I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
2
,
A
p
r
il
20
25
,
p
p
.
2
2
0
2
~
2
2
1
0
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
2
.
pp
2
2
0
2
-
2
2
1
0
2202
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
4
H
AN:
hy
perg
ra
ph
-
ba
sed hi
erarc
hica
l atten
tion ne
twork for
fake news
predi
ct
io
n
Alpa
na
A.
B
o
rse
1
,
G
a
j
a
na
n.
K
.
K
ha
ra
t
e
2
,
Na
m
r
a
t
a
G
.
K
ha
ra
t
e
3
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
En
g
i
n
e
e
r
i
n
g
,
M
a
t
o
sh
r
i
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
e
r
i
n
g
a
n
d
R
e
sea
r
c
h
C
e
n
t
e
r
,
P
u
n
e
U
n
i
v
e
r
si
t
y
,
M
a
h
a
r
a
s
h
t
r
a
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
M
a
t
o
s
h
r
i
C
o
l
l
e
g
e
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
R
e
se
a
r
c
h
C
e
n
t
e
r
,
P
u
n
e
U
n
i
v
e
r
si
t
y
,
M
a
h
a
r
a
s
h
t
r
a
,
I
n
d
i
a
3
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
En
g
i
n
e
e
r
i
n
g
,
V
i
sh
w
a
k
a
ma
I
n
st
i
t
u
t
e
o
f
I
n
f
o
r
ma
t
i
o
n
Te
c
h
n
o
l
o
g
y
,
P
u
n
e
U
n
i
v
e
r
si
t
y
,
M
a
h
a
r
a
sh
t
r
a
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 3
,
2
0
2
4
R
ev
is
ed
No
v
2
9
,
2
0
2
4
Acc
ep
ted
Dec
1
4
,
2
0
2
4
F
a
k
e
n
e
ws
p
re
se
n
ts
si
g
n
ifi
c
a
n
t
th
re
a
ts
to
b
o
t
h
s
o
c
iety
a
n
d
i
n
d
i
v
id
u
a
ls,
h
ig
h
li
g
h
ti
n
g
th
e
u
r
g
e
n
t
n
e
e
d
f
o
r
i
m
p
ro
v
e
d
n
e
ws
a
u
t
h
e
n
t
icity
v
e
rifi
c
a
ti
o
n
.
T
o
d
e
a
l
with
t
h
is
c
h
a
ll
e
n
g
e
,
we
p
r
o
v
i
d
e
a
n
o
v
e
l
stra
teg
y
c
a
ll
e
d
t
h
e
4
-
lev
e
l
h
iera
rc
h
ica
l
a
tt
e
n
ti
o
n
n
e
two
r
k
(
4
HA
N),
d
e
sig
n
e
d
to
e
n
h
a
n
c
e
f
a
k
e
n
e
ws
d
e
tec
ti
o
n
th
r
o
u
g
h
a
n
a
d
v
a
n
c
e
d
i
n
teg
ra
ti
o
n
o
f
h
y
p
e
rg
ra
p
h
c
o
n
v
o
l
u
ti
o
n
a
n
d
a
tt
e
n
ti
o
n
n
e
u
ra
l
n
e
two
r
k
m
e
c
h
a
n
ism
s.
Th
e
4
HA
N
m
o
d
e
l
o
p
e
ra
tes
a
c
ro
ss
fo
u
r
h
iera
rc
h
ica
l
lev
e
ls:
p
a
ra
g
ra
p
h
s,
se
n
ten
c
e
s,
wo
rd
s,
a
n
d
c
o
n
tex
tu
a
l
in
fo
rm
a
ti
o
n
(m
e
tad
a
ta).
At
t
h
e
h
i
g
h
e
st
le
v
e
l,
t
h
e
m
o
d
e
l
e
m
p
l
o
y
s h
y
p
e
rg
ra
p
h
-
b
a
se
d
a
tt
e
n
ti
o
n
a
n
d
c
o
n
v
o
lu
ti
o
n
n
e
u
ra
l
n
e
two
r
k
s
to
c
re
a
te
a
c
o
n
tex
t
u
a
l
in
fo
rm
a
ti
o
n
v
e
c
to
r,
u
ti
li
z
in
g
a
S
o
ftM
a
x
a
c
ti
v
a
ti
o
n
f
u
n
c
ti
o
n
.
T
h
is
v
e
c
to
r
is
th
e
n
c
o
m
b
in
e
d
wit
h
a
n
e
ws
c
o
n
ten
t
v
e
c
to
r
g
e
n
e
ra
ted
th
r
o
u
g
h
wo
rd
a
n
d
se
n
ten
c
e
-
lev
e
l
a
tt
e
n
ti
o
n
m
e
c
h
a
n
ism
s.
Th
is
a
rc
h
it
e
c
tu
re
e
n
a
b
les
th
e
4
HA
N
m
o
d
e
l
t
o
e
ffe
c
ti
v
e
ly
p
ri
o
rit
ize
t
h
e
re
lev
a
n
c
e
o
f
s
p
e
c
ifi
c
wo
rd
s
a
n
d
c
o
n
tex
t
u
a
l
in
fo
rm
a
ti
o
n
,
th
e
re
b
y
imp
ro
v
in
g
t
h
e
o
v
e
ra
ll
re
p
re
se
n
tati
o
n
a
n
d
a
c
c
u
ra
c
y
o
f
n
e
ws
c
o
n
ten
t
.
We
e
v
a
lu
a
te
t
h
e
4
HA
N
m
o
d
e
l
u
si
n
g
th
e
LIAR
d
a
tas
e
t
to
d
e
m
o
n
stra
te
it
s
e
ffica
c
y
in
e
n
h
a
n
c
in
g
fa
k
e
n
e
ws
p
re
d
ictio
n
a
c
c
u
ra
c
y
.
Co
m
p
a
ra
ti
v
e
a
n
a
l
y
sis
sh
o
ws
th
a
t
th
e
4
HA
N
m
o
d
e
l
o
u
t
p
e
rfo
rm
s
s
e
v
e
ra
l
o
f
c
u
tt
in
g
-
e
d
g
e
tec
h
n
iq
u
e
s,
li
k
e
re
c
u
rre
n
t
n
e
u
ra
l
n
e
tw
o
rk
s
(RNN
),
e
n
se
m
b
le
tec
h
n
iq
u
e
s,
a
n
d
a
tt
e
n
ti
o
n
m
e
c
h
a
n
ism
s
tec
h
n
iq
u
e
s.
O
u
r
re
su
l
ts
in
d
ica
te
4
HA
N
m
o
d
e
l
a
c
c
o
m
p
l
ish
e
s
a
n
o
tab
le
a
c
c
u
ra
c
y
o
f
9
6
%
,
sh
o
wc
a
sin
g
it
s
p
o
ten
ti
a
l
fo
r
sig
n
ifi
c
a
n
tl
y
a
d
v
a
n
c
i
n
g
fa
k
e
n
e
ws
p
re
d
icti
o
n
.
K
ey
w
o
r
d
s
:
Atten
tio
n
C
las
s
if
icatio
n
Fak
e
n
ews
Hy
p
er
g
r
a
p
h
Pre
d
ictio
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
:
Alp
an
a
Ar
u
n
B
o
r
s
e
D
e
p
a
r
tm
e
n
t
o
f
C
o
m
p
u
t
er
E
n
g
i
n
e
e
r
in
g
,
M
a
to
s
h
r
i
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
er
i
n
g
an
d
R
e
s
e
a
r
ch
C
e
n
t
e
r
,
P
u
n
e
U
n
i
v
er
s
i
t
y
E
k
lah
ar
e
R
o
ad
,
A
u
r
an
g
ab
ad
H
ig
h
way
,
Nasik
Dis
tr
ict,
Ma
h
ar
ash
tr
a
4
2
2
1
0
,
I
n
d
ia
E
m
ail:
alp
an
a.
b
o
r
s
e@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
I
n
th
e
ev
er
y
d
ay
wo
r
l
d
,
n
ews
is
ess
en
tial
b
y
k
ee
p
in
g
th
e
p
u
b
lic
in
f
o
r
m
e
d
a
b
o
u
t
cu
r
r
e
n
t
ev
en
ts
,
h
elp
in
g
p
e
o
p
le
m
ak
e
in
f
o
r
m
e
d
d
ec
is
io
n
s
,
an
d
s
h
ap
in
g
s
o
ci
etal
o
p
in
io
n
s
.
I
t
is
ess
en
tial
f
o
r
u
n
d
er
s
tan
d
i
n
g
th
e
wo
r
ld
an
d
p
a
r
ticip
atin
g
ef
f
ec
t
iv
ely
in
civ
ic
life
.
No
wad
a
y
s
,
th
e
u
s
e
o
f
s
o
cial
m
e
d
ia
is
r
is
in
g
as
th
e
i
n
ter
n
et
g
r
o
ws
[
1
]
;
th
e
f
a
k
e
n
ews
s
p
r
e
ad
h
as
b
ec
o
m
e
m
o
r
e
p
r
ev
alen
t
[
1
]
.
T
h
is
r
is
e
is
p
ar
tly
d
u
e
t
o
t
h
e
ea
s
e
with
wh
ich
in
f
o
r
m
atio
n
ca
n
b
e
s
h
ar
e
d
an
d
th
e
d
esire
f
o
r
s
en
s
atio
n
al
c
o
n
ten
t.
W
h
y
is
f
a
k
e
n
ews
s
o
tr
o
u
b
lin
g
?
B
ec
au
s
e
ea
ch
o
f
u
s
is
in
f
lu
en
ce
d
b
y
b
o
th
p
o
s
itiv
e
an
d
n
eg
ativ
e
f
o
r
ce
s
[
2
]
,
it
is
cr
u
cial
to
h
a
v
e
r
ap
id
p
r
e
d
ictio
n
m
ec
h
an
is
m
s
to
ef
f
ec
tiv
ely
cu
r
b
th
e
s
p
r
ea
d
o
f
m
is
in
f
o
r
m
atio
n
.
Au
to
m
atic
class
if
icatio
n
[
3
]
o
f
n
ews
to
p
ics
an
d
au
th
en
ticity
o
f
n
ews
s
im
u
ltan
eo
u
s
ly
[
4
]
p
r
esen
ts
a
s
ig
n
if
ican
t
ch
allen
g
e
a
n
d
h
as
r
ec
en
tly
g
ar
n
er
e
d
co
n
s
id
er
ab
le
atten
tio
n
f
r
o
m
b
o
th
th
e
p
u
b
lic
an
d
r
esear
ch
er
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
4
HA
N
:
Hyp
erg
r
a
p
h
-
b
a
s
ed
h
ie
r
a
r
ch
ica
l a
tten
tio
n
n
etw
o
r
k
fo
r
…
(
A
lp
a
n
a
A
.
B
o
r
s
e
)
2203
E
ar
ly
m
eth
o
d
s
f
o
r
d
etec
tin
g
f
ak
e
n
ews
ty
p
ically
r
elied
o
n
ex
ten
s
iv
e
h
an
d
c
r
af
ted
f
ea
t
u
r
e
s
,
s
u
ch
as
n
ews
co
n
ten
t
[
5
]
,
u
s
er
p
r
o
f
ile
s
,
an
d
n
ews
p
r
o
p
ag
atio
n
p
ath
s
,
to
tr
ain
class
if
ier
s
th
at
ass
es
s
th
e
tr
u
th
f
u
ln
ess
o
f
n
ews
s
to
r
ies
[
6
]
.
Ho
wev
e
r
,
cr
e
atin
g
a
co
m
p
r
eh
e
n
s
iv
e
s
et
o
f
f
ea
tu
r
es
[
7
]
is
d
if
f
icu
lt
d
u
e
to
t
h
e
d
iv
er
s
e
r
an
g
e
o
f
to
p
ics,
wr
itin
g
s
ty
les,
an
d
p
latf
o
r
m
s
ass
o
ciate
d
with
f
ak
e
n
ews
[
8
]
.
C
u
r
r
en
t
f
ak
e
n
ews
d
etec
tio
n
m
eth
o
d
o
l
o
g
ies
en
co
m
p
ass
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs)
[
9
]
,
lo
g
is
tic
r
eg
r
ess
i
o
n
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVMs),
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
[
1
0
]
,
a
n
d
g
r
ap
h
c
o
n
v
o
lu
tio
n
al
n
etwo
r
k
s
(
GC
Ns).
Desp
ite
th
eir
ef
f
icac
y
,
th
ese
ap
p
r
o
ac
h
es
en
co
u
n
ter
s
ig
n
if
ic
an
t
ch
allen
g
es,
s
u
ch
as
th
eir
lim
ited
ca
p
ac
ity
to
en
ca
p
s
u
late
in
tr
icate
r
elatio
n
s
h
ip
s
an
d
co
n
tex
tu
al
d
etails,
p
ar
ticu
lar
ly
with
in
s
h
o
r
t
-
f
o
r
m
co
n
ten
t
lik
e
s
o
cial
m
ed
ia
p
o
s
ts
.
Fu
r
th
er
m
o
r
e,
t
h
ey
o
f
te
n
lack
s
ca
lab
ilit
y
a
n
d
r
ea
l
-
tim
e
p
r
o
ce
s
s
in
g
ca
p
ab
ilit
ies,
wh
ich
ar
e
ess
en
tial
f
o
r
m
an
a
g
in
g
th
e
v
a
s
t
an
d
ev
e
r
-
ev
o
lv
in
g
lan
d
s
ca
p
e
o
f
s
o
cial
m
ed
ia
d
ata.
Desp
i
te
ad
v
an
ce
s
,
th
er
e
r
em
ain
s
a
n
ee
d
f
o
r
im
p
r
o
v
ed
a
cc
u
r
ac
y
in
p
r
ed
ictin
g
an
d
class
if
y
in
g
n
ews a
u
th
en
ticity
.
Ov
er
all,
d
e
v
elo
p
in
g
a
r
o
b
u
s
t
f
ak
e
n
ews
p
r
e
d
ictio
n
s
y
s
tem
is
ess
en
tial
f
o
r
e
n
s
u
r
in
g
th
e
cr
ed
ib
ilit
y
in
f
o
r
m
atio
n
o
n
s
o
cial
m
ed
ia
[
1
1
]
–
[
1
4
]
;
im
p
r
o
v
in
g
ac
cu
r
ac
y
o
f
n
ews
class
if
icatio
n
an
d
p
r
ed
ictio
n
[
1
5
]
.
T
h
e
p
r
o
p
o
s
ed
r
esear
ch
aim
s
to
d
e
v
elo
p
an
ac
cu
r
ate
f
ak
e
n
ews
p
r
ed
ictio
n
s
y
s
tem
to
class
if
y
f
a
k
e
n
ews
f
r
o
m
s
o
cial
m
ed
ia
s
o
u
r
ce
s
[
1
6
]
.
T
h
e
p
r
i
m
ar
y
o
b
jectiv
e
is
to
d
etec
t
a
n
d
h
i
g
h
lig
h
t
s
o
cial
m
ed
ia
s
ites
with
f
ak
e
n
ews
to
p
r
ev
en
t
au
t
h
en
ticity
o
f
n
ews
[
1
7
]
,
u
ltima
tely
en
h
an
cin
g
th
e
f
ak
e
n
ews
p
er
f
o
r
m
an
ce
o
f
p
r
ed
ictio
n
an
d
class
if
icatio
n
o
f
n
ews
to
p
ic
[
1
8
]
.
So
,
to
lear
n
d
is
tin
g
u
is
h
i
n
g
p
atter
n
s
au
to
m
atica
lly
;
o
u
r
ap
p
r
o
ac
h
em
p
l
o
y
s
d
ee
p
n
e
u
r
al
n
etwo
r
k
s
f
r
o
m
n
ews
co
n
ten
t,
n
ews
to
p
ics
an
d
co
n
tex
tu
al
in
f
o
r
m
atio
n
,
s
u
ch
as
au
th
o
r
p
r
o
f
iles
,
s
ig
n
if
ican
tly
im
p
ac
t
f
ak
e
n
e
ws
p
r
ed
ictio
n
[
1
9
]
.
Ou
r
r
e
s
ea
r
ch
is
g
u
id
e
d
b
y
two
p
r
im
ar
y
o
b
jectiv
es:
i)
p
r
ed
ictio
n
:
ac
cu
r
ately
p
r
e
d
ictin
g
wh
eth
er
n
ews
is
f
ak
e
o
r
tr
u
e
a
n
d
ii)
class
if
icatio
n
:
I
m
p
lem
en
tin
g
a
g
r
an
u
lar
class
if
icatio
n
s
y
s
te
m
th
at
ca
teg
o
r
izes
n
ews
in
to
f
iv
e
d
is
tin
ct
lab
els:
T
R
UE
,
FALSE
,
So
m
ewh
at
T
R
UE
,
m
o
s
tly
FALSE
,
an
d
p
a
n
ts
o
n
f
ir
e.
T
o
ad
d
r
ess
th
is
n
ews
au
th
en
ticity
b
y
p
r
ed
esti
n
in
g
f
a
k
e
n
e
ws,
we
p
r
o
p
o
s
e
a
d
ee
p
lear
n
in
g
-
b
ased
i)
n
o
v
el
a
p
p
r
o
ac
h
lev
el
b
ased
4
h
ier
a
r
ch
ical
atten
tio
n
n
et
wo
r
k
s
(
4
HAN)
m
o
d
el
[
2
0
]
–
[
2
2
]
th
at
lev
e
r
ag
es
p
r
o
p
o
s
ed
h
y
p
er
g
r
ap
h
n
eu
r
al
n
etwo
r
k
m
o
d
el
an
d
ii)
a
n
o
v
el
h
y
p
e
r
g
r
a
p
h
co
n
v
o
l
u
tio
n
n
eu
r
al
n
etwo
r
k
an
d
h
y
p
er
g
r
ap
h
atten
tio
n
n
eu
r
al
n
etwo
r
k
m
o
d
els
[
2
3
]
–
[
2
5
]
.
Ou
r
ap
p
r
o
ac
h
i
n
clu
d
es
a
n
e
w
n
ews
h
y
p
er
g
r
ap
h
(H
-
Gr
ap
h
)
m
et
h
o
d
a
n
d
HAN
to
ca
p
tu
r
e
r
elatio
n
s
h
ip
s
b
ased
o
n
b
o
th
tex
tu
al
a
n
d
co
n
tex
tu
al
in
f
o
r
m
atio
n
,
p
r
o
v
id
i
n
g
a
r
ich
er
r
ep
r
esen
tatio
n
o
f
n
ews.
T
h
is
m
o
d
e
l
co
n
s
id
er
s
th
e
in
tr
icate
an
d
o
f
ten
ir
r
elev
a
n
t
r
elatio
n
s
h
ip
s
in
th
e
h
y
p
er
g
r
a
p
h
as
p
a
r
t
o
f
th
e
lear
n
in
g
p
r
o
ce
s
s
an
d
em
p
lo
y
s
an
atten
tio
n
m
ec
h
a
n
is
m
to
v
is
u
alize
an
d
in
ter
p
r
et
th
ese
n
ews
r
elatio
n
s
.
W
e
also
in
tr
o
d
u
ce
a
d
y
n
a
m
ic
weig
h
tin
g
s
tr
ateg
y
to
b
alan
ce
m
u
ltip
le
task
s
ef
f
ec
tiv
ely
.
T
h
ese
im
p
r
o
v
em
e
n
ts
en
ab
le
th
e
m
o
d
elin
g
o
f
co
m
p
lex
,
m
u
lti
-
d
im
en
s
io
n
al
in
ter
ac
tio
n
s
an
d
d
e
p
en
d
e
n
cies
with
in
n
ews
d
ata,
th
e
r
eb
y
in
c
r
ea
s
in
g
th
e
s
y
s
tem
’
s
ca
p
ab
ilit
y
to
ac
cu
r
ately
an
d
co
n
tex
tu
ally
d
etec
t
f
ak
e
n
ew
s
.
Pre
v
io
u
s
ef
f
o
r
ts
h
av
e
im
p
r
o
v
ed
th
e
p
er
f
o
r
m
a
n
ce
o
f
f
a
k
e
n
ews
p
r
ed
ictio
n
m
o
d
els,
b
u
t
th
ese
im
p
r
o
v
em
e
n
ts
h
av
e
o
f
ten
d
im
in
is
h
ed
wh
en
d
ea
lin
g
with
s
h
o
r
t
n
ews
c
o
n
ten
t.
T
o
a
d
d
r
ess
th
is
,
we
tr
ain
o
u
r
m
o
d
el
u
s
in
g
s
h
o
r
t
co
n
ten
t
-
b
ased
L
I
AR
d
a
taset
to
an
d
p
r
ed
ict
o
f
n
ews
au
th
en
ticity
as
n
ews
is
f
ak
e
o
r
n
o
t.
T
h
e
r
est
o
f
th
is
p
ap
er
is
s
tr
u
ctu
r
ed
as
f
o
llo
ws:
s
ec
tio
n
2
p
r
o
v
id
es
a
f
ak
e
n
ews
class
if
i
ca
tio
n
an
d
p
r
ed
ictio
n
m
eth
o
d
-
4
HAN.
S
ec
tio
n
3
g
iv
es
th
e
ex
p
e
r
im
en
t
al
r
esu
lts
an
d
d
is
cu
s
s
es
th
em
in
d
etail,
in
clu
d
in
g
co
m
p
ar
is
o
n
s
with
b
aselin
es,
ex
is
tin
g
wo
r
k
s
,
an
d
s
tate
-
of
-
th
e
-
ar
t
s
y
s
tem
s
ac
r
o
s
s
v
ar
io
u
s
m
etr
ics.
Fin
ally
,
s
ec
tio
n
4
co
n
clu
d
es th
e
s
tu
d
y
an
d
s
u
g
g
ests
p
o
ten
tial d
i
r
ec
tio
n
s
f
o
r
f
u
tu
r
e
r
esear
c
h
.
2.
M
E
T
H
O
D
T
h
e
4
HAN
tec
h
n
iq
u
e
is
d
e
s
ig
n
ed
f
o
r
t
h
e
s
im
u
ltan
eo
u
s
class
if
icatio
n
an
d
p
r
e
d
ictio
n
o
f
n
ews
au
th
en
ticity
,
p
ar
ticu
lar
ly
in
d
e
tectin
g
d
is
in
f
o
r
m
atio
n
.
I
t
u
tili
ze
s
th
e
L
I
AR
d
atase
t
to
f
o
cu
s
o
n
id
en
tify
in
g
f
ak
e
n
ews
with
co
n
te
n
t
co
m
m
o
n
ly
f
o
u
n
d
in
s
o
cial
ch
an
n
els.
T
h
e
m
o
d
el
in
teg
r
ates
n
ews
to
p
i
cs,
co
n
tex
tu
al
an
d
tex
tu
al
in
f
o
r
m
atio
n
,
an
d
th
e
c
r
ed
ib
ilit
y
h
is
to
r
y
o
f
au
th
o
r
p
r
o
f
iles
.
B
y
estab
li
s
h
in
g
co
r
r
ela
tio
n
s
am
o
n
g
n
ews
to
p
ics,
au
th
o
r
cr
e
d
ib
ilit
y
d
is
tr
ib
u
tio
n
s
,
an
d
th
e
tr
u
th
f
u
ln
ess
o
f
n
ews,
4
HAN
ef
f
ec
ti
v
ely
m
an
ag
es
b
o
t
h
p
r
ed
ictio
n
a
n
d
class
if
icatio
n
s
im
u
ltan
eo
u
s
ly
.
T
h
e
4
HAN
f
r
am
ewo
r
k
o
p
e
r
a
tes
ac
r
o
s
s
f
o
u
r
h
ier
ar
ch
ical
lev
els:
wo
r
d
s
,
s
en
ten
ce
s
,
h
ea
d
lin
es,
an
d
m
etad
ata.
I
t
co
n
s
tr
u
cts
u
s
in
g
a
h
ier
ar
ch
ical
b
o
tto
m
-
u
p
p
r
o
ce
s
s
in
g
ap
p
r
o
ac
h
;
a
n
ews
ar
ticle
in
p
u
t
v
ec
to
r
-
s
tar
tin
g
f
r
o
m
th
e
wo
r
d
lev
el
,
m
o
v
in
g
to
s
en
ten
ce
s
,
th
en
to
h
ea
d
lin
es,
a
n
d
f
in
ally
to
t
h
e
m
etad
ata
lev
el,
m
etad
ata
is
p
r
o
ce
s
s
ed
u
s
in
g
a
So
f
tMa
x
o
p
er
ato
r
.
At
t
h
e
m
eta
d
ata
lev
el,
4
HAN
in
tr
o
d
u
ce
s
a
n
en
h
an
ce
d
g
r
ap
h
-
h
y
p
er
g
r
ap
h
n
eu
r
al
n
etwo
r
k
m
o
d
el,
u
tili
zin
g
a
h
y
p
er
g
r
ap
h
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etw
o
r
k
(
HGCN)
an
d
a
h
y
p
er
g
r
ap
h
atten
tio
n
n
eu
r
al
n
etwo
r
k
(
HGAN
)
f
o
r
d
ee
p
lea
r
n
in
g
o
n
g
r
ap
h
s
th
at
in
clu
d
es
to
p
ic
an
d
r
elate
d
n
ews
co
n
tex
tu
al
in
f
o
r
m
atio
n
,
as
illu
s
tr
ated
in
Fig
u
r
e
1
:
s
y
s
tem
ar
ch
itectu
r
e
f
o
r
4
HAN
u
s
in
g
h
y
p
er
g
r
a
p
h
.
Featu
r
e
tr
an
s
f
o
r
m
atio
n
,
atten
tio
n
co
e
f
f
icien
ts
,
atten
tio
n
weig
h
ts
th
at
in
c
o
r
p
o
r
ate
p
r
o
ce
s
s
in
g
d
o
cu
m
e
n
ts
in
a
b
o
tto
m
-
u
p
f
ash
io
n
,
en
s
u
r
in
g
th
e
m
o
s
t
r
ele
v
an
t
in
f
o
r
m
atio
n
is
em
p
h
asized
at
ea
ch
lev
el.
Featu
r
e
tr
an
s
f
o
r
m
atio
n
is
th
e
p
r
o
ce
s
s
o
f
c
o
n
v
e
r
tin
g
r
aw
d
ata
in
to
a
f
o
r
m
at
s
u
itab
le
f
o
r
m
ac
h
i
n
e
lear
n
in
g
m
o
d
els
th
r
o
u
g
h
n
o
r
m
aliza
tio
n
in
o
r
d
e
r
to
in
cr
ea
s
e
m
o
d
el
p
er
f
o
r
m
a
n
ce
an
d
ac
cu
r
ac
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
2
0
2
-
2
2
1
0
2204
Fig
u
r
e
1
.
Sy
s
tem
ar
c
h
itectu
r
e
f
o
r
4
HAN
–
4
h
ier
a
r
ch
ical
atte
n
tio
n
n
etwo
r
k
u
s
in
g
h
y
p
e
r
g
r
a
p
h
co
n
v
o
lu
tio
n
an
d
atten
tio
n
n
eu
r
al
n
etwo
r
k
−
Featu
r
e
tr
an
s
f
o
r
m
atio
n
is
ca
lcu
lated
as (
1
)
,
ℎ
′
=
.
ℎ
(
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
4
HA
N
:
Hyp
erg
r
a
p
h
-
b
a
s
ed
h
ie
r
a
r
ch
ica
l a
tten
tio
n
n
etw
o
r
k
fo
r
…
(
A
lp
a
n
a
A
.
B
o
r
s
e
)
2205
−
Atten
tio
n
c
o
ef
f
icie
n
ts
r
ep
r
ese
n
t
th
e
im
p
o
r
tan
ce
o
r
weig
h
t
as
s
ig
n
ed
to
s
p
ec
if
ic
p
ar
ts
o
f
th
e
in
p
u
t
d
ata
(
e.
g
.
,
wo
r
d
s
,
s
en
ten
ce
s
)
with
in
an
at
ten
tio
n
m
ec
h
a
n
is
m
.
I
t is ca
lcu
lated
as
(
2
)
,
=
(
[
ℎ
′
|
|
ℎ
′
]
)
(
2
)
wh
er
e
is
atten
tio
n
weig
h
t
b
etwe
en
n
eu
r
o
n
s
an
d
,
a
i
s
a
lear
n
ab
le
weig
h
t
v
ec
to
r
,
∥
d
en
o
te
s
co
n
ca
ten
atio
n
,
an
d
So
f
tMa
x
is
th
e
ac
tiv
atio
n
f
u
n
ctio
n
.
−
Atten
tio
n
w
eig
h
ts
d
eter
m
in
e
t
h
e
s
ig
n
if
ican
ce
o
f
ea
ch
elem
e
n
t
(
e.
g
.
,
wo
r
d
s
,
s
en
ten
ce
s
)
in
t
h
e
co
n
tex
t
o
f
th
e
task
,
allo
win
g
th
e
m
eth
o
d
to
f
o
cu
s
o
n
th
e
m
o
s
t im
p
o
r
tan
t in
f
o
r
m
atio
n
f
o
r
im
p
r
o
v
ed
p
r
ed
ict
io
n
ac
cu
r
ac
y
.
I
t
is
ca
lcu
lated
as
(
3
)
,
a
it
=
exp
(
e
ij
)
∑
exp
(
e
ik
)
∈
(
)
(
3
)
wh
er
e
,
(
)
s
ig
n
if
ies
th
e
n
eu
r
o
n
s
s
et
co
n
n
ec
ted
t
o
i
th
r
o
u
g
h
a
h
y
p
er
ed
g
e,
a
n
d
e
ij
is
th
e
n
o
r
m
alize
d
atten
tio
n
co
ef
f
icien
t.
T
o
im
p
lem
en
t th
e
4
HAN
s
y
s
tem
,
f
o
llo
win
g
h
ier
ar
c
h
ical
m
o
d
els ar
e
u
s
ed
.
2
.
1
.
At
t
ent
io
n
m
o
del
-
wo
rd
-
l
ev
el
I
n
4
HAN,
t
h
e
wo
r
d
-
lev
el
m
o
d
el
u
s
es
an
atten
tio
n
m
ec
h
an
is
m
to
co
m
p
u
te
th
e
im
p
o
r
tan
ce
o
f
in
d
iv
id
u
al
wo
r
d
s
,
h
ig
h
lig
h
tin
g
th
e
m
o
s
t in
f
o
r
m
ativ
e
o
n
es.
T
h
e
ag
g
r
eg
ated
wo
r
d
v
ec
to
r
ca
p
tu
r
es th
e
s
en
ten
ce
'
s
o
r
d
o
cu
m
en
t'
s
co
r
e
m
ea
n
in
g
,
s
er
v
in
g
as
in
p
u
t
f
o
r
h
ig
h
e
r
lev
els.
T
h
is
p
r
o
ce
s
s
b
u
ild
s
co
n
tex
tu
ally
r
ich
r
ep
r
esen
tatio
n
s
,
en
h
an
cin
g
task
s
lik
e
s
en
tim
en
t a
n
aly
s
is
,
s
u
m
m
ar
izatio
n
,
an
d
class
if
icatio
n
.
−
C
o
m
p
u
tin
g
th
e
atten
ti
o
n
s
co
r
e
T
h
e
atten
tio
n
s
co
r
e
q
u
an
tifie
s
th
e
r
elev
an
ce
o
r
im
p
o
r
tan
c
e
o
f
a
wo
r
d
in
a
s
en
ten
ce
.
T
h
is
s
co
r
e
is
u
s
ed
to
weig
h
th
e
co
n
tr
ib
u
tio
n
o
f
ea
ch
wo
r
d
wh
en
a
g
g
r
eg
atin
g
d
ata
at
th
e
s
en
ten
ce
l
ev
el.
T
h
e
v
ec
to
r
at
lev
el
wo
r
d
is
ac
h
iev
ed
b
y
ca
l
cu
latin
g
,
c
o
m
p
u
tin
g
th
e
atten
tio
n
s
co
r
e
an
d
ag
g
r
eg
atin
g
th
e
f
ea
tu
r
e
v
ec
to
r
s
.
=
ℎ
(
ℎ
+
)
(
4
)
a
it
=
∑
a
it
ℎ
it
=
1
(
5
)
wh
er
e
r
ep
r
esen
ts
th
e
atten
tio
n
weig
h
t
ass
ig
n
ed
to
th
e
wo
r
d
in
th
e
s
en
ten
ce
,
is
th
e
atte
n
tio
n
s
co
r
e
v
ec
to
r
,
an
d
th
e
d
en
o
m
i
n
ato
r
is
th
e
s
u
m
o
f
t
h
e
ex
p
o
n
e
n
tials
o
f
atten
tio
n
s
co
r
es f
o
r
all
n
o
d
es
at
tim
e
.
2
.
2
.
At
t
ent
io
n
m
o
del
-
s
ent
ence
-
lev
el
T
h
e
atten
tio
n
s
co
r
e
at
lev
el
s
e
n
ten
ce
,
q
u
an
tifie
s
th
e
weig
h
ta
g
e
o
f
ea
c
h
s
en
ten
ce
s
in
th
e
co
n
tex
t
o
f
th
e
d
o
cu
m
e
n
t
.
I
t is
ca
lcu
lated
b
y
ap
p
ly
i
n
g
a
weig
h
ted
s
u
m
o
f
s
en
ten
ce
an
d
atten
tio
n
weig
h
t
em
b
ed
d
in
g
s
.
T
h
en
it is
g
et
n
o
r
m
alize
d
th
r
o
u
g
h
a
So
f
tMa
x
f
u
n
ctio
n
to
em
p
h
asize
k
ey
s
en
ten
ce
s
.
−
C
o
m
p
u
tin
g
th
e
atten
ti
o
n
s
co
r
e
an
d
atten
ti
o
n
weig
h
t
u
is
=
t
a
n
h(
W
s
v
i
+
b
s
)
(
6
)
B
is
=
e
xp
(
u
s
)
∑
e
xp
(
u
s
)
S
=
1
(
7
)
wh
er
e
,
is
a
weig
h
t
m
atr
ix
,
is
th
e
f
ea
tu
r
e
v
ec
t
o
r
o
f
n
o
d
e
,
a
b
ias
ter
m
atten
tio
n
s
co
r
e
v
ec
to
r
an
d
th
e
d
en
o
m
i
n
ato
r
is
th
e
s
u
m
o
f
th
e
e
x
p
o
n
en
tials
o
f
atten
t
io
n
s
co
r
es f
o
r
all
n
o
d
es
in
th
e
s
et
.
2
.
3
.
At
t
ent
io
n
m
o
del
-
hea
dli
ne
-
lev
el
T
h
is
m
o
d
el
p
r
o
ce
s
s
es
an
d
as
s
ig
n
s
atten
tio
n
to
h
ea
d
lin
es,
cr
ea
tin
g
r
ep
r
esen
tatio
n
s
f
r
o
m
s
en
ten
ce
-
lev
el
o
u
tp
u
ts
an
d
tu
n
in
g
s
p
ec
if
ically
f
o
r
h
ea
d
lin
e
d
ata.
I
t
l
ev
er
ag
es
s
im
ilar
m
ec
h
an
is
m
s
as
p
r
ev
io
u
s
m
o
d
els
wh
ile
o
p
tim
izin
g
f
o
r
b
r
ev
ity
an
d
r
elev
an
ce
in
h
ea
d
lin
e
co
n
tex
ts
.
B
y
ca
p
t
u
r
in
g
th
e
m
o
s
t
s
alien
t
f
ea
tu
r
es
o
f
h
ea
d
lin
es,
it e
n
s
u
r
es im
p
r
o
v
ed
p
er
f
o
r
m
an
ce
in
task
s
lik
e
class
if
icatio
n
o
r
s
u
m
m
ar
izatio
n
co
n
tex
ts
.
−
Ag
g
r
eg
atin
g
th
e
f
ea
tu
r
e
v
ec
to
r
s
at
h
ea
d
lin
e
lev
el
ℎ
∗
=
∑
B
is
v
i
=
1
(
8
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
2
0
2
-
2
2
1
0
2206
wh
er
e
ℎ
∗
is
th
e
u
p
d
ated
f
ea
tu
r
e
v
ec
to
r
f
o
r
n
o
d
e
,
an
d
th
e
s
u
m
is
tak
en
o
v
er
all
atten
tio
n
wei
g
h
ts
B
is
an
d
co
r
r
esp
o
n
d
in
g
f
ea
tu
r
e
v
ec
to
r
s
.
2
.
4
.
M
et
a
da
t
a
-
lev
el
m
o
del w
it
h hy
perg
ra
ph
co
nv
o
lutio
n a
nd
a
t
t
ent
io
n
A
h
eter
o
g
en
e
o
u
s
g
r
a
p
h
co
n
v
o
lu
tio
n
al
n
etwo
r
k
(
HGCN)
p
r
o
ce
s
s
es
n
ews
m
etad
ata,
ca
p
tu
r
in
g
r
elatio
n
s
h
ip
s
b
etwe
en
elem
e
n
ts
lik
e
to
p
ics,
k
e
y
wo
r
d
s
,
a
u
th
o
r
s
,
an
d
p
u
b
licatio
n
d
ates.
Op
e
r
atin
g
at
th
e
f
o
u
r
t
h
lev
el,
it
en
co
d
es
co
m
p
lex
in
t
er
ac
tio
n
s
u
s
in
g
h
eter
o
g
en
eo
u
s
g
r
ap
h
s
to
en
h
an
ce
r
ep
r
esen
tatio
n
an
d
im
p
r
o
v
e
task
s
s
u
ch
as
class
if
icatio
n
,
r
ec
o
m
m
en
d
atio
n
,
a
n
d
clu
s
t
er
in
g
.
I
ts
ca
p
ac
ity
t
o
in
teg
r
ate
m
u
lti
-
r
elatio
n
al
in
f
o
r
m
atio
n
m
ak
es it id
ea
l f
o
r
u
n
d
er
s
tan
d
i
n
g
in
tr
icate
m
eta
d
ata
r
elatio
n
s
h
ip
s
in
n
ews d
atasets
.
2
.
4
.
1
.
H
y
perg
ra
ph
A
h
y
p
er
g
r
ap
h
is
a
u
n
iq
u
e
ty
p
e
o
f
g
r
a
p
h
wh
e
r
e
a
s
in
g
le
ed
g
e,
ca
lled
a
h
y
p
er
ed
g
e,
ca
n
co
n
n
ec
t
m
u
ltip
le
n
o
d
es.
W
e
d
ev
elo
p
e
d
a
m
o
d
el
f
o
r
a
n
ews
au
th
e
n
ticity
as
a
f
ak
e
n
ews
p
r
ed
ictio
n
as
a
clas
s
if
icatio
n
task
o
n
th
e
h
y
p
e
r
g
r
a
p
h
.
T
h
i
s
in
v
o
lv
es
u
s
in
g
a
f
ea
tu
r
e
v
ec
to
r
o
f
lab
ele
d
n
ews
ar
ticles
with
in
th
e
n
ews
h
y
p
er
g
r
ap
h
in
ter
ac
tio
n
s
am
o
n
g
n
ews
ar
ticles,
as
s
h
o
wn
in
T
ab
le
1
:
r
elatio
n
tab
le
b
etwe
en
n
ews
with
co
n
tex
tu
al
in
f
o
r
m
atio
n
illu
s
tr
ates th
e
r
elatio
n
s
h
ip
s
f
r
o
m
n
ew
s
1
to
n
ews
6.
T
ab
le
1
.
R
elatio
n
s
h
ip
tab
le
b
etwe
en
n
ews a
n
d
its
co
n
tex
t
u
al
in
f
o
r
m
atio
n
Ed
g
e
E1
E2
E3
E4
E5
N
o
d
e
/
V
e
r
t
e
x
Lo
c
a
t
i
o
n
C
r
e
d
i
t
h
i
st
o
r
y
A
u
t
h
o
r
P
u
b
l
i
s
h
e
r
C
o
n
t
e
n
t
f
e
a
t
u
r
e
V
1
(
N
e
w
s
1
)
1
1
1
0
1
V
2
(
N
e
w
s
2
)
1
1
1
1
1
V
3
(
N
e
w
s
3
)
1
1
0
1
1
V
4
(
N
e
w
s
4
)
0
0
1
0
0
V
5
(
N
e
w
s
5
)
1
0
1
0
1
V
6
(
N
e
w
s
6
)
0
1
1
1
0
V
7
(
N
e
w
s
7
)
1
0
0
0
1
V
3
(
N
e
w
s
3
)
1
1
0
1
1
2
.
4
.
2
.
G
ra
ph
co
nv
o
lutio
n o
n
hy
perg
ra
ph
E
x
ten
d
GC
N
to
o
p
er
ate
o
n
h
y
p
e
r
g
r
ap
h
-
s
tr
u
ctu
r
e
d
d
ata.
Her
e,
we
co
m
p
u
te
n
o
d
e
e
m
b
ed
d
in
g
s
co
n
s
id
er
in
g
i
n
ter
ac
tio
n
s
th
r
o
u
g
h
h
y
p
er
ed
g
es.
ℎ
+
1
=
(
∑
1
|
ɛ
is
|
∑
1
|V
|
ℎ
∈
∈
ɛ
is
(
9
)
wh
er
e
ℎ
r
ep
r
esen
ts
n
o
d
e
's
r
ep
r
esen
tatio
n
at
lay
er
,
ɛ
I
d
en
o
t
es
th
e
h
y
p
er
ed
g
es
in
cid
en
t
o
n
n
o
d
e
,
V
r
ep
r
esen
ts
th
e
n
o
d
es
co
n
n
ec
te
d
b
y
h
y
p
er
e
d
g
e
,
an
d
is
th
e
s
o
f
tm
ax
ac
tiv
atio
n
f
u
n
ctio
n
a
p
p
lied
elem
en
t
-
wis
e
to
n
o
r
m
alize
th
e
o
u
tp
u
t.
2
.
4
.
3
.
G
ra
ph
a
t
t
ent
i
o
n o
n hy
perg
ra
ph
A
h
y
p
er
g
r
ap
h
atten
tio
n
n
eu
r
al
n
etwo
r
k
(
HGAN
)
th
at
a
p
p
lies
an
atten
tio
n
m
ec
h
a
n
is
m
to
th
e
h
y
p
er
g
r
ap
h
,
f
o
cu
s
in
g
o
n
im
p
o
r
tan
t
m
etad
ata
r
elatio
n
s
h
ip
s
.
T
h
is
m
o
d
el
is
u
s
ed
in
tan
d
em
with
th
e
HG
C
N
to
en
h
an
ce
th
e
f
in
al
class
if
icatio
n
an
d
p
r
e
d
ictio
n
p
er
f
o
r
m
a
n
ce
.
I
n
HGAN
,
th
e
g
o
al
is
to
lear
n
th
e
im
p
o
r
tan
ce
o
f
n
o
d
es a
n
d
h
y
p
er
ed
g
es in
a
h
y
p
er
g
r
a
p
h
.
Her
e
a
r
e
th
e
k
ey
f
o
r
m
u
las in
v
o
lv
ed
:
α
ve
=
e
xp
(
a
T
[
W
x
v
Wc
v
]
)
∑
e
xp
(
µ
(
a
T
[
W
x
v
Wc
v
]
)
u
∈
e
(
10)
wh
er
e
,
is
w
eig
h
t
m
atr
ix
,
x
v
is
f
ea
tu
r
e
v
ec
to
r
o
f
n
o
d
e
v
,
c
v
is
f
e
a
t
u
r
e
v
e
c
t
o
r
r
e
p
r
e
s
e
n
t
i
n
g
t
h
e
h
y
p
e
r
e
d
g
e
e
(
o
f
t
e
n
c
o
m
p
u
t
e
d
a
s
a
n
a
g
g
r
e
g
a
t
i
o
n
o
f
n
o
d
e
f
e
a
t
u
r
e
s
w
i
t
h
i
n
e
)
,
a
is
lear
n
ab
le
atten
tio
n
v
e
cto
r
,
µ
is
n
o
n
-
lin
ea
r
ac
tiv
atio
n
f
u
n
cti
o
n
2
.
4
.
4
.
H
y
perg
ra
ph
inte
g
ra
t
io
n wit
h
hea
dli
ne
v
ec
t
o
r
I
n
teg
r
atin
g
a
f
o
u
r
th
lev
el
in
to
th
e
m
o
d
el,
co
m
b
i
n
in
g
h
y
p
er
g
r
a
p
h
,
tr
ip
le
h
ier
ar
ch
ical
atten
tio
n
n
etwo
r
k
s
(
3
HAN)
,
a
n
d
So
f
t
Ma
x
ac
tiv
atio
n
f
u
n
ctio
n
,
ca
n
f
u
r
t
h
er
en
h
an
ce
th
e
m
o
d
e
l's
ab
ilit
y
to
d
etec
t
co
m
p
lex
p
atter
n
s
in
f
ak
e
n
e
ws
d
ete
ctio
n
o
n
d
atasets
lik
e
L
I
AR
.
Ap
p
ly
So
f
tMa
x
ac
t
iv
atio
n
to
co
m
p
u
te
p
r
o
b
a
b
ilit
ies o
v
er
n
o
d
e
em
b
e
d
d
in
g
s
h
I
in
th
e
f
in
al
lay
er
̅
o
f
t
h
e
m
o
d
el
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
4
HA
N
:
Hyp
erg
r
a
p
h
-
b
a
s
ed
h
ie
r
a
r
ch
ica
l a
tten
tio
n
n
etw
o
r
k
fo
r
…
(
A
lp
a
n
a
A
.
B
o
r
s
e
)
2207
(
(
ℎ
(
)
)
=
e
xp
(
(
ℎ
(
)
)
∑
e
xp
(
(
ℎ
(
)
)
∈
(
1
1
)
wh
er
e
(
ℎ
(
)
)
is
th
e
v
ec
to
r
o
f
n
o
d
e
e
m
b
ed
d
in
g
s
f
o
r
n
o
d
e
at
th
e
f
in
al
lay
er
,
an
d
is
th
e
s
et
o
f
all
n
o
d
es.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
ch
o
ice
o
f
ac
tiv
atio
n
f
u
n
ct
io
n
s
ig
n
if
ican
tly
im
p
ac
ts
m
o
d
el
p
er
f
o
r
m
an
ce
,
with
So
f
tMa
x
lead
in
g
to
en
h
an
ce
d
p
r
ed
ictio
n
ac
c
u
r
ac
y
,
f
aster
co
n
v
er
g
en
ce
,
an
d
m
o
r
e
s
tab
le
g
r
ad
ien
t
p
r
o
p
ag
atio
n
.
T
h
is
u
n
d
er
s
co
r
es
th
e
im
p
o
r
tan
ce
o
f
s
elec
tin
g
a
p
p
r
o
p
r
iate
ac
tiv
atio
n
f
u
n
ctio
n
s
f
o
r
o
p
tim
izin
g
n
eu
r
al
n
etw
o
r
k
ef
f
icien
c
y
an
d
p
r
ed
ictio
n
r
eliab
ilit
y
in
class
if
icatio
n
task
s
.
Activ
atio
n
f
u
n
ctio
n
s
lik
e
R
eL
U
an
d
its
v
ar
ian
ts
m
ay
b
e
m
o
r
e
ef
f
ec
tiv
e
in
h
id
d
en
lay
er
s
d
u
e
to
th
eir
ab
ilit
y
to
m
itig
ate
v
an
is
h
in
g
g
r
a
d
ien
ts
,
wh
ile
S
o
f
tMa
x
is
id
ea
l
f
o
r
m
u
lti
-
class
o
u
tp
u
ts
,
as
it
p
r
o
d
u
ce
s
n
o
r
m
alize
d
p
r
o
b
ab
ilit
y
d
is
tr
ib
u
tio
n
s
.
T
h
er
ef
o
r
e,
u
n
d
er
s
tan
d
i
n
g
th
e
ch
ar
ac
ter
is
tics
o
f
ea
ch
ac
tiv
atio
n
f
u
n
ctio
n
is
cr
u
cial
f
o
r
d
esig
n
in
g
m
o
d
els
th
at
ac
h
iev
e
d
esire
d
p
er
f
o
r
m
an
ce
m
etr
ics.
3
.
1
.
Wo
rd
co
un
t
m
o
del
o
n L
I
AR
da
t
a
s
et
W
e
ev
alu
ated
v
ar
io
u
s
wo
r
d
co
u
n
t
m
eth
o
d
s
to
d
eter
m
in
e
th
eir
im
p
ac
t
o
n
th
e
p
er
f
o
r
m
i
n
g
o
f
th
e
4
HAN
m
eth
o
d
,
f
o
c
u
s
in
g
o
n
k
ey
m
etr
ics
s
u
ch
as
ac
cu
r
ac
y
,
co
m
p
u
tatio
n
al
e
f
f
icien
c
y
,
an
d
co
n
tex
tu
al
R
elev
an
ce
o
n
t
h
e
L
I
AR
d
atas
et.
T
ab
le
2
s
u
m
m
ar
izes
th
e
r
esu
lts
o
f
th
ese
ev
alu
atio
n
s
.
A
m
o
n
g
th
e
m
et
h
o
d
s
test
ed
,
T
F
-
I
DF
em
er
g
ed
as
th
e
m
o
s
t
ef
f
ec
tiv
e,
o
f
f
er
in
g
a
b
alan
ce
d
co
m
b
in
atio
n
o
f
p
r
ec
is
io
n
,
r
elev
an
ce
,
a
n
d
co
m
p
u
tatio
n
al
e
f
f
icien
cy
.
B
ased
o
n
o
u
r
co
m
p
ar
ativ
e
a
n
aly
s
is
with
in
th
e
4
HAN
f
r
am
ewo
r
k
,
T
F
-
I
DF
was
s
elec
ted
as
th
e
p
r
ef
er
r
e
d
wo
r
d
co
u
n
t
m
eth
o
d
.
I
t
d
em
o
n
s
tr
ated
s
u
p
e
r
io
r
p
er
f
o
r
m
an
ce
in
o
p
tim
izin
g
b
o
th
th
e
lea
r
n
in
g
d
y
n
am
ics
an
d
p
r
ed
ictiv
e
ac
c
u
r
ac
y
ac
r
o
s
s
all
h
ier
ar
ch
ical
lev
els
o
f
th
e
4
HAN
m
o
d
el
wh
en
ap
p
lied
to
t
h
e
L
I
AR
d
ataset.
T
F
-
I
DF’
s
ab
ili
ty
to
b
alan
ce
p
r
ec
is
io
n
,
r
elev
an
ce
,
an
d
co
m
p
u
tatio
n
al
ef
f
ic
ien
cy
m
ak
es
it
th
e
m
o
s
t su
itab
le
ch
o
ice
f
o
r
en
h
a
n
cin
g
th
e
ef
f
ec
tiv
en
ess
o
f
o
u
r
f
ak
e
n
ews d
etec
tio
n
f
r
am
ewo
r
k
.
T
ab
le
2
.
Acc
u
r
ac
y
p
e
r
f
o
r
m
an
c
e
f
o
r
v
ar
io
u
s
wo
r
d
co
u
n
t m
o
d
els
W
o
r
d
c
o
u
n
t
me
t
h
o
d
A
c
c
u
r
a
c
y
R
e
mar
k
R
a
w
w
o
r
d
c
o
u
n
t
85
-
8
8
%
B
a
si
c
c
o
u
n
t
i
n
g
o
f
w
o
r
d
s
i
n
t
h
e
t
e
x
t
.
L
i
mi
t
e
d
i
mp
a
c
t
a
s
i
t
l
a
c
k
s s
e
ma
n
t
i
c
u
n
d
e
r
st
a
n
d
i
n
g
.
TF
-
I
D
F
92
-
9
5
%
W
e
i
g
h
s
w
o
r
d
f
r
e
q
u
e
n
c
y
a
g
a
i
n
s
t
i
t
s
o
v
e
r
a
l
l
o
c
c
u
r
r
e
n
c
e
,
e
mp
h
a
si
z
i
n
g
i
m
p
o
r
t
a
n
t
t
e
r
ms
.
Li
k
e
l
y
t
o
si
g
n
i
f
i
c
a
n
t
l
y
i
m
p
r
o
v
e
a
c
c
u
r
a
c
y
b
y
f
o
c
u
si
n
g
o
n
r
e
l
e
v
a
n
t
w
o
r
d
s.
B
a
g
o
f
w
o
r
d
s
(
B
o
W
)
88
-
9
1
%
R
e
p
r
e
se
n
t
s
t
e
x
t
a
s a
c
o
l
l
e
c
t
i
o
n
o
f
w
o
r
d
s wi
t
h
o
u
t
c
o
n
s
i
d
e
r
i
n
g
o
r
d
e
r
o
r
c
o
n
t
e
x
t
.
P
r
o
v
i
d
e
s
mo
d
e
r
a
t
e
i
m
p
r
o
v
e
m
e
n
t
b
u
t
m
a
y
mi
s
s
c
o
n
t
e
x
t
u
a
l
n
u
a
n
c
e
s.
W
o
r
d
e
m
b
e
d
d
i
n
g
s
96
-
9
8
%
D
e
n
se
v
e
c
t
o
r
r
e
p
r
e
se
n
t
a
t
i
o
n
s
c
a
p
t
u
r
i
n
g
se
ma
n
t
i
c
r
e
l
a
t
i
o
n
sh
i
p
s.
H
i
g
h
i
m
p
a
c
t
o
n
a
c
c
u
r
a
c
y
d
u
e
t
o
e
n
h
a
n
c
e
d
c
o
n
t
e
x
t
u
a
l
u
n
d
e
r
s
t
a
n
d
i
n
g
.
F
r
e
q
u
e
n
c
y
-
b
a
se
d
m
e
t
h
o
d
s
86
-
8
9
%
F
o
c
u
ses
o
n
h
i
g
h
-
f
r
e
q
u
e
n
c
y
w
o
r
d
s
o
r
n
-
g
r
a
m
s.
C
a
n
i
n
t
r
o
d
u
c
e
n
o
i
se,
l
e
a
d
i
n
g
t
o
mi
x
e
d
r
e
su
l
t
s
i
n
a
c
c
u
r
a
c
y
.
P
o
si
t
i
o
n
a
l
e
n
c
o
d
i
n
g
94
-
9
7
%
A
d
d
s
i
n
f
o
r
m
a
t
i
o
n
a
b
o
u
t
w
o
r
d
p
o
s
i
t
i
o
n
s,
e
s
p
e
c
i
a
l
l
y
e
f
f
e
c
t
i
v
e
i
n
s
e
q
u
e
n
c
e
-
b
a
s
e
d
mo
d
e
l
s
l
i
k
e
Tr
a
n
sf
o
r
mers
.
H
y
b
r
i
d
m
e
t
h
o
d
s
98
-
9
9
%
C
o
m
b
i
n
e
s m
u
l
t
i
p
l
e
a
p
p
r
o
a
c
h
e
s
(
e
.
g
.
,
TF
-
I
D
F
+
Emb
e
d
d
i
n
g
s)
.
P
r
o
v
i
d
e
s
t
h
e
b
e
st
o
v
e
r
a
l
l
a
c
c
u
r
a
c
y
b
y
l
e
v
e
r
a
g
i
n
g
t
h
e
s
t
r
e
n
g
t
h
s
o
f
d
i
f
f
e
r
e
n
t
me
t
h
o
d
s.
3
.
2
.
Act
i
v
a
t
io
n
f
un
ct
io
n
o
n
L
I
AR
da
t
a
s
et
I
n
o
u
r
ex
p
e
r
im
en
ts
,
as
d
etailed
in
T
ab
le
3
,
we
s
y
s
tem
atica
lly
ev
alu
ated
th
e
p
er
f
o
r
m
a
n
ce
m
etr
ics,
in
clu
d
in
g
ac
c
u
r
ac
y
,
c
o
n
v
e
r
g
e
n
ce
s
p
ee
d
(
in
e
p
o
ch
s
)
,
a
n
d
g
r
ad
ien
t
p
r
o
p
a
g
atio
n
,
o
f
th
e
So
f
tMa
x
ac
tiv
atio
n
f
u
n
ctio
n
a
g
ain
s
t
v
ar
io
u
s
o
t
h
er
ac
tiv
atio
n
f
u
n
ctio
n
s
o
n
th
e
L
I
AR
d
ataset.
No
tab
ly
,
So
f
tMa
x
ac
h
iev
ed
s
u
p
e
r
io
r
ac
cu
r
ac
y
,
ex
h
i
b
ited
f
aster
co
n
v
er
g
e
n
ce
,
a
n
d
f
ac
ilit
ated
m
o
r
e
ef
f
icien
t
g
r
ad
ien
t
p
r
o
p
ag
atio
n
.
C
o
n
s
eq
u
e
n
tly
,
So
f
tMa
x
was
s
elec
ted
as
th
e
ac
tiv
atio
n
f
u
n
ctio
n
f
o
r
s
u
b
s
eq
u
en
t
co
m
p
u
tatio
n
al
p
r
o
c
ess
es,
o
win
g
to
its
d
em
o
n
s
tr
ated
ef
f
icac
y
in
o
p
ti
m
izin
g
b
o
th
t
h
e
lear
n
in
g
d
y
n
am
ics
an
d
p
r
ed
ictiv
e
p
er
f
o
r
m
an
ce
o
n
th
e
L
I
AR
d
ataset.
Ou
r
m
eth
o
d
4
HAN
h
i
g
h
lig
h
t
th
at
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
en
h
a
n
ce
s
ac
cu
r
ac
y
o
n
th
e
L
I
AR
d
ataset
b
y
u
tili
zin
g
th
e
So
f
tMa
x
ac
tiv
atio
n
f
u
n
ctio
n
,
wh
ich
s
ig
n
if
ican
tly
im
p
r
o
v
es
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
o
f
n
ews
to
p
ics
an
d
p
r
e
d
ictio
n
o
f
f
ak
e
n
ews.
T
h
e
f
o
llo
win
g
p
r
e
d
ictio
n
s
an
d
class
if
icatio
n
s
wer
e
m
ad
e
u
s
in
g
th
e
So
f
tMa
x
ac
tiv
atio
n
f
u
n
ctio
n
o
n
th
e
L
I
AR
d
ataset,
r
esu
ltin
g
i
n
n
o
tab
le
g
ain
s
in
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
s
co
r
e
s
h
o
w
n
in
T
ab
le
4
:
c
o
m
p
ar
is
o
n
o
f
p
r
o
p
o
s
ed
m
et
h
o
d
4
HAN
an
d
h
y
p
er
g
r
ap
h
.
F
ig
u
r
e
2
p
r
esen
ts
a
v
is
u
al
an
aly
s
is
o
f
th
e
p
r
ed
ictio
n
ac
cu
r
ac
y
,
F1
s
co
r
e,
r
ec
all,
an
d
p
r
ec
is
io
n
m
etr
ics
f
o
r
th
e
4
HAN
m
o
d
el
ac
r
o
s
s
d
if
f
er
en
t
wo
r
d
co
u
n
t
m
eth
o
d
s
.
T
h
is
co
m
p
ar
is
o
n
h
ig
h
lig
h
ts
th
e
p
er
f
o
r
m
an
ce
im
p
r
o
v
e
m
en
t
s
ac
h
iev
ed
b
y
ea
ch
m
eth
o
d
,
with
T
F
-
I
DF
co
n
s
is
ten
tly
lead
in
g
in
all
ev
alu
a
ted
m
etr
ics,
d
em
o
n
s
tr
atin
g
i
ts
ef
f
ec
tiv
en
ess
in
en
h
an
cin
g
th
e
m
o
d
el’
s
p
r
e
d
ictiv
e
ca
p
ab
ilit
ies
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
2
0
2
-
2
2
1
0
2208
T
ab
le
3
.
Acc
u
r
ac
y
,
c
o
n
v
er
g
en
ce
s
p
e
ed
(
E
p
o
ch
s
)
an
d
g
r
a
d
ien
t p
r
o
p
a
g
atio
n
(
G
No
r
m
)
o
f
v
a
r
io
u
s
ac
tiv
atio
n
f
u
n
ctio
n
o
n
L
I
AR
d
ataset
A
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
A
c
c
u
r
a
c
y
(
%)
C
o
n
v
e
r
g
e
n
c
e
s
p
e
e
d
(
E
p
o
c
h
s)
G
r
a
d
i
e
n
t
p
r
o
p
a
g
a
t
i
o
n
R
e
LU
3
4
.
5
0
30
1
.
2
0
Le
a
k
y
R
e
LU
3
5
.
1
0
28
1
.
1
5
Ta
n
h
3
3
.
8
0
32
1
.
2
5
S
i
g
m
o
i
d
3
2
.
9
0
35
1
.
3
0
S
o
f
t
ma
x
3
6
.
2
0
25
1
.
1
0
T
ab
le
4
.
C
o
m
p
a
r
is
o
n
o
f
p
r
o
p
o
s
ed
m
eth
o
d
4
HAN
an
d
h
y
p
er
g
r
ap
h
with
o
t
h
er
m
eth
o
d
s
g
iv
in
g
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
e
ca
ll a
n
d
F1
s
co
r
e
f
o
r
f
a
k
e
n
ews p
r
e
d
ictio
n
s
y
s
tem
P
r
o
p
o
se
d
m
e
t
h
o
d
s
A
c
c
u
r
a
c
y
%
P
r
e
c
i
s
i
o
n
%
R
e
c
a
l
l
%
F1
%
3
H
A
N
8
6
.
9
2
8
5
.
6
3
8
2
.
6
2
8
3
.
7
1
H
y
p
e
r
g
r
a
p
h
7
9
.
3
1
7
7
.
8
3
7
4
.
6
6
7
5
.
9
5
H
y
p
e
r
g
r
a
p
h
+
HAN
8
7
.
6
3
8
6
.
1
9
8
1
.
8
3
8
3
.
8
2
H
y
p
e
r
g
r
a
p
h
+
HAN
+
G
C
N
9
0
.
6
1
8
9
.
3
5
8
5
.
8
4
8
7
.
5
6
H
y
p
e
r
g
r
a
p
h
+
HAN
+
GAN
9
1
.
0
3
8
9
.
7
8
8
6
.
2
7
8
8
.
0
1
H
y
p
e
r
g
r
a
p
h
+
HAN
+
G
C
N
+
GAN
9
2
.
5
6
9
1
.
2
6
8
8
.
8
4
9
0
.
0
6
H
G
C
N
(
H
y
p
e
r
g
r
a
p
h
C
o
n
v
o
l
u
t
i
o
n
N
N
)
8
9
.
1
8
8
7
.
9
6
8
4
.
1
6
8
5
.
9
7
H
G
A
N
(
H
y
p
e
r
g
r
a
p
h
A
t
t
e
n
t
i
o
n
N
N
)
9
0
.
2
7
8
9
.
0
4
8
5
.
6
9
8
7
.
2
5
H
G
C
N
+
H
G
A
N
9
1
.
0
5
8
9
.
9
1
8
6
.
1
2
8
7
.
9
3
3
H
A
N
+
H
y
p
e
r
g
r
a
p
h
9
2
.
8
9
1
.
5
1
8
8
.
9
9
0
.
0
7
3
H
A
N
+
H
G
C
N
9
3
.
7
4
9
2
.
5
3
8
9
.
9
8
9
1
.
2
8
3
H
A
N
+
H
G
A
N
9
4
.
1
4
9
2
.
9
7
9
0
.
4
2
9
1
.
8
2
4
H
A
N
-
(
3
H
A
N
+
H
G
A
N
+
H
G
C
N
+
S
o
f
t
M
a
x
)
96
9
4
.
7
1
9
2
.
1
6
9
3
.
0
9
Fig
u
r
e
2
.
An
al
y
s
is
o
f
p
r
o
p
o
s
e
d
m
eth
o
d
4
HAN
an
d
h
y
p
e
r
g
r
a
p
h
with
o
th
e
r
m
eth
o
d
s
g
iv
in
g
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all
a
n
d
F1
s
co
r
e
f
o
r
f
a
k
e
n
ews p
r
e
d
ictio
n
s
y
s
tem
4.
CO
NCLU
SI
O
N
T
h
e
ap
p
licatio
n
o
f
a
d
v
an
c
ed
n
eu
r
al
n
etwo
r
k
ar
ch
itec
tu
r
es,
p
ar
ticu
lar
ly
t
h
e
4
HA
N
m
o
d
el
in
co
r
p
o
r
atin
g
h
y
p
er
g
r
ap
h
co
n
v
o
lu
tio
n
al
n
etwo
r
k
s
(
HGCN)
an
d
h
y
p
e
r
g
r
a
p
h
atten
tio
n
n
et
wo
r
k
s
(
HGAN
)
,
h
as
s
ig
n
if
ican
tly
im
p
r
o
v
e
d
th
e
ac
c
u
r
ac
y
o
f
f
ak
e
n
ews
class
if
icatio
n
a
n
d
p
r
e
d
ictio
n
task
s
o
n
th
e
L
I
AR
d
ataset.
B
y
lev
er
ag
in
g
th
e
h
ier
ar
ch
ical
n
a
tu
r
e
o
f
4
HAN
ac
r
o
s
s
m
u
ltip
le
lev
els
-
wo
r
d
,
s
en
ten
ce
,
h
ea
d
lin
e,
an
d
m
etad
ata
-
co
m
b
in
ed
with
th
e
n
o
r
m
aliz
atio
n
ca
p
a
b
ilit
ies
o
f
th
e
So
f
t
Ma
x
ac
tiv
atio
n
f
u
n
ctio
n
,
th
e
s
y
s
tem
ef
f
ec
tiv
ely
en
h
an
ce
s
th
e
in
ter
p
r
etab
ilit
y
a
n
d
r
eliab
ilit
y
o
f
p
r
e
d
ictio
n
s
.
T
h
e
in
teg
r
atio
n
o
f
t
h
ese
ad
v
a
n
ce
d
m
et
h
o
d
o
lo
g
ies
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
4
HA
N
:
Hyp
erg
r
a
p
h
-
b
a
s
ed
h
ie
r
a
r
ch
ica
l a
tten
tio
n
n
etw
o
r
k
fo
r
…
(
A
lp
a
n
a
A
.
B
o
r
s
e
)
2209
en
ab
les
co
n
tex
tu
ally
awa
r
e
an
aly
s
is
o
f
d
is
in
f
o
r
m
atio
n
,
l
ea
d
in
g
to
a
s
u
b
s
tan
tial
in
cr
ea
s
e
in
p
r
ed
ictiv
e
ac
cu
r
ac
y
.
T
h
ese
f
in
d
in
g
s
v
alid
ate
th
e
r
o
b
u
s
tn
ess
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
an
d
u
n
d
e
r
s
co
r
e
th
e
im
p
o
r
tan
ce
o
f
b
o
th
s
o
p
h
is
ticated
n
etwo
r
k
ar
ch
itectu
r
es
an
d
t
h
e
ju
d
icio
u
s
s
elec
tio
n
o
f
ac
tiv
atio
n
f
u
n
ctio
n
s
in
o
p
tim
izin
g
th
e
p
er
f
o
r
m
an
ce
o
f
class
if
icatio
n
task
s
in
co
m
p
lex
d
o
m
ain
s
s
u
ch
as
f
ak
e
n
ews d
etec
tio
n
.
RE
F
E
R
E
NC
E
S
[
1
]
S
.
A
r
a
l
,
S
.
V
o
s
o
u
g
h
i
,
a
n
d
D
.
R
o
y
,
“
T
h
e
s
p
r
e
a
d
o
f
t
r
u
e
a
n
d
f
a
l
se
n
e
w
s
o
n
l
i
n
e
,
”
S
c
i
e
n
c
e
,
v
o
l
.
3
5
9
,
p
p
.
1
1
4
6
–
1
1
5
1
,
2
0
1
8
.
[
2
]
Y
.
C
h
e
n
,
N
.
J.
C
o
n
r
o
y
,
a
n
d
V
.
L.
R
u
b
i
n
,
“
M
i
s
l
e
a
d
i
n
g
o
n
l
i
n
e
c
o
n
t
e
n
t
:
r
e
c
o
g
n
i
z
i
n
g
c
l
i
c
k
b
a
i
t
a
s
‘
f
a
l
se
n
e
w
s
,
’
”
i
n
W
MDD
2
0
1
5
-
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
AC
M
W
o
rks
h
o
p
o
n
M
u
l
t
i
m
o
d
a
l
D
e
c
e
p
t
i
o
n
D
e
t
e
c
t
i
o
n
,
c
o
-
l
o
c
a
t
e
d
w
i
t
h
I
C
MI
2
0
1
5
,
2
0
1
5
,
p
p
.
1
5
–
1
9
,
d
o
i
:
1
0
.
1
1
4
5
/
2
8
2
3
4
6
5
.
2
8
2
3
4
6
7
.
[
3
]
S
.
G
h
o
sh
a
n
d
C
.
S
h
a
h
,
“
To
w
a
r
d
a
u
t
o
ma
t
i
c
f
a
k
e
n
e
w
s
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
An
n
u
a
l
H
a
w
a
i
i
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
y
st
e
m
S
c
i
e
n
c
e
s
,
2
0
1
9
,
v
o
l
.
2
0
1
9
-
J
a
n
u
a
,
p
p
.
2
2
5
4
–
2
2
6
3
,
d
o
i
:
1
0
.
2
4
2
5
1
/
h
i
c
ss
.
2
0
1
9
.
2
7
3
.
[
4
]
Q
.
Li
a
o
e
t
a
l
.
,
“
A
n
i
n
t
e
g
r
a
t
e
d
m
u
l
t
i
-
t
a
s
k
mo
d
e
l
f
o
r
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
,
”
I
EEE
T
ra
n
s
a
c
t
i
o
n
s
o
n
K
n
o
w
l
e
d
g
e
a
n
d
D
a
t
a
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
3
4
,
n
o
.
1
1
,
p
p
.
5
1
5
4
–
5
1
6
5
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
T
K
D
E.
2
0
2
1
.
3
0
5
4
9
9
3
.
[
5
]
D
.
D
i
F
r
a
n
z
o
a
n
d
K
.
G
l
o
r
i
a
-
G
a
r
c
i
a
,
“
F
i
l
t
e
r
b
u
b
b
l
e
s
a
n
d
f
a
k
e
n
e
w
s,
”
XRD
S
:
C
ro
ssr
o
a
d
s
,
T
h
e
AC
M
M
a
g
a
z
i
n
e
f
o
r S
t
u
d
e
n
t
s
,
v
o
l
.
2
3
,
n
o
.
3
,
p
p
.
3
2
–
3
5
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
4
5
/
3
0
5
5
1
5
3
.
[
6
]
K
.
S
h
u
,
D
.
M
a
h
u
d
e
sw
a
r
a
n
,
a
n
d
H
.
Li
u
,
“
F
a
k
e
N
e
w
sTr
a
c
k
e
r
:
A
t
o
o
l
f
o
r
f
a
k
e
n
e
w
s c
o
l
l
e
c
t
i
o
n
,
d
e
t
e
c
t
i
o
n
,
a
n
d
v
i
s
u
a
l
i
z
a
t
i
o
n
,
”
i
n
2
0
1
8
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
o
c
i
a
l
C
o
m
p
u
t
i
n
g
,
Be
h
a
v
i
o
r
a
l
-
C
u
l
t
u
r
a
l
M
o
d
e
l
i
n
g
,
a
n
d
Pr
e
d
i
c
t
i
o
n
a
n
d
Be
h
a
v
i
o
r
Re
p
res
e
n
t
a
t
i
o
n
i
n
Mo
d
e
l
i
n
g
a
n
d
S
i
m
u
l
a
t
i
o
n
,
BR
i
M
S
2
0
1
8
,
2
0
1
8
,
p
p
.
6
0
–
7
1
.
[
7
]
P
.
N
o
r
d
b
e
r
g
,
J.
K
ä
v
r
e
s
t
a
d
,
a
n
d
M
.
N
o
h
l
b
e
r
g
,
“
A
u
t
o
ma
t
i
c
d
e
t
e
c
t
i
o
n
o
f
f
a
k
e
n
e
w
s
,
”
C
EU
R
W
o
r
k
sh
o
p
Pro
c
e
e
d
i
n
g
s
,
v
o
l
.
2
7
8
9
,
p
p
.
1
6
8
–
1
7
9
,
2
0
2
0
.
[
8
]
S
.
F
e
n
g
,
R
.
B
a
n
e
r
j
e
e
,
a
n
d
Y
.
C
h
o
i
,
“
S
y
n
t
a
c
t
i
c
st
y
l
o
m
e
t
r
y
f
o
r
d
e
c
e
p
t
i
o
n
d
e
t
e
c
t
i
o
n
,
”
i
n
5
0
t
h
An
n
u
a
l
Me
e
t
i
n
g
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
s
t
i
c
s,
AC
L
2
0
1
2
-
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
C
o
n
f
e
re
n
c
e
,
2
0
1
2
,
v
o
l
.
2
,
p
p
.
1
7
1
–
1
7
5
.
[
9
]
Z.
Ji
n
,
J.
C
a
o
,
H
.
G
u
o
,
Y
.
Z
h
a
n
g
,
a
n
d
J.
L
u
o
,
“
M
u
l
t
i
m
o
d
a
l
f
u
si
o
n
w
i
t
h
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
r
u
mo
r
d
e
t
e
c
t
i
o
n
o
n
mi
c
r
o
b
l
o
g
s,
”
i
n
MM
2
0
1
7
-
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
1
7
AC
M
M
u
l
t
i
m
e
d
i
a
C
o
n
f
e
re
n
c
e
,
2
0
1
7
,
p
p
.
7
9
5
–
8
1
6
,
d
o
i
:
1
0
.
1
1
4
5
/
3
1
2
3
2
6
6
.
3
1
2
3
4
5
4
.
[
1
0
]
K
.
S
h
u
,
A
.
S
l
i
v
a
,
S
.
W
a
n
g
,
J.
Ta
n
g
,
a
n
d
H
.
Li
u
,
“
F
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
o
n
s
o
c
i
a
l
m
e
d
i
a
:
a
d
a
t
a
mi
n
i
n
g
p
e
r
s
p
e
c
t
i
v
e
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
I
n
n
o
v
a
t
i
v
e
T
e
c
h
n
o
l
o
g
y
a
n
d
Ex
p
l
o
ri
n
g
E
n
g
i
n
e
e
ri
n
g
,
2
0
2
0
.
[
1
1
]
S
.
Y
a
n
g
,
K
.
S
h
u
,
S
.
W
a
n
g
,
R
.
G
u
,
F
.
W
u
,
a
n
d
H
.
L
i
u
,
“
U
n
su
p
e
r
v
i
s
e
d
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
o
n
s
o
c
i
a
l
me
d
i
a
:
a
g
e
n
e
r
a
t
i
v
e
a
p
p
r
o
a
c
h
,
”
i
n
3
3
rd
AA
AI
C
o
n
f
e
re
n
c
e
o
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
AAAI
2
0
1
9
,
3
1
st
I
n
n
o
v
a
t
i
v
e
A
p
p
l
i
c
a
t
i
o
n
s
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
C
o
n
f
e
re
n
c
e
,
I
AAI
2
0
1
9
a
n
d
t
h
e
9
t
h
AAAI
S
y
m
p
o
s
i
u
m
o
n
E
d
u
c
a
t
i
o
n
a
l
Ad
v
a
n
c
e
s
i
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
EAAI
2
0
1
9
,
2
0
1
9
,
p
p
.
5
6
4
4
–
5
6
5
1
,
d
o
i
:
1
0
.
1
6
0
9
/
a
a
a
i
.
v
3
3
i
0
1
.
3
3
0
1
5
6
4
4
.
[
1
2
]
Y
.
Li
u
a
n
d
Y
.
F
.
B
.
W
u
,
“
Ea
r
l
y
d
e
t
e
c
t
i
o
n
o
f
f
a
k
e
n
e
w
s
o
n
s
o
c
i
a
l
m
e
d
i
a
t
h
r
o
u
g
h
p
r
o
p
a
g
a
t
i
o
n
p
a
t
h
c
l
a
ssi
f
i
c
a
t
i
o
n
w
i
t
h
r
e
c
u
r
r
e
n
t
a
n
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
t
w
o
r
k
s,
”
i
n
3
2
n
d
AAAI
C
o
n
f
e
re
n
c
e
o
n
A
rt
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
A
AAI
2
0
1
8
,
2
0
1
8
,
p
p
.
3
5
4
–
3
6
1
,
d
o
i
:
1
0
.
1
6
0
9
/
a
a
a
i
.
v
3
2
i
1
.
1
1
2
6
8
.
[
1
3
]
J.
Z
h
a
n
g
,
L.
C
u
i
,
Y
.
F
u
,
a
n
d
F
.
B
.
G
o
u
z
a
,
“
F
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
w
i
t
h
d
e
e
p
d
i
f
f
u
s
i
v
e
n
e
t
w
o
r
k
m
o
d
e
l
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
-
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
D
a
t
a
E
n
g
i
n
e
e
ri
n
g
,
2
0
1
8
,
v
o
l
.
2
0
2
0
-
A
p
r
i
l
,
p
p
.
1
8
2
6
–
1
8
2
9
.
[
1
4
]
N
.
J.
C
o
n
r
o
y
,
V
.
L.
R
u
b
i
n
,
a
n
d
Y
.
C
h
e
n
,
“
A
u
t
o
m
a
t
i
c
d
e
c
e
p
t
i
o
n
d
e
t
e
c
t
i
o
n
:
M
e
t
h
o
d
s
f
o
r
f
i
n
d
i
n
g
f
a
k
e
n
e
w
s
,
”
i
n
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
Asso
c
i
a
t
i
o
n
f
o
r
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
2
0
1
5
,
v
o
l
.
5
2
,
n
o
.
1
,
p
p
.
1
–
4
,
d
o
i
:
1
0
.
1
0
0
2
/
p
r
a
2
.
2
0
1
5
.
1
4
5
0
5
2
0
1
0
0
8
2
.
[
1
5
]
K
.
S
h
u
,
L
.
C
u
i
,
S
.
W
a
n
g
,
D
.
L
e
e
,
a
n
d
H
.
L
i
u
,
“
D
e
f
e
n
d
:
e
x
p
l
a
i
n
a
b
l
e
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
AC
M
S
I
G
K
D
D
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
K
n
o
w
l
e
d
g
e
D
i
sc
o
v
e
ry
a
n
d
D
a
t
a
M
i
n
i
n
g
,
2
0
1
9
,
p
p
.
3
9
5
–
4
0
5
,
d
o
i
:
1
0
.
1
1
4
5
/
3
2
9
2
5
0
0
.
3
3
3
0
9
3
5
.
[
1
6
]
F
.
Y
a
n
g
,
X
.
Y
u
,
Y
.
Li
u
,
a
n
d
M
.
Y
a
n
g
,
“
A
u
t
o
m
a
t
i
c
d
e
t
e
c
t
i
o
n
o
f
r
u
mo
r
o
n
S
i
n
a
W
e
i
b
o
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
A
C
M
S
I
G
K
D
D
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
K
n
o
w
l
e
d
g
e
D
i
sc
o
v
e
ry
a
n
d
D
a
t
a
M
i
n
i
n
g
,
2
0
1
2
,
p
p
.
1
–
7
,
d
o
i
:
1
0
.
1
1
4
5
/
2
3
5
0
1
9
0
.
2
3
5
0
2
0
3
.
[
1
7
]
Z.
L
i
n
e
t
a
l
.
,
“
H
i
e
r
a
r
c
h
i
c
a
l
a
t
t
e
n
t
i
o
n
n
e
t
w
o
r
k
s
f
o
r
d
o
c
u
me
n
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
i
n
2
0
1
6
C
o
n
f
e
re
n
c
e
o
f
t
h
e
N
o
rt
h
Am
e
ri
c
a
n
C
h
a
p
t
e
r
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s:
H
u
m
a
n
L
a
n
g
u
a
g
e
T
e
c
h
n
o
l
o
g
i
e
s,
N
AAC
L
H
L
T
2
0
1
6
-
P
ro
c
e
e
d
i
n
g
s
o
f
t
h
e
C
o
n
f
e
re
n
c
e
,
2
0
1
6
,
v
o
l
.
2
0
1
7
-
D
e
c
e
m,
n
o
.
Lr
e
c
,
p
p
.
6
5
4
–
6
5
8
.
[
1
8
]
S
.
S
i
n
g
h
a
n
i
a
,
N
.
F
e
r
n
a
n
d
e
z
,
a
n
d
S
.
R
a
o
,
“
3
H
A
N
:
a
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
F
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
,
”
i
n
L
e
c
t
u
re
N
o
t
e
s
i
n
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
i
n
c
l
u
d
i
n
g
s
u
b
ser
i
e
s
L
e
c
t
u
r
e
N
o
t
e
s
i
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
L
e
c
t
u
re
N
o
t
e
s
i
n
Bi
o
i
n
f
o
rm
a
t
i
c
s)
,
2
0
1
7
,
v
o
l
.
1
0
6
3
5
LN
C
S
,
p
p
.
5
7
2
–
5
8
1
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
3
1
9
-
7
0
0
9
6
-
0
_
5
9
.
[
1
9
]
Y
.
R
e
n
a
n
d
J.
Z
h
a
n
g
,
“
F
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
o
n
n
e
w
s
-
o
r
i
e
n
t
e
d
h
e
t
e
r
o
g
e
n
e
o
u
s
i
n
f
o
r
ma
t
i
o
n
n
e
t
w
o
r
k
s
t
h
r
o
u
g
h
h
i
e
r
a
r
c
h
i
c
a
l
g
r
a
p
h
a
t
t
e
n
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
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
re
n
c
e
o
n
N
e
u
r
a
l
N
e
t
w
o
r
k
s
,
2
0
2
1
,
v
o
l
.
2
0
2
1
-
Ju
l
y
,
p
p
.
1
–
8
,
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
4
3
6
2
.
[
2
0
]
H
.
K
a
r
i
mi
,
P
.
C
.
R
o
y
,
S
.
S
a
b
a
-
S
a
d
i
y
a
,
a
n
d
J.
Ta
n
g
,
“
M
u
l
t
i
-
so
u
r
c
e
mu
l
t
i
-
c
l
a
ss
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
,
”
i
n
C
O
L
I
N
G
2
0
1
8
-
2
7
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s,
Pr
o
c
e
e
d
i
n
g
s
,
2
0
1
8
,
p
p
.
1
5
4
6
–
1
5
5
7
.
[
2
1
]
N
.
R
u
c
h
a
n
s
k
y
,
S
.
S
e
o
,
a
n
d
Y
.
L
i
u
,
“
C
S
I
:
a
h
y
b
r
i
d
d
e
e
p
m
o
d
e
l
f
o
r
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
n
f
o
rm
a
t
i
o
n
a
n
d
K
n
o
w
l
e
d
g
e
M
a
n
a
g
e
m
e
n
t
,
Pro
c
e
e
d
i
n
g
s
,
v
o
l
.
P
a
r
t
F
1
3
1
8
.
p
p
.
7
9
7
–
8
0
6
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
4
5
/
3
1
3
2
8
4
7
.
3
1
3
2
8
7
7
.
[
2
2
]
E.
T
a
c
c
h
i
n
i
,
G
.
B
a
l
l
a
r
i
n
,
M
.
L.
D
e
l
l
a
V
e
d
o
v
a
,
S
.
M
o
r
e
t
,
a
n
d
L
.
d
e
A
l
f
a
r
o
,
“
S
o
m
e
l
i
k
e
i
t
H
o
a
x
:
a
u
t
o
m
a
t
e
d
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
i
n
so
c
i
a
l
n
e
t
w
o
r
k
s
,
”
C
EU
R
W
o
rks
h
o
p
P
ro
c
e
e
d
i
n
g
s
,
v
o
l
.
1
9
6
0
,
2
0
1
7
.
[
2
3
]
U
.
Je
o
n
g
,
K
.
D
i
n
g
,
L
.
C
h
e
n
g
,
R
.
G
u
o
,
K
.
S
h
u
,
a
n
d
H
.
L
i
u
,
“
N
o
t
h
i
n
g
st
a
n
d
s
a
l
o
n
e
:
r
e
l
a
t
i
o
n
a
l
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
w
i
t
h
h
y
p
e
r
g
r
a
p
h
n
e
u
r
a
l
n
e
t
w
o
r
k
s,”
i
n
Pr
o
c
e
e
d
i
n
g
s
-
2
0
2
2
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
B
i
g
D
a
t
a
,
B
i
g
D
a
t
a
2
0
2
2
,
2
0
2
2
,
p
p
.
5
9
6
–
6
0
5
,
d
o
i
:
1
0
.
1
1
0
9
/
B
i
g
D
a
t
a
5
5
6
6
0
.
2
0
2
2
.
1
0
0
2
0
2
3
4
.
[
2
4
]
S
.
B
a
i
,
F
.
Zh
a
n
g
,
a
n
d
P
.
H
.
S
.
T
o
r
r
,
“
H
y
p
e
r
g
r
a
p
h
c
o
n
v
o
l
u
t
i
o
n
a
n
d
h
y
p
e
r
g
r
a
p
h
a
t
t
e
n
t
i
o
n
,
”
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
v
o
l
.
1
1
0
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
a
t
c
o
g
.
2
0
2
0
.
1
0
7
6
3
7
.
[2
5
]
W
.
Y
.
W
a
n
g
,
“
‘
LI
A
R
,
LI
A
R
p
a
n
t
s
o
n
f
i
r
e
’
:
A
n
e
w
b
e
n
c
h
mar
k
d
a
t
a
s
e
t
f
o
r
f
a
k
e
n
e
w
s
d
e
t
e
c
t
i
o
n
,
”
a
r
Xi
v
p
re
p
ri
n
t
a
rXi
v
:
1
7
0
5
.
0
0
6
4
8
,
2
0
1
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
2
0
2
-
2
2
1
0
2210
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Alp
a
n
a
A.
Bo
r
se
re
c
e
iv
e
d
th
e
B.
En
g
.
d
e
g
re
e
in
c
o
m
p
u
ter
e
n
g
in
e
e
rin
g
fro
m
KBTCOE,
Na
sik
,
I
n
d
ia,
in
2
0
0
8
a
n
d
t
h
e
M
.
E
.
a
n
d
P
h
.
D.
d
e
g
re
e
p
e
ru
si
n
g
in
c
o
m
p
u
te
r
e
n
g
in
e
e
rin
g
fro
m
M
CERC,
Na
sik
,
I
n
d
ia.
Cu
rre
n
tl
y
,
s
h
e
is
a
n
a
ss
istan
t
p
r
o
fe
ss
o
r
a
t
th
e
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
o
n
Tec
h
n
o
l
o
g
y
,
P
CCOE
,
P
u
n
e
.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
d
e
e
p
lea
rn
in
g
,
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
s
(CNN
s
),
re
c
u
rre
n
t
n
e
u
ra
l
n
e
two
rk
s
(RNN
s),
a
n
d
g
e
n
e
ra
ti
v
e
a
d
v
e
rsa
rial
n
e
tw
o
rk
s
(
G
A
Ns
),
n
a
t
u
ra
l
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
(NLP
),
tran
sfe
r
lea
rn
in
g
,
n
e
u
ra
l
n
e
two
rk
o
p
ti
m
iza
ti
o
n
,
a
u
to
n
o
m
o
u
s
s
y
ste
m
s,
c
o
m
p
u
ter
v
isio
n
,
p
re
d
icti
v
e
a
n
a
l
y
ti
c
s,
e
x
p
lain
a
b
le
AI (XAI).
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
a
lp
a
n
a
.
b
o
rse
@
g
m
a
il
.
c
o
m
.
G
a
ja
n
a
n
K
h
a
r
a
te
is
th
e
Dire
c
to
r
o
f
M
a
to
sh
ri
C
o
ll
e
g
e
o
f
En
g
i
n
e
e
rin
g
a
n
d
Re
se
a
rc
h
Ce
n
tre,
Ek
la
h
a
re
,
Na
sh
ik
.
He
is
a
c
ti
v
e
ly
in
v
o
l
v
e
d
in
f
o
ste
rin
g
re
se
a
rc
h
a
n
d
in
n
o
v
a
ti
o
n
with
in
th
e
in
stit
u
ti
o
n
.
U
n
d
e
r
h
is
lea
d
e
rsh
ip
,
t
h
e
c
o
ll
e
g
e
h
a
s
e
sta
b
li
sh
e
d
a
n
in
n
o
v
a
ti
o
n
a
n
d
i
n
c
u
b
a
ti
o
n
c
e
n
ter
a
ime
d
a
t
b
ri
d
g
i
n
g
t
h
e
g
a
p
b
e
tw
e
e
n
a
c
a
d
e
m
ic
re
se
a
rc
h
a
n
d
in
d
u
stry
n
e
e
d
s
b
y
p
ro
v
id
in
g
m
o
d
e
rn
to
o
ls
a
n
d
re
so
u
rc
e
s
f
o
r
e
x
p
e
ri
m
e
n
tatio
n
.
Ad
d
it
i
o
n
a
ll
y
,
Dr.
Kh
a
ra
te
h
a
s
g
u
id
e
d
se
v
e
ra
l
stu
d
e
n
ts
to
c
o
m
p
lete
th
e
ir
P
h
.
D.
d
e
g
re
e
s,
c
o
n
tri
b
u
ti
n
g
sig
n
ifi
c
a
n
tl
y
to
th
e
a
c
a
d
e
m
ic
g
ro
wt
h
a
n
d
d
e
v
e
lo
p
m
e
n
t
o
f
h
is
stu
d
e
n
ts.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
g
k
k
h
a
ra
te@
re
d
iffma
il
.
c
o
m
.
Na
m
r
a
t
a
G
.
K
h
a
r
a
te
re
c
e
iv
e
d
a
P
h
.
D
.
d
e
g
re
e
in
c
o
m
p
u
ter
e
n
g
i
n
e
e
rin
g
fr
o
m
S
P
P
U,
I
n
d
ia,
i
n
2
0
2
2
.
S
h
e
h
a
s
b
e
e
n
a
p
ro
fe
ss
o
r
o
f
c
o
m
p
u
ter
e
n
g
in
e
e
rin
g
wi
th
Vish
wa
k
a
rm
a
Un
iv
e
rsity
,
si
n
c
e
2
0
1
7
.
S
h
e
h
a
s
a
u
th
o
re
d
a
n
d
c
o
-
a
u
th
o
re
d
n
u
m
e
ro
u
s
re
se
a
rc
h
p
a
p
e
rs,
j
o
u
r
n
a
l
a
rti
c
les
,
a
n
d
c
o
n
fe
re
n
c
e
p
re
se
n
tatio
n
s.
He
r
wo
rk
o
fte
n
a
d
d
re
ss
e
s
c
o
n
tem
p
o
ra
ry
c
h
a
ll
e
n
g
e
s
i
n
tec
h
n
o
l
o
g
y
a
n
d
e
n
g
in
e
e
rin
g
,
re
flec
ti
n
g
h
e
r
e
x
p
e
rti
se
a
n
d
d
e
d
ica
ti
o
n
to
a
d
v
a
n
c
in
g
h
e
r
f
ield
.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
a
re
a
s
su
c
h
a
s
so
ftwa
re
e
n
g
in
e
e
ri
n
g
,
a
n
d
d
a
ta
a
n
a
ly
ti
c
s.
S
h
e
h
a
s
p
u
b
li
sh
e
d
se
v
e
ra
l
p
a
p
e
rs
i
n
th
e
se
d
o
m
a
in
s,
c
o
n
tri
b
u
ti
n
g
t
o
b
o
t
h
a
c
a
d
e
m
ic
k
n
o
wle
d
g
e
a
n
d
p
ra
c
ti
c
a
l
a
p
p
li
c
a
ti
o
n
s i
n
h
e
r
fiel
d
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
n
a
m
ra
tak
h
a
ra
te1
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.