I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
9
,
No
.
2
,
A
p
r
il
201
9
,
p
p
.
1
3
1
3
~
1
3
2
0
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v9
i
2
.
pp
1
3
1
3
-
1320
1313
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e
.
co
m/
jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JE
C
E
Feature
select
io
n
,
o
pti
m
i
z
a
tion
a
nd
clusteri
ng
strateg
ies
o
f
text
do
cu
m
ents
A.
K
o
us
a
r
Nikh
a
t
h
1
,
K
.
Su
b
ra
h
m
a
ny
a
m
2
1
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter S
c
ien
c
e
a
n
d
E
n
g
in
e
e
rin
g
,
Ko
n
e
ru
L
a
k
sh
a
m
a
iah
Ed
u
c
a
ti
o
n
F
o
u
n
d
a
ti
o
n
,
In
d
ia
2
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter S
c
ien
c
e
a
n
d
E
n
g
in
e
e
rin
g
,
Ko
n
e
ru
La
k
sh
a
m
a
iah
Ed
u
c
a
ti
o
n
F
o
u
n
d
a
ti
o
n
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
2
8
,
2
0
1
7
R
ev
i
s
ed
Sep
1
6
,
2
0
1
8
A
cc
ep
ted
O
ct
1
,
2
0
1
8
Clu
ste
rin
g
is
o
n
e
o
f
th
e
m
o
st
re
s
e
a
rc
h
e
d
a
re
a
s
o
f
d
a
ta
m
in
in
g
a
p
p
li
c
a
ti
o
n
s
in
th
e
c
o
n
tem
p
o
ra
ry
li
tera
tu
re
.
T
h
e
n
e
e
d
f
o
r
e
ff
icie
n
t
c
lu
ste
rin
g
is
o
b
se
rv
e
d
a
c
ro
ss
w
id
e
s
e
c
to
rs
in
c
lu
d
in
g
c
o
n
su
m
e
r
se
g
m
e
n
tatio
n
,
c
a
teg
o
riza
ti
o
n
,
sh
a
re
d
f
il
terin
g
,
d
o
c
u
m
e
n
t
m
a
n
a
g
e
m
e
n
t,
a
n
d
in
d
e
x
in
g
.
T
h
e
re
s
e
a
rc
h
o
f
c
lu
ste
rin
g
tas
k
is
to
b
e
p
e
r
f
o
r
m
e
d
p
rio
r
to
it
s
a
d
a
p
tatio
n
in
th
e
tex
t
e
n
v
iro
n
m
e
n
t.
Co
n
v
e
n
ti
o
n
a
l
a
p
p
r
o
a
c
h
e
s
ty
p
ica
ll
y
e
m
p
h
a
siz
e
d
o
n
th
e
q
u
a
n
ti
tativ
e
in
f
o
rm
a
ti
o
n
w
h
e
re
th
e
se
l
e
c
ted
fe
a
tu
re
s
a
re
n
u
m
b
e
rs.
E
ff
o
rts
a
lso
h
a
v
e
b
e
e
n
p
u
t
f
o
rw
a
rd
f
o
r
a
c
h
iev
in
g
e
ff
ici
e
n
t
c
lu
ste
rin
g
in
th
e
c
o
n
tex
t
o
f
c
a
teg
o
rica
l
in
f
o
rm
a
ti
o
n
w
h
e
re
th
e
s
e
lec
ted
f
e
a
tu
re
s
c
a
n
a
ss
u
m
e
n
o
m
in
a
l
v
a
lu
e
s.
T
h
is
m
a
n
u
sc
rip
t
p
re
se
n
ts
a
n
in
-
d
e
p
th
a
n
a
ly
sis
o
f
c
h
a
ll
e
n
g
e
s
o
f
c
lu
ste
rin
g
in
th
e
tex
t
e
n
v
iro
n
m
e
n
t.
F
u
rth
e
r,
th
is
p
a
p
e
r
a
lso
d
e
tai
ls
p
ro
m
in
e
n
t
m
o
d
e
ls
p
ro
p
o
se
d
f
o
r
c
lu
ste
rin
g
a
lo
n
g
w
it
h
th
e
p
ro
s
a
n
d
c
o
n
s
o
f
e
a
c
h
m
o
d
e
l.
In
a
d
d
it
io
n
,
it
a
lso
f
o
c
u
se
s
o
n
v
a
rio
u
s
late
st
d
e
v
e
lo
p
m
e
n
ts
in
th
e
c
lu
ste
rin
g
tas
k
in
th
e
so
c
ial
n
e
tw
o
rk
a
n
d
a
ss
o
c
iate
d
e
n
v
iro
n
m
e
n
ts.
K
ey
w
o
r
d
s
:
Featu
r
e
ex
tr
ac
tio
n
Featu
r
e
s
elec
t
io
n
Se
m
i
-
s
u
p
er
v
i
s
ed
lear
n
i
n
g
Un
s
u
p
er
v
i
s
ed
lear
n
i
n
g
Co
p
y
rig
h
t
©
2
0
1
9
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
A
.
Ko
u
s
ar
Nik
h
at
h
,
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
i
n
ee
r
in
g
,
Ko
n
er
u
L
ak
s
h
a
m
aia
h
E
d
u
ca
ti
o
n
Fo
u
n
d
atio
n
,
Gu
n
tu
r
-
5
2
2
5
0
2
,
A
P
,
I
n
d
ia
.
E
m
ail:
k
o
u
s
ar
n
i
k
h
a
th
@
v
n
r
v
j
iet.
in
1.
I
NT
RO
D
UCT
I
O
N
C
lu
s
ter
i
n
g
o
f
d
o
cu
m
e
n
t
s
i
s
a
n
e
s
s
e
n
tial
p
r
o
ce
s
s
a
n
d
e
f
f
ici
en
t
al
g
o
r
it
h
m
s
n
ee
d
to
b
e
e
m
p
lo
y
ed
to
en
s
u
r
e
ef
f
ec
ti
v
e
d
o
cu
m
e
n
t
clu
s
ter
in
g
.
T
h
e
p
r
o
ce
s
s
o
f
clu
s
te
r
in
g
i
n
cl
u
d
es
ca
teg
o
r
izatio
n
o
f
g
i
v
e
n
d
o
cu
m
e
n
ts
in
to
i
n
d
iv
id
u
al
g
r
o
u
p
s
.
T
h
ese
clu
s
ter
s
s
h
o
u
ld
b
e
m
ea
n
in
g
f
u
l
an
d
p
r
o
v
id
e
r
i
g
h
t
d
escr
ip
tio
n
o
f
th
e
d
o
cu
m
en
t
s
.
Ho
w
e
v
er
,
f
o
r
an
e
f
f
icie
n
t
cl
u
s
ter
in
g
,
r
esear
ch
er
s
o
f
te
n
f
ac
e
th
e
c
h
alle
n
g
e
o
f
co
m
p
le
x
it
y
i
n
ter
m
s
o
f
lar
g
e
n
u
m
b
er
o
f
w
o
r
d
s
.
W
h
en
th
e
clu
s
ter
in
g
is
d
o
n
e
in
th
e
f
o
r
m
o
f
m
atr
ices,
ea
ch
an
d
ev
er
y
d
o
cu
m
e
n
t
i
s
co
n
s
id
er
ed
as
an
in
s
ta
n
ce
an
d
all
th
e
ter
m
s
ass
o
ciate
d
w
ill
b
e
f
ea
tu
r
es.
I
n
g
e
n
er
al,
th
e
v
o
lu
m
e
o
f
f
ea
t
u
r
es
is
al
m
o
s
t
e
q
u
al
to
a
d
ictio
n
ar
y
,
p
o
s
in
g
s
tr
o
n
g
c
h
alle
n
g
e
s
f
o
r
alg
o
r
ith
m
d
e
v
elo
p
er
s
.
T
h
e
clu
s
ter
i
n
g
al
g
o
r
ith
m
ef
f
icien
c
y
lar
g
el
y
f
l
u
ctu
a
tes
w
it
h
in
cr
ea
s
in
g
n
u
m
b
er
o
f
w
o
r
d
s
.
A
cc
o
r
d
in
g
l
y
,
r
esear
ch
er
s
s
ea
r
ch
f
o
r
n
o
n
-
co
n
tex
t
r
elate
d
,
r
ed
u
n
d
an
t
wo
r
d
s
an
d
s
to
p
w
o
r
d
s
a
n
d
at
t
e
m
p
t
to
ig
n
o
r
e
o
r
r
e
m
o
v
e
s
u
ch
w
o
r
d
s
to
b
o
o
s
t
ef
f
icien
c
y
o
f
al
g
o
r
ith
m
.
Do
cu
m
e
n
t
cl
u
s
ter
i
n
g
co
n
tai
n
s
p
ar
ticu
lar
m
et
h
o
d
s
an
d
alg
o
r
ith
m
s
b
u
il
t
o
n
u
n
s
u
p
er
v
is
ed
d
o
cu
m
en
t
m
an
a
g
e
m
e
n
t
[
1
]
.
I
n
clu
s
ter
in
g
th
e
as
s
ets,
m
e
m
b
er
s
h
ip
s
,
a
n
d
n
u
m
b
er
o
f
t
h
e
clas
s
es
n
o
t
r
ec
o
g
n
i
ze
d
i
n
ad
v
a
n
ce
.
Do
cu
m
e
n
t
s
ca
n
g
r
o
u
p
to
g
et
h
er
b
u
ilt
o
n
an
e
x
ac
t
t
y
p
e,
s
u
ch
a
s
leg
al,
ec
o
n
o
m
ic,
an
d
m
ed
ical.
Ma
ch
i
n
e
lear
n
in
g
alg
o
r
it
h
m
s
h
a
v
e
b
ec
o
m
e
p
r
ev
a
len
t
i
n
n
u
m
er
o
u
s
d
o
m
ai
n
s
,
i
m
p
ac
ti
n
g
a
w
id
e
d
i
v
er
s
it
y
o
f
ap
p
licatio
n
s
.
I
n
th
e
p
ast
p
er
io
d
s
,
th
e
m
ac
h
i
n
e
-
l
ea
r
n
i
n
g
co
m
m
u
n
it
y
h
as
el
ab
o
r
ated
to
d
ec
r
ea
s
e
th
e
lab
elin
g
w
o
r
k
d
o
n
e
b
y
th
e
h
u
m
a
n
f
o
r
s
u
p
er
v
i
s
ed
m
ac
h
in
e
lear
n
in
g
p
r
o
ce
d
u
r
es
o
r
to
d
ev
elo
p
u
n
s
u
p
er
v
is
ed
lear
n
in
g
w
it
h
o
n
l
y
s
m
al
lest
s
u
p
er
v
is
io
n
.
Nev
er
t
h
eless
,
t
h
er
e
ar
e
s
till
s
e
v
er
al
ca
s
es
t
h
at
n
eit
h
er
s
e
m
i
-
su
p
er
v
is
ed
lear
n
in
g
n
o
r
tr
an
s
f
er
lear
n
i
n
g
ca
n
h
elp
.
P
r
o
v
id
en
tiall
y
,
w
it
h
t
h
e
p
r
o
p
ag
atio
n
o
f
g
en
er
a
l
-
d
eter
m
i
n
atio
n
k
n
o
w
led
g
e
b
ase
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
2
,
A
p
r
il 2
0
1
9
:
1
3
1
3
-
1320
1314
(
o
r
k
n
o
w
led
g
e
g
r
ap
h
s
)
,
e.
g
.
,
W
ik
iT
ax
o
n
o
m
y
W
ik
ip
ed
ia,
F
r
ee
b
ase
,
P
r
o
b
ase
,
T
ex
tR
u
n
n
er
,
DB
p
ed
ia
,
NE
L
L
an
d
Kn
o
w
led
g
e
Va
u
lt
,
w
e
h
av
e
p
r
o
f
u
s
io
n
o
f
a
v
ailab
le
w
o
r
ld
k
n
o
w
led
g
e.
W
e
ca
ll
th
ese
k
n
o
w
led
g
e
b
ase
s
w
o
r
ld
k
n
o
w
led
g
e
.
T
h
e
co
n
ce
p
t
o
f
r
ep
r
esen
tin
g
t
h
e
g
i
v
en
d
o
cu
m
en
t
b
y
th
e
g
r
o
u
p
o
f
wo
r
d
s
in
clu
d
ed
in
th
e
d
o
cu
m
en
t
is
u
n
d
er
l
y
in
g
m
an
y
tex
t
m
in
i
n
g
s
t
u
d
ies.
O
f
ten
r
e
f
er
r
ed
to
as
b
ag
-
of
-
w
o
r
d
s
,
t
h
e
co
n
ce
p
t
r
eq
u
ir
es
ac
cu
r
ate
d
escr
ip
tio
n
o
f
th
e
wo
r
d
p
o
s
itio
n
in
t
h
e
g
iv
e
n
d
o
cu
m
e
n
t.
A
cc
o
r
d
in
g
l
y
,
r
esear
ch
er
s
o
p
t
f
o
r
v
ec
to
r
r
ep
r
esen
tatio
n
o
f
t
h
e
w
o
r
d
p
o
s
itio
n
a
n
d
as
s
i
g
n
an
„
i
m
p
o
r
tan
c
e‟
v
al
u
e
to
ea
c
h
w
o
r
d
.
T
h
e
v
e
cto
r
s
p
ac
e
m
o
d
el
i
s
v
er
s
atile
b
ec
au
s
e
v
ec
t
o
r
r
ep
r
esen
tatio
n
ca
n
u
s
e
a
s
a
f
e
atu
r
e
v
ec
to
r
f
o
r
a
lar
g
e
n
u
m
b
er
o
f
cl
u
s
ter
in
g
alg
o
r
ith
m
s
.
T
h
e
v
ec
to
r
-
b
ased
d
o
cu
m
e
n
t
m
o
d
el
s
d
o
n
o
t
h
a
v
e
t
h
e
in
f
o
r
m
a
tio
n
ab
o
u
t
t
h
e
o
r
d
er
b
y
w
h
ic
h
t
h
e
w
o
r
d
s
o
cc
u
r
in
a
d
o
cu
m
e
n
t.
I
n
p
r
ev
io
u
s
ar
ticle
s
,
r
esear
ch
er
s
d
ev
elo
p
ed
a
m
u
ch
-
ad
v
a
n
ce
d
d
o
cu
m
e
n
t
m
o
d
el
ter
m
ed
„
ST
D
m
o
d
el
‟
.
T
h
e
ap
p
r
o
ac
h
is
b
ased
o
n
s
to
r
in
g
co
m
p
lete
w
o
r
d
s
eq
u
e
n
ce
d
ata.
Ov
er
lap
p
in
g
b
et
w
ee
n
s
tr
in
g
s
i
n
th
e
co
m
b
i
n
ed
s
u
f
f
i
x
tr
ee
is
u
s
ed
to
r
ep
r
esen
t
th
e
d
o
cu
m
e
n
t
s
i
m
ilar
i
t
y
.
A
n
o
v
el
m
o
d
el
r
ely
i
n
g
o
n
lin
ea
r
co
n
v
ex
m
i
x
o
f
d
o
cu
m
en
ts
i
s
s
t
u
d
ied
b
y
r
esear
c
h
er
s
in
.
T
o
en
ab
le
f
ea
tu
r
e
b
asi
s
as
th
is
m
ix
tu
r
e,
co
n
v
e
x
-
NM
F
ap
p
r
o
ac
h
is
p
r
o
p
o
s
ed
.
T
h
e
m
o
d
el
also
attain
ed
s
i
m
ilar
f
ac
to
r
izatio
n
a
s
attain
ed
b
y
C
F
f
ac
to
r
izatio
n
ap
p
r
o
ac
h
.
2.
T
AXO
NO
M
Y
2
.
1
.
Sub
s
ec
t
io
n
1
f
ea
t
ure
ex
t
ra
ct
io
n
Feat
u
r
e
E
x
tr
ac
tio
n
(
FE)
p
r
o
ce
s
s
is
ca
teg
o
r
ized
in
to
th
r
ee
t
y
p
es
in
clu
d
i
n
g
S
y
n
tactica
l,
Se
m
an
tic
an
d
Mo
r
p
h
o
lo
g
ical
An
al
y
s
i
s
.
O
f
t
h
ese,
M
A
i
s
p
r
i
m
ar
il
y
e
n
g
ag
e
d
in
d
ea
lin
g
w
it
h
ea
ch
a
n
d
ev
er
y
w
o
r
d
(
in
d
i
v
id
u
al
w
o
r
d
s
)
o
f
t
h
e
g
i
v
en
te
x
t
d
o
cu
m
e
n
t.
P
r
ed
o
m
i
n
an
t
l
y
,
it
c
o
m
p
r
i
s
es
to
k
e
n
izatio
n
,
s
to
p
wo
r
d
eli
m
in
a
tio
n
an
d
s
te
m
m
i
n
g
[
2
]
.
I
n
to
k
en
izatio
n
p
r
o
ce
s
s
,
th
e
tex
t
d
o
cu
m
en
t
i
s
o
f
ten
co
n
s
id
er
ed
as
w
o
r
d
s
tr
in
g
s
w
h
ic
h
ar
e
w
o
r
d
s
eq
u
en
ce
s
a
n
d
d
iv
id
es
t
h
e
m
b
y
eli
m
i
n
ati
n
g
p
u
n
ct
u
atio
n
s
[
3
].
T
h
e
r
esear
ch
er
s
in
[
4
]
atte
m
p
ted
to
u
n
d
er
s
tan
d
th
e
ex
ac
t
lo
g
ic
r
ep
r
esen
ted
b
y
a
p
ar
ticu
lar
s
e
n
te
n
ce
.
T
h
at
is
,
a
s
en
ten
ce
s
h
o
u
ld
h
a
v
e
p
r
o
p
er
g
r
am
m
at
ical
co
n
n
ec
ti
v
es.
S
A
ca
ter
s
u
n
d
er
s
tan
d
in
g
o
f
th
e
g
r
a
m
m
atica
l
ar
r
an
g
e
m
e
n
t
o
f
a
ce
r
tain
lan
g
u
ag
e,
o
f
ten
r
ef
er
r
ed
to
as
“
s
y
n
tax
”.
F
u
r
th
er
,
P
OS
T
ag
g
i
n
g
p
r
o
ce
s
s
allo
w
s
ad
d
in
g
o
f
co
n
te
x
t
u
al
g
r
a
m
m
ar
k
n
o
w
l
ed
g
e
f
o
r
a
s
p
ec
i
f
ic
w
o
r
d
in
t
h
e
g
iv
e
n
s
en
te
n
ce
.
B
y
id
e
n
ti
f
y
in
g
t
h
e
o
p
en
wo
r
d
class
,
lin
g
u
i
s
tic
a
n
al
y
s
i
s
ca
n
b
e
p
er
f
o
r
m
ed
ea
s
il
y
[
5
]
.
Nu
m
er
o
u
s
ap
p
r
o
ac
h
es
w
er
e
p
r
o
p
o
s
ed
in
s
cien
tific
liter
atu
r
e
ai
m
i
n
g
to
i
m
p
le
m
en
t
P
OS
T
ag
g
i
n
g
p
r
o
ce
s
s
d
ep
en
d
in
g
o
n
t
h
e
d
icti
o
n
ar
ies [
6
]
.
2
.
1
.
1
.
F
ea
t
ure
s
elec
t
io
n
A
f
ea
t
u
r
e
r
ef
er
s
to
an
in
d
i
v
id
u
al
m
ea
s
u
r
ab
le
p
r
o
p
er
ty
o
f
a
p
r
o
ce
s
s
,
w
h
ic
h
is
b
ein
g
o
b
s
er
v
ed
.
T
h
r
o
u
g
h
th
e
u
s
e
o
f
a
s
et
o
f
f
e
atu
r
es,
an
y
m
ac
h
i
n
e
lear
n
i
n
g
alg
o
r
ith
m
is
ca
p
ab
le
o
f
p
er
f
o
r
m
i
n
g
cla
s
s
i
f
icatio
n
.
Ov
er
th
e
p
ast
y
ea
r
s
i
n
th
e
ap
p
licatio
n
s
o
f
p
atter
n
r
ec
o
g
n
it
io
n
o
r
m
ac
h
i
n
e
lear
n
i
n
g
,
th
e
d
o
m
ai
n
o
f
f
ea
t
u
r
es
h
as
g
en
er
all
y
ex
te
n
d
ed
f
r
o
m
ten
s
to
h
u
n
d
r
ed
s
o
f
f
ea
t
u
r
es
o
r
v
ar
iab
les
w
h
ic
h
ar
e
em
p
lo
y
ed
in
th
o
s
e
ap
p
licatio
n
s
.
Nu
m
er
o
u
s
tec
h
n
iq
u
e
s
h
av
e
b
e
en
i
n
v
e
n
ted
s
o
a
s
to
ef
f
ec
ti
v
el
y
ad
d
r
ess
t
h
e
p
r
o
b
le
m
o
f
r
ed
u
cin
g
ir
r
elev
a
n
t,
a
s
w
ell
as
r
ed
u
n
d
an
t
v
ar
iab
les
th
at
ar
e
a
b
u
r
d
e
n
o
n
ch
a
llen
g
i
n
g
ta
s
k
s
[
7
]
.
I
t
i
s
i
m
p
er
ativ
e
t
h
at
Feat
u
r
e
Selec
tio
n
(
v
ar
iab
le
eli
m
i
n
atio
n
)
is
h
i
g
h
l
y
b
en
e
f
icial
i
n
u
n
d
er
s
ta
n
d
i
n
g
d
ata,
m
i
n
i
m
izi
n
g
co
m
p
u
t
atio
n
r
eq
u
ir
e
m
e
n
t,
m
i
n
i
m
izi
n
g
t
h
e
ef
f
ec
t o
f
cu
r
s
e
o
f
d
i
m
en
s
io
n
al
it
y
b
es
id
es e
n
h
an
cin
g
t
h
e
p
r
ed
icto
r
p
er
f
o
r
m
a
n
ce
.
2
.
1
.
2
.
F
ilte
r
m
et
ho
d
s
Fil
ter
tech
n
iq
u
e
s
u
s
e
v
ar
iab
l
e
r
an
k
i
n
g
ap
p
r
o
ac
h
es
as
t
h
e
m
ain
s
ta
n
d
ar
d
s
f
o
r
v
ar
iab
le
s
elec
tio
n
th
r
o
u
g
h
o
r
d
er
in
g
.
R
a
n
k
in
g
te
ch
n
iq
u
es
ar
e
e
m
p
lo
y
ed
b
ec
au
s
e
o
f
t
h
eir
s
i
m
p
lic
it
y
.
A
t
t
h
e
s
a
m
e
ti
m
e,
g
o
o
d
s
u
cc
e
s
s
is
o
f
te
n
r
ep
o
r
ted
f
o
r
p
r
ac
tical
ap
p
licatio
n
s
.
A
h
i
g
h
l
y
ap
p
r
o
p
r
iate
r
an
k
in
g
p
r
in
ci
p
le
is
e
m
p
lo
y
ed
i
n
s
co
r
in
g
th
e
v
ar
iab
les.
Ag
ai
n
,
a
th
r
es
h
o
ld
is
o
f
te
n
e
m
p
lo
y
ed
f
o
r
th
e
r
e
m
o
v
al
o
f
v
ar
iab
les
b
elo
w
t
h
e
t
h
r
es
h
o
ld
.
R
an
k
i
n
g
tec
h
n
iq
u
es
ar
e
f
i
lter
m
et
h
o
d
s
b
ec
au
s
e
t
h
e
y
ar
e
u
s
ed
p
r
io
r
to
clas
s
if
icatio
n
f
o
r
f
ilter
i
n
g
o
u
t
t
h
e
v
ar
iab
les,
w
h
ich
ar
e
le
s
s
r
el
ev
an
t.
A
s
i
m
p
le
p
r
o
p
er
ty
o
f
a
u
n
iq
u
e
f
ea
tu
r
e
i
s
to
h
a
v
e
h
ig
h
l
y
b
en
e
f
icia
l
in
f
o
r
m
atio
n
r
e
g
ar
d
in
g
th
e
d
i
v
e
r
s
e
class
es i
n
t
h
e
g
i
v
e
n
d
ata.
2
.
1
.
3
.
Wra
pp
er
m
et
ho
ds
W
r
ap
p
e
r
tech
n
iq
u
es
g
e
n
er
all
y
e
m
p
lo
y
th
e
p
r
ed
icto
r
as
a
b
lack
b
o
x
a
n
d
t
h
e
p
r
ed
icto
r
p
r
esen
tatio
n
a
s
o
b
j
ec
tiv
e
f
u
n
ctio
n
f
o
r
t
h
e
e
v
a
l
u
atio
n
o
f
t
h
e
v
ar
iab
le
s
u
b
s
et.
B
ec
au
s
e
t
h
e
e
v
alu
at
io
n
o
f
2
N
s
u
b
s
et
s
h
as
b
ec
o
m
e
an
NP
-
h
ar
d
p
r
o
b
lem
,
s
u
b
o
p
tim
al
s
u
b
s
et
s
ca
n
b
e
g
o
t
t
h
r
o
u
g
h
t
h
e
u
s
e
o
f
s
ea
r
ch
al
g
o
r
ith
m
s
,
w
h
ic
h
f
in
d
a
s
u
b
s
et
h
eu
r
i
s
ticall
y
.
N
u
m
er
o
u
s
s
ea
r
ch
alg
o
r
it
h
m
s
m
a
y
b
e
ad
o
p
ted
f
o
r
f
in
d
i
n
g
a
s
u
b
s
et
o
f
v
ar
iab
les,
w
h
ic
h
m
ax
i
m
izes
th
e
o
b
j
ec
tiv
e
f
u
n
ctio
n
t
h
at
i
s
t
h
e
c
lass
if
icatio
n
p
r
ese
n
tatio
n
[
8
].
W
e
g
en
er
all
y
ca
te
g
o
r
ize
t
h
e
W
r
ap
p
e
r
tech
n
iq
u
es
in
to
Seq
u
en
tial
Selectio
n
A
l
g
o
r
ith
m
s
,
a
s
w
ell
a
s
He
u
r
is
t
ic
Sear
c
h
A
l
g
o
r
ith
m
s
.
Seq
u
e
n
tia
l
s
elec
tio
n
al
g
o
r
it
h
m
s
co
m
m
e
n
ce
w
it
h
a
n
e
m
p
t
y
s
e
t
(
f
u
ll
s
e
t
)
.
I
t
th
er
ea
f
ter
ad
d
s
f
ea
t
u
r
es
(
r
e
m
o
v
e
f
ea
t
u
r
es)
u
p
to
th
e
p
o
in
t o
f
ac
h
iev
e
m
e
n
t o
f
m
ax
i
m
u
m
o
b
j
ec
tiv
e
f
u
n
ctio
n
.
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:
2
0
8
8
-
8708
F
ea
tu
r
e
s
elec
tio
n
,
o
p
timiz
a
tio
n
a
n
d
clu
s
teri
n
g
s
tr
a
teg
ies o
f te
xt
d
o
cu
men
ts
(
A
.
Ko
u
s
ar
Nik
h
ath
)
1315
2
.
1
.
4
.
E
m
be
dd
ed
m
et
ho
ds
T
h
e
e
m
b
ed
d
ed
m
et
h
o
d
s
ai
m
t
o
p
er
f
o
r
m
f
ea
t
u
r
e
s
elec
tio
n
t
h
r
o
u
g
h
o
u
t
t
h
e
tr
ai
n
i
n
g
p
r
o
ce
d
u
r
e
an
d
ar
e
ess
e
n
tial
a
n
d
d
is
tin
c
t
to
th
e
v
ar
io
u
s
m
ac
h
i
n
e
lear
n
i
n
g
alg
o
r
ith
m
s
i
m
p
le
m
e
n
ted
.
E
m
b
ed
d
ed
tech
n
iq
u
e
s
[
9
]
w
a
n
t
to
m
i
n
i
m
ize
t
h
e
co
m
p
u
t
atio
n
ti
m
e
w
h
ic
h
is
ta
k
e
n
u
p
in
t
h
e
r
ec
las
s
i
f
icatio
n
o
f
d
iv
er
s
e
s
u
b
s
e
ts
t
h
at
i
s
d
o
n
e
in
w
r
ap
p
er
tech
n
iq
u
es.
T
h
e
m
aj
o
r
ap
p
r
o
ac
h
en
tails
t
h
e
i
n
co
r
p
o
r
atio
n
o
f
th
e
f
ea
t
u
r
e
s
elec
tio
n
as
a
n
ele
m
e
n
t o
f
t
h
e
p
r
o
ce
s
s
o
f
tr
ain
in
g
.
2
.
1
.
5
.
H
y
brid a
pp
ro
a
ches
T
h
e
ap
p
r
o
ac
h
co
m
b
i
n
es
f
ilter
,
as
w
el
l
as
t
h
e
w
r
ap
p
er
-
b
ase
d
tech
n
iq
u
es.
Fil
ter
ap
p
r
o
ac
h
s
elec
ts
a
clu
s
ter
o
f
ca
n
d
id
ate
f
ea
t
u
r
es
f
r
o
m
h
ig
h
d
i
m
en
s
io
n
al
a
n
d
ef
f
icien
t
o
r
ig
i
n
al
f
ea
t
u
r
e
s
et.
T
h
en
,
b
y
u
tili
z
in
g
a
w
r
ap
p
er
tech
n
iq
u
e,
th
is
ca
n
d
id
ate
f
ea
tu
r
e
s
et
w
ill
b
e
r
ef
i
n
ed
.
I
t
g
en
er
all
y
ex
p
lo
it
s
th
e
v
ar
io
u
s
k
i
n
d
s
o
f
ad
v
an
ta
g
es
w
h
ic
h
ar
e
b
r
o
u
g
h
t a
b
o
u
t b
y
t
h
e
u
s
e
o
f
t
h
e
t
w
o
m
eth
o
d
s
.
Featu
r
e
s
elec
tio
n
[
5
]
g
en
er
all
y
p
la
y
s
h
u
g
e
r
o
le
in
t
h
e
d
etec
tio
n
o
f
th
e
an
o
m
alie
s
o
f
n
et
w
o
r
k
s
.
I
n
t
h
e
a
n
o
m
al
y
b
ase
d
d
etec
tio
n
s
y
s
te
m
s
,
b
y
m
o
n
ito
r
i
n
g
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
r
eg
u
lar
d
ata
th
o
r
o
u
g
h
l
y
in
co
n
tr
a
s
t
w
it
h
th
e
o
n
es
w
h
ic
h
ar
e
ir
r
eg
u
lar
,
in
co
n
s
i
s
te
n
c
y
w
i
ll
b
e
id
en
tifie
d
w
it
h
in
th
e
n
et
wo
r
k
.
T
h
u
s
,
t
h
is
k
i
n
d
o
f
d
etec
t
io
n
s
y
s
te
m
w
ill
p
la
y
a
v
ital
r
o
le
in
r
ec
o
g
n
izi
n
g
v
ar
io
u
s
i
n
tr
u
s
io
n
s
d
ep
en
d
in
g
o
n
th
e
d
is
ti
n
ct
c
h
ar
ac
ter
is
tic
s
o
f
n
et
w
o
r
k
tr
a
f
f
ic.
2
.
2
.
Si
m
ila
rit
y
m
ea
s
ure
s
P
r
io
r
to
clu
s
ter
in
g
,
t
h
er
e
is
t
h
e
n
ee
d
f
o
r
th
e
d
eter
m
i
n
atio
n
o
f
a
s
i
m
ilar
it
y
o
r
a
d
is
tan
c
e
m
ea
s
u
r
e.
Gen
er
all
y
,
t
h
e
m
ea
s
u
r
e
r
e
f
le
cts
t
h
e
p
r
o
x
i
m
it
y
o
f
th
e
tar
g
eted
o
b
j
ec
ts
o
r
t
h
e
d
eg
r
ee
o
f
v
ec
to
r
s
ep
ar
atio
n
.
I
t
s
h
o
u
ld
r
elate
d
i
f
f
er
en
t
ch
ar
ac
ter
is
tics
u
s
ed
to
s
ep
ar
at
e
th
e
clu
s
ter
s
.
I
n
s
ev
er
al
ci
r
cu
m
s
tan
ce
s
,
t
h
ese
ch
ar
ac
ter
is
tic
s
v
ar
y
in
ac
co
r
d
an
ce
w
i
th
d
ata
an
d
ca
n
also
d
ep
en
d
o
n
t
h
e
p
r
o
b
le
m
co
n
te
x
t
.
Ho
w
e
v
er
,
as
ea
ch
clu
s
ter
i
n
g
p
r
o
b
le
m
d
if
f
er
s
f
r
o
m
o
th
er
,
n
o
s
u
ch
m
ea
s
u
r
e
is
e
x
is
t
in
g
to
s
a
tis
f
y
ev
er
y
k
i
n
d
o
f
clu
s
ter
in
g
p
r
o
b
le
m
.
Fu
r
t
h
er
,
s
elec
ti
n
g
a
n
ap
p
r
o
p
r
i
ate
s
i
m
ilar
it
y
m
ea
s
u
r
e
w
ill
b
e
a
k
e
y
d
r
iv
er
in
C
l
u
s
ter
An
al
y
s
i
s
,
p
r
ed
o
m
i
n
an
tl
y
f
o
r
s
p
ec
if
ied
clu
s
ter
i
n
g
m
o
d
el
s
[
10
]
.
T
h
u
s
,
r
ea
lizin
g
th
e
s
i
g
n
i
f
ican
ce
a
n
d
ef
f
ic
ien
c
y
o
f
v
a
r
io
u
s
m
ea
s
u
r
es
w
ill
s
u
p
p
o
r
t
th
e
s
elec
tio
n
o
f
th
e
m
o
s
t
s
u
itab
le
o
n
e.
T
h
is
v
al
u
e
in
-
tu
r
n
r
elies
o
n
t
w
o
d
is
ti
n
ct
f
ac
to
r
s
s
u
ch
a
s
t
h
e
p
r
o
p
er
ties
o
f
b
o
th
o
b
j
ec
ts
an
d
o
n
th
e
m
ea
s
u
r
e
m
e
n
t
m
etr
ic
s
.
T
h
e
f
iv
e
m
ea
s
u
r
es
h
a
v
e
b
ee
n
d
is
cu
s
s
ed
b
elo
w
.
T
h
e
d
if
f
er
en
t
m
ea
s
u
r
e
b
r
in
g
s
ab
o
u
t
d
if
f
er
e
n
t
f
in
a
l
p
ar
titi
o
n
.
A
t
t
h
e
s
a
m
e
ti
m
e,
it
als
o
im
p
o
s
e
s
d
iv
er
s
e
r
eq
u
ir
e
m
en
ts
f
o
r
s
i
m
i
lar
clu
s
te
r
in
g
al
g
o
r
ith
m
.
2
.
2
.
1
.
E
ucli
dea
n d
is
t
a
nce
E
u
clid
ea
n
d
is
ta
n
ce
r
ef
er
s
to
a
s
tan
d
ar
d
m
etr
ic
u
s
ed
f
o
r
g
eo
m
e
tr
i
ca
l
p
r
o
b
lem
s
.
A
t
th
e
s
a
m
e
ti
m
e,
it
ca
n
b
e
d
ef
i
n
ed
as
t
h
e
o
r
d
in
ar
y
d
i
s
tan
ce
b
et
w
ee
n
t
w
o
p
o
in
ts
.
Me
a
s
u
r
i
n
g
it
ca
n
ea
s
il
y
b
e
d
o
n
e
th
r
o
u
g
h
t
h
e
u
s
e
o
f
a
r
u
ler
in
t
w
o
-
o
r
in
th
r
ee
-
d
i
m
e
n
s
io
n
al
s
p
ac
e.
I
n
ad
d
itio
n
,
it
is
also
o
b
s
er
v
ed
th
at
E
u
clid
ea
n
d
is
ta
n
ce
w
il
l a
ls
o
b
e
s
elec
ted
in
cl
u
s
ter
in
g
p
r
o
b
le
m
s
,
w
h
ic
h
co
m
p
r
is
e
s
clu
s
ter
in
g
te
x
t.
I
t
is
s
ati
s
f
y
i
n
g
all
t
h
e
f
o
u
r
m
a
in
co
n
d
itio
n
s
w
h
ic
h
h
a
v
e
b
ee
n
g
i
v
en
ab
o
v
e
a
n
d
a
s
a
r
esu
lt,
it
i
s
a
tr
u
e
m
etr
ic.
A
t
th
e
s
a
m
e
ti
m
e,
it
i
s
th
e
d
ef
a
u
lt
d
is
ta
n
ce
m
ea
s
u
r
e
th
at
is
u
s
ed
w
it
h
k
-
m
ea
n
s
a
lg
o
r
ith
m
.
R
e
s
o
lv
i
n
g
th
e
d
i
s
tan
ce
m
ea
s
u
r
e
b
et
w
ee
n
tex
t
d
o
cu
m
en
ts
x
d
a
n
d
y
d
w
ill
b
e
d
en
o
ted
b
y
th
e
ir
r
esp
ec
ti
v
e
t
er
m
v
ec
to
r
s
ca
lled
x
t
an
d
y
t
.
Hen
ce
,
th
e
E
u
clid
ea
n
m
etr
ic
o
f
th
e
s
e
t
w
o
d
o
cu
m
en
ts
co
u
ld
b
e
d
ef
i
n
ed
as:
1
/
2
2
,,
1
,
n
E
x
y
t
x
t
y
t
D
t
t
w
w
,
(
1
)
I
n
w
h
ic
h
th
e
ter
m
s
et
is
1
{
,
.
.
.
.
}
n
T
t
t
.
As
d
is
cu
s
s
ed
in
ab
o
v
e
s
ec
tio
n
,
t
f
i
d
f
v
a
lu
e
ca
n
b
e
co
n
s
id
er
ed
as
ter
m
w
ei
g
h
ts
,
i.e
.
,
,
(
,
)
t
x
x
w
t
f
i
d
f
d
t
.
2
.
2
.
2
.
Co
s
ine s
i
m
ila
rit
y
As
p
o
in
ted
ab
o
v
e,
th
e
tex
t
d
o
cu
m
e
n
t
s
ar
e
in
d
icate
d
as
ter
m
v
ec
to
r
s
.
I
n
th
is
s
ce
n
ar
io
,
th
e
s
i
m
ilar
it
y
m
ea
s
u
r
e
b
et
w
ee
n
2
tex
t
d
o
cu
m
en
ts
i
m
p
l
ies
th
e
as
s
o
ciatio
n
in
b
et
w
ee
n
th
e
s
elec
ted
v
ec
to
r
s
.
I
n
g
e
n
er
al,
t
h
i
s
i
s
ev
alu
a
ted
as
t
h
e
C
o
s
i
n
e
f
u
n
ct
io
n
s
b
et
w
ee
n
g
i
v
e
n
ter
m
v
ec
t
o
r
s
an
d
is
ca
lled
C
o
s
in
e
Si
m
ilar
it
y
.
I
t
is
w
o
r
t
h
p
o
in
tin
g
o
u
t t
h
at
co
s
i
n
e
s
i
m
ila
r
it
y
f
o
r
m
s
p
ar
t o
f
t
h
e
m
o
s
t p
o
p
u
lar
m
ea
s
u
r
e
o
f
s
i
m
ilar
it
y
th
at
is
u
s
ed
i
n
o
r
d
er
to
tex
t d
o
cu
m
en
t
s
.
T
h
e
C
o
s
in
e
Si
m
ilar
it
y
(
C
S)
m
ea
s
u
r
e
f
o
r
x
t
d
o
cu
m
e
n
t a
n
d
y
t
d
o
c
u
m
e
n
t i
s
d
ep
icted
:
.
,
xy
Z
x
y
xy
tt
S
IM
t
t
tt
,
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
2
,
A
p
r
il 2
0
1
9
:
1
3
1
3
-
1320
1316
W
h
er
e
x
t
an
d
y
t
ar
e
ca
lled
m
u
ltid
i
m
en
s
io
n
a
l
v
ec
to
r
s
o
f
th
e
Vec
t
o
r
-
ter
m
s
et
1
{
,
.
.
.
.
}
n
T
t
t
.
E
ac
h
d
im
en
s
io
n
co
n
tain
s
it
s
o
w
n
w
eig
h
t
an
d
co
r
r
esp
o
n
d
s
to
a
ter
m
s
e
t.
T
h
e
v
al
u
e
o
f
t
h
ese
d
i
m
en
s
io
n
s
i
s
al
w
a
y
s
m
o
r
e
th
a
n
ze
r
o
.
Hen
ce
,
th
e
C
S
h
o
ld
s
p
o
s
itiv
e
v
alu
e
s
an
d
w
il
l a
l
w
a
y
s
b
e
b
o
u
n
d
b
et
w
ee
n
[
0
,
1
]
.
A
n
o
te
w
o
r
th
y
p
r
o
p
er
ty
o
f
th
i
s
k
i
n
d
o
f
s
i
m
ilar
it
y
is
t
h
at
i
t
is
in
d
ep
en
d
e
n
t
o
f
d
o
cu
m
e
n
t
l
en
g
t
h
.
Fo
r
in
s
ta
n
ce
,
b
y
m
er
g
i
n
g
t
w
o
co
p
i
es
o
f
a
p
ar
ticu
lar
te
x
t
d
o
cu
m
e
n
t
d
to
g
e
n
er
ate
a
p
s
e
u
d
o
-
d
o
cu
m
en
t
'
d
,
th
e
C
S
v
alu
e
co
m
p
u
ted
b
et
w
ee
n
d
an
d
'
d
w
i
ll
b
e
eq
u
al
to
1
.
T
h
is
r
ef
er
s
th
a
t,
m
atc
h
i
n
g
s
h
o
u
ld
b
e
ca
r
r
ied
o
u
t
a
m
o
n
g
t
w
o
d
o
cu
m
e
n
t
s
.
W
h
e
n
f
ed
w
it
h
a
n
o
th
er
d
o
cu
m
e
n
t
m
,
d
an
d
'
d
w
o
u
ld
lik
el
y
to
r
esu
l
t
in
s
a
m
e
s
i
m
ilar
it
y
to
m
an
d
is
'
,,
d
m
m
d
s
i
m
t
t
s
i
m
t
t
.
On
th
e
o
th
er
h
an
d
,
it
ca
n
also
b
e
ex
p
r
ess
ed
as,
f
o
r
tex
t
d
o
cu
m
en
ts
w
it
h
s
i
m
ilar
co
n
te
n
t
o
r
w
o
r
d
s
,
d
iv
er
s
e
to
tals
w
il
l
b
e
m
a
n
ag
ed
id
en
tical
l
y
.
Ho
w
e
v
er
,
th
is
i
s
u
n
ab
le
to
s
atis
f
y
th
e
m
etr
ic
‟
s
s
ec
o
n
d
co
n
d
itio
n
b
ec
au
s
e
w
ith
t
h
e
co
n
s
o
lid
atio
n
o
f
t
w
o
s
i
m
ilar
co
p
ies,
a
co
m
p
letel
y
d
is
s
i
m
ilar
o
b
j
ec
t
w
ill
b
e
o
b
tain
ed
f
r
o
m
o
r
ig
i
n
al
tex
t
d
o
cu
m
en
t.
I
n
ad
d
itio
n
,
it
is
ess
e
n
tial
to
n
o
te
th
at,
if
th
e
v
ec
to
r
s
ar
e
n
o
r
m
al
ized
to
a
f
i
x
ed
u
n
it le
n
g
th
,
t
h
is
ca
s
e
r
ef
lect
s
s
i
m
ilar
n
o
tatio
n
s
f
o
r
b
o
th
d
a
n
d
d
0
.
2
.
2
.
3
.
J
a
cc
a
rd
c
o
ef
f
icient
J
ac
ca
r
d
C
o
ef
f
icien
t
o
r
T
an
im
o
to
C
o
ef
f
icie
n
t
i
s
also
p
r
o
p
o
s
ed
to
ca
lcu
late
s
i
m
ilar
it
y
.
Acc
o
r
d
in
g
to
th
i
s
co
m
p
u
tatio
n
,
s
i
m
ilar
it
y
is
m
e
asu
r
ed
a
s
th
e
“
i
n
ter
s
ec
t
io
n
to
co
m
b
i
n
ed
s
p
ec
if
ied
o
b
j
ec
ts
r
a
tio
”.
Fo
r
th
e
g
iv
e
n
tex
t
d
o
cu
m
en
t,
t
h
i
s
co
ef
f
icie
n
t
ev
alu
ate
s
t
h
e
to
tal
w
eig
h
t
o
f
t
h
e
m
u
t
u
al
ter
m
s
e
x
is
ti
n
g
i
n
b
o
th
d
o
cu
m
en
t
s
w
it
h
th
e
to
tal
w
ei
g
h
t
o
f
a
ll
ter
m
s
ex
is
t
in
g
i
n
at
lea
s
t
o
n
e
o
f
t
h
e
t
w
o
d
o
cu
m
e
n
ts
b
u
t
u
n
iq
u
e
t
er
m
s
.
B
ased
o
n
t
h
is
co
m
p
u
tatio
n
,
m
atc
h
i
n
g
a
m
o
n
g
th
e
d
o
cu
m
e
n
t
s
w
ill
b
e
ca
r
r
ied
o
u
t.
T
h
e
g
en
er
al
co
m
p
u
tatio
n
f
o
r
m
u
la
h
a
s
b
ee
n
d
ep
icted
:
22
.
,
.
xy
H
x
y
x
y
x
y
tt
S
IM
t
t
t
t
t
t
(
3
)
J
ac
ca
r
d
co
ef
f
icien
t
i
s
a
s
i
m
ila
r
it
y
m
ea
s
u
r
e
an
d
it
b
o
u
n
d
s
b
et
w
ee
n
0
an
d
1
.
T
h
e
m
ea
s
u
r
e
w
i
ll
b
e
1
if
b
o
th
th
e
d
o
cu
m
en
ts
ar
e
s
i
m
i
l
ar
an
d
0
w
h
en
t
h
e
y
ar
e
d
is
s
i
m
ilar
.
I
n
g
en
er
al,
co
e
f
f
icien
t
v
alu
e
o
f
1
r
ep
r
esen
ts
th
at
b
o
th
g
i
v
e
n
o
b
j
ec
ts
ar
e
s
a
m
e,
w
h
er
ea
s
,
co
ef
f
icie
n
t
v
a
lu
e
o
f
0
d
en
o
tes
th
at
t
h
e
s
p
ec
if
ied
o
b
j
ec
ts
ar
e
ex
tr
e
m
e
l
y
d
if
f
er
en
t.
I
n
ad
d
itio
n
,
d
is
s
i
m
ilar
it
y
s
h
o
u
ld
also
b
e
o
b
s
er
v
ed
in
th
is
s
i
m
ilar
it
y
m
ea
s
u
r
e
-
th
e
J
ac
ca
r
d
d
is
tan
ce
m
ea
s
u
r
e
[
11
].
T
h
e
d
is
s
i
m
ilar
it
y
a
m
o
n
g
t
h
e
g
i
v
e
n
o
b
j
ec
ts
w
i
ll
b
e
co
m
p
u
ted
u
s
i
n
g
d
i
s
tan
ce
m
etr
ic
s
an
d
is
1
HH
D
S
I
M
.
H
D
ca
n
also
b
e
u
s
ed
as a
n
alter
n
ati
v
e
in
f
o
llo
w
in
g
e
x
p
er
i
m
e
n
ts
.
2
.
3
.
All a
bo
ut
c
lus
t
er
ing
Data
m
in
in
g
r
e
f
er
s
to
t
h
e
p
r
o
ce
s
s
w
h
ich
m
ain
l
y
en
tails
th
e
ex
tr
ac
tio
n
o
f
i
m
p
licit,
p
r
ev
io
u
s
l
y
u
n
k
n
o
w
n
as
w
e
ll
as
p
o
ten
tial
l
y
b
e
n
e
f
icial
i
n
f
o
r
m
atio
n
f
r
o
m
d
ata.
I
t
is
im
p
er
ati
v
e
th
a
t
d
o
cu
m
e
n
t
cl
u
s
ter
in
g
,
w
h
ic
h
is
a
s
u
b
g
r
o
u
p
o
f
d
ata
clu
s
ter
i
n
g
,
r
ef
er
s
to
a
d
ata
m
in
i
n
g
ap
p
r
o
ac
h
th
a
t
in
c
lu
d
es
v
ar
i
o
u
s
co
n
ce
p
ts
f
r
o
m
in
f
o
r
m
atio
n
r
etr
iev
al,
n
at
u
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
i
n
g
,
as
w
e
ll
as
m
ac
h
i
n
e
lear
n
i
n
g
f
ie
ld
s
[1
2
].
T
h
e
h
i
g
h
-
qu
alit
y
an
d
ef
f
icie
n
t
d
o
cu
m
e
n
t
cl
u
s
te
r
in
g
m
et
h
o
d
s
p
la
y
a
v
ital
r
o
l
e
in
s
u
p
p
o
r
tin
g
t
h
e
clie
n
ts
in
ter
m
s
o
f
e
f
f
ec
ti
v
e
n
av
i
g
atio
n
,
s
u
m
m
ar
izi
n
g
an
d
o
r
g
an
izi
n
g
d
iv
er
s
if
ied
s
et
o
f
i
n
f
o
r
m
atio
n
ef
f
ec
t
iv
el
y
.
A
s
p
ec
if
ied
d
o
cu
m
e
n
t
w
i
ll
al
w
a
y
s
h
a
v
e
a
p
r
o
b
ab
ilit
y
to
o
cc
u
r
in
m
u
ltip
le
cl
u
s
ter
s
[
1
3
]
in
t
h
e
o
v
er
lap
p
in
g
p
ar
titi
o
n
.
Fu
r
t
h
er
,
in
d
is
j
o
in
t
p
ar
titi
o
n
,
th
e
te
x
t d
o
cu
m
e
n
t
will a
p
p
ea
r
in
o
n
l
y
o
n
e
cl
u
s
ter
.
As
p
o
in
ts
o
u
t,
d
o
cu
m
en
t
cl
u
s
ter
in
g
ca
n
b
e
g
r
o
u
p
ed
in
to
tw
o
m
a
in
s
u
b
ca
teg
o
r
ies,
w
h
ic
h
in
cl
u
d
es:
So
f
t
(
o
v
er
lap
p
in
g
)
a
n
d
Har
d
C
lu
s
ter
in
g
.
Ov
er
lap
p
in
g
C
l
u
s
ter
in
g
is
cl
u
s
ter
ed
in
to
Hier
ar
ch
ical
clu
s
ter
in
g
,
P
ar
titi
o
n
in
g
an
d
I
te
m
s
et
-
b
ased
C
lu
s
ter
in
g
.
a.
Dis
j
o
in
t
(
Har
d
)
:
I
t
w
il
l
co
m
p
u
te
d
is
j
o
in
t
ass
i
g
n
m
e
n
t
s
o
f
a
s
p
ec
if
ied
t
ex
t
d
o
cu
m
en
t
to
w
ar
d
s
a
clu
s
ter
.
T
h
at
is
,
a
s
m
en
tio
n
ed
ab
o
v
e,
h
ar
d
clu
s
ter
i
n
g
w
ill
al
w
a
y
s
as
s
ig
n
a
d
o
cu
m
en
t
to
s
i
n
g
le
cl
u
s
ter
,
w
h
ich
t
h
en
ca
ter
s
a
s
et
o
f
d
if
f
er
en
t c
lu
s
ter
s
.
b.
Ov
er
lap
p
in
g
(
So
f
t
C
l
u
s
ter
in
g
)
:
T
h
is
t
y
p
e
o
f
cl
u
s
ter
i
n
g
p
r
o
ce
s
s
s
o
f
t
a
s
s
i
g
n
m
e
n
ts
w
ill
b
e
ca
r
r
ied
o
u
t.
T
h
at
is
,
ev
er
y
te
x
t
d
o
cu
m
e
n
t
is
ca
n
b
e
p
r
esen
ted
in
d
is
ti
n
ct
clu
s
ter
s
.
Hen
ce
,
s
o
f
t
cl
u
s
ter
in
g
p
r
o
d
u
ce
s
m
u
ltip
le
o
v
er
lap
p
in
g
clu
s
ter
s
.
c.
P
ar
titi
o
n
in
g
:
I
t
is
p
r
im
ar
il
y
en
g
a
g
ed
in
ass
ig
n
i
n
g
d
o
cu
m
en
ts
i
n
to
a
s
p
ec
if
ic
v
o
lu
m
e
o
f
No
n
-
E
m
p
t
y
C
lu
s
ter
s
.
I
n
p
ar
ticu
lar
,
k
-
m
ea
n
s
alo
n
g
w
it
h
it
s
alter
n
ativ
e
s
a
r
e
h
ig
h
l
y
r
ep
u
d
iated
p
ar
titi
o
n
in
g
tec
h
n
iq
u
e
s
as p
er
[
1
]
.
d.
Hier
ar
ch
ical:
I
t
i
n
v
o
l
v
es
d
ev
elo
p
in
g
d
e
n
d
r
o
g
r
a
m
s
,
w
h
er
e
clu
s
ter
s
ar
e
o
r
g
an
ized
in
h
ier
ar
ch
ical
tr
e
e
p
atter
n
s
.
I
n
th
e
tr
ee
,
th
e
L
ea
f
n
o
d
e
r
ep
r
esen
ts
th
e
s
u
b
-
s
et
o
f
g
iv
e
n
d
o
cu
m
e
n
t
co
llecti
o
n
.
B
o
th
HAC
clu
s
ter
i
n
g
an
d
UP
GM
A
clu
s
te
r
in
g
ar
e
g
r
o
u
p
ed
in
t
h
e
h
ier
ar
ch
ical
s
tr
u
ct
u
r
e
[
1
4
].
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:
2
0
8
8
-
8708
F
ea
tu
r
e
s
elec
tio
n
,
o
p
timiz
a
tio
n
a
n
d
clu
s
teri
n
g
s
tr
a
teg
ies o
f te
xt
d
o
cu
men
ts
(
A
.
Ko
u
s
ar
Nik
h
ath
)
1317
2
.
3
.
1
.
Do
cu
m
ent
c
lus
t
er
i
ng
Do
cu
m
e
n
t
C
l
u
s
ter
i
n
g
p
la
y
s
a
v
ital
r
o
le
i
n
cl
u
s
ter
i
n
g
t
h
e
g
iv
e
n
d
o
cu
m
e
n
ts
in
to
n
u
m
er
o
u
s
to
p
ics
w
it
h
o
u
t
h
a
v
i
n
g
a
n
y
i
n
f
o
r
m
ati
o
n
o
f
t
h
e
s
tr
u
ctu
r
e
o
f
th
e
ca
t
eg
o
r
y
av
a
ilab
le
in
a
g
i
v
en
d
o
cu
m
e
n
t
co
llect
io
n
.
E
ac
h
an
d
ev
er
y
Se
m
a
n
tic
I
n
f
o
r
m
at
io
n
is
o
b
tai
n
ed
f
r
o
m
w
i
t
h
i
n
th
e
g
i
v
e
n
d
o
cu
m
e
n
t
s
an
d
is
Un
-
s
u
p
er
v
i
s
ed
.
On
t
h
e
o
t
h
er
s
id
e,
d
o
cu
m
e
n
t
class
i
f
icat
io
n
i
s
co
n
ce
r
n
ed
w
it
h
a
s
s
i
g
n
i
n
g
th
e
tex
t
d
o
cu
m
en
ts
to
p
r
e
-
d
ef
i
n
ed
ca
teg
o
r
ies,
w
h
er
e
lab
eled
in
s
tan
ce
s
f
o
r
lear
n
i
n
g
f
r
o
m
t
h
e
clu
s
ter
i
n
g
f
o
r
cla
s
s
i
f
icatio
n
is
ca
lled
s
u
p
er
v
i
s
ed
lear
n
in
g
in
w
h
ic
h
a
g
iv
e
n
clas
s
if
ier
is
lear
n
ed
f
r
o
m
t
h
e
lab
eled
ex
a
m
p
les.
I
t
is
th
en
u
s
ed
f
o
r
p
r
e
d
ictin
g
clas
s
es
o
f
u
n
s
ee
n
d
o
cu
m
e
n
t
s
.
Do
cu
m
en
t
clu
s
ter
in
g
i
s
e
m
p
lo
y
ed
in
n
u
m
er
o
u
s
d
i
v
er
s
e
co
n
te
x
ts
,
lik
e
ex
p
lo
r
in
g
t
h
e
s
tr
u
ct
u
r
e
i
n
a
g
i
v
e
n
d
o
cu
m
en
t
co
llec
tio
n
f
o
r
t
h
e
d
is
co
v
er
y
o
f
k
n
o
w
led
g
e
[
8
]
,
d
i
m
en
s
io
n
a
lit
y
co
n
tr
ac
tio
n
f
o
r
all
o
th
er
ta
s
k
s
s
u
c
h
as
clas
s
i
f
icatio
n
[
1
5
]
,
g
r
o
u
p
in
g
s
ea
r
c
h
o
u
tco
m
es
to
r
an
k
ed
lis
t
[
9
]
f
o
r
ex
ec
u
ti
n
g
an
alter
n
ati
v
e
p
r
esen
tatio
n
an
d
al
s
o
e
m
p
lo
y
ed
f
o
r
p
s
eu
d
o
-
r
ele
v
an
ce
f
ee
d
b
ac
k
.
2
.
4
.
Clus
t
er
ev
a
lua
t
io
n
m
ea
s
ures
E
v
alu
a
tio
n
o
f
d
o
cu
m
e
n
t
cl
u
s
t
er
in
g
i
s
a
d
i
f
f
ic
u
lt
tas
k
.
B
u
ilt
-
in
q
u
al
it
y
m
ea
s
u
r
es
li
k
e
d
is
to
r
tio
n
o
r
lo
g
p
o
s
s
ib
ilit
ies
i
m
p
l
y
h
o
w
a
ce
r
t
ain
al
g
o
r
ith
m
o
p
ti
m
izes
a
g
iv
en
r
ep
r
esen
tat
io
n
.
Me
a
n
w
h
ile
,
in
ter
n
a
l
m
ea
s
u
r
es
co
u
ld
n
o
t
b
e
co
m
p
ar
ed
a
m
o
n
g
d
if
f
er
en
t
r
ep
r
esen
tatio
n
s
.
I
n
ad
d
itio
n
,
it‟s
a
n
o
te
w
o
r
th
y
p
o
in
t
th
at
ex
ter
n
al
v
ie
w
s
o
f
tr
u
t
h
ar
e
h
u
m
a
n
-
m
a
d
e.
T
h
e
y
co
n
ti
n
u
e
to
s
u
f
f
er
f
r
o
m
t
h
e
m
aj
o
r
s
h
i
f
t
f
o
r
h
u
m
an
s
to
u
n
d
er
s
tan
d
d
if
f
er
e
n
t
d
o
cu
m
e
n
t
to
p
ics
in
a
d
is
tin
ct
m
a
n
n
er
.
P
r
e
d
o
m
i
n
an
t
l
y
,
w
h
et
h
er
th
e
ce
r
tain
d
o
cu
m
en
t
b
elo
n
g
s
to
th
at
p
ar
ticu
lar
to
p
ic
o
r
n
o
t
m
i
g
h
t
b
e
s
u
b
j
ec
tiv
e.
Ho
w
e
v
er
,
a
s
cl
u
s
ter
in
g
o
f
a
d
o
cu
m
en
t
h
as
f
ea
s
i
b
ilit
y
to
ex
ec
u
te
in
a
n
u
m
b
er
o
f
w
a
y
s
,
ab
o
v
e
m
e
n
t
io
n
ed
s
ce
n
ar
io
co
u
ld
ev
e
n
co
m
p
licate
t
h
e
co
n
d
itio
n
s
.
T
h
e
m
aj
o
r
ad
v
an
tag
e
o
f
t
h
is
m
ea
s
u
r
e
i
n
co
m
p
ar
ed
to
ev
al
u
atio
n
th
r
o
u
g
h
te
x
t
cla
s
s
i
f
icat
io
n
is
th
a
t
th
er
e
is
n
o
n
ee
d
o
f
s
u
c
h
co
n
d
i
tio
n
s
w
h
ich
ar
e
d
ep
icted
ab
o
v
e.
T
h
is
m
ea
s
u
r
e
d
o
es
n
o
t
in
clu
d
e
eith
er
a
test
b
ed
p
latf
o
r
m
(
co
m
p
r
i
s
es
lab
eled
d
o
cu
m
e
n
ts
)
o
r
co
n
s
i
s
ten
c
y
f
ac
t
o
r
a
m
id
cl
u
s
ter
s
an
d
tar
g
eted
ca
teg
o
r
ies.
O
n
t
h
e
o
th
er
h
a
n
d
,
it
ap
p
r
o
x
i
m
atel
y
e
v
alu
a
tes
th
e
o
u
tco
m
e
o
f
te
x
t
c
lu
s
ter
i
n
g
[1
3
]
,
o
n
l
y
w
h
en
t
h
e
lab
eled
d
o
cu
m
e
n
t
s
ar
e
u
tili
ze
d
a
s
te
s
t
b
ed
.
T
ex
t
c
lass
i
f
icatio
n
p
ar
a
m
eter
s
li
k
e
a
cc
u
r
ac
y
,
r
e
-
ca
ll,
F1
a
n
d
p
r
ec
is
io
n
m
ea
s
u
r
es
w
er
e
u
s
ed
f
o
r
esti
m
ati
n
g
t
h
e
p
r
esen
tatio
n
o
f
tex
t
cl
u
s
ter
i
n
g
i
n
[
1
4
]
,
[
1
6
]
.
B
ased
o
n
p
r
o
p
er
l
y
class
i
f
ied
tex
t
d
o
cu
m
en
ts
a
n
d
ea
ch
a
n
d
ev
e
r
y
d
o
cu
m
e
n
t
p
r
ese
n
t
i
n
t
h
e
t
est
b
ed
,
th
e
r
ate
o
f
ac
c
u
r
ac
y
w
i
ll
b
e
co
m
p
u
ted
.
Fu
r
t
h
er
,
th
e
m
ea
s
u
r
e
is
th
e
s
i
m
p
lest
m
ea
s
u
r
i
n
g
p
ar
a
m
eter
i
n
ass
o
ciate
d
clas
s
i
f
icatio
n
p
r
o
b
le
m
s
.
T
h
is
m
ea
s
u
r
e
is
d
ir
ec
tl
y
ap
p
licab
le
to
th
e
Mu
lti
-
C
lass
if
icatio
n
P
r
o
b
lem
s
.
Ho
w
e
v
er
,
s
ig
n
i
f
ica
n
t
m
ea
s
u
r
es
lik
e
p
r
ec
is
io
n
,
r
e
-
ca
ll
an
d
F1
ca
n
b
e
d
ir
ec
tl
y
ap
p
lied
to
th
e
b
in
ar
y
class
i
f
icatio
n
tas
k
s
.
Hen
ce
,
t
o
ev
alu
ate
t
h
e
class
if
icatio
n
p
e
r
f
o
r
m
an
ce
b
y
m
a
k
in
g
t
h
e
u
s
e
o
f
th
o
s
e
m
ea
s
u
r
es,
th
e
r
esp
ec
ti
v
e
p
r
o
b
le
m
i
s
to
b
e
s
p
lit
i
n
to
b
in
ar
y
clas
s
i
f
icatio
n
p
r
o
b
le
m
s
.
E
ac
h
an
d
e
v
er
y
c
lass
co
r
r
esp
o
n
d
s
to
a
s
p
ec
if
ic
b
in
ar
y
clas
s
i
f
icati
o
n
tas
k
i
n
M
u
lt
i
-
C
las
s
i
f
icati
o
n
tas
k
.
O
f
t
h
e
cla
s
s
e
s
,
p
o
s
i
tiv
e
o
n
es
r
ep
r
ese
n
t
“
B
elo
n
g
i
n
g
to
th
e
clas
s
”
a
n
d
th
e
No
n
-
P
o
s
iti
v
e
o
n
es
r
ep
r
es
en
t
“
No
t
-
B
elo
n
g
in
g
to
th
e
cl
ass
”.
T
h
e
ev
o
l
u
tio
n
m
ea
s
u
r
e
m
aj
o
r
l
y
co
n
ce
n
tr
ates
o
n
th
e
p
o
s
iti
v
e
clas
s
.
I
n
th
e
tex
t
ca
te
g
o
r
izatio
n
,
r
e
-
ca
ll
m
ea
s
u
r
e
w
i
ll
b
e
o
b
ta
in
ed
b
y
th
e
r
atio
o
f
th
e
s
p
ec
i
f
ic
tr
u
e
p
o
s
itiv
e
d
o
cu
m
en
t
to
all
d
o
cu
m
e
n
t
s
th
at
ar
e
tr
u
e.
P
r
ec
is
io
n
m
ea
s
u
r
e
r
ef
er
r
ed
as
th
e
r
ate
o
f
class
if
ied
tr
u
e
p
o
s
iti
v
e
d
o
cu
m
en
ts
to
e
v
er
y
clas
s
if
ied
p
o
s
itiv
e
d
o
cu
m
e
n
t i
n
cl
u
d
es b
o
th
tr
u
e
p
o
s
iti
v
es a
n
d
f
alse p
o
s
i
tiv
es.
W
h
er
ea
s
,
F
1
is
u
s
ed
to
d
eter
m
i
n
e
a
v
al
u
e
u
s
in
g
b
o
th
R
e
-
ca
ll
R
an
d
P
r
ec
is
io
n
P
m
ea
s
u
r
es b
y
u
s
i
n
g
(
4
).
2
1
RP
F
m
e
a
s
u
r
e
RP
(
4
)
Var
io
u
s
m
etr
ics
li
k
e
F1
,
ac
cu
r
ac
y
,
d
etec
tio
n
co
s
ts
ar
e
e
m
p
lo
y
ed
i
n
tex
t
ca
teg
o
r
izatio
n
.
T
h
ese
ar
e
p
r
im
ar
il
y
e
m
p
lo
y
ed
to
ca
lcu
l
ate
th
e
p
er
f
o
r
m
a
n
ce
m
etr
ic
i
n
te
x
t
cl
u
s
ter
in
g
.
W
h
e
n
th
e
s
e
m
ea
s
u
r
es
ar
e
u
s
ed
th
er
e
al
w
a
y
s
ex
i
s
t
t
w
o
co
n
d
i
tio
n
s
.
E
ac
h
a
n
d
ev
er
y
g
i
v
en
d
o
cu
m
en
t
i
n
s
id
e
th
e
s
p
ec
i
f
i
ed
test
b
ed
s
h
o
u
ld
co
n
tain
tar
g
e
t
ca
teg
o
r
ie
s
an
d
m
u
s
t
b
e
lab
eled
.
I
t
is
s
o
m
e
w
h
at
cr
itical
in
r
ea
l
-
ti
m
e
in
ter
m
s
o
f
g
etti
n
g
lab
eled
d
o
cu
m
en
t
w
h
en
co
m
p
ar
ed
to
th
e
d
o
cu
m
en
t
w
h
ich
i
s
u
n
lab
e
led
.
Me
an
w
h
i
le,
th
e
p
r
o
ce
s
s
w
h
ic
h
is
e
n
g
ag
ed
i
n
lab
elin
g
d
o
cu
m
e
n
t
s
f
o
llo
w
s
i
n
p
r
ac
tice
w
it
h
clu
s
ter
i
n
g
d
o
c
u
m
e
n
t
s
.
I
n
ad
d
itio
n
,
it
is
al
s
o
s
ig
n
i
f
ica
n
t
to
n
o
te
th
at,
v
as
t
ti
m
e
w
ill
b
e
co
n
s
u
m
ed
b
y
t
h
e
p
r
o
ce
s
s
w
h
ich
i
s
e
n
g
a
g
ed
i
n
e
v
al
u
ati
n
g
t
h
e
ap
p
r
o
ac
h
es
to
tex
t
clu
s
ter
i
n
g
w
h
en
p
r
ep
ar
atio
n
o
f
lab
eled
d
o
cu
m
en
t
s
i
s
o
n
g
o
in
g
.
Seco
n
d
l
y
,
t
h
e
cl
u
s
ter
n
u
m
b
er
m
u
s
t
b
e
co
n
s
ta
n
t
w
it
h
tar
g
et
ca
teg
o
r
ies
n
u
m
b
er
.
Fo
r
ex
a
m
p
le,
w
h
en
a
s
eq
u
en
c
e
o
f
d
o
cu
m
en
ts
h
a
v
in
g
s
a
m
e
t
ar
g
et
ca
te
g
o
r
y
w
i
ll
b
e
p
ar
titi
o
n
ed
in
to
t
w
o
cl
u
s
ter
s
,
th
e
n
th
e
ev
o
l
u
tio
n
m
ea
s
u
r
es
o
f
te
x
t
c
h
ar
ac
ter
izatio
n
w
ill
n
o
t
b
e
ap
p
licab
le
in
s
u
c
h
ca
s
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
2
,
A
p
r
il 2
0
1
9
:
1
3
1
3
-
1320
1318
3.
RE
VI
E
W
O
F
L
I
T
E
RA
T
UR
E
3
.
1
.
F
e
a
t
ure
ex
t
ra
ct
io
n str
a
t
eg
ies
T
h
ese
tech
n
iq
u
es
ar
e
in
tr
o
d
u
ce
d
b
ased
o
n
k
e
y
w
o
r
d
s
.
T
h
ese
k
e
y
w
o
r
d
s
ar
e
e
m
p
lo
y
e
d
to
d
e
p
ict
v
ar
io
u
s
e
m
o
t
io
n
s
w
h
ich
e
x
is
t
in
s
id
e
th
e
te
x
t
[
1
7
]
.
I
n
co
n
tr
as
t,
th
e
m
ain
d
is
ad
v
a
n
ta
g
e
o
f
t
h
is
m
et
h
o
d
is
th
at
it
r
elies
o
n
p
r
esen
ce
o
f
v
ar
io
u
s
a
f
f
ec
t
iv
e
w
o
r
d
s
i
n
t
h
e
te
x
t
.
T
o
o
v
er
co
m
e
s
u
c
h
d
r
a
w
b
a
ck
s
a
n
d
to
ac
h
ie
v
e
ac
cu
r
ate
ex
tr
ac
tio
n
s
an
d
o
u
tc
o
m
e
s
,
th
e
a
u
t
h
o
r
s
p
r
o
p
o
s
ed
a
n
o
v
el
m
o
d
el
ca
lled
Se
m
a
n
tic
Net
w
o
r
k
s
i
n
[
1
5
]
.
T
h
ese
n
et
w
o
r
k
s
r
ep
r
esen
t
e
v
e
n
ts
,
r
elatio
n
s
h
ip
s
an
d
v
ar
io
u
s
co
n
ce
p
ts
a
m
o
n
g
t
h
e
m
.
U
n
li
k
e
f
ea
tu
r
e
ex
tr
ac
tio
n
,
th
ese
s
e
m
an
t
ic
n
et
w
o
r
k
s
ar
e
i
n
d
ep
en
d
en
t
o
n
k
e
y
w
o
r
d
s
to
d
ep
ict
th
e
h
u
m
a
n
e
m
o
t
io
n
s
in
t
h
e
te
x
t.
Hen
ce
,
[
1
8
]
m
ad
e
v
al
u
ab
le
co
n
cl
u
s
io
n
s
a
b
o
u
t
th
e
p
r
o
ce
s
s
o
f
ac
h
iev
in
g
en
h
a
n
ce
d
p
er
f
o
r
m
an
ce
i
n
d
etec
tin
g
t
h
e
h
u
m
a
n
e
m
o
tio
n
s
th
r
o
u
g
h
s
e
m
a
n
tic
n
et
w
o
r
k
s
.
I
n
t
h
ese
n
et
w
o
r
k
s
,
h
u
m
a
n
e
m
o
tio
n
s
w
i
ll
b
e
id
en
ti
f
ied
th
r
o
u
g
h
co
n
tex
t
u
al
i
n
f
o
r
m
atio
n
.
I
n
p
ar
ticu
lar
,
a
n
d
p
r
esen
ted
a
r
a
n
g
e
o
f
ex
p
la
n
atio
n
s
o
f
th
is
ap
p
r
o
ac
h
.
Ho
w
ev
er
,
t
h
e
y
f
ailed
to
e
x
p
lain
th
e
r
e
s
p
ec
tiv
e
o
u
tco
m
e
s
o
f
th
e
e
x
p
er
i
m
e
n
t
s
.
Mo
r
eo
v
er
,
th
er
e
is
a
n
ec
es
s
it
y
o
f
h
u
g
e
d
atab
ases
lik
e
Se
n
tiW
o
r
d
Net
an
d
W
o
r
d
Net
-
Af
f
ec
t to
i
m
p
r
o
v
e
t
h
e
ac
c
u
r
ac
y
o
f
r
es
u
lt
s
.
3
.
2
.
F
e
a
t
ure
s
elec
t
io
n s
t
ra
t
e
g
ies
Mu
ltip
le
f
ea
t
u
r
e
id
e
n
ti
f
icatio
n
p
r
o
g
r
a
m
s
,
ar
e
i
m
p
le
m
e
n
ted
f
o
r
class
i
f
icatio
n
.
Ho
w
ev
er
,
a
ll
p
r
o
j
ec
te
d
alg
o
r
ith
m
s
h
a
v
e
co
m
m
o
n
g
o
a
l,
i.e
.
,
s
ea
r
ch
i
n
g
f
o
r
ef
f
icie
n
t
f
ea
t
u
r
es
s
et
w
h
ich
ca
ter
s
r
es
u
lts
i
n
ter
m
s
o
f
b
est
class
i
f
icatio
n
s
.
I
n
g
e
n
er
al,
v
ar
i
o
u
s
alg
o
r
it
h
m
s
in
v
o
l
v
ed
in
f
ea
tu
r
e
s
elec
tio
n
e
m
p
lo
y
d
is
ti
n
ct
ev
alu
a
tio
n
m
etr
ics
lik
e
i
n
f
o
r
m
atio
n
g
ai
n
a
n
d
co
r
r
e
latio
n
.
I
n
ad
d
itio
n
,
t
h
e
y
o
f
ten
u
s
e
p
o
p
u
latio
n
-
b
ased
h
e
u
r
is
tics
s
u
c
h
as
an
t
co
lo
n
y
o
p
ti
m
izatio
n
,
s
i
m
u
late
d
an
n
ea
li
n
g
,
p
ar
ticle
-
s
w
ar
m
o
p
ti
m
izatio
n
,
an
d
g
e
n
etic
alg
o
r
ith
m
s
.
A
cc
o
r
d
in
g
to
b
y
u
s
i
n
g
f
ea
t
u
r
e
s
i
m
ilar
it
y
,
Un
-
Su
p
er
v
i
s
ed
Featu
r
e
S
u
b
-
Set
Select
io
n
T
ec
h
n
iq
u
e
wer
e
p
r
o
p
o
s
ed
[
1
9
].
T
h
is
ap
p
r
o
ac
h
is
u
s
ed
to
av
o
i
d
th
e
d
u
p
licatio
n
s
a
m
o
n
g
th
e
s
elec
ted
f
ea
t
u
r
es.
T
h
is
ap
p
r
o
a
ch
u
s
e
s
n
e
w
m
etr
ics
ca
lled
MI
C
I
n
d
ex
f
o
r
ca
lcu
l
atin
g
th
e
s
i
m
ilar
it
y
m
ea
s
u
r
e
b
et
w
ee
n
t
w
o
d
i
f
f
er
en
t
v
ar
i
ab
les
f
o
r
s
elec
ti
n
g
a
f
ea
t
u
r
e.
In
F
u
zz
y
r
o
u
g
h
s
e
t
th
eo
r
y
i
s
e
m
p
lo
y
ed
f
o
r
th
e
s
elec
tio
n
o
f
f
ea
t
u
r
e
b
y
co
n
s
id
er
in
g
t
h
e
n
at
u
r
al
p
r
o
p
er
ties
o
f
b
o
th
f
u
zz
y
lo
g
i
c
t
-
n
o
r
m
s
an
d
t
-
co
n
o
r
m
s
.
A
d
d
itio
n
all
y
,
i
n
MI
FS
-
U
al
g
o
r
ith
m
i
s
in
tr
o
d
u
ce
d
to
h
an
d
le
r
estrictio
n
s
li
n
k
ed
w
it
h
MI
F
S.
T
h
e
p
r
i
m
ar
y
o
b
j
ec
tiv
e
o
f
t
h
is
ap
p
r
o
ac
h
is
to
g
et
i
m
p
r
o
v
ed
s
i
m
ilar
in
f
o
r
m
atio
n
a
m
o
n
g
in
p
u
t
ch
ar
ac
ter
is
tics
an
d
o
u
tp
u
t
class
e
s
o
f
th
e
MI
FS
.
Si
m
ilar
l
y
,
[
1
2
]
a
ls
o
p
r
o
p
o
s
ed
f
ea
tu
r
e
s
elec
tio
n
tec
h
n
iq
u
e
ca
lled
M
ax
-
R
ele
v
an
ce
a
n
d
Min
-
R
ed
u
n
d
an
c
y
(
MRM
R
)
b
ased
o
n
m
u
tu
al
i
n
f
o
r
m
atio
n
co
n
ce
p
t.
I
n
g
e
n
er
al,
th
is
tec
h
n
iq
u
e
m
in
i
m
ize
s
t
h
e
r
ed
u
n
d
a
n
c
y
b
et
w
ee
n
th
e
f
ea
t
u
r
es
a
s
w
ell
as
m
a
x
i
m
ize
s
th
e
d
ep
en
d
en
c
y
b
et
w
ee
n
a
cla
s
s
la
b
el
an
d
s
u
b
-
s
et
o
f
f
ea
t
u
r
es.
3
.
3
.
Clus
t
er
ing
t
ec
hn
iqu
es
C
lu
s
ter
i
n
g
Alg
o
r
it
h
m
s
ar
e
c
h
a
r
ac
ter
ized
b
ased
o
n
t
w
o
m
aj
o
r
p
r
o
p
er
ties
.
T
h
e
f
ir
s
t
p
r
o
p
er
ty
p
r
im
ar
il
y
d
ea
ls
w
it
h
w
h
e
th
er
ce
r
tain
m
e
m
b
er
s
h
ip
o
f
clu
s
ter
is
d
is
tin
ct.
T
h
e
h
ar
d
o
r
d
is
jo
in
t
clu
s
ter
i
n
g
al
g
o
r
ith
m
s
allo
ca
te
ea
ch
an
d
e
v
er
y
d
o
cu
m
en
t
to
j
u
s
ti
f
y
a
s
in
g
le
c
lu
s
ter
.
T
h
e
o
th
er
s
id
e,
t
h
e
s
o
f
t
o
r
o
v
er
lap
p
in
g
cl
u
s
ter
in
g
alg
o
r
ith
m
s
allo
ca
te
d
is
s
i
m
i
lar
d
o
cu
m
en
ts
to
s
in
g
le
o
r
m
u
ltip
l
e
clu
s
ter
s
in
d
i
s
cr
ete
m
e
m
b
er
s
h
ip
d
eg
r
ee
.
On
t
h
e
o
th
er
h
a
n
d
,
t
h
e
s
ec
o
n
d
p
r
o
p
er
t
y
co
n
tr
o
ls
t
h
e
cl
u
s
ter
s
s
tr
u
ct
u
r
e.
I
n
g
e
n
er
al,
t
h
e
s
tr
u
ct
u
r
e
m
a
y
b
e
o
b
s
er
v
ed
i
n
eith
er
f
lat
o
r
h
ier
ar
ch
ical.
On
f
lat
cl
u
s
ter
in
g
tec
h
n
iq
u
e
f
r
o
n
t,
i
t
g
e
n
er
ates
r
ig
id
cl
u
s
ter
s
,
w
it
h
o
u
t
a
n
y
co
r
r
elatio
n
b
et
w
ee
n
th
e
m
.
O
n
th
e
co
n
tr
ar
y
,
t
h
e
h
ier
ar
ch
ical
a
lg
o
r
ith
m
s
ar
e
e
n
g
ag
ed
i
n
g
e
n
e
r
atin
g
c
lu
s
ter
s
i
n
a
tr
ee
s
tr
u
ct
u
r
e.
I
t
f
o
llo
w
s
b
o
tt
o
m
-
u
p
ap
p
r
o
ac
h
,
as
it
i
n
v
o
l
v
es
ex
ec
u
ti
n
g
t
h
e
p
r
o
ce
d
u
r
e
f
r
o
m
it
s
b
o
tto
m
m
o
s
t
clu
s
ter
(
at
t
h
e
r
o
o
t)
o
f
th
e
tr
ee
s
tr
u
ct
u
r
e.
3
.
3
.
1
.
P
a
rt
i
t
io
nin
g
a
nd
hiera
rc
hica
l do
cu
m
ent
clu
s
t
er
ing
T
h
e
m
aj
o
r
it
y
o
f
tr
ad
itio
n
al
cl
u
s
ter
i
n
g
alg
o
r
it
h
m
s
ar
e
ca
te
g
o
r
ized
in
to
t
w
o
m
ai
n
g
r
o
u
p
s
in
cl
u
d
in
g
p
ar
titi
o
n
in
g
al
g
o
r
ith
m
s
a
n
d
h
i
er
ar
ch
ical
al
g
o
r
ith
m
s
[
1
8
]
.
T
h
e
h
ier
ar
ch
ica
l
cl
u
s
ter
i
n
g
al
g
o
r
ith
m
s
ar
e
p
r
i
m
ar
il
y
in
v
o
l
v
ed
in
d
ec
o
m
p
o
s
in
g
a
s
p
ec
if
ied
d
ataset
h
ier
ar
ch
ical
l
y
.
Hen
ce
,
it
f
o
r
m
s
a
d
en
d
r
o
g
r
am
tr
ee
w
h
er
e
g
iv
e
n
d
ataset
is
s
p
li
t
r
ep
ea
ted
l
y
i
n
to
s
m
all
s
u
b
-
s
e
ts
.
T
h
u
s
,
th
e
d
o
cu
m
e
n
t
s
w
i
ll
b
e
r
ep
r
esen
ted
in
Mu
lti
-
L
ev
el
s
tr
u
ct
u
r
e
as
d
ep
icted
.
T
h
ese
alg
o
r
ith
m
s
ar
e
o
f
te
n
g
r
o
u
p
ed
in
to
eith
er
d
iv
is
i
v
e
alg
o
r
it
h
m
s
o
r
ag
g
lo
m
er
ati
v
e
alg
o
r
ith
m
s
,
On
t
h
e
o
t
h
er
h
a
n
d
,
in
ag
g
lo
m
er
ativ
e
p
r
o
ce
d
u
r
e
,
ea
ch
d
o
cu
m
en
t
is
allo
ca
ted
to
a
s
ep
ar
ate
clu
s
ter
.
L
at
er
,
th
e
p
r
o
ce
d
u
r
e
in
v
o
l
v
e
s
m
er
g
in
g
s
i
m
ilar
cl
u
s
ter
s
r
e
p
ea
ted
ly
u
n
t
il
ter
m
i
n
atio
n
cr
iter
io
n
is
o
b
s
er
v
ed
.
W
h
ile
o
n
t
h
e
d
iv
i
s
i
v
e
alg
o
r
it
h
m
s
f
r
o
n
t,
it
i
n
cr
ea
s
e
s
th
e
n
u
m
b
er
o
f
clu
s
ter
s
at
ea
c
h
iter
ati
v
e
s
tag
e
b
y
s
p
litt
i
n
g
th
e
w
h
o
le
d
o
cu
m
e
n
t
i
n
to
a
s
p
ec
if
ied
q
u
a
n
tit
y
o
f
cl
u
s
ter
s
.
I
n
ad
d
itio
n
,
an
o
th
er
clu
s
ter
i
n
g
a
lg
o
r
ith
m
b
ased
o
n
P
ar
titi
o
n
in
g
is
o
n
e
o
f
th
e
m
o
s
t
s
tu
d
ied
ca
teg
o
r
ies
[1
7
]
.
I
t
u
p
h
o
ld
s
ex
tr
e
m
e
r
ea
lis
tic
tec
h
n
iq
u
es
to
clu
s
ter
b
i
g
d
atasets
as
r
ep
r
esen
ted
,
u
n
l
ik
e
d
en
d
r
o
g
r
a
m
tr
ee
s
tr
u
c
tu
r
e,
th
ese
tec
h
n
iq
u
e
s
clu
s
ter
d
ata
in
a
s
i
n
g
le
lev
e
l.
I
n
g
en
er
al,
t
h
ese
ap
p
r
o
ac
h
es
ar
e
en
g
ag
ed
i
n
d
iv
id
in
g
a
g
i
v
en
d
o
cu
m
e
n
t
co
llectio
n
in
to
d
is
ti
n
ct
clu
s
ter
s
,
w
h
ic
h
in
-
tu
r
n
i
n
cr
ea
s
es
t
h
e
p
r
e
-
d
e
f
in
ed
o
b
j
ec
tiv
e
v
al
u
e.
B
y
co
n
tai
n
in
g
e
f
f
icie
n
t
cl
u
s
ter
i
n
g
ch
ar
ac
t
er
is
tics
i
n
ter
m
s
o
f
q
u
alit
y
a
n
d
ac
cu
r
ac
y
,
th
e
s
e
h
i
er
ar
ch
ical
clu
s
ter
i
n
g
alg
o
r
ith
m
s
d
o
n
o
t
o
f
f
er
r
e
-
allo
ca
tio
n
o
f
d
o
cu
m
en
ts
.
T
h
is
is
a
m
aj
o
r
d
r
a
w
b
ac
k
o
f
th
i
s
ap
p
r
o
ac
h
an
d
h
e
n
ce
t
h
er
e
ca
n
b
e
c
h
an
ce
s
o
f
p
o
o
r
p
er
f
o
r
m
an
ce
m
ea
s
u
r
es
i
n
t
h
e
ea
r
l
y
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:
2
0
8
8
-
8708
F
ea
tu
r
e
s
elec
tio
n
,
o
p
timiz
a
tio
n
a
n
d
clu
s
teri
n
g
s
tr
a
teg
ies o
f te
xt
d
o
cu
men
ts
(
A
.
Ko
u
s
ar
Nik
h
ath
)
1319
s
tag
e
s
o
f
cl
u
s
ter
in
g
.
A
cc
o
r
d
in
g
l
y
,
in
v
ar
io
u
s
d
ata
o
b
j
ec
ts
,
th
e
ti
m
e
ta
k
e
n
to
ex
ec
u
te
a
h
ier
a
r
ch
ical
al
g
o
r
ith
m
is
Qu
ad
r
atic.
I
n
t
h
e
r
ec
en
t
p
ast,
it
w
a
s
d
e
m
o
n
s
tr
ated
t
h
at
t
h
e
p
ar
tit
io
n
tec
h
n
iq
u
es
ar
e
b
est
s
u
itab
le
f
o
r
a
p
p
licatio
n
s
w
h
ic
h
co
m
p
r
is
e
b
i
g
d
ataset
s
d
u
e
to
it
s
Mi
n
i
m
ized
C
o
m
p
u
tatio
n
al
C
o
m
p
lex
it
y
,
T
i
m
e
co
m
p
le
x
it
y
i
s
co
m
p
ar
ati
v
el
y
les
s
as
co
m
p
ar
ed
to
h
ier
ar
ch
ical
tech
n
iq
u
e
s
an
d
is
lin
e
ar
.
T
h
u
s
,
p
ar
titi
o
n
in
g
tec
h
n
iq
u
e
s
ar
e
h
ig
h
l
y
ad
ap
tab
le
f
o
r
b
ig
s
c
ale
clu
s
ter
in
g
.
I
n
ad
d
itio
n
,
t
o
less
en
th
e
l
i
m
i
tatio
n
s
b
r
o
u
g
h
t
ab
o
u
t
b
y
t
h
e
co
n
v
e
n
tio
n
al
p
ar
titi
o
n
clu
s
ter
i
n
g
tech
n
iq
u
es
d
i
s
cu
s
s
ed
ea
r
li
er
,
m
u
ltip
le
m
o
d
els
ar
e
i
n
tr
o
d
u
ce
d
in
th
e
r
ec
en
t
f
e
w
y
e
ar
s
.
T
h
ese
m
et
h
o
d
s
f
o
c
u
s
ed
o
n
i
m
p
le
m
e
n
ti
n
g
o
p
ti
m
iz
atio
n
tec
h
n
iq
u
es o
v
er
a
p
r
e
-
d
eter
m
i
n
ed
clu
s
ter
in
g
th
r
o
u
g
h
o
b
j
ec
tiv
e
f
u
n
c
tio
n
.
3
.
3
.
2
.
M
a
chine
lea
rning
ba
s
ed
do
cu
m
ent
clu
s
t
er
ing
Nu
m
er
o
u
s
k
n
o
w
led
g
e
b
ases
lik
e
C
y
c
p
r
o
j
ec
t,
Fre
eb
ase
,
Kn
o
w
I
t
All,
W
ik
ip
ed
ia,
T
e
x
tR
u
n
n
er
,
W
ik
iT
ax
o
n
o
m
y
,
P
r
o
b
ase,
DB
p
ed
ia,
YA
G
O,
NE
L
L
[
8
]
as
w
ell
as
K
n
o
w
led
g
e
Vau
lt
g
e
n
er
all
y
p
la
y
a
h
i
g
h
l
y
v
ital r
o
le
in
t
h
e
p
r
o
ce
s
s
o
f
d
o
cu
m
e
n
t c
l
u
s
ter
in
g
w
it
h
r
eg
ar
d
s
to
co
n
tex
t,
co
n
ce
p
t a
n
d
s
e
m
a
n
tic
r
elatio
n
s
.
So
,
as
to
n
o
tice
all
o
f
t
h
ese
r
elatio
n
s
b
e
t
w
ee
n
th
e
d
o
cu
m
e
n
ts
,
a
p
r
io
r
k
n
o
w
led
g
e
i
s
v
ital.
T
h
is
w
il
l
p
o
in
t
o
u
t
t
h
e
n
ee
d
o
f
h
i
g
h
l
y
s
o
p
h
i
s
ticated
lear
n
i
n
g
tech
n
iq
u
e
s
to
n
o
tify
t
h
e
r
e
latio
n
s
.
O
n
t
h
e
co
n
tr
ar
y
,
t
h
e
a
f
o
r
esaid
k
n
o
w
led
g
e
b
ases
h
a
v
e
t
h
e
ab
ilit
y
o
f
tr
ai
n
in
g
t
h
e
lear
n
i
n
g
ap
p
r
o
ac
h
es
s
o
as
to
cl
u
s
ter
t
h
e
g
iv
e
n
d
o
c
u
m
e
n
t
s
b
ased
o
n
a
s
in
g
le
o
r
ad
d
itio
n
al
co
n
tex
t,
c
o
n
ce
p
t
as
w
ell
a
s
s
e
m
a
n
tic
r
el
atio
n
s
.
T
h
e
ar
g
u
m
en
t
ill
u
s
tr
at
es
th
e
h
u
g
e
r
o
le
o
f
m
ac
h
in
e
lear
n
in
g
w
h
e
n
it
co
m
es
to
Do
cu
m
e
n
t
C
l
u
s
ter
in
g
.
Usag
e
o
f
m
u
ltip
le
e
x
is
ti
n
g
k
n
o
w
led
g
e
b
ases
i
s
p
r
im
ar
il
y
ai
m
ed
at
en
h
an
c
in
g
d
o
cu
m
en
t
‟
s
f
ea
t
u
r
es
o
f
m
u
lti
-
s
et
o
f
w
o
r
d
s
r
ep
r
esen
ta
tio
n
.
Fo
r
ex
a
m
p
le,
u
s
in
g
W
o
r
d
Net,
a
lin
g
u
i
s
tic
k
n
o
w
l
ed
g
e
b
ase,
r
eso
l
v
es
s
y
n
o
n
y
m
s
w
h
ile
in
tr
o
d
u
ci
n
g
v
ar
io
u
s
W
o
r
d
Net
co
n
ce
p
ts
.
Utilizatio
n
o
f
s
u
c
h
i
n
n
o
v
ati
v
e
k
n
o
w
led
g
e
b
ase
co
n
ce
p
ts
i
m
p
r
o
v
es
th
e
q
u
ali
t
y
o
f
tex
t
d
o
cu
m
e
n
t
a
s
d
ep
icted
in
[
4
]
.
B
y
m
ap
p
in
g
th
e
g
i
v
e
n
co
n
ten
t
to
th
e
s
e
m
an
tic
s
p
ac
e
w
h
ic
h
is
o
f
f
er
ed
th
r
o
u
g
h
W
ik
i
p
ed
ia
p
ag
es,
it
h
as
b
ee
n
p
r
o
v
ed
as
an
ef
f
icien
t
k
n
o
w
led
g
e
b
ase
an
d
is
b
est
s
u
i
tab
le
f
o
r
Sh
o
r
t
T
ex
t
C
lass
i
f
ica
tio
n
an
d
D
o
cu
m
e
n
t
C
lu
s
ter
i
n
g
[
9
]
,
[
20
]
.
I
n
ad
d
itio
n
,
in
[
1
6
]
,
o
th
er
t
w
o
k
n
o
w
le
d
g
e
b
ases
in
cl
u
d
in
g
P
r
o
b
ase
an
d
T
ax
o
n
o
m
y
ar
e
in
tr
o
d
u
ce
d
.
T
h
ese
k
n
o
w
led
g
e
b
ases
ar
e
m
aj
o
r
l
y
in
v
o
l
v
ed
in
en
h
a
n
ci
n
g
th
e
ad
k
e
y
w
o
r
d
s
f
ea
tu
r
es
i
n
o
r
d
er
to
b
u
ild
a
n
o
v
e
l
tax
o
n
o
m
y
o
f
k
e
y
w
o
r
d
s
w
h
i
c
h
ar
e
d
o
m
ai
n
d
e
p
en
d
en
t.
T
h
u
s
,
i
t
m
ig
h
t
b
e
s
ig
n
i
f
ican
t
to
co
n
s
id
er
th
e
k
n
o
w
led
g
e
a
s
“
S
u
p
er
v
i
s
io
n
”
to
d
ir
ec
t
t
h
e
o
t
h
er
Ma
c
h
i
n
e
L
ea
r
n
in
g
T
ec
h
n
iq
u
e
s
a
n
d
d
is
tin
c
t
tas
k
s
.
Di
s
tan
t
Su
p
er
v
i
s
io
n
lear
n
i
n
g
s
c
h
e
m
e
e
m
p
lo
y
s
in
f
o
r
m
a
tio
n
e
n
tit
ies
a
n
d
r
esp
ec
tiv
e
r
elatio
n
s
f
r
o
m
F
r
ee
b
ase
k
n
o
w
led
g
e
b
ases
as
s
u
p
er
v
is
io
n
to
ex
ec
u
t
e
en
tit
y
an
d
r
elatio
n
e
x
tr
ac
tio
n
[
1
5
]
,
[
1
3
]
an
d
[
1
4
]
.
I
n
ad
d
itio
n
,
it
also
e
m
p
lo
y
s
k
n
o
w
led
g
e
s
u
p
er
v
is
io
n
f
o
r
ex
tr
ac
tin
g
m
o
r
e
e
n
titi
e
s
an
d
r
ela
tio
n
s
h
ip
s
f
r
o
m
t
h
e
n
o
v
el
co
n
t
en
t
o
r
also
u
s
ed
f
o
r
g
en
er
ati
n
g
a
n
e
f
f
icie
n
t
in
s
tall
atio
n
o
f
b
o
th
en
t
ities
an
d
r
ela
tio
n
s
.
T
h
u
s
,
e
x
p
lo
itatio
n
o
f
d
i
r
ec
t
s
u
p
er
v
is
io
n
i
s
r
estricte
d
to
k
n
o
w
led
g
e
e
n
titi
e
s
an
d
r
elatio
n
s
a
m
o
n
g
t
h
e
m
.
4.
CO
NCLU
SI
O
N
T
h
is
r
esear
ch
p
ap
er
d
is
c
u
s
s
e
s
a
d
etailed
s
u
r
v
e
y
o
f
d
i
f
f
er
en
t
clu
s
ter
i
n
g
ap
p
r
o
ac
h
es
f
o
r
d
ata
m
i
n
i
n
g
i
n
th
e
te
x
t
en
v
ir
o
n
m
e
n
t.
An
ef
f
i
cien
t
te
x
t
cl
u
s
ter
i
n
g
ap
p
r
o
ac
h
m
u
s
t
c
h
o
o
s
e
o
p
ti
m
al
attr
ib
u
t
es
alo
n
g
w
it
h
th
e
r
ig
h
t
al
g
o
r
ith
m
f
o
r
e
x
ec
u
tio
n
.
Of
v
ar
io
u
s
t
y
p
e
s
o
f
al
g
o
r
ith
m
s
f
o
u
n
d
in
liter
at
u
r
e,
d
is
ta
n
ce
-
b
ased
ap
p
r
o
ac
h
es
ar
e
o
b
s
er
v
ed
to
b
e
b
o
th
ef
f
ici
en
t
a
n
d
w
id
el
y
i
m
p
le
m
en
ted
ac
r
o
s
s
d
if
f
er
en
t
d
o
m
a
in
s
.
Ov
er
th
e
p
ast
f
e
w
y
ea
r
s
,
r
esear
ch
er
s
w
o
r
k
i
n
g
o
n
tex
t c
l
u
s
ter
i
n
g
f
o
cu
s
ed
o
n
t
w
o
t
y
p
es
o
f
ap
p
licatio
n
s
.
a.
D
y
n
a
m
ic
:
Hu
g
e
v
o
lu
m
i
n
o
u
s
i
n
f
o
r
m
atio
n
g
e
n
er
ated
in
d
y
n
a
m
ic
e
n
v
ir
o
n
m
e
n
t
s
in
cl
u
d
i
n
g
s
o
cial
n
et
w
o
r
k
i
n
g
p
latf
o
r
m
s
o
r
o
n
li
n
e
c
h
at
r
es
u
l
ted
in
a
s
tr
o
n
g
r
eq
u
ir
e
m
en
t
f
o
r
s
tr
ea
m
in
g
i
n
f
o
r
m
atio
n
.
T
h
ese
ap
p
licatio
n
s
s
h
o
u
ld
b
e
ad
ap
tab
le
to
s
ce
n
ar
io
s
w
h
er
e
th
e
te
x
t is o
f
ten
n
o
t c
lear
lik
e
th
e
s
o
cial
n
et
w
o
r
k
in
g
p
latf
o
r
m
s
.
b.
Hete
r
o
g
en
eo
u
s
:
I
n
t
h
ese
ap
p
licatio
n
s
,
t
h
e
te
x
t
is
o
f
te
n
p
r
ese
n
t
a
s
li
n
k
s
a
n
d
d
i
f
f
er
e
n
t
m
u
lti
m
ed
ia
f
o
r
m
at
s
.
Fo
r
in
s
tan
ce
,
in
p
lat
f
o
r
m
s
li
k
e
Fli
c
k
r
,
te
x
t
cl
u
s
ter
in
g
s
h
o
u
l
d
b
e
ad
ap
ted
.
A
cc
o
r
d
in
g
l
y
,
it
is
i
m
p
o
r
tan
t
to
ef
f
icien
tl
y
i
m
p
le
m
e
n
t te
x
t
m
i
n
in
g
ap
p
r
o
ac
h
es i
n
th
is
e
n
v
ir
o
n
m
en
t.
T
h
is
m
a
n
u
s
cr
ip
t
o
b
s
er
v
es
t
h
at
th
e
ar
ea
o
f
cl
u
s
ter
in
g
i
n
t
ex
t
m
i
n
in
g
ap
p
licatio
n
s
is
wid
e
an
d
is
ch
alle
n
g
i
n
g
to
co
m
p
letel
y
p
r
esen
t
i
n
o
n
e
p
ap
er
.
C
er
tain
ap
p
r
o
ac
h
es
lik
e
co
m
m
ittee
-
d
r
iv
en
clu
s
ter
in
g
co
u
ld
n
o
t
b
e
clea
r
ly
ca
te
g
o
r
ized
in
t
o
an
y
g
r
o
u
p
s
as
t
h
e
y
i
n
co
r
p
o
r
ate
m
u
l
tip
le
clu
s
ter
in
g
tech
n
i
q
u
es
to
g
en
er
ate
t
h
e
f
i
n
al
o
u
tco
m
e.
T
h
e
in
ten
t
io
n
o
f
th
is
w
o
r
k
is
to
p
u
t
f
o
r
w
ar
d
a
co
m
p
lete
b
r
ief
o
f
p
r
o
m
in
e
n
t
ap
p
r
o
ac
h
es
p
r
o
p
o
s
ed
f
o
r
tex
t
m
i
n
i
n
g
,
to
s
e
r
v
e
as a
n
i
n
itial
s
tep
f
o
r
o
th
er
p
r
o
p
o
s
ed
r
esear
ch
co
n
tr
ib
u
tio
n
s
.
RE
F
E
R
E
NC
E
S
[1
]
Ja
in
,
A
n
il
K.
,
"
Da
ta
Clu
ste
rin
g
:
5
0
Ye
a
rs
b
e
y
o
n
d
K
-
M
e
a
n
s
,
"
Pa
tt
e
rn
re
c
o
g
n
it
io
n
letter
s
,
v
o
l.
31
(
8
)
,
p
p
.
6
5
1
-
6
6
6
2
0
1
0
.
[2
]
V
ij
a
y
S
o
n
a
w
a
n
e
,
D.
Ra
jes
h
w
a
ra
Ra
o
,
“
A
n
Op
ti
m
isti
c
A
p
p
ro
a
c
h
f
o
r
Clu
ste
rin
g
M
u
lt
i
-
v
e
rsio
n
X
M
L
D
o
c
u
m
e
n
ts
Us
in
g
Co
m
p
re
ss
e
d
De
lt
a
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
El
e
c
c
trica
l
a
n
d
Co
m
p
u
ter
En
g
in
e
e
rin
g
,
v
o
l.
5
(
6
)
,
p
p
.
1
4
7
2
-
1
4
7
9
,
De
c
2
0
1
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
2
,
A
p
r
il 2
0
1
9
:
1
3
1
3
-
1320
1320
[3
]
S
riv
id
y
a
S
iv
a
sa
n
k
a
r,
S
ru
th
i
Na
ir,
M
.
V.
Ju
d
y
,
“
F
e
a
tu
re
Re
d
u
c
ti
o
n
in
Cli
n
ica
l
Da
ta Cl
a
ss
i
f
ica
ti
o
n
u
si
n
g
a
u
g
m
e
n
ted
-
G
e
n
e
ti
c
A
l
g
o
rit
h
m
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
E
lec
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
,
v
ol
.
5
(
6
)
,
p
p
.
1
5
1
6
-
1
5
2
4
,
De
c
2
0
1
5
.
[4
]
M
u
g
u
n
t
h
a
d
e
v
i
K.,
e
t
a
l.
,
"
S
u
rv
e
y
o
n
F
e
a
tu
re
S
e
lec
ti
o
n
i
n
Do
c
u
m
e
n
t
Clu
ste
rin
g
,
"
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
n
Co
mp
u
ter
S
c
ien
c
e
a
n
d
E
n
g
in
e
e
rin
g
,
v
o
l.
3
(
3
)
,
p
p
.
1
2
4
0
-
1
2
4
1
,
2
0
1
1
.
[5
]
F
o
rm
a
n
,
G
e
o
rg
e
,
a
n
d
Ev
a
n
Ki
rsh
e
n
b
a
u
m
,
"
Ex
tre
m
e
l
y
F
a
st
T
e
x
t
F
e
a
tu
re
Ex
tr
a
c
ti
o
n
f
o
r
Clas
s
if
ica
ti
o
n
a
nd
In
d
e
x
in
g
,
"
Pro
c
e
e
d
in
g
s o
f
th
e
1
7
t
h
ACM
c
o
n
fer
e
n
c
e
o
n
I
n
fo
rm
a
ti
o
n
a
n
d
k
n
o
wle
d
g
e
ma
n
a
g
e
me
n
t.
A
CM
,
2
0
0
8
.
[6
]
L
i,
Ya
n
ju
n
,
S
o
o
n
M
.
Ch
u
n
g
,
a
n
d
Jo
h
n
D.
Ho
lt
,
"
T
e
x
t
Do
c
u
m
e
n
t
C
lu
ste
rin
g
Ba
se
d
o
n
F
re
q
u
e
n
t
W
o
rd
M
e
a
n
in
g
S
e
q
u
e
n
c
e
s
,
"
Da
ta
&
Kn
o
wled
g
e
En
g
i
n
e
e
rin
g
,
v
o
l.
64
(
1
),
p
p
.
3
8
1
-
404
,
2
0
0
8
.
[7
]
Ch
a
n
d
ra
sh
e
k
a
r,
G
iri
sh
,
a
n
d
F
e
r
a
tS
a
h
in
,
"
A
S
u
rv
e
y
o
n
F
e
a
tu
re
S
e
lec
ti
o
n
M
e
th
o
d
s
,
"
Co
mp
u
ter
s
&
El
e
c
trica
l
En
g
i
n
e
e
rin
g
.
V
o
l.
40
(
1
)
,
p
p
.
16
-
28
,
2
0
0
8
.
[8
]
P
e
d
ra
m
V
a
h
d
a
n
i
Am
o
li
,
Om
id
S
o
jo
o
d
i
S
h
,
“
S
c
ien
ti
f
ic
Do
c
u
m
e
n
ts
Clu
ste
rin
g
Ba
se
d
o
n
T
e
x
t
S
u
m
m
e
ri
z
a
ti
o
n
,
”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
E
lec
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
V
o
l
.
5
(
4
),
p
p
.
7
8
2
-
7
8
7
,
A
u
g
2
0
1
5
.
[9
]
L
a
w
,
M
a
rti
n
HC
.
,
M
a
rio
A
T
.
F
ig
u
e
ired
o
,
a
n
d
A
n
il
K.
Ja
i
n
,
"
S
im
u
lt
a
n
e
o
u
s
F
e
a
tu
re
S
e
lec
ti
o
n
a
nd
Cl
u
ste
rin
g
u
sin
g
M
ix
tu
re
M
o
d
e
ls,
"
IEE
E
T
ra
n
s
a
c
ti
o
n
s
o
n
P
a
tt
e
rn
An
a
lys
is
a
n
d
M
a
c
h
in
e
I
n
telli
g
e
n
c
e
,
v
o
l.
26
(
9
),
1
1
5
4
-
1
1
6
6
,
2
0
0
4
.
[1
0
]
G
a
b
ril
o
v
ich
,
Ev
g
e
n
i
y
,
a
n
d
S
h
a
u
l
M
a
rk
o
v
it
c
h
.
"
F
e
a
tu
re
G
e
n
e
r
a
ti
o
n
f
o
r
T
e
x
t
Ca
te
g
o
riza
ti
o
n
Us
in
g
W
o
rld
Kn
o
w
led
g
e
,
"
IJ
CAI
,
v
o
l.
5
,
2
0
0
5
.
[1
1
]
A.
Ko
u
sa
r
Nik
h
a
th
,
K.
S
u
b
ra
h
m
a
n
y
a
m
,
“
In
c
re
m
e
n
tal
Ev
o
lu
ti
o
n
a
ry
Ge
n
e
ti
c
A
lg
o
rit
h
m
Ba
se
d
O
p
ti
m
a
l
Do
c
u
m
e
n
t
Clu
ste
rin
g
,
”
J
o
u
r
n
a
l
o
f
T
h
e
o
riti
c
a
l
a
n
d
Ap
p
li
e
d
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
lo
g
y
,
v
ol
.
87
(
3
)
,
M
a
y
2018
.
[1
2
]
A.
Ko
u
sa
r
Ni
k
h
a
th
,
K.
S
u
b
ra
h
m
a
n
a
y
m
,
“
Co
n
c
e
p
tu
a
l
Re
lev
a
n
c
e
Ba
se
d
Do
c
u
m
e
n
t
Clu
ste
r
in
g
Us
in
g
Co
n
c
e
p
t
Util
it
y
S
c
a
le,
”
Asia
n
J
o
u
r
n
a
l
o
f
S
c
ien
ti
fi
c
Res
e
a
rc
h
,
v
o
l
.
1
1
(1
)
,
p
p
.
22
-
31,
2
0
1
8
.
[1
3
]
W
h
isse
ll
,
Jo
h
n
S
.
,
a
n
d
Ch
a
rles
LA
.
Clark
e
.
"
I
m
p
ro
v
in
g
Do
c
u
m
e
n
t
Clu
ste
rin
g
u
sin
g
Ok
a
p
i
B
M
2
5
F
e
a
tu
re
W
e
ig
h
ti
n
g
,
"
In
fo
rm
a
ti
o
n
re
triev
a
l
,
v
o
l.
14
(
5
),
p
p
.
4
6
6
-
4
8
7
,
2
0
1
1
.
[1
4
]
Kw
a
k
,
No
ju
n
,
a
n
d
Ch
o
n
g
-
Ho
C
h
o
i
,
"
In
p
u
t
F
e
a
tu
re
S
e
lec
ti
o
n
f
o
r
C
las
sif
ic
a
ti
o
n
P
ro
b
lem
s
,
"
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Ne
u
ra
l
Ne
two
rk
s
,
v
o
l.
13
(
1
),
p
p
.
143
-
1
5
9
,
2
0
0
2
.
[1
5
]
L
e
e
,
K
y
u
n
g
S
o
o
n
,
W
.
Bru
c
e
Cro
ft,
a
n
d
Ja
m
e
s
A
ll
a
n
.
"
A
C
lu
ste
r
-
b
a
se
d
Re
sa
m
p
li
n
g
M
e
th
o
d
f
o
r
P
se
u
d
o
-
re
lev
a
n
c
e
F
e
e
d
b
a
c
k
,
"
Pro
c
e
e
d
in
g
s o
f
th
e
3
1
st a
n
n
u
a
l
in
ter
n
a
ti
o
n
a
l
ACM
S
IGIR
c
o
n
fer
e
n
c
e
o
n
Res
e
a
rc
h
a
n
d
d
e
v
e
lo
p
me
n
t
in
in
fo
rm
a
ti
o
n
re
triev
a
l.
ACM
,
2
0
0
8
.
[1
6
]
P
e
n
g
,
Ha
n
c
h
u
a
n
,
F
u
h
u
i
L
o
n
g
,
a
n
d
Ch
ris
Din
g
.
"
F
e
a
tu
re
S
e
lec
ti
o
n
Ba
se
d
on
M
u
t
u
a
l
In
f
o
rm
a
ti
o
n
C
rit
e
ria
o
f
M
a
x
-
De
p
e
n
d
e
n
c
y
,
M
a
x
-
re
le
v
a
n
c
e
,
a
n
d
M
in
-
r
e
d
u
n
d
a
n
c
y
,
"
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
p
a
t
ter
n
a
n
a
lys
is
a
n
d
ma
c
h
i
n
e
in
telli
g
e
n
c
e
,
v
o
l.
27
(
8
),
p
p
.
1
2
2
6
-
1
2
3
8
,
2
0
0
5
.
[1
7
]
S
u
n
Y.,
C.
F
.
Ba
b
b
s,
a
n
d
E.
J.
De
lp
.
"
A
Co
m
p
a
riso
n
o
f
F
e
a
tu
re
S
e
lec
ti
o
n
M
e
th
o
d
s
f
o
r
th
e
De
t
e
c
t
io
n
o
f
Bre
a
st
Ca
n
c
e
rs
in
M
a
m
m
o
g
r
a
m
s:
A
d
a
p
t
iv
e
S
e
q
u
e
n
ti
a
l
F
lo
a
ti
n
g
S
e
a
rc
h
v
s.
G
e
n
e
ti
c
A
lg
o
rit
h
m
,
"
En
g
in
e
e
rin
g
in
M
e
d
icin
e
a
n
d
Bi
o
lo
g
y
S
o
c
iety
,
2
0
0
5
.
IE
EE
-
EM
B
S
2
0
0
5
.
2
7
t
h
A
n
n
u
a
l
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
f
th
e
.
IEE
E,
2
0
0
6
.
[1
8
]
L
u
,
Yiju
a
n
,
e
t
a
l
.
"
F
e
a
tu
re
S
e
lec
ti
o
n
u
si
n
g
P
r
in
c
ip
a
l
F
e
a
tu
re
A
n
a
l
y
sis,
"
Pro
c
e
e
d
in
g
s
o
f
th
e
1
5
th
ACM
in
ter
n
a
t
io
n
a
l
c
o
n
fer
e
n
c
e
o
n
M
u
lt
i
me
d
ia
.
ACM
,
2
0
0
7
.
[1
9
]
W
o
lf
,
L
io
r,
a
n
d
Am
n
o
n
S
h
a
sh
u
a
.
"
F
e
a
tu
re
S
e
lec
ti
o
n
f
o
r
Un
su
p
e
rv
i
se
d
a
n
d
S
u
p
e
rv
ise
d
In
f
e
re
n
c
e
:
Th
e
e
m
e
r
g
e
n
c
e
o
f
S
p
a
rsit
y
in
a
Weig
h
t
-
b
a
se
d
Ap
p
r
o
a
c
h
,
"
J
o
u
rn
a
l
o
f
M
a
c
h
in
e
L
e
a
rn
in
g
Res
e
a
rc
h
,
v
o
l.
6
,
p
p
.
1
8
5
5
-
1
8
8
7
,
No
v
2
0
0
5
.
[2
0
]
S
u
n
,
Z
h
a
n
q
u
a
n
,
e
t
a
l.
"
A
P
a
ra
ll
e
l
Clu
ste
rin
g
M
e
th
o
d
C
o
m
b
in
e
d
In
f
o
rm
a
ti
o
n
Bo
tt
len
e
c
k
T
h
e
o
ry
a
n
d
Ce
n
tro
i
d
-
Ba
se
d
Clu
ste
rin
g
,
"
T
h
e
J
o
u
rn
a
l
o
f
S
u
p
e
rc
o
mp
u
ti
n
g
,
v
o
l
.
69
(
1
),
p
p
.
4
5
2
-
4
6
7
,
2
0
1
4
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
A.
K
o
u
sa
r
Ni
k
h
a
t
h
is
c
u
rre
n
tl
y
w
o
rk
in
g
a
s
A
s
st.
P
ro
f
e
ss
o
r
in
Co
m
p
u
ter
S
c
ien
c
e
&
En
g
in
e
e
rin
g
De
p
a
rtme
n
t
a
t
V
NR
V
JIET
,
H
y
d
e
ra
b
a
d
.
S
h
e
is
i
n
t
o
tea
c
h
in
g
p
r
o
f
e
s
sio
n
f
o
r
th
e
p
a
st
1
3
y
e
a
rs.
S
h
e
is
c
u
rre
n
tl
y
p
u
rsu
in
g
P
h
.
D.
a
t
Ko
n
e
ru
L
a
k
sh
a
m
a
iah
Ed
u
c
a
ti
o
n
F
o
u
n
d
a
ti
o
n
,
G
u
n
tu
r
.
S
h
e
h
a
s
p
u
b
l
ish
e
d
n
e
a
rly
a
b
o
u
t
1
0
p
a
p
e
rs
in
v
a
rio
u
s
Jo
u
rn
a
ls/
In
tern
a
t
i
o
n
a
l
c
o
n
f
e
re
n
c
e
s.
He
r
re
se
a
rc
h
a
re
a
in
tere
st
in
c
lu
d
e
d
T
e
x
t
m
in
in
g
,
Da
ta m
in
in
g
,
Do
c
u
m
e
n
t
Clu
ste
rin
g
,
A
rti
f
icia
l
In
telli
g
e
n
c
e
a
n
d
Ne
u
ra
l
Ne
tw
o
rk
.
Dr
.
K
S
u
b
r
a
h
m
a
n
y
a
m
,
a
G
o
ld
M
e
d
a
li
st
f
ro
m
A
n
d
h
ra
Un
iv
e
rsit
y
(1
9
9
2
-
9
3
)
is
c
u
rre
n
tl
y
w
o
rk
in
g
a
s
a
P
r
o
f
e
ss
o
r
in
Co
m
p
u
ter
S
c
ien
c
e
&
En
g
in
e
e
rin
g
De
p
a
rt
m
e
n
t
o
f
Ko
n
e
ru
L
a
k
sh
a
m
a
iah
Ed
u
c
a
ti
o
n
F
o
u
n
d
a
ti
o
n
,
G
u
n
tu
r.
He
is
in
te
a
c
h
in
g
p
ro
f
e
ss
io
n
f
o
r
th
e
p
a
st
2
5
y
e
a
rs
a
n
d
p
ri
o
r
to
jo
i
n
in
g
Ko
n
e
ru
L
a
k
sh
a
m
a
iah
Ed
u
c
a
ti
o
n
F
o
u
n
d
a
ti
o
n
h
e
w
o
rk
e
d
a
s
P
ro
g
ra
m
m
e
Lea
d
e
r
in
th
e
S
c
h
o
o
l
o
f
En
g
in
e
e
rin
g
,
S
c
ien
c
e
&
Tec
h
n
o
lo
g
y
a
t
KD
U
Un
iv
e
rsit
y
,
M
a
la
y
sia
f
o
r
a
b
o
u
t
1
0
y
e
a
rs.
He
h
a
s
p
u
b
li
s
h
e
d
m
o
re
th
a
n
4
0
p
a
p
e
rs
in
b
o
th
n
a
ti
o
n
a
l
a
n
d
i
n
ter
n
a
ti
o
n
a
l
j
o
u
r
n
a
ls
a
n
d
c
o
n
f
e
re
n
c
e
s
a
n
d
a
tt
e
n
d
e
d
v
a
rio
u
s
w
o
rk
sh
o
p
s
in
M
a
la
y
sia
,
S
in
g
a
p
o
re
,
USA
&
In
d
ia.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
Kn
o
w
led
g
e
&
S
o
f
t
w
a
re
En
g
in
e
e
r
in
g
,
Da
ta
M
in
in
g
,
S
o
f
t
S
y
ste
m
s
M
e
th
o
d
o
lo
g
ies
.
He
h
a
s
g
u
i
d
e
d
1
0
0
o
v
e
r
stu
d
e
n
ts
to
w
a
rd
s
th
e
ir
M
a
ste
r‟s
a
n
d
Ba
c
h
e
lo
r
Diss
e
rtatio
n
s,
a
n
d
c
u
rre
n
tl
y
g
u
id
in
g
8
to
w
a
rd
s th
e
ir
P
h
D.
Evaluation Warning : The document was created with Spire.PDF for Python.