I
n
t
ern
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
E
l
ect
ri
ca
l
a
n
d
C
o
m
p
u
t
er E
n
g
i
n
eeri
n
g
(
I
J
E
C
E
)
V
o
l.
8
,
No
.
5,
O
c
t
obe
r
20
1
8,
p
p.
32
04~
3
213
I
S
S
N
:
2088
-
8708
,
D
O
I
:
10.
11
591/
i
j
ece
.
v8
i
5
.
pp
320
4
-
3213
3204
Jou
r
n
al
h
om
e
p
age
:
h
ttp
:
//ia
e
s
c
o
r
e
.
c
o
m/
j
our
nal
s
/
i
nde
x
.
php/
I
J
E
C
E
O
p
t
i
mi
s
at
i
on
t
o
w
a
rds
L
a
t
ent
Diric
hlet
Al
lo
ca
t
io
n
:
I
t
s
T
o
pic
Nu
m
ber
a
nd
Co
ll
a
ps
ed
G
ibbs
Sa
m
pling
I
nf
ere
nce
P
ro
ces
s
B
a
m
ba
ng
Su
be
no
1
,
R
e
t
no
K
us
u
m
a
ni
ng
r
u
m
2
,
F
a
r
i
khi
n
3
1
I
nf
or
m
a
t
i
on S
y
s
t
e
m
,
U
ni
v
e
r
s
i
t
a
s
D
i
pone
g
or
o
,
I
nd
one
s
i
a
2
D
ep
ar
t
m
en
t
o
f
I
n
f
o
r
m
at
i
cs
,
U
n
i
v
e
r
s
i
t
a
s
D
i
pone
g
or
o
,
I
n
do
ne
s
i
a
3
D
ep
ar
t
m
en
t
o
f
M
at
h
em
at
i
cs
,
U
n
i
v
er
s
i
t
as
D
i
p
o
n
eg
o
r
o
,
I
nd
one
s
i
a
A
rt
i
cl
e I
n
f
o
AB
S
T
RAC
T
A
r
tic
le
h
is
to
r
y
:
R
e
c
e
i
v
e
d N
ov
20,
2017
R
e
v
i
s
e
d J
a
n
19,
20
18
A
ccep
t
ed
A
u
g
23,
2018
L
a
t
e
nt
D
i
r
i
c
hl
e
t
A
l
l
oc
a
t
i
on (
L
D
A
)
i
s
a
pr
oba
bi
l
i
t
y
m
ode
l
f
or
g
r
oupi
ng
hi
d
de
n
t
opi
c
s
i
n d
oc
um
e
nt
s
b
y
t
he
num
be
r
of
pr
e
de
f
i
ne
d t
opi
c
s
.
I
f
co
n
d
u
ct
ed
in
c
o
r
r
e
c
tly
,
d
e
te
r
m
in
in
g
th
e
a
m
ount
o
f
K
to
p
ic
s
w
ill r
e
s
u
lt in
li
m
ite
d
w
o
r
d
c
or
r
e
l
a
t
i
on w
i
t
h t
op
i
c
s
.
T
oo l
a
r
ge
or
t
oo s
m
a
l
l
nu
m
be
r
of
K
t
opi
c
s
c
a
us
e
s
i
na
c
c
ur
a
c
i
e
s
i
n g
r
oupi
ng
t
o
pi
c
s
i
n t
he
f
or
m
a
t
i
on of
t
r
a
i
ni
ng
m
o
de
l
s
.
T
hi
s
s
t
udy
a
i
m
s
t
o de
t
e
r
m
i
ne
t
he
opt
i
m
a
l
nu
m
be
r
of
c
or
pus
t
o
pi
c
s
i
n
t
he
LD
A
m
e
t
hod us
i
ng
t
he
m
a
x
i
m
u
m
l
i
ke
l
i
ho
od a
n
d M
i
ni
m
um
D
e
s
c
r
i
pt
i
on L
e
ng
t
h
(
M
D
L
)
a
ppr
oa
c
h.
T
h
e e
x
p
er
i
m
e
n
t
al
p
r
o
ces
s
u
s
es
I
ndone
s
i
a
n
n
e
w
s
a
r
tic
le
s
w
i
t
h t
he
num
be
r
of
doc
um
e
nt
s
a
t
25,
5
0,
9
0,
a
nd
60
0;
i
n e
a
c
h doc
um
e
nt
,
t
h
e
num
be
r
s
of
w
or
ds
ar
e
389
8,
77
6
0,
13
00
5,
a
n
d
43
65.
T
he
r
e
s
ul
t
s
s
how
t
ha
t
t
he
m
a
x
i
m
u
m
l
i
k
e
l
i
hood a
n
d M
D
L
a
ppr
oa
c
h r
e
s
ul
t
i
n t
he
s
a
m
e
num
be
r
of
opt
i
m
a
l
t
opi
c
s
.
T
he
o
pt
i
m
a
l
num
be
r
of
t
opi
c
s
i
s
i
nf
l
ue
nc
e
d by
a
l
pha
a
nd
be
t
a
p
ar
am
et
er
s
.
I
n a
ddi
t
i
on
,
t
he
nu
m
be
r
o
f
doc
um
e
nt
s
doe
s
n
o
t
af
f
ect
t
h
e
c
om
put
a
t
io
n
tim
e
s
but
t
he
num
be
r
of
w
or
ds
doe
s
.
C
om
put
a
t
i
ona
l
t
i
m
e
s
f
or
e
a
c
h of
t
hos
e
da
t
a
s
e
t
s
a
r
e
2.
97
21
,
6.
4
96
37
,
1
3.
29
67,
a
nd
3.
71
52 s
e
c
onds
.
T
he
o
p
tim
i
s
a
t
i
on m
ode
l
ha
s
r
e
s
ul
t
e
d
in
ma
n
y
LD
A
to
p
ic
s
a
s
a
c
la
s
s
if
ic
a
tio
n
m
o
d
el
.
T
h
i
s
ex
p
er
i
m
en
t
s
h
o
w
s
t
h
at
t
h
e
h
i
g
h
es
t
av
er
ag
e ac
cu
r
ac
y
i
s
6
1
%
w
i
t
h
a
l
pha
0
.
1 a
nd
be
t
a
0.
00
1.
Ke
y
wo
rd
:
L
a
te
n
t D
ir
ic
h
le
t a
llo
c
a
tio
n
L
i
ke
l
i
ho
o
d
M
in
i
m
u
m
d
e
s
c
r
ip
tio
n
le
n
g
t
h
N
um
be
r
of
t
opi
c
s
O
p
ti
m
i
s
a
tio
n
C
opy
r
i
g
ht
©
201
8
I
ns
t
i
t
ut
e
o
f
A
d
v
anc
e
d E
ngi
ne
e
r
i
ng
an
d
Sc
i
e
nc
e
.
A
l
l
ri
g
h
t
s re
se
rv
e
d
.
Co
rre
sp
o
n
d
i
n
g
Au
t
h
o
r
:
R
e
t
n
o K
u
s
um
a
ni
ng
r
um
,
D
ep
ar
t
m
en
t
o
f
I
n
f
o
r
m
at
i
c
s
,
U
ni
ve
r
s
i
t
a
s
D
i
p
o
ne
go
r
o
,
J
l
.
P
r
of
.
S
oe
da
r
t
o,
S
H
,
T
e
m
ba
l
a
n
g
,
S
e
m
a
r
a
n
g
,
50275
,
I
nd
o
ne
s
i
a
.
E
m
a
il:
r
et
n
o
@
l
i
v
e.
u
n
d
i
p
.
ac.
i
d
1.
I
NT
RO
D
UCT
I
O
N
N
o
wa
d
a
y
s
,
t
e
x
t
m
i
ni
n
g
i
s
w
i
d
el
y
i
m
p
l
e
m
e
n
t
ed
d
u
e
t
o
a
w
i
d
e v
ar
i
et
y
o
f
t
ex
t
t
y
p
es
,
s
u
c
h
as
n
e
w
s
a
r
tic
le
s
,
s
c
ie
n
t
if
ic
a
r
tic
le
s
,
b
o
o
k
s
,
e
m
a
il
m
e
s
s
a
g
e
s
,
e
tc
.
F
u
r
th
e
r
m
o
r
e
,
it
en
co
u
r
a
g
es
an
i
n
cr
eas
ed
n
eed
t
o
ex
t
r
act
t
h
e i
n
f
o
r
m
a
t
i
o
n c
o
nt
a
i
ne
d
i
n a
d
o
c
u
m
e
nt
.
F
u
r
t
h
er
m
o
r
e,
i
t
en
co
u
r
a
g
es
a
n
i
n
cr
eas
ed
n
eed
t
o
ex
t
r
act
t
h
e
i
nf
or
m
a
t
i
on c
on
t
a
i
n
e
d i
n
a
doc
um
e
n
t
t
o
g
e
n
e
r
a
t
e
us
e
f
u
l
kn
o
w
l
e
d
g
e
[1
],
[2
],
[3
],
[4
]
.
T
h
e d
i
f
f
er
e
n
ce b
et
w
ee
n
n
e
ws
a
r
tic
le
s
o
r
te
x
tu
a
l a
r
tic
le
s
d
is
s
e
m
i
n
a
te
d
th
r
o
u
g
h
e
le
c
tr
o
n
ic
m
e
d
ia
w
it
h
o
th
e
r
d
o
c
u
m
e
n
ts
i
s
th
e
m
o
d
e
l o
f
i
n
f
o
r
m
a
t
i
o
n
f
l
o
w
.
T
he
ne
w
s
f
l
o
w
i
s
a
d
yna
m
i
c
a
nd
c
o
nt
i
nuo
us
l
y
up
d
a
t
e
d
s
tr
e
a
m
; th
e
m
o
r
e
th
e
n
e
w
s
a
r
tic
le
in
el
ect
r
o
n
i
c
m
ed
i
a
i
s
,
t
h
e
mo
r
e
e
xt
e
n
s
i
ve
t
h
e
d
at
a
co
l
l
ect
i
o
n
as
i
t
al
w
a
y
s
i
n
cr
eas
es
[5
]
.
W
i
t
h
e
no
r
m
o
us
d
a
t
a
va
r
i
a
t
i
o
n
s
,
p
r
o
b
l
e
m
s
o
c
c
ur
w
he
n
ne
e
d
i
ng t
o
t
a
ke
o
n
t
he
d
i
f
f
e
r
e
nt
ne
w
s
w
h
i
l
e
ha
vi
n
g t
he
s
a
m
e
t
he
m
e
.
S
o
,
t
o
f
a
c
ilita
te
n
a
v
ig
a
tio
n
,
n
e
w
s
a
r
ti
c
le
s
m
u
s
t b
e
g
r
o
u
p
e
d
b
y
t
h
e
s
a
m
e
to
p
ic
.
O
ne
w
a
y
t
o
ge
t
t
he
to
p
ic
in
f
o
r
m
a
tio
n
c
o
n
ta
i
n
e
d
in
t
h
e
c
o
r
p
u
s
o
f
a
n
e
w
s
a
r
tic
le
d
o
c
u
m
e
n
t is
to
u
s
e
to
p
ic
m
ode
l
l
i
ng
.
L
a
te
n
t D
ir
ic
h
le
t
A
llo
c
a
tio
n
(
L
D
A
)
is
a
to
p
i
c
m
o
d
e
llin
g
t
e
c
hni
q
ue
t
ha
t
c
a
n gr
o
up
w
o
r
d
s
i
n
t
o
s
p
e
c
if
ic
to
p
ic
s
f
r
o
m
v
a
r
io
u
s
m
a
te
r
ia
ls
[6
]
.
T
h
e
n
um
be
r
of
t
opi
c
s
c
on
t
a
i
n
e
d i
n
t
h
e
c
or
pu
s
w
i
t
h
m
u
ltip
le
v
ar
i
at
i
o
n
s
i
s
n
eces
s
ar
y
t
o
o
p
tim
i
s
e
t
he
nu
m
b
e
r
o
f
to
p
ic
s
lis
t
e
d
w
i
t
hi
n
t
he
c
o
r
p
us
.
T
h
er
e
ar
e
s
e
v
e
r
a
l e
s
t
i
m
a
tio
n
a
l
go
r
i
t
h
m
s
us
e
d
i
n L
D
A
i
nc
lu
d
i
n
g
E
x
p
e
c
ta
tio
n
-
M
a
x
i
m
iz
a
tio
n
a
lg
o
r
it
h
m
[6
]
,
E
x
p
e
c
ta
tio
n
-
P
r
opa
g
a
t
i
on
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
I
S
S
N
:
2088
-
8708
O
p
tim
is
a
tio
n
to
w
a
r
d
s
L
a
te
n
t D
ir
ic
h
le
t A
llo
c
a
tio
n
:
I
t
s
T
opi
c
N
um
be
r
and C
ol
l
aps
e
d
…
(
B
a
m
bang Sube
no
)
3205
a
lg
o
r
ith
m
to
o
b
ta
i
n
b
e
tte
r
a
c
c
u
r
a
c
y
[7
]
,
a
s
w
e
l
l
a
s
C
ol
l
a
ps
e
d G
i
bbs
S
a
m
pl
i
ng
[8
]
.
E
M
v
a
ri
a
t
i
o
ns
r
e
q
ui
r
e
hi
gh
c
o
m
p
ut
a
t
i
o
n a
nd
l
ear
n
i
n
g
m
o
d
el
s
t
o
b
e
b
i
as
ed
an
d
i
n
accu
r
at
e.
Al
s
o
, a
ll o
f
th
e
s
e
a
l
g
o
r
ith
m
s
a
nd
t
h
e
num
be
r
of
t
opi
c
s
s
h
oul
d be
s
e
t
be
f
or
e
h
a
nd.
D
e
t
e
r
m
i
ni
ng t
he
n
um
be
r
of
K
t
opi
c
s
i
s
v
e
r
y
i
m
por
t
a
nt
i
n
L
D
A
.
I
n
c
or
r
e
c
t
l
y
id
e
n
ti
f
y
in
g
t
h
e
n
um
be
r
of
K
to
p
ic
s
c
a
n
r
e
s
u
lt i
n
li
m
ite
d
w
o
r
d
c
o
r
r
e
la
tio
n
w
it
h
t
h
e
t
opi
c
[9
]
.
T
oo l
a
r
g
e
or
t
oo s
m
a
l
l
num
be
r
o
f
t
h
e
to
p
ic
w
i
l
l
a
f
f
ect
t
h
e
i
n
f
er
en
ce
p
r
o
ces
s
a
nd
c
a
us
e
i
na
c
c
ur
a
c
i
e
s
i
n
gr
o
up
i
n
g t
o
p
i
c
s
i
n
t
he
t
r
a
i
ni
ng
m
o
d
e
l
[
10]
.
T
he
us
e
o
f
B
a
y
es
i
an
no
np
a
r
a
m
e
t
r
i
c
m
e
t
ho
d
s
,
s
uc
h a
s
H
i
e
r
ar
ch
i
a
l
D
i
r
i
ch
l
et
P
r
o
ces
s
(
H
D
P
)
i
n
d
et
er
m
i
n
i
n
g
t
h
e n
u
m
b
er
of
t
opi
c
s
,
e
x
pe
r
i
e
n
c
e
d bot
t
l
e
ne
c
k
s
du
r
i
ng h
i
gh
c
o
m
p
u
t
a
t
i
o
n
[
11]
.
T
h
e u
s
e o
f
s
t
o
c
h
as
t
i
c
v
ar
i
at
i
o
n
al
i
n
f
er
en
ce
a
n
d
p
a
r
a
lle
l s
a
m
p
li
n
g
i
s
n
o
t c
o
n
s
i
s
te
n
t
wi
t
h
th
e
d
e
te
r
m
in
a
t
io
n
of
t
h
e
num
be
r
of
t
opi
c
s
i
n
t
he
L
D
A
m
ode
l
[
12]
.
I
n
th
is
s
tu
d
y
,
w
e
o
p
ti
m
i
s
e
t
he
n
um
be
r
of
t
opi
c
L
D
A
u
s
i
ng
m
a
x
i
m
um
l
i
k
e
l
i
h
ood a
n
d
M
i
n
i
m
u
m
D
es
cr
i
p
t
i
o
n
L
en
g
t
h
(
M
D
L
)
t
o
w
ar
d
s
t
h
e u
s
ag
e I
n
d
o
n
es
i
a
n
n
e
w
s
ar
t
i
cl
e
s
.
B
as
i
cal
l
y
,
L
D
A
C
o
l
l
ap
s
ed
G
i
bbs
S
a
m
pl
i
n
g
(
C
G
S
)
r
un
s
ba
s
e
d
on
t
h
e
n
um
be
r
of
doc
um
e
n
t
s
[
13]
,
[
1
4]
,
[
15]
,
s
o
th
a
t
th
e
re
p
o
rt
s
d
r
am
at
i
cal
l
y
a
f
f
e
c
ts
th
e
c
o
m
p
u
ta
tio
n
ti
m
e
.
I
n
t
h
is
s
t
u
d
y
,
th
e
n
u
m
b
e
r
o
f
d
o
c
u
m
e
n
ts
d
o
e
s
n
o
t a
f
f
e
c
t
t
h
e
c
o
m
p
u
ta
tio
n
ti
m
e
,
w
hi
l
e
t
h
e
num
be
r
of
w
or
ds
gr
e
a
t
l
y
a
f
f
e
c
t
s
t
h
e
c
o
m
p
u
t
i
ng
t
i
m
e
.
T
o obt
a
i
n
t
h
e
opt
i
m
a
l
n
um
be
r
of
t
opi
c
K
ba
s
e
d on
l
i
k
e
l
i
h
ood,
L
D
A
C
G
S
w
ill r
u
n
f
r
o
m
th
e
s
m
a
lle
s
t a
m
o
u
n
t
of
K
t
o t
h
e
m
o
s
t s
i
g
n
if
ic
a
n
t
n
um
be
r
of
K
.
F
o
r
each
K
,
w
e
w
i
l
l
cal
cu
l
at
e
l
o
g
-
l
i
ke
l
i
ho
o
d
va
l
ue
a
nd
p
e
r
p
le
x
it
y
w
it
h
s
p
e
c
i
f
ic
ite
r
a
tio
n
.
T
h
e
ite
r
a
tio
n
w
ill
s
t
o
p
i
t
s
el
f
i
f
p
er
p
l
ex
i
t
y
v
al
u
e
co
n
v
er
g
e
n
ces
.
T
h
e o
p
t
i
m
al
n
u
m
b
er
o
f
t
he
to
p
ic
w
ill a
u
to
m
a
tic
a
ll
y
b
e
o
b
ta
in
e
d
ba
s
e
d on
t
he
m
a
xi
m
um
l
og
-
l
i
k
e
l
i
h
ood v
a
l
u
e
of
t
h
e
K
r
a
ng
e
.
F
or
M
D
L
a
s
oppos
e
d
t
o l
i
k
e
l
i
h
ood,
L
D
A
C
G
S
w
i
l
l
r
un f
r
o
m
m
a
xi
m
um
num
be
r
of
K
t
o
m
i
n
i
m
um
num
be
r
of
K
.
T
h
e
s
m
a
l
l
e
s
t
M
D
L
v
a
l
u
e
o
f
t
h
e
K
r
a
ng
e
r
e
pr
e
s
e
n
t
s
t
h
e
opt
i
m
a
l
num
be
r
of
t
opi
c
s
.
2.
R
ES
EA
R
C
H
M
ETH
O
D
T
h
i
s
s
e
c
t
i
on
di
s
c
us
s
e
s
t
h
e
i
m
pl
e
m
e
n
t
a
t
i
on
o
f
l
i
k
e
l
i
h
ood
a
nd
M
D
L
t
o
f
i
n
d t
h
e
opt
i
m
a
l
num
be
r
of
t
opi
c
L
D
A
.
T
h
e
pr
oc
e
s
s
of
opt
i
m
i
s
i
ng
t
h
e
num
be
r
o
f
t
opi
c
L
D
A
i
s
a
o
n
e
-
t
i
m
e
e
xe
c
ut
i
o
n.
T
he
o
p
ti
m
i
s
a
tio
n
p
r
o
ces
s
s
t
ag
e
s
ar
e d
o
cu
m
e
n
t
ed
w
it
h
th
e
ir
i
n
put
,
pr
e
-
p
r
o
c
e
s
s
i
ng,
B
a
g o
f
W
o
r
d
(
B
o
W
)
,
d
e
t
e
r
m
i
ni
ng t
he
m
a
xi
m
um
num
be
r
of
t
opi
c
K
,
a
n
d opt
i
m
i
s
i
n
g
n
um
be
r
of
t
opi
c
.
T
h
e
pr
oc
e
s
s
of
o
pt
i
m
i
s
i
ng t
h
e
num
be
r
of
t
opi
c
L
D
A
ca
n
b
e s
een
i
n
F
i
g
u
r
e 1
.
F
i
g
ur
e
1
.
P
r
o
c
e
ss
of
opt
i
m
i
s
a
t
i
on
num
be
r
of
t
opi
c
L
D
A
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
20
88
-
8708
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
,
V
o
l.
8
, N
o
.
5
,
O
ct
o
b
er
2
01
8
:
320
4
-
3213
3206
2
.1
.
M
ax
i
m
u
m
N
u
m
b
e
r
of
T
op
i
c
B
a
g
of
W
or
d (
B
oW
)
pr
e
-
pr
oc
e
s
s
i
n
g r
e
s
u
l
t
s
s
t
i
l
l
c
o
m
e
i
n
r
a
ndom
da
t
a
,
w
hi
c
h
c
a
n
be
m
a
de
i
n
t
o g
r
ou
p
d
at
a.
L
i
s
t
s
co
n
t
ai
n
i
n
g
g
r
o
u
p
ed
d
at
a b
y
a s
p
eci
f
i
c i
n
t
er
v
al
cl
as
s
o
r
b
y
a p
ar
t
i
cu
l
ar
cat
eg
o
r
y
ar
e cal
l
ed
f
r
e
q
u
e
n
c
y
d
is
tr
ib
u
tio
n
[1
6
],
[1
7
]
.
T
h
e f
o
r
m
u
l
a
f
o
r
cal
cu
l
at
i
n
g
t
h
e
n
u
m
b
er
o
f
g
r
o
u
p
s
i
s
as
f
o
l
l
o
w
s
[1
6
],
[1
7
]
:
=
1
+
3
.
3
2
2
1
0
(
)
≈
1
+
2
(
)
(1
)
W
he
r
e
N
i
s
t
h
e
num
be
r
of
da
t
a
.
F
or
e
x
a
m
pl
e
,
t
h
e
r
e
s
u
l
t
e
d
w
or
ds
a
r
e
“
m
ak
an”
,
“
j
e
r
uk
”
,
“
m
angga”
,
“
be
l
i
”
,
“j
e
r
u
k
”,
“a
p
e
l
”,
“t
a
r
i
f
”,
“s
o
p
i
r
”,
“a
n
g
k
u
t
”,
“m
a
h
a
l
”,
“b
b
m
”,
“n
a
i
k
”,
“b
b
m
”,
“s
o
l
a
r
”,
and “
m
a
hal
”
.
B
as
ed
o
n
e
qu
a
t
i
on
1,
t
h
e
da
t
a
c
a
n
be
g
r
ou
pe
d i
n
t
o 4 or
5 g
r
o
u
ps
.
2
.2
.
L
D
A
C
o
l
l
a
ps
e
d G
i
b
bs
Sa
m
pl
i
ng
L
a
te
n
t D
ir
ic
h
le
t
A
llo
c
a
tio
n
i
s
a
to
p
ic
m
o
d
e
llin
g
t
ec
h
n
i
q
u
e t
h
at
d
es
cr
i
b
es
t
h
e
pr
oba
bi
l
i
t
y
pr
oc
e
du
r
e
of
doc
um
e
nt
[6
]
.
A
ppl
y
i
n
g
t
opi
c
m
o
d
e
llin
g
t
o a
doc
um
e
n
t
w
i
l
l
be
a
bl
e
t
o pr
o
du
c
e
a
s
e
t
of
l
o
w
-
di
m
e
ns
i
on
a
l
pol
y
n
o
m
i
a
l
di
s
t
r
i
b
u
t
i
ons
c
a
l
l
e
d t
opi
c
.
E
a
c
h
t
opi
c
w
i
l
l
be
us
e
d t
o c
o
m
bi
n
e
s
o
m
e
i
n
f
or
m
a
t
i
on
f
r
o
m
doc
um
e
nt
s
t
h
at
h
av
e t
h
e s
a
m
e
w
o
r
d
r
el
at
i
o
n
s
h
i
p
.
T
h
e r
es
u
l
t
ed
t
o
p
i
c can
b
e ex
t
r
act
ed
i
n
t
o
a s
e
m
an
t
i
c s
t
r
u
ct
u
r
e
w
i
t
h
co
m
p
r
e
h
en
s
i
v
e r
es
u
l
t
s
,
ev
e
n
i
n
l
ar
g
e d
at
a
[
18]
,
[
19]
.
L
D
A
m
ode
l
i
s
a
pr
oba
bi
l
i
t
y
m
ode
l
t
h
a
t
c
a
n
e
x
pl
a
i
n
t
h
e
c
or
r
e
l
a
t
i
on
be
t
w
e
e
n
w
or
ds
w
i
t
h
h
i
dde
n
t
opi
c
s
i
n
t
h
e
doc
um
e
n
t
,
f
i
n
d t
opi
c
s
,
a
n
d s
um
m
a
r
i
z
e
t
e
x
t
doc
um
e
n
t
s
[
20]
.
T
h
e
m
ai
n
i
d
ea
o
f
t
o
p
i
c
m
ode
l
l
i
ng
as
s
u
m
e
s
t
h
at
each
d
o
cu
m
e
n
t
can
b
e r
ep
r
es
en
t
ed
as
a
di
s
t
r
i
bu
t
i
on
of
s
e
v
e
r
a
l
t
opi
c
s
w
he
r
e
by
eac
h
t
o
p
i
c
is
t
he
pr
oba
bi
l
i
t
y
di
s
t
r
i
bu
t
i
on
of
t
he
w
or
ds
[
21]
.
T
h
e d
ev
el
o
p
m
en
t
o
f
L
D
A
m
e
t
h
od u
s
e
d t
o
da
y
i
s
L
D
A
a
s
a
g
en
er
at
i
v
e
m
o
d
el
a
n
d
L
D
A
as
i
n
f
er
e
n
ce
m
o
d
el
,
wh
i
c
h
can
b
e s
ee
n
i
n
F
i
g
u
r
e 2
[
22]
.
P
s
e
u
do c
ode
of
C
G
S
S
t
a
n
da
r
d,
P
s
e
u
do c
ode
of
E
f
f
i
c
i
e
nt
C
G
S
-
S
h
or
t
c
u
t
,
P
s
e
udo c
ode
of
C
ol
l
a
ps
e
d G
i
bbs
S
a
m
pl
i
n
g (
C
G
S
)
o
p
tim
is
a
tio
n
[
13]
a
s
s
h
o
w
n
i
n
F
i
gu
r
e
3,
4,
5.
F
i
g
ur
e
2
.
L
D
A r
ep
r
es
en
t
at
i
o
n
m
o
d
el
L
D
A
a
s
a
g
e
n
e
r
a
t
i
v
e
m
ode
l
i
s
u
s
e
d t
o
g
e
n
e
r
a
t
e
a
doc
um
e
nt
ba
s
e
d on
t
h
e
pr
oba
bi
l
i
t
y
v
a
l
ue
of
w
or
d
to
p
ic
(
)
a
n
d
pr
op
or
t
i
on
t
opi
c
of
doc
um
e
n
t
(
θ
)
.
L
D
A
a
s
a
n
i
nf
e
r
e
n
c
e
m
ode
l
us
i
ng
C
ol
l
a
ps
e
d G
i
bbs
S
a
m
p
l
i
n
g
(
C
G
S
)
i
s
t
h
e r
ev
er
s
e o
f
g
en
er
at
i
v
e p
r
o
ces
s
as
i
t
a
i
m
s
t
o
d
et
er
m
i
n
e o
r
f
i
n
d
h
i
d
d
en
v
al
u
e
v
ar
i
ab
l
es
,
i.
e
.
,
pr
o
ba
bi
l
i
t
y
w
or
d t
opi
c
(
)
a
n
d
pr
opor
t
i
on
t
opi
c
of
doc
um
e
nt
s
(
)
f
r
o
m
t
h
e
p
r
ed
ef
i
n
ed
obs
e
r
v
a
t
i
o
n
d
at
a
[
22
]
.
I
n
C
G
S
p
r
o
ces
s
es
,
ev
er
y
w
o
r
d
i
n
t
h
e d
o
cu
m
e
n
t
w
i
l
l
b
e d
et
er
m
i
n
ed
at
r
an
d
o
m
at
t
h
e b
eg
i
n
n
i
n
g
o
f
t
h
e t
o
p
i
c.
T
h
en
,
each
w
o
r
d
w
i
l
l
b
e p
r
o
ces
s
ed
t
o
d
et
er
m
i
n
e a
n
e
w
t
o
p
i
c b
as
ed
o
n
t
h
e p
r
o
b
ab
i
l
i
t
y
v
al
u
e o
f
eac
h
to
p
ic
.
T
o
c
a
lc
u
la
te
th
e
p
r
o
b
a
b
i
lit
y
v
a
lu
e
,
t
h
e
f
o
llo
w
in
g
f
o
r
m
u
la
is
u
s
e
d
[
14]
:
(
=
|
⃑
−
,
)
=
,
−
(
)
+
,
−
(
.
)
+
(
∗
)
∗
,
−
(
)
+
(2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
I
S
S
N
:
2088
-
8708
O
p
tim
is
a
tio
n
to
w
a
r
d
s
L
a
te
n
t D
ir
ic
h
le
t A
llo
c
a
tio
n
:
I
t
s
T
opi
c
N
um
be
r
and C
ol
l
aps
e
d
…
(
B
a
m
bang Sube
no
)
3207
W
he
r
e
V
i
s
n
um
be
r
of
v
oc
a
bu
l
a
r
y
;
n
k
,
−
i
(
)
i
s
t
h
e
n
um
be
r
of
w
or
ds
w
on
t
opi
c
k
,
e
x
c
e
pt
t
ok
e
n
i
;
,
−
(
)
is
th
e
n
um
be
r
of
w
or
ds
i
n
doc
um
e
nt
d s
pe
c
i
f
i
e
d a
s
t
o
pi
c
k
,
e
x
c
e
p
t
t
ok
e
n
i
;
a
n
d
,
−
(
.
)
is
th
e
to
ta
l w
o
r
d
o
n
t
o
p
ic
k
,
e
x
c
e
pt
t
h
e
t
ok
e
n
i
.
T
o
de
t
e
r
m
i
n
e
t
h
e
pr
oba
bi
l
i
t
y
w
or
ds
t
opi
c
a
n
d pr
o
por
t
i
on
t
opi
c
of
t
he
doc
um
e
n
t
a
f
t
e
r
g
oi
ng
t
h
r
oug
h
t
h
e
G
i
bbs
S
a
m
pl
i
n
g
p
r
oc
e
s
s
,
t
h
e
f
ol
l
o
w
i
ng
f
or
m
u
l
a
i
s
us
e
d
[
22]
:
,
=
=
(
)
+
∑
(
(
)
+
)
=
1
(3
)
,
=
=
(
)
+
∑
(
(
)
+
)
=
1
(4
)
f
o
r
(
d=
1 t
o D
)
do
fo
r
(
i=
1
to
)
do
←
,
←
fo
r
(
j =
1
to
)
do
k =
←
−
1
, N
←
−
1
fo
r
(
k =
1
t
o K
)
do
=
(
+
)
x (
+
)
/
(
∑
+
)
~
(
0
,
)
k
←
ℎ
(
:
−
1
<
<
)
←
+
1
, N
←
+
1
= k
F
i
g
ur
e
3
.
P
s
e
u
do c
ode
of
C
G
S
S
t
a
n
da
r
d
[
13]
f
o
r
(
d=
1 t
o D
)
do
fo
r
(
i=
1
to
)
do
←
,
k =
←
−
1
,
←
−
1
fo
r
(
k
=
1 t
o K
)
do
=
(
+
)
x (
+
)
/
(
∑
+
)
~
(
0
,
)
k
←
ℎ
(
:
−
1
<
<
)
←
+
1
, N
←
+
1
= k
F
i
g
ur
e
4
.
P
s
e
u
do c
o
d
e
o
f E
ffi
c
i
e
n
t
C
G
S
-
S
ho
r
t
c
ut
[
13]
fo
r
(
i=
1
to
)
do
←
,
_
←
fo
r
(
k
=
1 t
o K
)
do
if
(
k =
)
t
he
n
←
−
1
,
←
−
1
=
(
+
)
x (
+
)
/
(
∑
+
)
_
←
_
(
ma
x
(
)
)
= k
if
(
_
=
_
)
t
he
n
←
+
1
,
←
+
1
F
i
g
ur
e
5
.
P
s
e
u
do c
ode
of
C
ol
l
a
ps
e
d G
i
bbs
S
a
m
pl
i
ng
(
C
G
S
)
o
p
tim
is
a
tio
n
2
.3
.
L
ik
e
li
h
o
o
d
M
ax
i
m
u
m
L
i
k
el
i
h
o
o
d
i
s
t
h
e
es
t
i
m
a
t
ed
s
t
an
d
ar
d
u
s
ed
t
o
d
et
er
m
i
n
e t
h
e p
o
i
n
t
es
t
i
m
a
t
i
o
n
o
f
an
u
nkn
o
w
n pa
r
a
m
e
t
e
r
of
a
p
r
o
b
a
b
ilit
y
d
is
tr
ib
u
tio
n
w
it
h
m
a
x
i
m
u
m
p
r
o
b
a
b
ilit
y
.
P
s
e
u
do c
ode
of
l
i
k
e
l
i
h
ood
s
t
an
d
ar
d
,
a
nd
ps
e
u
do c
ode
of
l
i
k
e
l
i
h
ood opt
i
m
i
s
a
t
i
on
a
s
s
ho
w
n
i
n
F
i
gur
e
6
a
nd
F
i
g
ur
e
7
.
T
h
e
e
s
ti
m
a
tio
n
obt
a
i
n
e
d by
t
h
e
l
i
k
e
l
i
h
ood
m
a
x
i
m
um
m
e
t
h
od i
s
c
a
l
l
e
d l
i
k
e
l
i
h
ood
m
a
x
i
m
u
m
e
s
t
i
m
at
e
[
23]
.
T
h
er
e ar
e s
ev
er
al
l
i
k
e
l
i
h
ood s
a
m
pl
e
m
ode
l
s
d
e
v
e
l
ope
d f
or
e
s
t
i
m
a
t
i
o
n
on
t
opi
c
m
o
d
e
llin
g
s
uc
h a
s
I
m
p
o
r
t
a
nc
e
S
a
m
p
l
i
n
g,
H
ar
m
o
n
i
c M
ea
n
,
Mea
n
F
i
el
d
A
p
p
r
o
x
i
m
at
i
o
n
,
L
e
f
t
-
to
-
R
i
g
h
t
S
a
m
p
l
e
r
s
,
L
e
ft
-
to
-
R
i
g
h
t P
a
r
tic
ip
a
n
t S
a
m
p
le
r
s
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
20
88
-
8708
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
,
V
o
l.
8
, N
o
.
5
,
O
ct
o
b
er
2
01
8
:
320
4
-
3213
3208
L
e
ft
-
to
-
R
i
ght
S
e
q
ue
nt
i
a
l
S
a
m
p
l
e
r
s
[
24]
.
T
he
l
o
g
-
l
i
k
e
l
i
h
ood f
u
n
c
t
i
on on
t
opi
c
L
D
A
m
o
d
e
llin
g
i
s
a
s
f
o
l
l
o
w
s
[
14]
:
(
|
)
=
∑
(
)
l
og
(
∑
,
.
,
=
1
)
=
1
(5
)
f
o
r
( v
=
1
t
o
V
)
do
fo
r
( d
=
1
to
D
)
do
fo
r
(
k
=
1 t
o K
)
do
// c
a
lc
u
la
te
_
m
a
tr
ix
,
=
,
+
(
,
,
)
=
,
l
og
(
,
)
=
+
F
i
g
ur
e
6
.
P
s
e
u
do c
ode
o
f
L
i
k
e
l
i
h
ood
s
t
an
d
ar
d
←
n
e
w
Bo
W
f
o
rea
ch
B
∈
{
i=
1
to
}
←
_
←
fo
r
(
k
=
1 t
o K
)
do
// c
a
lc
u
l
a
te
_
m
a
tr
ix
,
=
,
+
(
,
,
)
=
,
l
og
(
,
)
=
+
Fi
g
ur
e
7
.
P
s
e
u
do c
ode
o
f
L
i
ke
l
i
ho
o
d
o
p
tim
i
s
a
tio
n
2
.4
.
M
i
ni
m
u
m
D
e
s
c
r
i
pt
i
o
n
L
e
ng
t
h
M
i
n
i
m
u
m
D
es
cr
i
p
t
i
o
n
L
en
g
t
h
(
M
D
L
)
i
s
a
m
et
h
o
d
u
s
ed
t
o
o
p
t
i
m
i
ze p
ar
a
m
et
er
es
t
i
m
at
i
o
n
o
f
a
s
ta
tis
t
ic
a
l d
is
tr
ib
u
tio
n
a
n
d
m
o
d
e
l s
e
le
c
tio
n
i
n
a
m
ode
l
l
i
ng
p
r
o
ces
s
.
I
n
t
h
i
s
M
D
L
p
r
i
n
ci
p
l
e,
t
he
B
a
y
e
s
i
a
n t
he
o
r
y
is
u
s
e
d
to
d
e
te
r
m
i
n
e
e
s
ti
m
a
tio
n
b
y
c
o
n
s
id
e
r
a
tio
n
o
f
th
e
l
i
k
e
l
i
h
ood da
t
a
a
n
d e
x
i
s
t
i
n
g
kn
ow
l
e
d
g
e
of
t
h
e
pr
i
or
pr
oba
bi
l
i
t
y
[
25]
.
I
m
p
le
m
e
n
ta
ti
o
n
o
f
th
e
M
D
L
p
r
i
n
c
ip
le
c
o
m
e
s
f
r
o
m
t
h
e
n
o
r
m
a
liz
a
tio
n
o
f
m
a
x
i
m
u
m
li
k
e
li
h
o
o
d
t
o
m
eas
u
r
e t
h
e
m
o
d
el
co
m
p
l
e
x
i
t
y
o
f
t
h
e d
at
a s
et
s
[
26]
.
T
h
e f
o
r
m
u
l
a
f
o
r
cal
cu
l
at
i
n
g
t
h
e
MD
L
i
s
as
f
o
l
l
o
w
s
[
27]
:
=
−
l
og
(
|
)
+
1
2
l
og
(
)
,
(6
)
=
1
1
0
0
1
+
+
(
+
1
)
2
−
1
W
he
r
e
l
og
(
|
)
i
s
lo
g
-
lik
e
lih
o
o
d
v
a
lu
e
,
T
is
th
e
n
u
m
b
e
r
o
f
to
p
ic
s
us
e
d
,
a
nd
N
i
s
t
he
n
um
be
r
of
w
or
ds
i
n
t
h
e
doc
um
e
n
t
.
2
.5
.
P
erp
l
ex
i
t
y
P
er
p
l
ex
i
t
y
i
s
an
o
t
h
er
w
a
y
t
o
cal
cu
l
at
e t
h
e l
i
k
el
i
h
o
o
d
u
s
ed
t
o
m
eas
u
r
e t
h
e p
er
f
o
r
m
a
n
ce o
f
t
h
e L
D
A
m
o
d
e
l.
T
h
e
s
m
a
lle
s
t p
e
r
p
le
x
it
y
v
a
lu
e
i
s
t
h
e
b
e
s
t
L
D
A
m
o
d
e
l
[
14]
.
T
he
f
o
r
m
ul
a
f
o
r
cal
cu
l
at
i
n
g
t
h
e p
er
p
l
ex
i
t
y
is
a
s
f
o
llo
w
s
:
=
ex
p
−
∑
l
o
g
(
|
)
=
1
∑
=
1
(7
)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
I
S
S
N
:
2088
-
8708
O
p
tim
is
a
tio
n
to
w
a
r
d
s
L
a
te
n
t D
ir
ic
h
le
t A
llo
c
a
tio
n
:
I
t
s
T
opi
c
N
um
be
r
and C
ol
l
aps
e
d
…
(
B
a
m
bang Sube
no
)
3209
W
he
r
e
D
i
s
t
h
e
num
be
r
o
f
doc
um
e
n
t
s
,
l
og
(
|
)
is
lo
g
-
l
i
k
e
l
i
h
ood a
c
c
or
di
n
g
t
o t
h
e
e
qu
a
t
i
o
n
(
5)
,
a
nd N
i
s
t
h
e
num
be
r
of
w
or
ds
i
n
t
h
e
do
c
um
e
n
t
.
3.
RE
S
U
L
T
S
AND AN
AL
Y
S
I
S
S
ect
i
o
n
I
V
co
n
s
i
s
t
s
o
f
t
h
r
ee s
u
b
s
e
c
tio
n
s
,
i.
e
.
,
ex
p
er
i
m
en
t
s
s
et
up
,
t
he
s
ce
n
ar
i
o
o
f
ex
p
er
i
m
e
n
t
s
,
e
xp
e
r
i
m
e
nt
s
r
e
s
ul
t
,
a
nd
a
na
l
ys
i
s
.
3
.1
.
E
x
p
eri
m
en
t
s
Se
t
U
p
I
n t
hi
s
s
t
ud
y
,
w
e
u
s
e
I
nd
o
ne
s
i
a
n ne
w
s
a
r
t
i
c
l
e
s
f
r
o
m
o
nl
i
ne
p
o
r
t
a
l
o
f
d
e
t
i
k.
c
o
m
a
nd
R
a
d
a
r
S
e
m
a
r
a
ng.
T
h
e
n
um
be
r
s
o
f
doc
um
e
nt
s
w
e
u
s
e
a
r
e
25,
50,
90,
a
n
d 600
w
i
t
h
t
h
e
num
be
r
s
of
pr
e
-
pr
oc
e
s
s
i
n
g
w
or
ds
of
e
a
c
h
doc
um
e
nt
a
r
e
3898,
776
0,
13
0
05,
a
n
d 4365
.
I
m
p
le
m
e
n
ta
tio
n
o
f
e
xp
e
r
i
m
e
nt
s
us
e
P
H
P
p
r
o
gr
a
m
m
i
n
g l
a
n
gua
ge
,
M
y
S
Q
L
d
at
ab
as
e,
an
d
h
ar
d
w
a
r
e s
p
eci
f
i
cat
i
o
n
s
as
f
o
l
l
o
w
s
:
a.
I
n
te
l® C
o
r
e
™
i3
1
.
8
G
H
z
b.
4 G
B
of
m
e
m
or
y
c.
500 G
B
of
h
a
r
d di
s
k
dr
i
v
e
T
h
e
a
l
g
or
i
t
hm
s
i
n
F
i
g
u
r
e
4 a
n
d F
i
gu
r
e
6 of
t
h
e
doc
u
m
e
n
t
l
oopi
n
g
pr
oc
e
s
s
a
r
e
om
i
t
t
e
d be
c
a
u
s
e
doc
um
e
nt
i
n
de
x
i
nf
or
m
a
t
i
o
n
a
ppe
a
r
s
i
n
B
oW
r
e
s
u
l
t
s
.
O
p
ti
m
is
a
tio
n
pr
oc
e
s
s
ba
s
e
d on
m
a
x
i
m
um
l
i
k
e
l
i
h
ood
an
d
M
D
L
o
n
ce e
x
ec
u
t
ed
w
i
l
l
au
t
o
m
a
t
i
cal
l
y
ear
n
t
h
e o
p
t
i
m
al
n
u
m
b
er
of
t
opi
c
K
,
a
l
ong
w
i
t
h
t
h
e
v
a
l
u
e
of
pe
r
pl
e
x
i
t
y
,
pr
oba
bi
l
i
t
y
w
or
d t
opi
c
,
pr
o
por
t
i
on
t
opi
c
f
or
doc
u
m
e
nt
,
a
n
d pr
oba
bi
l
i
t
y
t
opi
c
of
e
a
c
h
c
l
a
s
s
3
.2
.
S
cen
a
ri
o
of
E
xp
e
r
i
m
e
n
t
s
B
as
ed
o
n
ex
p
er
i
m
en
t
s
s
et
u
p,
w
e
pe
r
f
or
m
f
ou
r
e
x
pe
r
i
m
e
n
t
a
l
s
c
e
n
a
r
i
os
us
i
n
g c
o
m
bi
n
a
t
i
on
s
of
a
l
ph
a
0.
1,
0.
0
01 a
n
d
be
t
a
0.
1,
0.
0
0
1.
S
cen
ar
i
o
1
ai
m
s
t
o
co
m
p
ar
e t
h
e
ex
ecu
t
i
o
n
t
i
m
e
b
et
w
ee
n
s
ta
n
d
a
r
d
a
lg
o
r
ith
m
a
nd
CG
S
o
p
ti
m
is
a
tio
n
,
w
h
er
e
w
e
u
s
ed
s
e
v
er
al
d
at
as
e
t
s
f
or
a
l
ph
a
0.
1 a
n
d be
t
a
0.
1.
T
h
e d
at
as
et
s
co
n
s
i
s
t
o
f
a
v
a
r
i
ou
s
num
be
r
o
f
doc
um
e
n
t
s
,
i
.
e
.
,
25,
90
,
a
nd
600.
S
cen
ar
i
o
2
ai
m
s
t
o
k
n
o
w
t
h
e p
ar
a
m
et
er
s
t
h
at
af
f
ect
t
h
e
t
i
me
o
f
o
p
ti
m
i
s
a
tio
n
o
f
t
h
e
n
u
m
b
er
o
f
t
o
p
i
cs
.
S
ce
n
ar
i
o
3
ai
m
s
t
o
k
n
o
w
t
h
e p
ar
a
m
et
er
s
t
h
at
af
f
ect
t
h
e o
p
t
i
m
a
l
n
um
be
r
of
t
opi
c
s
b
y
u
s
i
ng
L
i
k
e
l
i
h
ood
a
n
d
M
D
L
.
S
c
e
n
a
r
i
o
4
a
i
m
s
t
o kn
o
w
t
h
e
a
ppl
i
c
a
t
i
on
of
t
he
r
e
s
u
lte
d
opt
i
m
a
l
num
be
r
of
t
he
to
p
ic
w
ith
L
D
A
C
G
S
a
s
t
h
e
c
la
s
s
i
f
y
in
g
m
o
d
e
l.
L
D
A
C
G
S
i
m
p
le
m
e
n
ta
tio
n
r
e
s
u
lt
s
i
n
t
h
e o
p
t
i
m
al
n
u
m
b
er
o
f
t
o
p
i
cs
as
a cl
as
s
i
f
i
cat
i
o
n
m
o
d
el
.
W
e u
s
e
100 a
r
t
i
c
l
e
s
di
v
i
de
d i
n
t
o 90%
,
or
90
doc
um
e
n
t
a
r
t
i
c
l
e
s
a
s
t
r
a
i
n
i
n
g
da
t
a
a
n
d 10%
,
or
10 a
r
t
i
c
l
e
doc
um
e
n
t
s
a
s
te
s
ti
n
g
d
a
ta
.
T
h
e
a
r
tic
le
d
o
c
u
m
e
n
t is
d
i
v
id
e
d
in
to
fi
v
e
cl
as
s
es
:
eac
h
cl
as
s
f
o
r
t
r
ai
n
i
n
g
d
a
t
a co
n
s
i
s
t
i
n
g
o
f
1
8
n
e
w
s
ar
t
i
cl
es
.
I
n
t
h
e t
es
t
i
n
g
p
r
o
ces
s
,
w
e u
s
e K
u
l
l
b
ac
k
-
L
ei
b
l
er
D
i
v
er
g
en
ce (
K
L
D
)
t
o
m
ea
s
u
r
e t
h
e d
i
s
t
r
ib
u
tio
n
s
i
m
i
l
a
r
i
t
y
be
t
w
e
e
n
t
h
e
pr
opo
r
t
i
on
of
doc
um
e
n
t
t
e
s
t
i
n
g
t
o
pi
c
s
a
n
d t
h
e
pr
opor
t
i
on
of
t
opi
c
s
f
or
e
a
c
h
c
l
a
s
s
p
r
o
d
u
ced
i
n
t
h
e t
r
ai
n
i
n
g
p
r
o
ces
s
.
T
h
e p
r
ed
i
ct
i
o
n
o
f
t
h
e d
o
cu
m
e
n
t
t
es
t
i
n
g
cl
as
s
i
s
t
a
k
en
f
r
o
m
t
h
e
s
m
al
l
es
t
v
a
lu
e
o
f
K
L
D
.
D
e
ta
ile
d
in
f
o
r
m
a
tio
n
o
f
K
L
D
c
a
n b
e
f
o
und
i
n
[
22]
.
3
.3
.
E
x
pe
r
i
m
e
nt
s
R
e
s
ul
t
a
nd
A
na
l
y
s
i
s
T
h
e r
es
u
l
t
s
o
f
t
h
e ex
p
er
i
m
en
t
a
l
s
cen
ar
i
o
1
can
b
e
s
een
i
n
F
i
g
ur
e
8,
a
n
d F
i
g
ur
e
9.
W
h
ile
t
he
r
e
s
ul
t
s
o
f
t
he
ex
p
er
i
m
e
n
t
al
s
ce
n
ar
i
o
2
ca
n
b
e s
een
i
n
T
ab
l
e 1
,
F
i
g
ur
e
10,
a
n
d F
i
g
ur
e
1
1
.
T
he
r
e
s
ul
t
s
o
f
t
he
e
xp
e
r
i
m
e
nt
a
l
s
cen
ar
i
o
3
ca
n
b
e s
ee
n
i
n
T
ab
l
e 2
an
d
F
i
g
ur
e
1
2
.
F
u
r
t
h
er
m
o
r
e,
t
h
e r
es
u
l
t
o
f
e
x
p
er
i
m
e
n
t
a
l
s
ce
n
ar
i
o
4
can
b
e
s
een
i
n
T
ab
l
e 3
an
d
F
i
g
ur
e
13.
F
i
g
ur
e
8
.
C
o
m
pa
r
i
s
on
C
G
S
S
t
a
n
da
r
d a
n
d
O
p
ti
m
is
a
t
io
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
20
88
-
8708
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
,
V
o
l.
8
, N
o
.
5
,
O
ct
o
b
er
2
01
8
:
320
4
-
3213
3210
F
i
g
ur
e
9
.
C
o
m
pa
r
i
s
on
of
E
x
e
c
u
t
i
on
T
i
m
e
S
t
a
n
da
r
d a
n
d
O
p
tim
is
a
tio
n
f
o
r
A
l
l
P
r
o
ces
s
es
R
e
s
ul
t
s
i
n
T
a
bl
e
1,
F
i
g
u
r
e
10,
a
n
d F
i
gu
r
e
11
s
h
o
w
s
t
h
a
t
t
he
n
um
be
r
of
w
or
ds
u
s
e
d
w
i
l
l
a
f
f
e
c
t
t
h
e
co
m
p
u
t
at
i
o
n
al
t
i
m
e:
t
h
e
g
r
eat
er
t
he
n
um
be
r
of
w
or
ds
is
,
th
e
lo
n
g
e
r
th
e
c
o
m
p
u
ta
tio
n
a
l ti
m
e
w
ill i
n
c
r
e
a
s
e
.
T
h
e
n
u
m
b
er
o
f
d
o
cu
m
e
n
t
s
an
d
co
m
b
i
n
at
i
o
n
s
o
f
al
p
h
a,
b
et
a
d
o
e
s
n
o
t
af
f
ect
t
h
e
co
m
p
u
t
at
i
o
n
a
l
t
i
m
e.
T
h
e
u
s
e
o
f
a
l
go
r
i
t
h
m
s
s
ho
w
n
i
n
F
i
gur
e
5
a
nd
F
i
g
ur
e
7
g
r
e
a
t
l
y
c
on
c
e
r
n
s
t
h
e
opt
i
m
i
s
a
t
i
on of
t
h
e
e
x
e
c
ut
i
on
t
i
m
e
.
L
oopi
ng
doc
um
e
nt
i
s
r
e
m
o
v
e
d be
c
a
us
e
t
h
e
B
a
g
o
f
W
or
d (
B
oW
)
pr
e
-
p
r
o
c
e
s
s
i
ng r
e
s
ul
t
s
s
ho
w
a
d
o
c
u
m
e
nt
i
nd
e
x.
T
h
is
is
s
ho
w
n b
y t
he
e
xp
e
r
i
m
e
nt
a
l
r
e
s
ul
t
s
o
f
th
e
f
ir
s
t
s
c
e
n
a
r
io
,
w
h
ic
h
is
ill
u
s
tr
a
te
d
in
F
i
g
u
r
e
8
a
n
d
F
ig
u
r
e
9.
T
a
b
le
1
.
T
i
me
O
p
ti
m
is
a
tio
n
P
r
o
c
e
ss
Re
s
u
l
t
No
D
oc
W
o
r
d
s
A
l
p
h
a
B
et
a
C
om
pu
t
i
n
g
Ti
m
e
(
s
e
c
on
d
)
L
i
k
e
l
i
h
oo
d
M
DL
1
2
5
3
8
9
8
0
.1
0
.1
2
.
9
7
2
1
6
2
.
9
7
2
1
6
2
2
5
3
8
9
8
0
.1
0
.
0
0
1
2
.
9
6
7
1
7
2
.
9
6
7
1
7
3
2
5
3
8
9
8
0
.
0
0
1
0
.1
2
.
9
5
5
1
6
2
.
9
5
5
1
6
4
2
5
3
8
9
8
0
.
0
0
1
0
.
0
0
1
2
.
9
7
8
1
6
2
.
9
7
8
1
6
5
5
0
7
7
6
0
0
.1
0
.1
6
.
4
96
3
7
1
6
.
4
96
3
7
1
6
5
0
7
7
6
0
0
.1
0
.
0
0
1
6
.
4
67
3
7
0
6
.
4
67
3
7
0
7
5
0
7
7
6
0
0
.
0
0
1
0
.1
6
.
4
76
3
7
0
6
.
4
77
3
7
7
8
5
0
7
7
6
0
0
.
0
0
1
0
.
0
0
1
6
.
4
57
3
6
9
6
.
4
58
3
6
9
9
9
0
1
3
0
0
5
0
.1
0
.1
1
3
.
29
6
7
6
1
3
.
29
6
7
6
1
0
9
0
1
3
0
0
5
0
.1
0
.
0
0
1
1
3
.
31
6
7
6
1
3
.
31
6
7
6
1
1
9
0
1
3
0
0
5
0
.
0
0
1
0
.1
1
3
.
30
9
7
5
1
3
.
30
9
7
5
1
2
9
0
1
3
0
0
5
0
.
0
0
1
0
.
0
0
1
1
3
.
30
4
7
6
1
3
.
30
4
7
6
1
3
6
0
0
4
3
6
5
0
.1
0
.1
3
.
7
15
2
0
8
3
.
7
25
2
0
8
1
4
6
0
0
4
3
6
5
0
.1
0
.
0
0
1
3
.
7
15
2
1
2
3
.
7
15
2
1
2
1
5
6
0
0
4
3
6
5
0
.
0
0
1
0
.1
3
.
7
16
2
1
2
3
.
7
16
2
1
2
1
6
6
0
0
4
3
6
5
0
.
0
0
1
0
.
0
0
1
3
.
7
15
2
1
2
3
.
7
15
2
1
2
Fi
g
ur
e
10
.
Co
m
p
a
r
i
s
o
n
of
L
i
ke
l
i
h
ood a
n
d M
D
L
c
o
m
p
u
t
a
t
i
o
n
t
i
m
e
t
o
w
or
d c
ou
n
t
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
I
S
S
N
:
2088
-
8708
O
p
tim
is
a
tio
n
to
w
a
r
d
s
L
a
te
n
t D
ir
ic
h
le
t A
llo
c
a
tio
n
:
I
t
s
T
opi
c
N
um
be
r
and C
ol
l
aps
e
d
…
(
B
a
m
bang Sube
no
)
3211
Fi
g
ur
e
11
.
T
h
e ef
f
ect
o
f
a co
m
b
i
n
at
i
o
n
o
f
al
p
h
a
-
b
et
a v
al
u
es
o
n
o
p
ti
m
is
a
t
io
n
t
i
me
B
a
s
e
d
o
n t
he
e
xp
e
r
i
m
e
nt
a
l
r
e
s
ul
t
s
i
n
T
ab
l
e 2
an
d
F
i
g
u
r
e 1
0
,
hy
pe
r
-
p
ar
a
m
et
er
al
p
h
a,
b
et
a can
af
f
ect
t
h
e
opt
i
m
a
l
num
be
r
of
t
opi
c
s
on
l
i
k
e
l
i
h
ood a
n
d M
D
L
.
A
l
t
hou
gh
t
h
e
us
e
of
a
l
ph
a
,
be
t
a
v
a
l
u
es
m
a
y
af
f
ect
t
h
e
n
um
be
r
of
t
opi
c
s
,
t
h
e
L
i
k
e
l
i
h
o
od a
n
d M
D
L
pr
oc
e
s
s
e
s
w
i
l
l
r
e
s
u
l
t
i
n
t
h
e
s
a
m
e
opt
i
m
a
l
n
um
b
e
r
of
t
opi
c
s
.
T
a
b
le
3
s
h
o
w
s
t
h
e r
es
u
l
t
o
f
L
D
A
C
G
S
i
m
p
l
e
m
e
n
t
at
i
o
n
as
a c
l
as
s
i
f
i
cat
i
o
n
m
o
d
el
u
s
i
n
g
1
0
f
o
l
d
.
T
h
e h
i
g
h
es
t
accu
r
ac
y
of
doc
um
e
n
t
c
l
a
s
s
i
f
i
c
a
t
i
on
i
s
0.
80 o
r
80%
w
i
t
h
a
l
ph
a
0.
1 a
n
d
be
t
a
0.
0
01.
T
a
b
le
2
.
O
pt
i
m
a
l
N
um
be
r
of
T
opi
c
s
B
a
s
e
d on
L
i
k
e
l
i
h
ood a
nd M
D
L
No
D
oc
W
o
r
d
s
A
l
p
h
a
B
et
a
O
p
t
im
a
l
N
u
m
b
e
r
o
f
T
o
p
i
c
L
i
k
e
l
i
h
oo
d
M
DL
1
2
5
3
8
9
8
0
.1
0
.1
1
1
1
1
2
2
5
3
8
9
8
0
.1
0
.
0
0
1
1
2
1
2
3
2
5
3
8
9
8
0
.
0
0
1
0
.1
1
3
1
3
4
2
5
3
8
9
8
0
.
0
0
1
0
.
0
0
1
1
3
1
3
5
5
0
7
7
6
0
0
.1
0
.1
1
3
1
3
6
5
0
7
7
6
0
0
.1
0
.
0
0
1
1
4
1
4
7
5
0
7
7
6
0
0
.
0
0
1
0
.1
1
4
1
4
8
5
0
7
7
6
0
0
.
0
0
1
0
.
0
0
1
1
4
1
4
9
9
0
1
3
0
0
5
0
.1
0
.1
1
5
1
5
1
0
9
0
1
3
0
0
5
0
.1
0
.
0
0
1
1
5
1
5
1
1
9
0
1
3
0
0
5
0
.
0
0
1
0
.1
1
5
1
5
1
2
9
0
1
3
0
0
5
0
.
0
0
1
0
.
0
0
1
1
5
1
5
1
3
6
0
0
4
3
6
5
0
.1
0
.1
1
2
1
2
1
4
6
0
0
4
3
6
5
0
.1
0
.
0
0
1
1
2
1
2
1
5
6
0
0
4
3
6
5
0
.
0
0
1
0
.1
1
3
1
3
1
6
6
0
0
4
3
6
5
0
.
0
0
1
0
.
0
0
1
1
3
1
3
Fi
g
ur
e
12
.
T
h
e
i
n
f
l
u
e
n
c
e
of
a
l
ph
a
,
be
t
a
c
om
bi
n
a
t
i
ons
on
t
h
e
opt
i
m
a
l
num
be
r
of
to
p
ic
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
20
88
-
8708
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
,
V
o
l.
8
, N
o
.
5
,
O
ct
o
b
er
2
01
8
:
320
4
-
3213
3212
T
a
b
le
3
.
A
v
er
ag
e A
cc
u
r
ac
y
C
l
as
s
i
f
ic
a
tio
n
o
f
E
v
e
r
y F
o
l
d
F
i
g
ur
e
13
.
A
v
er
a
g
e co
m
p
ar
i
s
o
n
o
f
accu
r
ac
y
t
o
al
p
h
a
-
b
et
a v
a
l
u
e ch
a
n
g
es
B
a
s
e
d
o
n t
he
e
xp
e
r
i
m
e
nt
a
l
r
e
s
ul
t
i
n
T
ab
l
e 3
an
d
F
i
g
u
r
e 1
1
,
i
t
i
s
s
ho
w
n
t
ha
t
t
he
a
ve
r
a
ge
hi
ghe
s
t
cl
as
s
i
f
i
cat
i
o
n
accu
r
ac
y
o
f
each
f
o
l
d
i
s
6
1
%
w
i
t
h
h
y
p
er
-
p
ar
a
m
et
er
al
p
h
a 0
.
1
an
d
b
et
a 0
.
0
0
1
.
T
h
e u
s
e
of
a
l
ph
a
an
d
b
et
a g
r
eat
l
y
a
f
f
ec
t
s
t
h
e ac
cu
r
ac
y
o
f
d
o
cu
m
en
t
cl
as
s
i
f
i
ca
t
i
o
n
.
T
h
e
m
et
h
o
d
o
f
ap
p
r
o
p
r
i
a
t
e h
y
p
er
-
p
ar
a
m
et
er
al
p
h
a,
b
et
a
w
i
l
l
p
r
o
d
u
ce a h
i
g
h
d
eg
r
ee o
f
accu
r
ac
y
as
i
n
f
o
l
d
4
w
i
t
h
0
.
8
0
o
r
8
0
%
ef
f
i
ci
en
c
y
.
4.
CO
NCL
U
S
I
O
N
T
he
o
p
ti
m
is
a
tio
n
num
be
r
of
t
opi
c
wi
t
h
L
D
A
,
u
s
i
ng
L
i
k
e
l
i
hood a
n
d
M
D
L
,
y
i
e
l
ds
t
h
e
s
a
m
e
opt
i
m
a
l
n
um
be
r
of
t
opi
c
.
T
h
e
n
um
be
r
of
doc
um
e
n
t
s
doe
s
no
t
ha
ve
a
s
i
g
ni
f
i
c
a
nt
e
f
f
e
c
t
o
n
t
he
o
p
ti
m
is
a
tio
n
pr
oc
e
s
s
,
bu
t
t
h
e
num
be
r
of
w
or
ds
doe
s
.
T
h
e
m
or
e
n
um
be
r
t
h
e
w
or
ds
u
s
e
d,
t
h
e
l
on
g
e
r
t
h
e
c
o
m
p
u
ta
tio
n
a
l ti
m
e
wa
s.
C
o
m
b
i
n
at
i
o
n
o
f
al
p
h
a,
b
et
a v
al
u
es
w
i
l
l
co
n
d
u
ct
an
ef
f
ect
on t
h
e
opt
i
m
a
l
n
um
be
r
of
t
opi
c
b
ut
d
o
e
s
no
t
gi
ve
a
s
i
g
n
i
f
i
ca
n
t
ef
f
ect
o
n
co
m
p
u
t
at
i
o
n
al
t
i
m
e.
M
o
r
eo
v
er
,
o
p
tim
i
s
i
ng
t
h
e
n
um
be
r
of
t
opi
c
s
wi
t
h
L
D
A
,
w
e
h
av
e g
a
t
h
er
ed
t
h
at
C
G
S c
a
n
b
e ap
p
l
i
ed
as
a cl
as
s
i
f
i
cat
i
o
n
m
o
d
el
,
b
u
t
t
o
g
et
g
o
o
d
accu
r
ac
y
,
o
n
e s
h
o
u
l
d
d
o
s
ev
er
al
i
t
er
at
i
o
n
s
a
n
d
u
s
e ap
p
r
o
p
r
i
at
e al
p
h
a,
b
et
a v
al
u
es
.
T
h
e i
n
co
r
r
ect
u
s
e o
f
al
p
h
a,
b
et
a
v
al
u
e
s
w
i
l
l
af
f
ect
t
h
e o
p
t
i
m
al
n
u
m
b
er
o
f
t
o
p
i
cs
,
an
d
t
h
e
cl
as
s
i
f
i
cat
i
o
n
acc
u
r
ac
y
i
s
no
t
go
o
d
.
I
n t
hi
s
s
t
ud
y
,
t
he
h
i
g
he
s
t
m
e
a
n
va
l
ue
e
a
r
ne
d
f
o
r
1
0
-
f
ol
d i
s
0.
61 or
61%
w
i
t
h
al
p
h
a 0
.
1
an
d
b
et
a 0
.
0
0
1
.
T
h
e b
es
t
cl
as
s
i
f
i
cat
i
o
n
acc
u
r
a
c
y
i
s
s
h
o
w
n
i
n
f
ol
d 4
w
i
t
h
0.
8
0 or
8
0%
a
c
c
u
r
a
c
y
va
l
ue
.
ACK
NO
W
L
E
D
G
E
M
E
NT
S
T
he
a
ut
ho
r
s
w
o
ul
d
l
i
ke
t
o
a
c
k
no
w
l
e
d
ge
t
he
r
e
s
e
a
r
c
h f
u
nd
i
n
g s
up
p
o
r
t
e
d
b
y
U
ni
ve
r
s
i
t
a
s
D
i
p
o
ne
go
r
o
u
n
d
er
t
h
e
g
r
an
t
o
f
r
es
ea
r
ch
f
o
r
i
n
t
er
n
at
i
o
n
al
s
ci
e
n
t
i
f
i
c p
u
b
l
i
cat
i
o
n
-
Y
e
a
r
2017 (
n
um
be
r
276
-
36/
U
N
7.
5.
1/
P
G
/
20
17)
.
T
hi
s
r
es
ear
ch
f
u
n
d
i
n
g
i
s
g
r
an
t
ed
t
o
t
h
e s
eco
n
d
au
t
ho
r
.
Fo
l
d
A
ccu
r
acy
o
f
D
o
cu
m
e
n
t
C
l
as
s
i
f
i
ca
t
i
o
n
A
l
p
h
a
0
.
1
B
et
a
0
.1
A
l
p
h
a
0
.
1
B
et
a
0
.0
0
1
A
l
p
h
a
0
.
0
0
1
B
et
a
0
.1
A
l
p
h
a
0
.
0
0
1
B
et
a
0
.0
0
1
1
6
0
%
7
0
%
4
0
%
5
0
%
2
6
0
%
5
0
%
5
0
%
4
0
%
3
5
0
%
5
0
%
6
0
%
5
0
%
4
5
0
%
8
0
%
5
0
%
5
0
%
5
4
0
%
6
0
%
4
0
%
5
0
%
6
5
0
%
7
0
%
4
0
%
5
0
%
7
5
0
%
7
0
%
5
0
%
7
0
%
8
5
0
%
5
0
%
3
0
%
6
0
%
9
5
0
%
6
0
%
4
0
%
5
0
%
1
0
5
0
%
5
0
%
5
0
%
5
0
%
A
v
e
r
a
g
e
5
1
%
6
1
%
4
5
%
5
2
%
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t
J
E
l
e
c
&
C
o
m
p
E
n
g
I
S
S
N
:
2088
-
8708
O
p
tim
is
a
tio
n
to
w
a
r
d
s
L
a
te
n
t D
ir
ic
h
le
t A
llo
c
a
tio
n
:
I
t
s
T
opi
c
N
um
be
r
and C
ol
l
aps
e
d
…
(
B
a
m
bang Sube
no
)
3213
R
EF
ER
EN
C
ES
[
1]
S
.
M
or
o,
P
.
C
or
t
e
z
a
n
d
P
.
R
i
t
a
,
“
B
us
i
ne
s
s
i
nt
e
l
l
i
g
e
nc
e
i
n
ba
nk
i
n
g
:
A
l
i
t
e
r
a
t
ur
e
a
na
l
y
s
i
s
f
r
o
m
200
2 t
o 2
01
3
us
i
ng
T
e
x
t M
in
in
g
a
n
d
L
a
te
n
t D
ir
ic
h
le
t
A
llo
c
a
tio
n
,
”
Ex
p
e
r
t S
y
s
te
m
s
w
ith
Ap
p
lic
a
ti
o
n
s
,
vo
l
.
42
,
p
p.
13
14
-
132
4,
20
15
.
[
2]
N
.
N
aw
an
d
E
.
E
.
H
l
ai
n
g
,
“R
el
ev
an
t
W
o
r
d
s
E
x
t
r
act
i
o
n
M
et
h
o
d
f
o
r
R
eco
m
m
en
d
at
i
o
n
S
y
s
t
e
m
,
”
Bu
lle
tin
o
f
Ele
c
tr
ic
a
l
E
ngi
ne
e
r
i
ng an
d
I
nf
or
m
at
i
c
s
,
v
ol
.
2,
no.
3,
p
p.
16
9
-
1
76,
2
01
3.
[
3]
N
.
N
a
w
,
“
R
e
l
e
v
a
nt
W
or
ds
E
x
t
r
a
c
t
i
on M
e
t
ho
d i
n T
e
x
t
M
i
ni
ng
,
”
Bu
lle
tin
o
f Ele
c
tr
ic
a
l E
ngi
ne
e
r
i
ng a
nd
I
nf
or
m
at
i
c
s
,
v
ol
.
2,
n
o.
3,
pp.
1
77
-
18
1,
20
13.
[
4]
R
.
S
.
A
an
d
S
.
R
am
as
a
m
y
,
“
C
o
n
t
ex
t
B
as
ed
C
l
as
s
i
f
i
cat
i
o
n
o
f
R
ev
i
e
w
s
U
s
i
n
g
A
s
s
o
ci
at
i
o
n
R
u
l
e
M
i
n
i
n
g
,
F
u
zzy
L
og
i
c
s
a
nd O
nt
ol
og
y
,
”
Bu
lle
tin
o
f
Ele
c
tr
ic
a
l En
g
i
n
e
e
r
in
g
a
n
d
I
n
fo
r
m
a
tic
s
,
vo
l
.
6
,
n
o.
3,
pp
.
2
50
-
2
55
,
20
17.
[
5]
D
.
B
r
ace
w
el
l
,
Y
.
J
i
aj
u
n
an
d
R
.
F
u
j
i
,
“C
at
eg
o
r
y
C
l
as
s
i
f
i
cat
i
o
n
an
d
T
o
p
i
c D
i
s
co
v
er
y
o
f
J
ap
an
es
e an
d
E
n
g
l
i
s
h
N
ew
s
A
r
t
i
c
l
e
s
,
”
2009.
[
6]
D
.
M
.
B
le
i,
A
.
Y
.
N
g
a
n
d
M
.
I
.
J
o
r
d
a
n
,
“
L
a
te
n
t D
ir
ic
h
le
t A
llo
c
a
tio
n
,
”
J
o
u
r
n
a
l
o
f
M
ach
i
n
e L
ear
n
i
n
g
R
es
ear
ch
3
,
pp.
99
3
-
102
2,
20
03
.
[
7]
T
.
M
i
n
k
a an
d
J
.
L
af
f
er
t
y
,
“E
x
p
e
ct
at
i
o
n
-
pr
o
pa
g
a
t
i
on f
or
t
he
g
e
ne
r
a
t
i
v
e
a
s
pe
c
t
m
ode
l
,
”
I
n U
A
I
,
p.
3
52
–
35
9,
20
02
.
[
8]
T
.
L
.
G
r
i
f
f
i
t
h
s
an
d
M
.
S
t
ey
v
er
s
,
“F
i
n
d
i
n
g
s
ci
en
t
i
f
i
c t
o
p
i
cs
,
”
P
r
o
ceed
i
n
g
o
f
t
h
e N
at
i
o
n
al
A
cad
e
m
y
o
f
S
ci
en
ces
,
v
ol
.
1
01,
p
p.
52
28
-
5
23
5,
20
04.
[
9]
A
.
K
u
l
es
za,
N
.
R
.
R
ao
an
d
S
.
S
i
n
g
h
,
“L
o
w
-
R
an
k
S
p
ect
r
al
L
e
ar
n
i
n
g
,
” I
n
t
er
n
at
i
o
n
al
C
o
n
f
er
en
ce o
n
A
r
t
i
f
i
ci
al
I
n
te
llig
e
n
c
e
a
n
d
S
ta
tis
tic
s
(
A
I
S
TA
T
S
)
,
v
o
l.
3
3
,
2
0
1
4
.
[
1
0]
J
.
T
a
ng
,
Z
.
M
e
ng
,
X
.
N
g
u
y
e
n
a
nd Q
.
M
e
i
,
“
U
nde
r
s
t
a
ndi
ng
t
he
L
i
m
i
t
i
ng
F
a
c
t
or
s
of
T
opi
c
M
ode
l
i
n
g
v
i
a
P
os
t
e
r
i
or
C
ont
r
a
c
t
i
o
n A
na
l
y
s
i
s
,
”
I
nt
e
r
na
t
i
ona
l
C
o
nf
e
r
e
nc
e
on M
a
c
hi
ne
L
e
a
r
ni
ng
,
v
ol
.
3
2,
20
14
.
[
1
1]
D
.
C
he
ng
,
X
.
H
e
a
n
d Y
.
L
i
u,
“
M
ode
l
S
e
l
e
c
t
i
on
f
or
T
opi
c
M
od
e
l
s
v
i
a
S
pe
c
t
r
a
l
D
e
c
om
pos
i
t
i
on,
”
I
nt
e
r
na
t
i
o
na
l
C
o
n
f
e
r
e
n
c
e
o
n
A
r
tif
ic
ia
l I
n
te
llig
e
nc
e
a
nd S
t
a
t
i
s
t
i
c
s
,
v
ol
.
38
,
2
01
5.
[
1
2]
S
.
W
i
l
l
i
a
m
s
on,
A
.
D
ube
y
a
nd E
.
P
.
X
i
ng
,
“
P
a
r
a
l
l
e
l
M
a
r
k
ov
C
ha
i
n M
ont
e
C
a
r
l
o f
or
P
a
r
a
l
l
e
l
M
a
r
k
o
v
C
ha
i
n M
on
t
e
C
a
r
l
o,
”
J
our
na
l
of
M
a
c
hi
ne
L
e
a
r
ni
ng
R
e
s
e
a
r
c
h,
v
ol
.
2
8,
pp.
9
8
-
1
06,
20
13.
[
1
3]
T
.
S
.
X
i
ao
H
an
,
“E
f
f
i
ci
en
t
C
o
l
l
ap
s
ed
G
i
bbs
S
a
m
pl
i
ng
F
or
L
a
t
e
nt
D
i
r
i
c
hl
e
t
A
l
l
oc
a
t
i
on,
”
A
s
i
a
n
C
onf
e
r
e
nc
e
on
M
a
c
hi
ne
L
e
a
r
ni
ng
(
A
C
M
L
2010)
,
20
10.
[
1
4]
G
.
H
ei
n
r
i
ch
,
P
ar
am
et
er
es
t
i
m
at
i
o
n
f
o
r
t
ex
t
an
al
y
s
i
s
,
2
.
9
ed
.
,
D
ar
m
s
t
ad
t
,
G
er
m
an
y
:
F
r
au
n
h
o
f
er
I
G
D
,
2
0
0
9
.
[
1
5]
R
.
K
.
V
a
nd K
.
R
a
g
huv
e
e
r
,
“
L
e
ga
l
D
oc
um
e
nt
s
C
l
us
t
er
i
n
g
an
d
S
u
m
m
ar
i
zat
i
o
n
u
s
i
n
g
H
i
er
ar
ch
i
cal
L
at
en
t
D
i
r
i
ch
l
et
A
llo
c
a
tio
n
,
”
I
A
ES
I
n
te
r
n
a
ti
o
n
a
l J
o
u
r
n
a
l o
f
A
r
tif
i
c
ia
l I
n
te
llig
e
n
c
e
(
I
J
-
A
I
)
, v
o
l
. 2
, n
o
.
1
,
p
p
. 2
7
-
3
5,
20
13.
[
1
6]
H
.
S
t
u
r
g
es
,
“T
h
e ch
o
i
ce o
f
a cl
as
s
i
n
t
er
v
al
,
” J
o
u
r
n
al
o
f
t
h
e A
m
er
i
ca
n
S
t
at
i
s
t
i
cal
A
s
s
o
c
i
a
t
i
on,
p
p.
65
-
66
,
19
26
.
[
1
7]
D
.
W
.
S
c
o
tt,
“
S
t
u
r
g
e
s
R
u
le
,
”
W
ir
e
s
C
o
m
p
u
ta
tio
n
a
l S
ta
tis
tic
s
,
2
0
0
9
.
[
1
8]
S
.
A
r
or
a
,
R
.
G
e
a
nd A
.
M
oi
t
r
a
,
“
L
e
a
r
ni
ng
T
opi
c
M
ode
l
s
-
G
oi
ng
be
y
ond S
V
D
,
”
I
E
E
E
5
3r
d
A
nnua
l
S
y
m
pos
i
um
on
F
ou
nda
t
i
ons
of
C
om
put
e
r
S
c
i
e
nc
e
,
v
ol
.
2,
pp.
1
-
10,
2
01
2.
[
1
9]
Z
.
L
i
u,
H
i
g
h
P
e
r
f
or
m
a
nc
e
L
a
t
e
nt
D
i
r
i
c
hl
e
t
A
l
l
oc
a
t
i
on f
or
T
e
x
t
M
i
n
i
ng
,
L
ond
on:
B
r
u
ne
l
U
ni
v
e
r
s
i
t
y
,
2013
.
[
2
0]
D
.
M
.
B
l
e
i
,
“
P
r
o
ba
bi
l
i
s
t
i
c
T
o
pi
c
M
ode
l
s
,
”
C
om
m
uni
c
a
t
i
on
of
T
he
A
C
M
,
v
ol
.
55
,
n
o.
4,
pp.
7
7
-
84
,
20
1
2.
[
2
1]
Z
.
Q
i
na
,
Y
.
C
ong
a
n
d T
.
W
a
n,
“
T
opi
c
m
ode
l
i
ng
of
C
hi
ne
s
e
l
a
ng
ua
g
e
be
y
ond a
ba
g
-
of
-
w
or
ds
,
”
C
om
put
e
r
S
pe
e
c
h
a
nd L
a
ng
ua
g
e
,
v
ol
.
40
,
p
p.
60
-
78
,
20
16.
[
2
2]
R
.
K
us
um
a
ni
ng
r
um
,
W
.
H
ong
,
R
.
M
a
nur
u
ng
a
nd M
.
A
ni
a
t
i
,
“
I
nt
e
g
r
a
t
e
d V
i
s
ua
l
V
oc
a
bul
a
r
y
i
n
L
D
A
ba
s
e
s
c
e
ne
c
l
a
s
i
f
i
c
a
t
i
on f
or
I
K
O
N
O
S
i
m
a
g
e
s
,
”
J
our
na
l
of
A
ppl
e
d R
e
m
ot
e
S
e
ns
i
ng
,
v
ol
.
8,
20
14.
[
2
3]
I
.
J
.
M
y
ung
,
“
T
ut
or
i
a
l
on m
a
xi
m
u
m
l
i
ke
l
i
hoo
d e
s
t
i
m
a
t
i
on,
”
J
our
na
l
of
M
a
t
he
m
a
t
i
c
a
l
P
s
y
c
hol
og
y
,
v
ol
.
47
,
pp.
90
-
1
00,
20
02
.
[
2
4]
W
.
B
unt
i
ne
,
“
E
s
t
i
m
a
t
i
ng
L
i
k
e
l
i
ho
ods
f
or
T
opi
c
M
ode
l
s
,
”
T
he
1s
t
A
s
i
a
n C
onf
e
r
e
nc
e
on
M
a
c
hi
ne
L
e
a
r
ni
ng
,
20
09.
[
2
5]
J
.
I
.
M
y
ung
,
D
.
J
.
N
a
v
a
r
r
o a
nd M
.
A
.
P
i
t
t
,
“
M
ode
l
s
e
l
e
c
t
i
on
b
y
nor
m
a
l
i
z
e
d
m
a
x
i
m
u
m
l
i
k
e
l
i
hood,
”
J
our
na
l
of
M
a
t
he
m
a
t
i
c
a
l
P
s
y
c
hol
og
y
,
v
ol
.
50,
p
p.
16
7
-
17
9,
20
05
.
[
2
6]
D
.
W
.
H
e
c
k
,
M
.
M
os
ha
g
e
n a
nd
E
.
E
r
df
e
l
de
r
,
“
M
o
de
l
s
e
l
e
c
t
i
on
by
m
i
ni
m
um
de
s
c
r
i
pt
i
on
l
e
ng
t
h
:
L
ow
e
r
-
boun
d
s
a
m
pl
e
s
i
z
e
s
f
or
t
he
F
i
s
he
r
i
nf
or
m
a
t
i
on a
ppr
ox
i
m
a
t
i
on,
”
J
our
na
l
of
M
a
t
he
m
a
t
i
c
a
l
P
s
y
c
hol
ogy
,
v
ol
.
60,
p
p.
2
9
-
34,
201
4.
[
2
7]
W
.
X
i
a
or
u
,
D
.
J
un
pi
ng
,
W
.
S
h
uz
he
a
nd
L
.
F
u
,
“
A
da
pt
i
v
e
R
e
g
i
on C
l
us
t
e
r
i
ng
i
n
L
D
A
F
r
a
m
e
w
or
k
f
or
I
m
a
g
e
S
e
gm
e
nt
a
t
i
on,
”
P
r
oc
e
e
di
ng
s
of
2
013
C
hi
ne
s
e
I
nt
e
l
l
i
g
e
nt
A
ut
om
a
t
i
on C
onf
e
r
e
nc
e
,
pp.
5
91
-
60
2,
20
1
3.
Evaluation Warning : The document was created with Spire.PDF for Python.