a
s
e
m
p
lo
y
ed
f
o
r
o
p
tim
al
l
y
s
elec
ti
n
g
th
e
h
y
p
er
p
ar
am
eter
o
u
tp
u
t
s
.
Vin
o
t
h
an
d
P
r
a
b
h
av
ath
y
[
1
5
]
s
u
g
g
ested
an
in
tellec
tu
al
m
ac
h
in
e
lear
n
i
n
g
-
b
a
s
ed
s
ar
c
as
m
d
etec
tio
n
an
d
class
i
f
icatio
n
(
I
ML
B
-
SD
C
)
m
o
d
el.
T
h
e
s
tu
d
y
i
n
te
n
d
s
f
o
r
r
ec
o
g
n
izi
n
g
t
h
e
s
ar
ca
s
m
p
r
esen
ce
in
s
o
cial
n
et
w
o
r
k
s
.
In
ad
d
itio
n
,
th
e
f
ea
tu
r
e
en
g
in
e
er
in
g
m
e
th
o
d
o
cc
u
r
s
u
s
i
n
g
TF
-
I
DF.
B
esid
es,
a
2
-
f
ea
tu
r
e
s
elec
tio
n
(
FS
)
tech
n
iq
u
e
h
as
b
ee
n
i
m
p
le
m
en
ted
lik
e
d
at
a
g
ain
a
n
d
ch
i
-
s
q
u
ar
e.
Vin
o
t
h
an
d
P
r
ab
h
av
ath
y
[
1
6
]
d
esig
n
ed
an
a
u
to
m
ated
s
ar
ca
s
m
d
etec
tio
n
a
n
d
class
i
f
icat
io
n
m
et
h
o
d
th
r
o
u
g
h
a
h
y
p
er
p
ar
a
m
eter
t
u
n
ed
-
d
ee
p
lear
n
in
g
(
ASD
C
-
HP
T
DL
)
ap
p
r
o
ac
h
f
o
r
a
s
o
cial
Net
w
o
r
k
p
lat
f
o
r
m
.
In
th
e
f
o
llo
w
i
n
g
p
h
ase
of
p
r
ep
r
o
ce
s
s
i
n
g
,
th
e
p
r
e
-
p
r
o
ce
s
s
ed
in
f
o
r
m
at
io
n
can
be
tr
an
s
f
er
r
ed
to
th
e
f
ea
tu
r
e
v
ec
to
r
s
u
tili
ze
d
by
Glo
v
e
E
m
b
ed
d
in
g
'
s
al
g
o
r
it
h
m
.
Ne
x
t,
th
e
at
ten
t
i
o
n
b
id
ir
ec
tio
n
al
-
g
ated
r
ec
u
r
r
en
t
u
n
it
(
A
b
i
-
GR
U)
m
et
h
o
d
w
a
s
e
m
p
lo
y
ed
f
o
r
t
h
e
class
i
f
icatio
n
a
n
d
d
etec
tio
n
of
s
ar
ca
s
m
.
A
l
s
o
,
th
e
i
m
p
r
o
v
ed
ar
tif
ic
ial
f
lo
r
a
opt
im
izatio
n
(
I
AFO)
ap
p
r
o
ac
h
w
a
s
ex
p
lo
ited
f
o
r
h
y
p
er
p
ar
a
m
eter
t
u
n
in
g
.
Ash
o
k
et
al
.
[
1
7
]
r
ec
o
m
m
e
n
d
ed
a
d
is
tin
ct
iv
e
d
ee
p
n
eu
r
al
n
et
w
o
r
k
(
DNN)
al
g
o
r
ith
m
r
elate
d
to
Bi
-
L
ST
M
on
h
y
p
er
p
ar
am
e
ter
s
o
p
ti
m
izer
t
h
r
o
u
g
h
a
g
en
etic
al
g
o
r
ith
m
(
G
A
)
a
n
d
th
en
a
co
n
v
o
l
u
tio
n
n
eu
r
al
n
et
w
o
r
k
(
C
NN)
f
o
r
SD.
Sah
u
an
d
Hu
d
n
u
r
k
ar
[
1
8
]
p
r
e
s
en
t
s
a
SD
m
o
d
el
u
t
ilizi
n
g
c
h
i
-
s
q
u
ar
e,
g
ai
n
,
m
u
tu
a
l
in
f
o
r
m
atio
n
,
an
d
s
y
m
m
etr
ical
u
n
ce
r
tai
n
t
y
m
eth
o
d
s
.
A
g
r
o
u
p
class
i
f
ier
n
a
m
e
l
y
n
e
u
r
al
n
et
w
o
r
k
s
(
NN
)
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
s
(
SVM
)
,
r
an
d
o
m
f
o
r
est
(
RF
)
,
an
d
an
o
p
tim
ized
d
ee
p
c
o
n
v
o
lu
tio
n
al
n
e
u
r
al
n
et
w
o
r
k
(
DC
NN)
w
i
th
w
ei
g
h
t
s
elec
tio
n
is
a
ls
o
e
m
p
lo
y
ed
.
T
h
e
cla
n
u
p
g
r
ad
ed
g
r
e
y
w
o
l
f
o
p
ti
m
izatio
n
(
C
U
-
GW
O)
m
o
d
el
is
in
tr
o
d
u
ce
d
f
o
r
o
p
tim
a
l
FS
an
d
o
p
ti
m
ized
DC
NN.
A
cc
o
r
d
in
g
to
Ma
g
o
o
an
d
Sin
g
h
[
1
9
]
,
a
h
eu
r
is
tic
-
as
s
is
te
d
ca
p
s
u
le
n
et
w
o
r
k
(H
-
C
ap
Net)
tec
h
n
iq
u
e
is
p
r
o
p
o
s
ed
.
To
im
p
r
o
v
e
t
h
e
ca
p
s
u
le
n
et
w
o
r
k
,
a
f
u
s
io
n
m
eta
h
e
u
r
is
ti
c
m
o
d
el,
s
p
ec
if
ical
l
y
th
e
escap
i
n
g
e
n
er
g
y
s
ea
r
ch
ed
g
r
e
y
–
h
ar
r
i
s
h
a
w
k
s
’
al
g
o
r
ith
m
(EEG
-
HH
A
)
,
is
i
m
p
le
m
en
ted
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
S
a
r
ca
s
m
d
etec
tio
n
o
n
s
o
cia
l d
a
ta
:
h
e
u
r
is
tic
s
ea
r
ch
a
n
d
d
ee
p
lea
r
n
in
g
(
A
r
u
mu
g
h
a
m
P
a
la
n
i
a
mma
l
)
4697
3.
T
H
E
P
RO
P
O
SE
D
M
O
DE
L
In
th
e
p
r
esen
ted
ar
ticle,
a
n
o
v
el
F
S
A
D
L
-
SD
C
SN
D
ap
p
r
o
a
ch
f
o
r
au
to
m
a
tic
an
d
ac
c
u
r
at
e
SD
an
d
class
i
f
icatio
n
on
s
o
cial
n
et
w
o
r
k
in
g
d
ata
is
i
n
tr
o
d
u
ce
d
.
T
h
e
in
te
n
tio
n
of
t
h
e
FS
ADL
-
SDC
SND
ap
p
r
o
ac
h
is
in
th
e
e
f
f
ec
t
u
al
a
n
d
au
to
m
atic
r
ec
o
g
n
itio
n
of
s
ar
ca
s
tic
te
x
t.
In
t
h
e
p
r
o
p
o
s
ed
FS
A
D
L
-
SD
C
SND
m
eth
o
d
,
th
r
ee
p
h
ases
ar
e
in
v
o
lv
ed
n
a
m
e
l
y
d
ata
p
r
ep
r
o
ce
s
s
in
g
,
BS
-
P
L
ST
M
-
b
ased
class
if
ica
tio
n
,
an
d
FS
A
-
b
ased
tu
n
in
g
.
Fig
u
r
e
1
d
ef
in
e
s
th
e
co
m
p
le
te
p
r
o
ce
d
u
r
e
of
th
e
FS
A
D
L
-
S
D
C
SND
m
o
d
el.
Fig
u
r
e
1.
Ov
er
all
p
r
o
ce
s
s
of
F
S
A
D
L
-
SD
C
SND
m
et
h
o
d
3
.
1
.
Da
t
a
prepro
ce
s
s
ing
To
co
n
v
er
t
th
e
d
ata
in
to
a
m
ea
n
in
g
f
u
l
f
o
r
m
at,
d
ata
p
r
ep
r
o
ce
s
s
in
g
is
in
v
o
lv
ed
.
SD
in
s
o
cial
m
ed
ia
tex
t
is
cr
u
cial
d
u
e
to
t
h
e
co
n
te
x
t
-
d
ep
en
d
en
t
an
d
i
n
f
o
r
m
al
n
a
tu
r
e
of
o
n
li
n
e
co
m
m
u
n
icat
io
n
.
Ho
w
e
v
er
,
th
e
ac
c
u
r
ac
y
of
th
e
SD
m
o
d
el
can
be
en
h
an
ce
d
w
it
h
th
e
r
i
g
h
t
an
d
ca
r
ef
u
l
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es.
T
ex
t
lo
w
er
ca
s
i
n
g
:
co
n
v
er
t
t
h
e
tex
t
i
n
to
lo
w
er
ca
s
e.
T
h
is
allo
ws
th
e
m
o
d
el
th
at
d
o
es
n
o
t
tr
ea
t
s
i
m
ilar
w
o
r
d
s
w
it
h
d
is
s
i
m
ilar
ca
s
es
as
d
is
s
i
m
ilar
w
o
r
d
s
.
T
o
k
en
izatio
n
:
s
p
li
t
t
h
e
tex
ts
in
to
to
k
e
n
s
or
in
d
iv
id
u
al
w
o
r
d
s
.
T
o
k
en
izatio
n
is
a
ch
alle
n
g
i
n
g
ta
s
k
to
u
n
d
er
s
ta
n
d
th
e
s
tr
u
ct
u
r
e
of
te
x
t.
R
e
m
o
v
i
n
g
p
u
n
ctu
a
tio
n
:
r
e
m
o
v
e
s
p
ec
ial
ch
ar
ac
ter
s
,
p
u
n
ctu
ati
o
n
m
ar
k
s
,
an
d
e
m
o
j
is
.
T
h
ey
can
ad
d
n
o
is
e
to
th
e
d
ata
an
d
do
n
o
t
co
n
tr
ib
u
te
m
u
c
h
m
ea
n
in
g
to
th
e
S
D.
R
e
m
o
v
i
n
g
s
to
p
w
o
r
d
s
:
s
to
p
w
o
r
d
s
ar
e
g
en
er
al
w
o
r
d
s
li
k
e
"
th
e,
"
"
an
d
,
"
"
is
,
"
,
th
at
m
a
y
not
ca
r
r
y
m
u
c
h
m
ea
n
in
g
f
o
r
a
s
en
te
n
ce
.
T
h
is
m
i
g
h
t
be
u
s
e
f
u
l
f
o
r
r
ed
u
cin
g
n
o
is
e
in
t
h
e
tex
t.
Han
d
lin
g
co
n
tr
ac
tio
n
s
:
e
x
p
an
d
co
n
tr
ac
tio
n
s
s
u
ch
as
"
ca
n
'
t"
to
"
ca
n
n
o
t"
an
d
"
w
o
n
'
t"
to
"w
il
l
n
o
t"
.
T
h
is
w
il
l
en
s
u
r
e
t
h
e
m
o
d
el
i
m
p
r
o
v
e
s
its
u
n
d
er
s
ta
n
d
in
g
a
n
d
r
ec
o
g
n
izes
th
e
co
m
p
le
te
w
o
r
d
s
.
Sp
ell
ch
ec
k
in
g
an
d
co
r
r
ec
tio
n
:
co
r
r
ec
t
s
p
ellin
g
er
r
o
r
s
to
d
ev
elo
p
th
e
o
v
er
all
q
u
alit
y
of
t
h
e
t
ex
t.
E
m
o
tio
n
an
d
in
ten
s
it
y
m
ar
k
er
s
:
id
en
tify
p
h
r
ase
s
or
w
o
r
d
s
th
a
t
r
ep
r
esen
t
in
ten
s
it
y
or
e
m
o
tio
n
s
,
th
is
e
n
s
u
r
es
in
d
etec
tio
n
of
s
ar
ca
s
m
.
C
o
n
te
x
t
u
al
f
ea
tu
r
e
s
:
ex
tr
ac
t
f
ea
tu
r
es
s
u
c
h
as
p
ar
t
-
of
-
s
p
ee
c
h
ta
g
s
,
ca
lled
e
n
titi
e
s
,
an
d
s
e
n
ti
m
en
t
s
co
r
es.
T
h
is
f
ea
t
u
r
e
p
r
o
v
id
es
o
th
er
co
n
tex
ts
to
d
etec
t
s
ar
ca
s
m
.
Ste
m
m
i
n
g
or
lem
m
atiza
t
io
n
:
a
co
m
m
o
n
r
o
o
t
(
s
tem
m
i
n
g
)
or
R
ed
u
cin
g
w
o
r
d
s
to
th
e
ir
b
ase
f
o
r
m
(
le
m
m
atiza
tio
n
)
m
i
g
h
t
be
u
s
e
f
u
l
in
r
ec
o
g
n
izi
n
g
t
h
e
v
ar
iatio
n
of
w
o
r
d
s
a
n
d
in
cr
ea
s
i
n
g
th
e
ir
v
o
ca
b
u
lar
y
co
v
er
ag
e.
Han
d
lin
g
ab
b
r
ev
iatio
n
s
an
d
ac
r
o
n
y
m
s
:
r
ep
lace
co
m
m
o
n
ab
b
r
ev
iatio
n
s
an
d
ac
r
o
n
y
m
s
w
it
h
th
eir
f
u
ll
f
o
r
m
s
.
T
h
is
allo
w
s
us
to
u
n
d
er
s
ta
n
d
t
h
e
in
te
n
d
ed
m
ea
n
i
n
g
.
Han
d
lin
g
r
ep
ea
ted
ch
ar
ac
ter
s
:
n
o
r
m
alize
r
ep
ea
ted
ch
ar
ac
ter
s
lik
e
t
u
r
n
i
n
g
"
h
ah
a
h
ah
a"
in
to
"
h
ah
a.
"
T
h
is
en
ab
les
s
ta
n
d
ar
d
ized
ex
p
r
ess
i
o
n
an
d
d
ec
r
ea
s
es
n
o
is
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
13
,
No
.
4
,
Dec
em
b
er
2024
:
4
6
9
5
-
4
7
0
2
4698
User
m
e
n
tio
n
s
a
n
d
h
as
h
ta
g
s
:
co
n
s
id
er
w
h
e
th
er
to
r
ep
lace
or
r
e
m
o
v
e
h
as
h
ta
g
s
a
n
d
u
s
er
m
en
tio
n
s
(
e.
g
.
,
"
@
u
s
er
n
a
m
e"
)
,
th
is
m
a
y
or
m
a
y
n
o
t
co
n
tr
ib
u
te
to
SD.
C
o
n
te
x
t
u
al
in
f
o
r
m
a
tio
n
:
ca
p
tu
r
e
th
e
co
n
v
er
s
a
tio
n
co
n
te
x
t
by
in
cl
u
d
in
g
s
o
m
e
p
ast
an
d
s
u
cc
e
ed
in
g
p
o
s
ts
or
s
en
te
n
ce
s
.
It
w
ill
p
r
o
v
id
e
co
n
t
ex
t
f
o
r
d
etec
tin
g
s
ar
ca
s
m
.
Neg
atio
n
h
a
n
d
lin
g
:
id
en
t
if
y
n
eg
atio
n
s
s
u
ch
as
"
n
o
t,"
"
n
o
,
"
"
n
ev
er
,
"
an
d
so
o
n
,
w
h
ic
h
r
ep
r
esen
t
s
ar
ca
s
m
w
h
e
n
ap
p
lied
in
a
s
p
ec
if
ic
co
n
tex
t.
Featu
r
e
en
g
i
n
ee
r
i
n
g
:
cr
ea
te
c
u
s
to
m
f
ea
tu
r
es
t
h
at
m
a
y
ass
is
t
in
d
etec
ti
n
g
s
ar
ca
s
m
,
l
ik
e
t
h
e
p
r
esen
ce
of
s
p
ec
if
ic
w
o
r
d
s
or
p
h
r
ases
,
w
o
r
d
co
u
n
t,
an
d
s
e
n
ten
ce
le
n
g
t
h
.
3
.
2
.
S
a
rc
a
s
m
det
ec
t
io
n
u
s
in
g
bid
irec
t
io
na
l seria
l
-
pa
ra
llel lo
ng
s
ho
rt
-
t
er
m
m
e
m
o
ry
m
o
del
Fo
r
th
e
SD
p
r
o
ce
s
s
,
th
e
BS
-
P
L
ST
M
m
o
d
el
is
ap
p
lied
.
BS
-
P
L
ST
M
is
a
d
ev
elo
p
m
e
n
t
of
Bi
-
L
ST
M
[
2
0
]
.
T
h
e
m
aj
o
r
d
if
f
er
en
ce
a
m
o
n
g
s
er
ial
-
p
ar
allel
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
S‐P
L
ST
M)
an
d
ex
is
t
s
ev
er
al
L
ST
M
ce
lls
f
r
o
m
th
e
S‐P
L
ST
M
at
tim
e
s
te
p
.
Fo
r
h
an
d
lin
g
th
e
d
ata
tr
an
s
f
er
r
in
g
an
d
in
te
g
r
ati
n
g
i
s
s
u
es
f
r
o
m
th
e
ti
m
e
p
ath
of
p
ar
allel
p
r
o
ce
d
u
r
es,
2
g
ates
ar
e
d
esig
n
ed
th
at
ar
e
ca
lled
in
teg
r
ated
g
ates
(
I
Gs)
an
d
in
ter
n
al
in
teg
r
ated
g
ate
s
(
I
I
Gs).
IIG
h
as
d
ev
elo
p
ed
to
ac
q
u
ir
e
th
e
in
ter
n
al
co
m
m
u
n
icatio
n
s
a
m
o
n
g
ev
er
y
p
ar
allel
m
o
d
el
at
t
h
e
ex
i
s
ti
n
g
ti
m
e
s
tep
an
d
s
p
ec
if
ic
m
et
h
o
d
at
th
e
s
u
b
s
eq
u
e
n
t
ti
m
e
-
s
tep
s
.
I
I
G
p
r
o
ce
e
d
s
to
th
e
h
id
d
en
la
y
e
r
(
HL
)
ℎ
̃
,
,
an
d
ce
ll
la
y
er
̃
,
of
p
ar
allel
L
ST
M
ce
lls
at
th
e
ex
i
s
ti
n
g
ti
m
e
s
tep
as
th
e
in
p
u
t
p
ar
t.
In
th
e
m
ea
n
ti
m
e,
it
e
m
p
lo
y
s
t
h
e
q
u
alit
y
e
m
b
ed
d
ed
v
ec
to
r
+
1
,
of
ℎ
p
r
o
ce
d
u
r
e
at
th
e
f
o
llo
w
in
g
ti
m
e
s
tep
as
an
o
th
er
i
n
p
u
t
p
ar
t.
Af
te
r
w
ar
d
,
th
e
HL
ℎ
,
an
d
ce
ll
la
y
er
,
an
d
th
e
n
i
n
ter
ac
tiv
e
i
n
te
g
r
atio
n
is
t
h
e
o
u
t
co
m
e
as
t
h
e
i
n
p
u
t
v
ec
to
r
s
of
ℎ
L
ST
M
ce
ll
at
th
e
f
o
llo
w
i
n
g
ti
m
estep
.
IIG
h
as
b
ee
n
e
s
tab
lis
h
ed
by
th
e
s
u
b
s
eq
u
en
t
eq
u
a
tio
n
s
:
,
=
(
{
̃
,
+
1
,
}
+
)
(
1
)
,
=
ex
p
(
,
)
∑
ex
p
(
,
)
(
2
)
{
,
ℎ
,
}
=
∑
,
∗
{
̃
,
ℎ
̃
,
}
(
3
)
w
h
er
ea
s
∈
ℝ
(
ℎ
+
)
×
ℎ
an
d
∈
ℝ
(
ℎ
+
)
ar
e
lear
n
ab
le
p
ar
a
m
eter
s
.
It
p
r
o
ce
ed
s
th
e
HL
ℎ
̃
,
an
d
ce
ll
la
y
er
̃
,
of
L
ST
M
la
y
er
s
at
th
e
las
t
ti
m
estep
as
in
p
u
ts
.
Af
ter
w
ar
d
,
th
e
HL
ℎ
is
t
h
e
o
u
t
co
m
e
af
te
r
in
te
g
r
atio
n
.
=
(
̃
,
+
)
(
4
)
=
ex
p
(
)
∑
ex
p
(
)
(
5
)
ℎ
=
∑
∗
ℎ
̃
,
(
6
)
W
h
er
ea
s
∈
ℝ
ℎ
×
ℎ
an
d
∈
ℝ
ℎ
d
en
o
te
th
e
lear
n
ab
le
p
ar
am
eter
s
of
I
G.
T
h
e
co
n
n
ec
tio
n
b
et
w
ee
n
BS
-
P
L
ST
M
an
d
S
-
P
L
ST
M
is
th
e
s
a
m
e
as
t
h
at
a
m
o
n
g
B
iLST
M
as
w
el
l
as
L
ST
M.
BS
-
P
L
ST
M
ass
u
m
es
2
d
ir
ec
tio
n
s
of
ti
m
e
s
eq
u
e
n
ce
s
.
T
h
u
s
,
th
e
la
s
t
o
u
tco
m
e
of
BS
-
P
L
ST
M
is
ℎ
=
[
ℎ
⃗
⃗
⃗
⃗
;
ℎ
1
⃖
⃗
⃗
⃗
⃗
]
.
3
.
3
.
H
y
perpa
ra
m
et
er
t
un
ing
us
ing
f
uture
s
ea
rc
h a
lg
o
rit
hm
Fo
r
ad
j
u
s
tin
g
th
e
h
y
p
er
p
ar
a
m
eter
s
ass
o
ciate
d
to
th
e
BS
-
P
L
ST
M
m
et
h
o
d
,
th
e
FS
A
can
be
u
s
ed
.
E
ac
h
s
p
ec
ies
on
t
h
e
ea
r
t
h
d
eter
m
i
n
e
s
an
o
p
ti
m
u
m
m
a
n
n
er
.
W
h
e
n
an
y
liv
in
g
b
ein
g
f
i
n
d
s
t
h
at
it
s
w
a
y
of
liv
in
g
is
n
o
t
o
p
tim
u
m
,
th
e
n
it
can
atte
m
p
t
to
f
o
llo
w
o
th
er
li
v
i
n
g
li
f
e
in
th
e
w
o
r
ld
[
2
1
]
.
E
v
er
y
p
er
s
o
n
t
r
ies
to
f
in
d
a
b
etter
lif
e.
FS
A
e
x
p
r
ess
e
s
a
m
at
h
e
m
atica
l
eq
u
atio
n
t
h
at
u
s
e
s
lo
ca
l
an
d
g
lo
b
al
s
ea
r
c
h
in
g
b
et
w
ee
n
p
eo
p
le
an
d
t
h
e
h
is
to
r
ies
b
est
p
eo
p
le
an
d
u
p
d
ates
th
e
r
an
d
o
m
in
i
tial.
A
n
y
h
e
u
r
is
tic
ap
p
r
o
ac
h
es
b
eg
i
n
w
it
h
ar
b
itra
r
y
s
ta
g
es
a
n
d
cr
ea
te
iter
atio
n
s
d
ep
en
d
in
g
on
th
e
p
r
i
m
ar
y
o
p
ti
m
u
m
p
er
f
o
r
m
an
ce
.
T
h
e
p
o
s
s
ib
le
p
er
f
o
r
m
a
n
c
e
m
ig
h
t
be
f
ar
f
r
o
m
th
e
g
lo
b
al
o
p
ti
m
u
m
p
er
f
o
r
m
a
n
ce
,
m
a
k
i
n
g
t
h
e
o
th
er
m
et
h
o
d
s
tak
e
a
lo
n
g
er
iter
atio
n
ti
m
e
to
o
b
tain
th
e
o
p
ti
m
u
m
s
o
lu
tio
n
.
FS
A
e
x
p
lo
its
t
h
e
g
lo
b
al
an
d
lo
ca
l
o
p
tim
u
m
s
o
lu
tio
n
f
o
r
f
i
n
d
i
n
g
t
h
e
o
p
ti
m
u
m
s
o
l
u
tio
n
.
In
t
h
i
s
w
o
r
k
,
th
e
s
ea
r
c
h
s
p
ac
e
is
r
ea
lized
as
th
at
p
er
s
o
n
lo
o
k
in
g
f
o
r
th
e
o
p
tim
u
m
li
v
i
n
g
in
t
h
e
w
o
r
ld
.
T
h
e
lo
ca
l
o
p
tim
u
m
s
o
lu
tio
n
b
et
w
ee
n
th
e
o
th
er
p
eo
p
le
is
co
n
s
id
er
ed
as
o
n
e
p
e
r
s
o
n
ac
h
iev
i
n
g
th
e
b
etter
p
er
f
o
r
m
an
ce
.
T
h
e
g
lo
b
al
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
S
a
r
ca
s
m
d
etec
tio
n
o
n
s
o
cia
l d
a
ta
:
h
e
u
r
is
tic
s
ea
r
ch
a
n
d
d
ee
p
lea
r
n
in
g
(
A
r
u
mu
g
h
a
m
P
a
la
n
i
a
mma
l
)
4699
o
p
tim
u
m
p
er
f
o
r
m
an
ce
b
et
w
ee
n
t
h
e
o
th
er
p
eo
p
les
is
co
n
s
id
er
ed
as
o
n
e
p
er
s
o
n
ac
co
m
p
li
s
h
i
n
g
t
h
e
b
etter
o
u
tco
m
e
in
a
co
u
n
tr
y
o
v
er
m
an
y
y
ea
r
s
.
B
ased
on
a
m
at
h
e
m
atica
l
eq
u
a
tio
n
,
FS
A
i
s
b
u
ilt
a
n
d
s
tep
s
s
ta
r
t
b
ased
on
th
e
r
an
d
o
m
s
o
lu
tio
n
.
(
,
∶
)
=
+
(
−
)
∗
(
1
,
)
(
7
)
In
(
7
)
,
(
,
:
)
i
m
p
lie
s
th
e
ℎ
s
o
lu
tio
n
,
,
an
d
p
o
r
tr
ay
s
t
h
e
s
ea
r
c
h
s
p
ac
e
’
s
u
p
p
er
an
d
lo
w
er
b
o
u
n
d
ar
ies,
a
n
d
d
en
o
tes
t
h
e
u
n
i
f
o
r
m
d
is
tr
ib
u
t
i
o
n
r
an
d
o
m
n
u
m
b
er
in
t
h
e
d
im
en
s
io
n
.
T
h
e
b
est
s
o
l
u
tio
n
is
co
n
s
id
er
ed
a
g
lo
b
al
s
o
lu
tio
n
(
),
an
d
each
s
o
lu
tio
n
is
co
n
s
id
er
ed
a
lo
ca
l
s
o
lu
tio
n
(
)
af
ter
b
u
ild
in
g
all
s
o
lu
tio
n
s
th
i
s
p
r
o
ce
s
s
is
r
ep
ea
ted
u
n
til
an
o
p
ti
m
al
o
u
tco
m
e
is
ac
h
ie
v
ed
.
In
(
8
)
,
th
e
s
ea
r
ch
p
h
ase
is
b
ased
o
n
th
e
s
u
p
p
o
r
t
o
f
th
e
ex
p
lo
itatio
n
f
ea
t
u
r
e.
I
n
(
9
)
,
th
e
s
ea
r
ch
p
r
o
ce
s
s
is
b
ased
o
n
th
e
m
ain
ten
a
n
ce
o
f
t
h
e
ex
p
l
o
itatio
n
f
ea
t
u
r
e.
E
ac
h
s
o
lu
tio
n
is
u
p
d
ated
by
(
1
0
)
,
af
ter
co
m
p
u
t
in
g
th
e
g
lo
b
al
an
d
th
e
lo
ca
l
co
n
v
er
g
en
ce
.
In
(
1
1
)
is
u
tili
ze
d
to
p
r
o
ce
s
s
u
p
g
r
ad
es
an
d
.
(
,
∶
)
=
(
(
,
∶
)
−
(
,
∶
)
∗
)
(
8
)
(
,
∶
)
=
(
−
(
,
∶
)
)
∗
(
9
)
(
,
∶
)
=
(
,
∶
)
+
(
,
∶
)
+
(
,
∶
)
(
1
0
)
(
,
∶
)
=
+
(
−
(
,
∶
)
)
∗
(
1
1
)
At
la
s
t,
t
h
e
alg
o
r
it
h
m
ch
ec
k
s
an
d
b
ec
au
s
e
of
t
h
e
u
p
g
r
ad
in
g
of
r
an
d
o
m
i
n
itial
s
an
d
u
p
d
ates
th
e
m
if
th
er
e
is
a
b
etter
s
o
lu
tio
n
t
h
a
n
an
d
.
T
h
e
iter
ativ
e
p
r
o
ce
s
s
of
FS
A
is
g
i
v
en
a
s
f
o
llo
w
s
:
‒
Step
1.
A
p
r
im
ar
y
p
o
p
u
latio
n
s
ize,
th
e
s
ea
r
ch
i
n
g
s
p
ac
e,
an
d
th
e
m
ai
n
f
u
n
c
tio
n
ar
e
r
an
d
o
m
l
y
d
e
f
i
n
ed
.
Set
th
e
an
d
lo
w
er
an
d
u
p
p
er
lim
it
s
.
M
a
x
.
Set
=
1
,
an
d
s
et
th
e
m
a
x
i
m
al
a
m
o
u
n
t
of
iter
atio
n
s
.
C
o
m
p
u
t
e
th
e
an
d
.
I
n
itialize
by
(
7
)
.
‒
Step
2.
T
h
e
s
ea
r
ch
in
all
th
e
co
u
n
tr
y
a
n
d
o
v
er
all
w
o
r
ld
is
d
ep
en
d
en
t
u
p
o
n
an
d
in
(
8
)
an
d
(
9
)
.
T
h
e
s
o
lu
tio
n
of
each
p
er
s
o
n
is
d
eter
m
in
ed
in
(
1
0
)
af
ter
co
m
p
u
ti
n
g
th
e
g
lo
b
al
an
d
th
e
lo
ca
l
co
n
v
er
g
en
ce
.
‒
Step
3.
C
o
m
p
ar
e
th
e
f
it
n
es
s
v
a
lu
e
of
all
t
h
e
p
o
s
s
ib
le
s
o
lu
t
io
n
s
to
d
ef
in
e
an
d
in
th
e
p
r
ese
n
t
g
en
er
atio
n
.
Up
d
ate
th
e
an
d
,
if
t
h
er
e
is
a
b
est
s
o
lu
t
io
n
.
‒
Step
4.
C
o
m
p
u
te
=
+
1
.
J
u
d
g
e
if
e
q
u
iv
ale
n
t
to
M
a
x
.
If
n
o
t,
go
to
s
te
p
2.
Or
else,
s
to
p
an
d
g
iv
e
th
e
r
esu
lt.
T
h
e
o
p
tim
u
m
f
itn
e
s
s
is
a
cr
itical
f
ac
to
r
in
th
e
FS
A
ap
p
r
o
ac
h
.
An
en
co
d
er
o
u
tp
u
t
is
e
m
p
lo
y
ed
f
o
r
d
ev
elo
p
in
g
an
i
m
p
r
o
v
ed
o
u
tco
m
e
f
o
r
ca
n
d
id
ate
ef
f
icac
y
.
N
o
w
,
th
e
v
al
u
e
of
ac
cu
r
ac
y
is
t
h
e
m
o
s
t
i
m
p
o
r
ta
n
t
co
n
d
itio
n
u
t
ilized
in
d
esi
g
n
in
g
an
FF.
=
ma
x
(
)
(
1
2
)
=
+
(
1
3
)
W
h
er
e,
an
d
d
ep
icts
th
e
v
al
u
e
s
of
f
al
s
e
an
d
tr
u
e
p
o
s
iti
v
e.
4.
RE
SU
L
T
S
ANA
L
YS
I
S
T
h
e
in
v
esti
g
atio
n
al
ass
es
s
m
en
t
of
th
e
FS
A
D
L
-
SD
C
SND
m
et
h
o
d
can
be
ex
p
er
im
en
ted
on
tw
o
d
atasets
[
2
2
]
–
[
2
5
]
,
as
p
o
r
tr
ay
ed
in
T
ab
le
1.
T
h
e
d
ataset
co
m
p
r
is
es
t
w
ee
ts
an
d
u
s
er
r
ep
lies
a
cc
u
m
u
lated
b
y
th
e
@
o
n
li
n
esar
ca
s
m
T
w
itter
b
o
t,
w
it
h
u
s
er
m
o
o
d
d
ata
an
d
p
r
o
ce
s
s
ed
to
eli
m
i
n
ate
U
R
L
s
an
d
r
ep
lace
m
en
t
io
n
s
,
u
tili
zi
n
g
t
w
ee
ts
a
s
co
n
ten
t
an
d
r
ep
lies
as
co
n
tex
t.
I
t
also
f
ea
t
u
r
es
n
e
w
s
h
ea
d
lin
e
s
f
r
o
m
T
h
e
On
io
n
a
n
d
Hu
f
f
P
o
s
t,
g
iv
in
g
h
ig
h
-
q
u
ali
t
y
,
f
o
r
m
al
co
n
te
n
t
w
it
h
clea
r
s
ar
ca
s
m
lab
els
.
T
h
is
d
ataset
m
iti
g
ate
s
n
o
is
e
a
n
d
s
p
ar
s
it
y
r
elate
d
to
T
w
itter
d
atasets
,
a
n
d
ev
er
y
r
ec
o
r
d
en
co
m
p
ass
e
s
attr
ib
u
te
s
f
o
r
SD
,
th
e
h
ea
d
lin
e
tex
t,
an
d
a
lin
k
to
th
e
o
r
ig
in
a
l
ar
ticle.
Deta
iled
s
tatis
t
ics a
n
d
a
h
y
b
r
id
NN
m
o
d
el
tr
ain
ed
o
n
th
is
d
atase
t a
r
e
av
ailab
le
o
n
G
itHu
b
.
T
ab
le
1.
Data
s
et
d
escr
ip
tio
n
C
l
a
ss
D
a
t
a
se
t
Tw
i
t
t
e
r
-
2
0
1
3
H
e
a
d
l
i
n
e
s
-
2
0
1
9
S
a
r
c
a
st
i
c
3
0
8
13
,
634
N
o
n
-
sarca
s
t
i
c
1
,
6
4
8
14
,
985
O
v
e
r
a
l
l
i
n
st
a
n
c
e
n
u
mb
e
r
s
1
,
9
5
6
28
,
619
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
13
,
No
.
4
,
Dec
em
b
er
2024
:
4
6
9
5
-
4
7
0
2
4700
Fig
u
r
e
2
ill
u
s
tr
ate
s
th
e
c
lass
if
i
ca
tio
n
ac
cu
r
ac
y
o
f
th
e
FS
A
D
L
-
SD
C
SN
D
m
e
th
o
d
o
lo
g
y
f
o
r
s
a
r
ca
s
tic
an
d
n
o
n
-
s
ar
ca
s
tic
t
w
ee
ts
u
s
i
n
g
th
e
T
w
itter
-
2
0
1
3
d
ataset.
Fig
u
r
e
s
2
(
a
)
an
d
2
(
b
)
p
o
r
tr
ay
ed
1
0
0
%
ac
cu
r
ac
y
i
n
d
etec
tin
g
n
o
n
-
s
ar
ca
s
t
ic
t
w
ee
t
s
an
d
co
r
r
ec
tly
p
r
ed
icted
m
o
s
t
s
ar
ca
s
tic
t
w
ee
t
s
u
n
d
er
8
0
:2
0
.
Fig
u
r
e
s
2
(
c)
an
d
2
(
d
)
d
em
o
n
s
tr
ated
h
i
g
h
ac
cu
r
ac
y
a
cr
o
s
s
b
o
th
tr
ai
n
in
g
a
n
d
te
s
ti
n
g
p
h
ases
u
n
d
er
7
0
:3
0
,
s
h
o
w
in
g
r
o
b
u
s
t p
er
f
o
r
m
a
n
ce
in
t
w
ee
t
cla
s
s
i
f
icat
io
n
.
T
h
e
v
a
lu
es
al
s
o
r
ef
er
r
ed
to
th
e
e
f
f
ec
t
iv
e
r
ec
o
g
n
itio
n
o
f
t
h
e
s
ar
ca
s
ti
c
an
d
n
o
n
-
s
ar
ca
s
ti
c
in
s
ta
n
ce
s
u
n
d
er
o
v
er
al
l c
lass
e
s
.
(
a)
(
b
)
(
c)
(
d
)
F
i
g
u
r
e
2
.
C
o
n
f
u
s
i
o
n
m
a
t
r
i
c
e
s
o
n
T
w
i
t
t
e
r
-
2
0
1
3
d
a
t
a
s
e
t
,
(
a
)
8
0
:
2
0
,
(
b
)
8
0
:
2
0
,
(
c
)
7
0
:
3
0
,
a
n
d
(
d
)
7
0
:
3
0
o
f
T
R
/
T
S
s
e
t
s
T
h
e
r
elatio
n
al
s
tu
d
y
of
th
e
FS
A
D
L
-
SD
C
SN
D
tech
n
iq
u
e
w
it
h
p
r
esen
t
ap
p
r
o
ac
h
es
is
p
r
o
v
id
ed
in
T
ab
le
2
.
B
ased
on
,
th
e
FS
A
D
L
-
SD
C
SN
D
tech
n
iq
u
e
r
ea
ch
e
s
a
h
ig
h
er
of
9
9
.
6
1
%
w
h
ile
t
h
e
Van
ill
a
C
NN,
FS,
E
L
Mo
B
iLST
M,
E
L
Mo
B
iLST
MFu
ll,
an
d
A
2
T
ex
tNet
m
o
d
els
h
a
v
e
s
h
o
w
n
lo
w
er
v
alu
e
s
of
7
2
.
6
3
%,
9
0
.
1
5
%,
7
7
.
0
6
%,
7
8
.
4
9
%,
an
d
9
3
.
3
9
%.
B
ased
o
n
,
th
e
FS
A
D
L
-
S
DC
SND
m
et
h
o
d
r
ea
ch
es
m
ax
i
m
a
l
o
f
9
9
.
6
2
%
w
h
ile
t
h
e
Van
illa
C
NN,
FS
,
E
L
Mo
B
iL
ST
M,
E
L
Mo
B
iL
ST
MFu
ll,
an
d
A
2
T
ex
tNet
m
et
h
o
d
s
ex
h
ib
ited
less
er
v
al
u
es
o
f
7
1
.
4
8
%,
8
8
.
7
4
%,
7
6
.
3
4
%,
7
8
.
0
2
%,
an
d
9
2
.
4
1
%.
T
h
en
,
b
ased
o
n
,
th
e
FS
A
D
L
-
SD
C
SND
m
e
th
o
d
g
ain
s
h
i
g
h
er
o
f
9
9
.
6
1
%
w
h
ile
t
h
e
Van
i
lla
C
N
N,
FS
,
E
L
Mo
B
iLST
M,
E
L
Mo
B
iL
ST
MFu
ll,
a
n
d
A
2
T
ex
t
-
Ne
t
m
et
h
o
d
s
s
h
o
w
ed
m
i
n
i
m
al
v
alu
e
s
o
f
6
7
.
7
6
%,
8
8
.
6
3
%,
7
5
.
7
7
%,
7
4
.
2
2
%,
an
d
9
1
.
4
7
%.
Nex
t,
b
ased
o
n
,
th
e
FS
A
D
L
-
SD
C
SN
D
m
et
h
o
d
attain
s
s
u
p
er
io
r
o
f
9
9
.
6
1
%
w
h
ile
t
h
e
Va
n
illa
C
NN,
F
S,
E
L
Mo
B
iL
ST
M,
E
L
Mo
B
iLST
MFu
ll,
a
n
d
A
2
T
e
x
tNet
m
o
d
el
s
d
ep
icted
m
in
i
m
a
l
v
alu
es
o
f
6
8
.
8
8
%,
8
8
.
5
4
%,
7
6
.
1
3
%,
7
5
.
9
9
%,
an
d
9
0
.
5
3
%.
T
h
ese
o
u
tp
u
ts
s
h
o
w
ed
th
e
s
u
p
r
e
m
ac
y
of
t
h
e
FS
AD
L
-
SD
C
SND
m
et
h
o
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
A
r
ti
f
I
n
tell
I
SS
N:
2252
-
8938
S
a
r
ca
s
m
d
etec
tio
n
o
n
s
o
cia
l d
a
ta
:
h
e
u
r
is
tic
s
ea
r
ch
a
n
d
d
ee
p
lea
r
n
in
g
(
A
r
u
mu
g
h
a
m
P
a
la
n
i
a
mma
l
)
4701
T
ab
le
2
.
R
elatio
n
al
o
u
tp
u
t
of
FS
A
D
L
-
SD
C
SN
D
tech
n
iq
u
e
w
it
h
e
x
is
t
in
g
ap
p
r
o
ac
h
es
T
e
c
h
n
i
q
u
e
s
F
S
A
D
L
-
S
D
C
S
N
D
9
9
.
6
1
9
9
.
6
2
9
9
.
6
1
9
9
.
6
1
V
a
n
i
l
l
a
C
N
N
7
2
.
6
3
7
1
.
4
8
6
7
.
7
6
6
8
.
8
8
F
r
a
c
k
i
n
g
S
a
r
c
a
sm
9
0
.
1
5
8
8
.
7
4
8
8
.
6
3
8
8
.
5
4
EL
M
o
B
i
L
S
T
M
7
7
.
0
6
7
6
.
3
4
7
5
.
7
7
7
6
.
1
3
EL
M
o
B
i
L
S
T
M
F
U
LL
7
8
.
4
9
7
8
.
0
2
7
4
.
2
2
7
5
.
9
9
A
2
Te
x
t
N
e
t
9
3
.
3
9
9
2
.
4
1
9
1
.
4
7
9
0
.
5
3
5.
CO
NCLU
SI
O
N
In
th
i
s
m
a
n
u
s
cr
ip
t,
a
n
o
v
el
F
S
A
D
L
-
S
DC
SND
m
e
th
o
d
o
lo
g
y
w
a
s
p
r
esen
ted
f
o
r
au
to
m
ated
an
d
p
r
ec
is
e
SD
an
d
s
o
cial
n
et
w
o
r
k
i
n
g
d
ata
class
i
f
icatio
n
.
T
h
e
m
aj
o
r
in
ten
t
of
th
e
F
S
A
D
L
-
SD
C
SND
m
eth
o
d
o
lo
g
y
is
in
th
e
au
to
m
ated
a
n
d
ef
f
ic
ien
t
r
ec
o
g
n
itio
n
of
s
ar
ca
s
tic
tex
t.
In
t
h
e
p
r
o
p
o
s
ed
FS
A
D
L
-
S
DC
S
ND
ap
p
r
o
ac
h
,
th
r
ee
p
h
a
s
e
s
ar
e
e
m
p
lo
y
ed
n
a
m
e
l
y
d
ata
p
r
ep
r
o
ce
s
s
in
g
,
BS
-
P
L
ST
M
-
b
ased
class
i
f
ica
tio
n
,
a
n
d
FS
A
-
b
ase
d
tu
n
in
g
.
Her
e,
t
h
e
BS
-
P
L
ST
M
m
o
d
el
h
as
b
ee
n
ex
p
lo
ited
f
o
r
d
etec
tin
g
an
d
cla
s
s
i
f
y
in
g
t
h
e
s
ar
ca
s
m
.
T
h
e
t
u
n
i
n
g
p
r
o
ce
s
s
is
ac
h
iev
e
d
by
e
m
p
lo
y
i
n
g
th
e
FS
A
to
i
m
p
r
o
v
e
th
e
d
etec
tio
n
p
er
f
o
r
m
a
n
c
e
of
th
e
BS
-
PL
ST
M
m
et
h
o
d
o
lo
g
y
.
To
ex
h
ib
it
th
e
g
r
ea
ter
s
o
lu
tio
n
of
th
e
FS
A
D
L
-
SD
C
SN
D
s
y
s
te
m
,
a
s
eq
u
e
n
ce
of
s
i
m
u
lat
io
n
s
ca
n
be
ap
p
l
ied
.
T
h
e
ex
ten
s
i
v
e
r
esu
lt
s
h
i
g
h
li
g
h
ted
th
e
g
r
ea
ter
s
o
lu
tio
n
of
t
h
e
FS
A
D
L
-
SD
C
SND
m
eth
o
d
o
lo
g
y
w
it
h
o
th
er
co
m
p
ar
ed
m
eth
o
d
s
in
ter
m
s
of
d
if
f
er
e
n
t
p
er
f
o
r
m
an
ce
m
ea
s
u
r
es.
T
h
e
FS
ADL
-
SDC
S
ND
ap
p
r
o
ac
h
m
a
y
s
tr
u
g
g
le
w
it
h
n
u
a
n
ce
d
,
co
n
tex
t
-
d
ep
en
d
en
t
s
ar
ca
s
m
,
p
r
o
m
p
ti
n
g
f
u
tu
r
e
r
esear
ch
a
v
en
u
e
s
f
o
r
en
h
an
ci
n
g
ac
cu
r
ac
y
by
in
co
r
p
o
r
atin
g
co
n
tex
t
u
al
i
n
f
o
r
m
atio
n
a
n
d
u
s
er
-
s
p
ec
if
ic
n
u
an
ce
s
in
o
n
li
n
e
c
o
m
m
u
n
icatio
n
.
RE
F
E
R
E
NC
E
S
[
1
]
Y.
Y.
T
a
n
,
C.
O.
C
h
o
w
,
J.
K
a
n
e
san
,
J.
H.
C
h
u
a
h
,
a
n
d
Y.
L.
L
i
m,
“
S
e
n
t
i
me
n
t
a
n
a
l
y
si
s
a
n
d
s
a
r
c
a
sm
d
e
t
e
c
t
i
o
n
u
s
i
n
g
d
e
e
p
mu
l
t
i
-
t
a
s
k
l
e
a
r
n
i
n
g
,
”
W
i
re
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
2
9
,
n
o
.
3,
p
p
.
2
2
1
3
–
2
2
3
7
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
2
7
7
-
023
-
1
0
2
3
5
-
4.
[
2
]
P.
G
o
e
l
,
R.
Ja
i
n
,
A.
N
a
y
y
a
r
,
S.
S
i
n
g
h
a
l
,
a
n
d
M.
S
r
i
v
a
st
a
v
a
,
“
S
a
r
c
a
sm
d
e
t
e
c
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
e
n
s
e
mb
l
e
l
e
a
r
n
i
n
g
,
”
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
.
81,
n
o
.
3
0
,
p
p
.
4
3
2
2
9
–
4
3
2
5
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
4
2
-
0
2
2
-
1
2
9
3
0
-
z.
[
3
]
A.
K
u
mar,
S.
R.
S
a
n
g
w
a
n
,
A.
K.
S
i
n
g
h
,
a
n
d
G.
W
a
d
h
w
a
,
“
H
y
b
r
i
d
d
e
e
p
l
e
a
r
n
i
n
g
mo
d
e
l
f
o
r
sarca
sm
d
e
t
e
c
t
i
o
n
in
i
n
d
i
a
n
i
n
d
i
g
e
n
o
u
s
l
a
n
g
u
a
g
e
u
si
n
g
w
o
r
d
-
e
mo
j
i
e
mb
e
d
d
i
n
g
s,”
A
C
M
T
ra
n
s
a
c
t
i
o
n
s
on
As
i
a
n
and
L
o
w
-
R
e
so
u
rc
e
L
a
n
g
u
a
g
e
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ssi
n
g
,
v
o
l
.
2
2
,
n
o
.
5,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
4
5
/
3
5
1
9
2
9
9
.
[
4
]
S.
K.
B
h
a
r
t
i
,
R.
K.
G
u
p
t
a
,
P.
K.
S
h
u
k
l
a
,
W.
A.
H
a
t
a
ml
e
h
,
H.
T
a
r
a
z
i
,
a
n
d
S.
J.
N
u
a
g
a
h
,
“
M
u
l
t
i
mo
d
a
l
sar
c
a
sm
d
e
t
e
c
t
i
o
n
:
a
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
,
”
W
i
re
l
e
ss
C
o
m
m
u
n
i
c
a
t
i
o
n
s
a
n
d
Mo
b
i
l
e
C
o
m
p
u
t
i
n
g
,
v
o
l
.
2
0
2
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
2
/
1
6
5
3
6
9
6
.
[
5
]
H.
A
h
ma
d
,
W.
A
k
b
a
r
,
N.
A
sl
a
m,
A.
A
h
me
d
,
a
n
d
M.
K
h
u
r
sh
i
d
,
“
TF
-
I
D
F
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
-
b
a
se
d
sarc
a
sm
d
e
t
e
c
t
i
o
n
on
so
c
i
a
l
me
d
i
a
,
”
J
o
u
r
n
a
l
of
C
o
m
p
u
t
i
n
g
\
&
Bi
o
m
e
d
i
c
a
l
I
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
5,
n
o
.
1,
p
p
.
1
1
8
–
1
2
9
,
2
0
2
3
.
[
6
]
X.
Z
h
a
o
,
J.
H
u
a
n
g
,
a
n
d
H.
Y
a
n
g
,
“
C
A
N
s
:
c
o
u
p
l
e
d
-
a
t
t
e
n
t
i
o
n
n
e
t
w
o
r
k
s
f
o
r
sarca
sm
d
e
t
e
c
t
i
o
n
on
so
c
i
a
l
me
d
i
a
,
”
Pro
c
e
e
d
i
n
g
s
of
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
r
e
n
c
e
on
N
e
u
ra
l
N
e
t
w
o
rks
,
v
o
l
.
2
0
2
1
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
JC
N
N
5
2
3
8
7
.
2
0
2
1
.
9
5
3
3
8
0
0
.
[
7
]
P.
D
u
t
t
a
a
n
d
C.
K.
B
h
a
t
t
a
c
h
a
r
y
y
a
,
“
M
u
l
t
i
-
mo
d
a
l
s
a
r
c
a
s
m
d
e
t
e
c
t
i
o
n
in
so
c
i
a
l
n
e
t
w
o
r
k
s:
a
c
o
m
p
a
r
a
t
i
v
e
r
e
v
i
e
w
,
”
Pro
c
e
e
d
i
n
g
s
-
6
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
on
C
o
m
p
u
t
i
n
g
M
e
t
h
o
d
o
l
o
g
i
e
s
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
,
I
C
C
M
C
2
0
2
2
,
p
p
.
207
–
2
1
4
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
M
C
5
3
4
7
0
.
2
0
2
2
.
9
7
5
3
9
8
1
.
[
8
]
D.
K.
N
a
y
a
k
a
n
d
B.
K.
B
o
l
l
a
,
“
Ef
f
i
c
i
e
n
t
d
e
e
p
l
e
a
r
n
i
n
g
me
t
h
o
d
s
f
o
r
sarc
a
sm
d
e
t
e
c
t
i
o
n
of
n
e
w
s
h
e
a
d
l
i
n
e
s,”
S
m
a
r
t
I
n
n
o
v
a
t
i
o
n
,
S
y
s
t
e
m
s
a
n
d
T
e
c
h
n
o
l
o
g
i
e
s
,
v
o
l
.
2
6
9
,
p
p
.
371
–
3
8
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
16
-
7
9
9
6
-
4
_
2
6
.
[
9
]
R.
K
a
n
a
k
a
m
a
n
d
R.
K.
N
a
y
a
k
,
“
S
a
r
c
a
sm
d
e
t
e
c
t
i
o
n
on
so
c
i
a
l
n
e
t
w
o
r
k
s
u
si
n
g
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms:
a
sy
st
e
mat
i
c
r
e
v
i
e
w
,
”
Pro
c
e
e
d
i
n
g
s
of
t
h
e
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
on
T
r
e
n
d
s
in
El
e
c
t
ro
n
i
c
s
a
n
d
I
n
f
o
rm
a
t
i
c
s
,
I
C
O
EI
2
0
2
1
,
p
p
.
1
1
3
0
–
1
1
3
7
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
O
E
I
5
1
2
4
2
.
2
0
2
1
.
9
4
5
2
9
5
4
.
[
1
0
]
D.
K.
S
h
a
r
ma,
B.
S
i
n
g
h
,
S.
A
g
a
r
w
a
l
,
H.
K
i
m,
a
n
d
R.
S
h
a
r
m
a
,
“
S
a
r
c
a
sm
d
e
t
e
c
t
i
o
n
o
v
e
r
so
c
i
a
l
me
d
i
a
p
l
a
t
f
o
r
ms
u
si
n
g
h
y
b
r
i
d
a
u
t
o
-
e
n
c
o
d
e
r
-
b
a
se
d
mo
d
e
l
,
”
E
l
e
c
t
r
o
n
i
c
s
,
v
o
l
.
11,
n
o
.
1
8
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s1
1
1
8
2
8
4
4
.
[
1
1
]
A.
S
i
v
a
l
i
n
g
a
m,
K.
S
u
n
d
a
r
a
r
a
j
a
n
,
a
n
d
A.
P
a
l
a
n
i
sa
my
,
“
C
R
F
-
M
EM
:
c
o
n
d
i
t
i
o
n
a
l
r
a
n
d
o
m
f
i
e
l
d
mo
d
e
l
b
a
se
d
mo
d
i
f
i
e
d
e
x
p
e
c
t
a
t
i
o
n
max
i
mi
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m
f
o
r
sarc
a
sm
d
e
t
e
c
t
i
o
n
in
so
c
i
a
l
me
d
i
a
,
”
J
o
u
r
n
a
l
of
I
n
t
e
rn
e
t
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
2
4
,
n
o
.
1,
p
p
.
45
–
5
4
,
2
0
2
3
,
d
o
i
:
1
0
.
5
3
1
0
6
/
1
6
0
7
9
2
6
4
2
0
2
3
0
1
2
4
0
1
0
0
5
.
[
1
2
]
R.
P
a
n
d
e
y
a
n
d
J.
P.
S
i
n
g
h
,
“
B
E
R
T
-
L
S
T
M
mo
d
e
l
f
o
r
sarc
a
sm
d
e
t
e
c
t
i
o
n
in
c
o
d
e
-
mi
x
e
d
so
c
i
a
l
me
d
i
a
p
o
st
,
”
J
o
u
r
n
a
l
of
I
n
t
e
l
l
i
g
e
n
t
I
n
f
o
rm
a
t
i
o
n
S
y
st
e
m
s
,
v
o
l
.
6
0
,
n
o
.
1,
p
p
.
2
3
5
–
2
5
4
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
8
4
4
-
0
2
2
-
0
0
7
5
5
-
z.
[
1
3
]
A.
K
u
mar,
V.
T.
N
a
r
a
p
a
r
e
d
d
y
,
V.
A.
S
r
i
k
a
n
t
h
,
A.
M
a
l
a
p
a
t
i
,
a
n
d
L.
B.
M.
N
e
t
i
,
“
S
a
r
c
a
sm
d
e
t
e
c
t
i
o
n
u
s
i
n
g
m
u
l
t
i
-
h
e
a
d
a
t
t
e
n
t
i
o
n
b
a
se
d
b
i
d
i
r
e
c
t
i
o
n
a
l
L
S
T
M
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
8,
p
p
.
6
3
8
8
–
6
3
9
7
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
9
.
2
9
6
3
6
3
0
.
[
1
4
]
A.
P
a
l
a
n
i
a
m
mal
a
n
d
P.
A
n
a
n
d
a
b
a
b
u
,
“
C
h
a
o
s
s
i
n
e
c
o
s
i
n
e
a
l
g
o
r
i
t
h
m
w
i
t
h
g
r
a
p
h
c
o
n
v
o
l
u
t
i
o
n
n
e
t
w
o
r
k
f
o
r
sarca
sm
d
e
t
e
c
t
i
o
n
in
so
c
i
a
l
me
d
i
a
,
”
7
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
on
T
re
n
d
s
in
El
e
c
t
r
o
n
i
c
s
a
n
d
I
n
f
o
rm
a
t
i
c
s,
I
C
O
EI
2
0
2
3
,
p
p
.
96
–
1
0
2
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
O
E
I
5
6
7
6
5
.
2
0
2
3
.
1
0
1
2
6
0
5
2
.
[
1
5
]
D.
V
i
n
o
t
h
a
n
d
P.
P
r
a
b
h
a
v
a
t
h
y
,
“
A
n
i
n
t
e
l
l
i
g
e
n
t
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
-
b
a
se
d
sarca
sm
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
mo
d
e
l
on
so
c
i
a
l
n
e
t
w
o
r
k
s,”
J
o
u
r
n
a
l
of
S
u
p
e
rc
o
m
p
u
t
i
n
g
,
v
o
l
.
7
8
,
n
o
.
8,
p
p
.
1
0
5
7
5
–
1
0
5
9
4
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
2
2
7
-
022
-
0
4
3
1
2
-
x.
[
1
6
]
D.
V
i
n
o
t
h
a
n
d
P.
P
r
a
b
h
a
v
a
t
h
y
,
“
A
u
t
o
mat
e
d
sarc
a
sm
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
ssi
f
i
c
a
t
i
o
n
u
si
n
g
h
y
p
e
r
p
a
r
a
me
t
e
r
t
u
n
e
d
d
e
e
p
l
e
a
r
n
i
n
g
mo
d
e
l
f
o
r
so
c
i
a
l
n
e
t
w
o
r
k
s,”
E
x
p
e
rt
S
y
st
e
m
s
,
v
o
l
.
39,
n
o
.
1
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
1
1
/
e
x
sy
.
1
3
1
0
7
.
[
1
7
]
D.
M.
A
sh
o
k
,
A.
N
.
G
h
a
n
s
h
y
a
m,
S.
S.
S
a
l
i
m,
D.
B
.
M
a
z
a
h
i
r
,
a
n
d
B.
S.
T
h
a
k
a
r
e
,
“
S
a
r
c
a
sm
d
e
t
e
c
t
i
o
n
u
si
n
g
g
e
n
e
t
i
c
o
p
t
i
m
i
z
a
t
i
o
n
on
L
S
T
M
w
i
t
h
C
N
N
,
”
2
0
2
0
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
f
o
r
Em
e
rg
i
n
g
T
e
c
h
n
o
l
o
g
y
(
I
N
C
ET)
,
Be
l
g
a
u
m
,
I
n
d
i
a
,
2
0
2
0
,
p
p
.
1
-
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
N
C
ET
4
9
8
4
8
.
2
0
2
0
.
9
1
5
4
0
9
0
.
[
1
8
]
G.
A.
S
a
h
u
a
n
d
M.
H
u
d
n
u
r
k
a
r
,
“
M
e
t
a
h
e
u
r
i
s
t
i
c
-
a
ssi
s
t
e
d
d
e
e
p
e
n
se
mb
l
e
t
e
c
h
n
i
q
u
e
f
o
r
i
d
e
n
t
i
f
y
i
n
g
s
a
r
c
a
sm
f
r
o
m
so
c
i
a
l
me
d
i
a
d
a
t
a
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
of
W
i
re
l
e
ss
and
Mo
b
i
l
e
C
o
m
p
u
t
i
n
g
,
v
o
l
.
2
6
,
n
o
.
1,
p
p
.
25
–
3
8
,
2
0
2
4
,
d
o
i
:
1
0
.
1
5
0
4
/
i
j
w
mc.
2
0
2
4
.
1
0
0
6
2
1
2
2
.
[
1
9
]
C.
M
a
g
o
o
a
n
d
M.
S
i
n
g
h
,
“A
n
o
v
e
l
h
y
b
r
i
d
a
p
p
r
o
a
c
h
f
o
r
i
n
t
e
n
t
c
r
e
a
t
i
o
n
a
n
d
d
e
t
e
c
t
i
o
n
u
si
n
g
k
-
me
a
n
s
-
b
a
se
d
t
o
p
i
c
c
l
u
s
t
e
r
i
n
g
a
n
d
h
e
u
r
i
s
t
i
c
-
b
a
se
d
c
a
p
s
u
l
e
n
e
t
w
o
r
k
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
of
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
a
n
d
D
e
c
i
si
o
n
M
a
k
i
n
g
,
v
o
l
.
2
2
,
n
o
.
6,
p
p
.
1
9
2
3
–
1
9
6
0
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
4
2
/
S
0
2
1
9
6
2
2
0
2
2
5
0
0
9
2
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
I
n
t J
A
r
ti
f
I
n
tell
,
Vo
l.
13
,
No
.
4
,
Dec
em
b
er
2024
:
4
6
9
5
-
4
7
0
2
4702
[
2
0
]
Z.
L
i
u
,
D.
Z
h
a
n
g
,
W.
Ji
a
,
X.
L
i
n
,
a
n
d
H.
L
i
u
,
“
A
n
a
d
v
e
r
sari
a
l
b
i
d
i
r
e
c
t
i
o
n
a
l
s
e
r
i
a
l
–
p
a
r
a
l
l
e
l
L
S
T
M
-
b
a
se
d
Q
T
D
f
r
a
m
e
w
o
r
k
f
o
r
p
r
o
d
u
c
t
q
u
a
l
i
t
y
p
r
e
d
i
c
t
i
o
n
,
”
J
o
u
rn
a
l
of
I
n
t
e
l
l
i
g
e
n
t
M
a
n
u
f
a
c
t
u
r
i
n
g
,
v
o
l
.
3
1
,
n
o
.
6,
p
p
.
1
5
1
1
–
1
5
2
9
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
8
4
5
-
0
1
9
-
0
1
5
3
0
-
8.
[
2
1
]
Y.
F
a
n
,
S.
Z
h
a
n
g
,
H.
Y
a
n
g
,
D.
X
u
,
a
n
d
Y.
W
a
n
g
,
“
A
n
i
m
p
r
o
v
e
d
f
u
t
u
r
e
se
a
r
c
h
a
l
g
o
r
i
t
h
m
b
a
se
d
on
t
h
e
s
i
n
e
c
o
si
n
e
a
l
g
o
r
i
t
h
m
f
o
r
f
u
n
c
t
i
o
n
o
p
t
i
mi
z
a
t
i
o
n
p
r
o
b
l
e
ms,”
I
E
E
E
Ac
c
e
ss
,
v
o
l
.
1
1
,
p
p
.
3
0
1
7
1
–
3
0
1
8
7
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
3
.
3
2
5
8
9
7
0
.
[
2
2
]
I
.
L
eo
,
“
S
a
r
c
a
sm
i
n
T
w
i
t
t
e
r
:
d
e
t
e
c
t
i
n
g
sar
c
a
sm
i
n
t
w
e
e
t
s
,
”
K
a
g
g
l
e
,
2
0
1
9
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
w
w
w
.
k
a
g
g
l
e
.
c
o
m/
c
o
mp
e
t
i
t
i
o
n
s
/
g
se
0
0
2
/
d
a
t
a
[
2
3
]
E.
R
i
l
o
f
f
,
A.
Q
a
d
i
r
,
P.
S
u
r
v
e
,
L.
D
.
S
i
l
v
a
,
N.
G
i
l
b
e
r
t
,
a
n
d
R.
H
u
a
n
g
,
“
S
a
r
c
a
s
m
as
c
o
n
t
r
a
s
t
b
e
t
w
e
e
n
a
p
o
si
t
i
v
e
se
n
t
i
me
n
t
a
n
d
n
e
g
a
t
i
v
e
si
t
u
a
t
i
o
n
,
”
EM
N
L
P
2
0
1
3
-
2
0
1
3
C
o
n
f
e
ren
c
e
on
Em
p
i
r
i
c
a
l
M
e
t
h
o
d
s
in
N
a
t
u
ra
l
L
a
n
g
u
a
g
e
Pr
o
c
e
s
si
n
g
,
Pr
o
c
e
e
d
i
n
g
s
of
t
h
e
C
o
n
f
e
re
n
c
e
,
p
p
.
7
0
4
–
7
1
4
,
2
0
1
3
.
[
2
4
]
R.
M
i
sr
a
,
“
N
e
w
s
h
e
a
d
l
i
n
e
s
d
a
t
a
se
t
f
o
r
sar
c
a
sm
d
e
t
e
c
t
i
o
n
,
”
K
a
g
g
l
e
,
2
0
1
9
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
w
w
w
.
k
a
g
g
l
e
.
c
o
m/
d
a
t
a
se
t
s/
r
m
i
s
r
a
/
n
e
w
s
-
h
e
a
d
l
i
n
e
s
-
d
a
t
a
se
t
-
f
o
r
-
sarca
s
m
-
d
e
t
e
c
t
i
o
n
[
2
5
]
R.
M
i
sr
a
a
n
d
P.
A
r
o
r
a
,
“
S
a
r
c
a
sm
d
e
t
e
c
t
i
o
n
u
s
i
n
g
h
y
b
r
i
d
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
R
e
se
a
rc
h
G
a
t
e
,
p
p
.
1
-
5
,
2
0
1
9
,
d
o
i
:
1
0
.
1
3
1
4
0
/
R
G
.
2
.
2
.
3
2
4
2
7
.
3
9
2
0
4
.
B
I
O
G
RAP
H
I
E
S
OF
AUTH
O
RS
Ar
u
m
u
g
h
a
m
P
a
la
n
ia
m
m
a
l
o
b
tain
e
d
h
e
r
b
a
c
h
e
l
o
r'
s
d
e
g
re
e
f
ro
m
Bh
a
ra
th
id
a
sa
n
Un
iv
e
rsit
y
a
n
d
h
e
r
m
a
ste
r'
s
d
e
g
r
e
e
M
.
P
h
il
.
f
ro
m
T
h
iru
v
a
ll
u
v
a
r
Un
iv
e
rsit
y
,
V
e
ll
o
re
,
b
e
f
o
re
p
u
rsu
i
n
g
h
e
r
P
h
.
D.
at
T
h
e
A
n
n
a
m
a
lai
Un
iv
e
rsit
y
in
Ch
it
h
a
m
b
a
ra
m
.
P
re
se
n
tl
y
,
sh
e
se
rv
e
s
as
a
c
a
d
e
m
ic
st
a
ff
at
BC
A
P
a
rv
a
th
y
'
s
A
rt'
s
a
n
d
S
c
ien
c
e
Co
ll
e
g
e
-
Din
d
ig
u
l.
He
r
re
se
a
rc
h
f
o
c
u
se
s
on
b
ig
d
a
ta
a
n
d
se
n
ti
m
e
n
tal
a
n
a
ly
sis,
w
it
h
a
p
a
rti
c
u
lar
i
n
tere
st
in
T
w
it
t
e
r
d
a
ta
c
re
d
ib
il
it
y
a
n
a
ly
sis
to
e
n
h
a
n
c
e
in
f
o
rm
a
ti
o
n
su
rv
e
il
l
a
n
c
e
on
so
c
ial
m
e
d
ia.
S
h
e
c
a
n
be
c
o
n
tac
ted
at
e
m
a
il
:
re
v
a
th
y
2
0
k
@g
m
a
il
.
c
o
m
.
Pu
r
u
sho
t
h
a
m
a
n
Ana
n
d
a
b
a
b
u
h
o
l
d
s
a
M
a
ste
r
of
C
o
m
p
u
ter
A
p
p
li
c
a
ti
o
n
s
a
n
d
a
P
h
.
D.
f
ro
m
A
n
n
a
m
a
lai
Un
iv
e
rsity
,
In
d
ia.
P
re
se
n
tl
y
,
he
se
rv
e
s
as
an
A
ss
istan
t
P
ro
f
e
ss
o
r
in
t
h
e
De
p
a
rtme
n
t
of
Co
m
p
u
ter
a
n
d
I
n
f
o
r
m
a
ti
o
n
S
c
ien
c
e
at
A
n
n
a
m
a
lai
Un
iv
e
rsity
.
W
it
h
n
o
ta
b
le
e
x
p
e
rti
se
in
d
a
ta
m
in
in
g
a
n
d
im
a
g
e
p
ro
c
e
ss
in
g
,
he
h
a
s
m
a
d
e
si
g
n
ifi
c
a
n
t
c
o
n
tri
b
u
ti
o
n
s
in
th
e
se
f
ield
s.
He
can
be
c
o
n
tac
ted
a
t
e
m
a
il
:
d
rp
a
b
c
u
d
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.