I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
5
,
Octo
b
er
20
25
,
p
p
.
4
6
3
0
~
4
6
4
1
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
5
.
pp
4
6
3
0
-
4
6
4
1
4630
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Exploring
ensem
ble learning
for cl
a
ss
ifying
geo
metric
pa
tt
e
rns:
insig
hts f
ro
m q
ua
ternion ca
r
tesia
n
fractiona
l
H
a
hn
mo
ments
Z
o
uh
a
ir
O
ua
ze
ne
1
,
Aziz
K
ha
m
j
a
ne
2
1
La
b
o
r
a
t
o
i
r
e
d
e
Tr
a
n
sm
i
ssi
o
n
e
t
d
e
Tr
a
i
t
e
me
n
t
d
e
L’
i
n
f
o
r
ma
t
i
o
n
,
E
c
o
l
e
S
u
p
é
r
i
e
u
r
e
d
e
Te
c
h
n
o
l
o
g
i
e
,
S
i
d
i
M
o
h
a
me
d
B
e
n
A
b
d
e
l
l
a
h
U
n
i
v
e
r
si
t
y
,
F
e
z
,
M
o
r
o
c
c
o
2
La
b
o
r
a
t
o
r
y
o
f
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
,
EN
S
A
H
,
A
b
d
e
l
m
a
l
e
k
Ess
a
a
d
i
U
n
i
v
e
r
si
t
y
,
A
l
H
o
c
e
i
m
a
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
5
,
2
0
2
5
R
ev
is
ed
J
u
n
3
,
2
0
2
5
Acc
ep
ted
J
u
n
3
0
,
2
0
2
5
Th
e
c
las
sifica
ti
o
n
o
f
g
e
o
m
e
tri
c
p
a
tt
e
rn
s,
p
a
rti
c
u
larly
i
n
Isla
m
ic
a
rt,
p
re
se
n
ts
a
c
o
m
p
e
ll
in
g
c
h
a
l
len
g
e
f
o
r
th
e
f
i
e
ld
o
f
c
o
m
p
u
ter
v
isio
n
d
u
e
to
it
s
in
tr
ica
te
sy
m
m
e
try
a
n
d
sc
a
le
in
v
a
rian
c
e
.
Th
is
st
u
d
y
p
ro
p
o
se
s
a
n
e
n
se
m
b
l
e
lea
rn
in
g
fra
m
e
wo
rk
to
c
las
sify
g
e
o
m
e
tri
c
p
a
tt
e
rn
s,
lev
e
ra
g
i
n
g
t
h
e
n
o
v
e
l
q
u
a
tern
i
o
n
c
a
rtes
ian
fra
c
ti
o
n
a
l
Ha
h
n
m
o
m
e
n
ts
(QCFrHM
s)
a
s
a
ro
b
u
st
fe
a
tu
re
e
x
trac
ti
o
n
m
e
th
o
d
.
QCFrHM
s
i
n
t
e
g
ra
te
th
e
fra
c
ti
o
n
a
l
Ha
h
n
p
o
l
y
n
o
m
ial
a
n
d
q
u
a
tern
i
o
n
a
l
g
e
b
ra
to
p
ro
v
id
e
c
o
m
p
a
c
t,
in
v
a
rian
t
d
e
sc
rip
to
rs
f
o
r
g
e
o
m
e
tri
c
p
a
tt
e
rn
s.
C
o
m
b
in
e
d
wit
h
Zern
i
k
e
M
o
m
e
n
ts,
th
is
d
u
a
l
-
fe
a
tu
re
a
p
p
r
o
a
c
h
e
n
su
re
s
re
sili
e
n
c
e
a
g
a
i
n
st
r
o
tat
io
n
,
sc
a
li
n
g
,
a
n
d
n
o
ise
v
a
riatio
n
s.
Th
e
e
x
trac
ted
fe
a
tu
re
s
we
re
e
v
a
lu
a
ted
u
sin
g
s
u
p
p
o
rt
v
e
c
to
r
m
a
c
h
i
n
e
s
(S
VM),
ra
n
d
o
m
fo
re
st,
a
n
d
a
so
ft
-
v
o
t
in
g
e
n
se
m
b
le
c
las
sifier.
Ex
p
e
rime
n
ts
we
re
c
o
n
d
u
c
ted
o
n
a
d
a
tas
e
t
c
o
m
p
risin
g
1
,
2
0
4
g
e
o
m
e
tri
c
ima
g
e
s
c
a
teg
o
rize
d
in
t
o
two
sy
m
m
e
try
g
r
o
u
p
s
(p
4
m
a
n
d
p
6
m
).
Re
su
lt
s
d
e
m
o
n
stra
ted
th
a
t
th
e
e
n
se
m
b
le
c
las
sifier
o
u
tp
e
rfo
r
m
e
d
sta
n
d
a
lo
n
e
m
o
d
e
ls,
a
c
h
iev
i
n
g
a
c
las
sifica
ti
o
n
a
c
c
u
ra
c
y
o
f
8
2
.
1
5
%
.
T
h
e
in
te
g
ra
ti
o
n
o
f
QCFrHM
s
sig
n
ifi
c
a
n
t
ly
e
n
h
a
n
c
e
d
th
e
sy
st
e
m
'
s
ro
b
u
stn
e
ss
c
o
m
p
a
re
d
to
t
ra
d
it
io
n
a
l
Zern
ik
e
-
o
n
ly
a
p
p
ro
a
c
h
e
s,
wh
ic
h
a
li
g
n
s
with
fi
n
d
i
n
g
s
i
n
p
ri
o
r
stu
d
ies
.
Th
is
re
se
a
rc
h
c
o
n
tri
b
u
tes
to
th
e
field
s
o
f
ima
g
e
p
ro
c
e
ss
in
g
a
n
d
p
a
tt
e
rn
r
e
c
o
g
n
it
i
o
n
b
y
in
t
ro
d
u
c
in
g
a
n
e
fficie
n
t
fe
a
tu
re
e
x
trac
ti
o
n
tec
h
n
iq
u
e
c
o
m
b
in
e
d
with
e
n
se
m
b
le
lea
rn
in
g
f
o
r
p
re
c
ise
a
n
d
sc
a
lab
le
g
e
o
m
e
tri
c
p
a
tt
e
rn
c
las
sifica
ti
o
n
.
Th
e
imp
li
c
a
ti
o
n
s
e
x
ten
d
t
o
a
rt
p
re
se
rv
a
ti
o
n
,
a
rc
h
it
e
c
tu
ra
l
a
n
a
ly
sis,
a
n
d
a
u
to
m
a
ted
i
n
d
e
x
i
n
g
o
f
c
u
lt
u
ra
l
h
e
rit
a
g
e
ima
g
e
ry
.
K
ey
w
o
r
d
s
:
Featu
r
e
ex
tr
ac
tio
n
Geo
m
etr
ic
p
atter
n
s
Ma
ch
in
e
lear
n
in
g
Qu
ater
n
io
n
ca
r
tesi
an
f
r
ac
tio
n
al
Hah
n
m
o
m
e
n
ts
Sy
m
m
etr
y
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Z
o
u
h
air
Ou
az
en
e
L
ab
o
r
ato
ir
e
d
e
T
r
a
n
s
m
is
s
io
n
et
d
e
T
r
aitem
en
t d
e
L
’
in
f
o
r
m
ati
o
n
,
E
co
le
Su
p
ér
ieu
r
e
d
e
T
ec
h
n
o
lo
g
ie,
Sid
i M
o
h
am
ed
B
en
Ab
d
ella
h
Un
iv
er
s
it
y
Fez,
Mo
r
o
cc
o
E
m
ail: z
o
u
h
air
.
o
u
az
en
e@
u
s
m
b
a.
ac
.
m
a
1.
I
NT
RO
D
UCT
I
O
N
Ad
v
an
ce
s
in
co
m
p
u
tatio
n
al
tech
n
iq
u
es
an
d
m
ac
h
in
e
l
ea
r
n
in
g
h
a
v
e
r
ec
en
tly
u
n
lo
ck
ed
n
ew
o
p
p
o
r
tu
n
ities
f
o
r
a
n
aly
zin
g
co
m
p
lex
v
is
u
al
p
atter
n
s
,
p
ar
ticu
l
ar
ly
with
in
cu
ltu
r
al
h
er
itag
e
a
n
d
a
r
tis
tic
d
o
m
ain
s
[
1
]
.
Geo
m
etr
ic
p
atter
n
s
,
wit
h
th
eir
in
tr
icate
s
y
m
m
etr
y
,
m
ath
em
atica
l
p
r
ec
is
io
n
,
an
d
ae
s
th
etic
ap
p
ea
l,
h
allm
ar
k
s
as
th
ey
wer
e
o
f
I
s
la
m
ic
ar
t,
p
r
esen
t
a
u
n
iq
u
e
ch
al
len
g
e
an
d
o
p
p
o
r
tu
n
ity
.
T
h
ese
p
atter
n
s
h
av
e
lo
n
g
f
ascin
ated
m
ath
em
atician
s
,
ar
tis
ts
,
an
d
co
m
p
u
ter
s
cien
tis
ts
alik
e.
De
s
p
ite
th
eir
ap
p
ar
en
t
s
im
p
licity
,
th
e
class
if
icatio
n
o
f
s
u
ch
p
atter
n
s
is
a
co
m
p
lex
p
r
o
b
lem
r
e
q
u
ir
in
g
p
o
wer
f
u
l
m
eth
o
d
s
ca
p
ab
le
o
f
ad
d
r
ess
in
g
v
ar
iatio
n
s
in
s
ca
le,
r
o
tatio
n
,
an
d
n
o
is
e
[
2
]
–
[
7
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
xp
lo
r
in
g
en
s
emb
le
lea
r
n
in
g
fo
r
cla
s
s
ifyin
g
g
eo
metric p
a
tter
n
s
:
in
s
ig
h
ts
fr
o
m
…
(
Zo
u
h
a
ir
Ou
a
z
en
e
)
4631
Mo
r
o
cc
o
’
s
r
ich
ar
tis
tic
tr
ad
iti
o
n
r
ef
lects
its
p
o
s
itio
n
as
a
c
u
ltu
r
al
cr
o
s
s
r
o
ad
s
in
th
e
I
s
lam
ic
wo
r
ld
.
Fro
m
th
e
1
1
th
ce
n
t
u
r
y
o
n
wa
r
d
,
M
o
r
o
cc
a
n
o
r
n
am
e
n
tatio
n
d
ev
elo
p
ed
th
r
o
u
g
h
s
u
cc
ess
iv
e
d
y
n
asti
es,
ea
ch
leav
in
g
its
m
ar
k
o
n
th
e
co
u
n
tr
y
'
s
ar
ch
itectu
r
al
an
d
d
ec
o
r
ativ
e
h
e
r
itag
e
[
8
]
,
[
9
]
.
C
h
ar
ac
ter
ized
b
y
its
s
y
m
m
etr
y
,
v
ib
r
an
t
c
o
lo
r
s
,
a
n
d
in
tr
icate
d
esig
n
s
,
M
o
r
o
cc
a
n
o
r
n
am
en
tatio
n
is
s
ee
n
in
m
o
s
q
u
es,
m
ad
r
asas
,
p
alac
es,
an
d
p
u
b
lic
s
p
ac
es.
T
h
r
ee
p
r
im
ar
y
th
em
es d
o
m
in
ate
Mo
r
o
cc
an
o
r
n
am
e
n
tal
ar
t:
a.
Geo
m
etr
ic
p
atter
n
s
: Sh
o
wca
s
in
g
th
e
p
r
ec
is
io
n
an
d
in
g
e
n
u
ity
o
f
cr
af
ts
m
en
.
b.
Flo
r
al
p
atter
n
s
(
T
awr
iq
)
: Rep
r
esen
tin
g
s
ty
lized
n
atu
r
al
m
o
tif
s
.
c.
C
allig
r
ap
h
y
: Co
m
b
in
in
g
ar
tis
tr
y
with
s
cr
ip
tu
r
al
r
e
v
er
en
ce
.
T
h
ese
th
em
es
ar
e
e
x
p
r
ess
ed
t
h
r
o
u
g
h
d
iv
er
s
e
m
ater
ials
,
in
cl
u
d
in
g
p
last
er
,
wo
o
d
,
ze
llij
s
(
m
o
s
aics)
,
an
d
ca
r
v
ed
s
to
n
e.
T
r
ad
itio
n
al
p
atter
n
r
ec
o
g
n
itio
n
m
eth
o
d
s
h
a
v
e
r
elied
o
n
a
h
o
s
t
o
f
f
ea
tu
r
e
d
escr
ip
to
r
s
li
k
e
Z
er
n
ik
e
m
o
m
en
ts
,
wh
ich
s
u
p
p
o
r
t
r
o
tatio
n
al
in
v
ar
ia
n
ce
an
d
h
a
v
e
p
r
o
v
en
th
eir
wo
r
th
in
ex
t
r
ac
tin
g
th
e
m
o
s
t
v
alu
a
b
le
g
eo
m
etr
ic
f
ea
t
u
r
es.
Ho
wev
er
,
th
e
n
atu
r
al
lim
itatio
n
s
o
f
th
ese
d
escr
ip
to
r
s
,
an
d
in
p
ar
ticu
lar
t
h
eir
h
an
d
lin
g
with
h
ig
h
er
-
o
r
d
e
r
co
m
p
lex
ities
an
d
co
lo
r
im
a
g
es,
ca
ll
f
o
r
m
o
r
e
s
o
p
h
is
ticated
ap
p
r
o
ac
h
es
[
1
0
]
,
[
1
1
]
.
T
h
is
en
co
u
r
a
g
es
th
e
ex
p
lo
r
atio
n
o
f
n
o
v
el
d
escr
ip
to
r
s
b
ased
o
n
m
o
m
en
ts
,
with
q
u
ater
n
io
n
ca
r
tesi
an
f
r
ac
tio
n
al
Hah
n
m
o
m
en
ts
(
QC
FrH
Ms)
b
ein
g
o
n
e
o
f
th
e
m
o
s
t p
r
o
m
is
in
g
ca
n
d
id
ates
[
1
2
]
.
QC
FrH
M
s
g
en
er
alize
th
e
cla
s
s
ic
m
o
m
en
ts
r
ep
r
esen
tatio
n
ca
p
ab
ilit
ies
b
y
em
b
ed
d
in
g
q
u
ater
n
io
n
alg
eb
r
a
a
n
d
f
r
ac
tio
n
al
p
o
ly
n
o
m
ials
f
o
r
a
co
m
p
ac
t,
h
o
lis
tic
r
ep
r
esen
tatio
n
o
f
g
r
ay
s
ca
le
an
d
co
lo
r
p
atter
n
s
.
T
h
is
wo
r
k
f
u
r
th
e
r
p
r
o
p
o
s
es
an
ex
ten
d
ed
f
r
a
m
ewo
r
k
th
at
in
t
eg
r
ates
QC
FrH
M
s
wi
th
th
e
en
s
em
b
le
lear
n
in
g
tech
n
iq
u
e
f
o
r
class
if
icatio
n
i
n
th
e
co
n
tex
t
o
f
g
e
o
m
etr
ic
p
atter
n
s
.
B
y
in
teg
r
atin
g
en
s
em
b
le
lear
n
in
g
,
wh
ic
h
allo
ws
th
e
ag
g
r
eg
atio
n
o
f
th
e
ab
ilit
ies
o
f
m
u
ltip
le
class
if
ier
s
,
th
e
r
o
b
u
s
tn
ess
an
d
p
r
ec
is
io
n
o
f
th
e
s
y
s
tem
ar
e
in
cr
ea
s
ed
.
W
e
d
em
o
n
s
tr
ate
th
e
ef
f
ec
tiv
e
n
ess
o
f
th
is
ap
p
r
o
a
ch
u
s
in
g
r
a
n
d
o
m
f
o
r
ests
(
R
F
h
en
ce
f
o
r
th
)
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM
h
en
ce
f
o
r
th
)
,
an
d
a
s
o
f
t
-
v
o
tin
g
class
if
ier
o
n
a
d
ataset
o
f
I
s
lam
ic
g
eo
m
etr
ic
p
atter
n
s
ca
teg
o
r
ized
in
to
th
eir
r
esp
ec
ti
v
e
s
y
m
m
etr
y
g
r
o
u
p
s
.
Pre
v
io
u
s
wo
r
k
s
h
av
e
alr
ea
d
y
u
n
d
er
lin
e
d
th
e
im
p
o
r
tan
ce
o
f
a
r
o
b
u
s
t
d
escr
ip
to
r
t
o
en
s
u
r
e
r
o
tatio
n
a
n
d
s
ca
le
in
v
ar
ian
ce
.
L
ik
ewise,
s
o
m
e
r
ec
e
n
t
w
o
r
k
s
d
em
o
n
s
tr
ated
QC
FrH
M
s
o
n
co
lo
r
im
ag
e
an
aly
s
is
an
d
p
r
o
v
ed
th
eir
ap
p
l
icatio
n
in
wate
r
m
ar
k
in
g
an
d
p
atter
n
r
ec
o
g
n
itio
n
task
s
.
B
e
s
id
es,
th
e
s
y
m
m
etr
y
r
esear
ch
in
I
s
lam
ic
g
eo
m
e
tr
ic
p
atter
n
s
co
n
d
u
cte
d
b
y
Kap
lan
an
d
Salesi
n
d
em
o
n
s
tr
ates
th
at
m
ath
em
atic
al
m
o
d
els
ar
e
h
ig
h
ly
im
p
o
r
ta
n
t
in
th
e
co
m
p
r
eh
e
n
s
io
n
an
d
elab
o
r
atio
n
o
f
s
u
ch
co
m
p
lex
p
atter
n
s
.
All
th
ese
wo
r
k
s
co
n
f
i
r
m
th
e
u
r
g
e
n
t
n
ee
d
to
co
m
b
in
e
a
d
v
an
ce
d
d
escr
ip
to
r
s
,
s
u
ch
as
QC
FrH
M
s
,
with
m
ac
h
in
e
lea
r
n
in
g
m
eth
o
d
o
lo
g
ies
f
o
r
im
p
r
o
v
in
g
th
e
r
esu
lts
in
class
if
i
ca
tio
n
.
I
n
f
ac
t,
th
e
ex
p
er
im
en
tal
r
esu
lts
d
em
o
n
s
tr
ated
th
at
th
e
in
co
r
p
o
r
atio
n
o
f
QC
FrH
M
s
s
ig
n
if
ican
tly
im
p
r
o
v
ed
class
if
icatio
n
p
er
f
o
r
m
an
ce
,
esp
ec
ially
u
n
d
er
n
o
is
y
,
r
o
tated
,
an
d
s
ca
led
v
a
r
iatio
n
s
.
T
h
e
d
ev
elo
p
m
en
t
h
er
e
will
f
ill
n
o
t
o
n
ly
th
e
g
ap
s
in
th
e
ex
is
tin
g
p
atte
r
n
r
ec
o
g
n
itio
n
ar
e
n
a
b
u
t
also
ex
ten
d
to
m
o
r
e
g
en
er
al
ap
p
l
icatio
n
s
in
d
ig
ital
ar
ch
iv
in
g
,
cu
ltu
r
al
h
er
itag
e
p
r
eser
v
atio
n
,
an
d
au
to
m
ated
in
d
ex
in
g
o
f
ar
tis
tic
d
esig
n
s
.
T
h
is
wo
r
k
p
r
esen
ts
an
ex
am
p
le
o
f
h
o
w
ad
v
an
ce
d
m
o
m
en
t
d
escr
ip
to
r
s
a
n
d
e
n
s
em
b
l
e
lear
n
in
g
ca
n
m
er
g
e
th
eir
s
tr
en
g
th
s
s
u
cc
ess
f
u
lly
to
s
o
lv
e
co
m
p
le
x
co
m
p
u
tatio
n
al
p
r
o
b
lem
s
.
W
e
h
av
e
s
tr
u
ctu
r
ed
th
is
p
ap
e
r
as
f
o
llo
ws.
T
h
e
f
o
r
th
c
o
m
in
g
s
ec
tio
n
d
ea
ls
with
s
o
m
e
r
elate
d
wo
r
k
s
th
at
ex
is
t
r
eg
ar
d
in
g
g
eo
m
etr
i
c
p
atter
n
class
if
icatio
n
an
d
s
o
m
e
m
o
m
e
n
t
-
b
ased
d
escr
ip
to
r
s
.
Nex
t,
th
is
wo
r
k
d
etails
o
n
th
e
m
eth
o
d
o
lo
g
y
p
r
o
p
o
s
ed
,
r
elatin
g
th
e
QC
FrH
M'
s
im
p
lem
en
tatio
n
with
t
h
e
co
n
ce
p
t
o
f
th
e
en
s
em
b
le
lear
n
in
g
f
r
am
ewo
r
k
.
T
h
is
wo
r
k
th
e
n
wr
ap
s
u
p
b
y
d
is
cu
s
s
in
g
th
e
im
p
licatio
n
s
o
f
ex
p
er
im
e
n
tal
r
esu
lts
an
d
o
f
f
er
i
n
g
a
c
o
n
clu
d
in
g
r
em
a
r
k
th
at
s
tip
u
lates a
f
e
w
f
u
tu
r
e
d
i
r
ec
tio
n
s
f
o
r
f
u
r
t
h
er
r
esear
ch
.
2.
RE
L
AT
E
D
WO
RK
T
h
e
class
if
icatio
n
o
f
g
eo
m
etr
ic
p
atter
n
s
,
p
ar
ticu
lar
ly
I
s
lam
ic
g
eo
m
etr
ic
p
atter
n
s
,
h
as
b
ee
n
a
f
o
cu
s
o
f
r
esear
ch
in
im
ag
e
an
aly
s
is
a
n
d
co
m
p
u
ter
v
is
io
n
f
o
r
s
ev
er
al
d
ec
ad
es.
q
u
ate
r
n
io
n
ca
r
tesi
an
f
r
ac
tio
n
al
Hah
n
m
o
m
en
t
s
(
QC
FrH
Ms)
r
ep
r
esen
t
a
s
ig
n
if
ican
t
ad
v
an
ce
m
e
n
t
in
th
is
d
o
m
ain
b
y
o
f
f
er
in
g
r
o
b
u
s
t
d
escr
ip
to
r
s
f
o
r
b
o
th
g
r
ay
s
ca
le
an
d
co
lo
r
i
m
ag
es.
T
h
is
s
ec
tio
n
ex
p
lo
r
es
f
o
u
n
d
atio
n
al
r
esear
ch
an
d
r
e
ce
n
t
ad
v
an
ce
m
en
ts
lead
in
g
to
th
ese
in
n
o
v
ativ
e
m
e
th
o
d
s
.
Or
th
o
g
o
n
al
m
o
m
e
n
ts
,
s
u
ch
a
s
Z
er
n
ik
e
m
o
m
en
ts
,
h
av
e
lo
n
g
b
ee
n
u
tili
ze
d
f
o
r
s
h
ap
e
-
b
ased
im
ag
e
class
if
icatio
n
d
u
e
to
th
eir
r
o
tat
io
n
al
in
v
ar
ian
ce
an
d
r
o
b
u
s
tn
ess
.
Ah
ad
ian
an
d
B
astan
f
ar
d
[
1
3
]
d
em
o
n
s
tr
ated
th
e
ef
f
icac
y
o
f
Z
er
n
ik
e
m
o
m
en
ts
f
o
r
class
if
y
in
g
I
s
lam
ic
g
eo
m
etr
ic
p
atter
n
s
.
Usi
n
g
n
eu
r
al
n
etwo
r
k
s
an
d
K
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
class
if
ier
s
,
t
h
ey
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
6
.
0
3
%
b
y
o
p
tim
izin
g
p
r
e
-
p
r
o
ce
s
s
in
g
an
d
f
ea
tu
r
e
ex
tr
ac
tio
n
tech
n
iq
u
es.
No
is
e
r
ed
u
ctio
n
,
s
eg
m
en
tatio
n
,
an
d
Z
er
n
ik
e
m
o
m
e
n
t
-
b
ased
d
escr
ip
t
o
r
s
wer
e
ce
n
tr
al
to
th
eir
ap
p
r
o
ac
h
,
alth
o
u
g
h
th
e
m
eth
o
d
was
lim
ited
to
g
r
ay
s
c
ale
im
ag
es.
Z
er
n
ik
e
m
o
m
en
ts
ar
e
m
at
h
em
atica
lly
d
ef
in
ed
as
[
1
4
]
:
=
+
1
∫
(
,
)
(
,
)
{
2
+
2
≤
1
}
(
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
6
3
0
-
4
6
4
1
4632
in
d
ee
d
,
d
e
f
in
es
th
e
Z
er
n
i
k
e
m
o
m
en
ts
,
wh
er
e
Vn
m
(
x
,
y
)
(
o
f
ten
ex
p
r
ess
ed
in
p
o
lar
f
o
r
m
)
ar
e
th
e
Z
er
n
ik
e
p
o
ly
n
o
m
ials
.
On
e
o
f
th
e
m
ajo
r
ad
v
an
tag
es
o
f
Z
e
r
n
ik
e
m
o
m
en
ts
is
th
eir
r
o
tatio
n
al
in
v
ar
ian
ce
,
wh
ich
is
wh
y
th
ey
ar
e
co
m
m
o
n
l
y
u
s
ed
in
g
e
o
m
etr
ic
p
atter
n
class
if
icatio
n
an
d
s
h
ap
e
r
ec
o
g
n
itio
n
.
T
h
e
em
er
g
en
ce
o
f
q
u
ater
n
io
n
alg
eb
r
a
in
im
ag
e
d
escr
ip
to
r
s
h
as
m
ar
k
ed
a
p
ar
ad
ig
m
s
h
if
t
in
im
ag
e
r
ep
r
esen
tatio
n
.
Yam
n
i
[
1
5
]
in
tr
o
d
u
ce
d
QC
FrH
Ms
as
a
g
en
er
aliza
tio
n
o
f
class
ical
Hah
n
m
o
m
en
ts
,
ex
ten
d
in
g
th
eir
ca
p
ab
ilit
ies
to
f
r
ac
tio
n
al
o
r
d
er
s
an
d
lev
er
a
g
in
g
q
u
ater
n
io
n
th
eo
r
y
f
o
r
c
o
m
p
ac
t
an
d
h
o
lis
tic
co
lo
r
im
ag
e
p
r
o
ce
s
s
in
g
.
QC
FrH
Ms a
d
d
r
ess
ed
s
ev
er
al
lim
itatio
n
s
o
f
p
r
e
v
io
u
s
m
eth
o
d
s
b
y
:
a.
Utilizin
g
f
r
ac
tio
n
al
Hah
n
p
o
ly
n
o
m
ials
(
FrH
Ps
)
to
en
h
an
ce
a
cc
u
r
ac
y
an
d
f
lex
i
b
ilit
y
.
b.
E
n
co
d
in
g
co
lo
r
in
f
o
r
m
atio
n
s
e
am
less
ly
u
s
in
g
q
u
ater
n
io
n
r
ep
r
esen
tatio
n
.
c.
Fra
ctio
n
al
Hah
n
p
o
ly
n
o
m
ials
ar
e
d
ef
in
e
d
r
ec
u
r
s
iv
ely
as
[
1
6
]
:
ℎ
(
,
)
(
)
=
(
2
+
+
−
1
)
(
−
1
)
ℎ
−
1
(
,
)
(
)
−
(
+
−
1
)
(
+
)
ℎ
−
2
(
,
)
(
)
(
2
)
W
ith
th
e
in
itial c
o
n
d
itio
n
s
ℎ
0
(
)
=
1
an
d
ℎ
1
(
)
=
(
−
)
+
(
1
+
+
)
(
−
1
)
QC
FrH
M
s
ef
f
ec
tiv
ely
en
co
d
e
th
e
in
h
e
r
en
t
s
y
m
m
etr
y
in
g
eo
m
etr
ic
p
atter
n
s
b
y
p
r
o
ce
s
s
in
g
co
lo
r
im
ag
es
in
a
h
o
lis
tic
an
d
c
o
m
p
ac
t
m
an
n
er
,
ca
p
tu
r
in
g
b
o
th
g
l
o
b
al
an
d
lo
ca
l
s
y
m
m
etr
ies.
T
h
ese
ad
v
an
ce
m
e
n
ts
h
av
e
b
r
o
a
d
en
e
d
th
e
s
co
p
e
o
f
a
p
p
licatio
n
s
f
o
r
QC
FrH
Ms,
in
clu
d
in
g
im
ag
e
wate
r
m
a
r
k
in
g
,
ed
g
e
d
etec
tio
n
,
a
n
d
p
atter
n
r
ec
o
g
n
itio
n
.
C
o
m
p
ar
e
d
to
tr
ad
itio
n
al
Hah
n
m
o
m
e
n
t
s
,
QC
FrH
Ms
d
em
o
n
s
tr
ate
r
ed
u
ce
d
c
o
m
p
u
tatio
n
al
co
m
p
lex
ity
an
d
en
h
a
n
ce
d
n
u
m
er
ical
s
tab
ilit
y
.
T
h
eir
r
o
b
u
s
tn
ess
to
g
eo
m
etr
ic
tr
an
s
f
o
r
m
atio
n
s
s
u
ch
as
r
o
tatio
n
an
d
s
ca
lin
g
m
ak
es th
e
m
p
ar
tic
u
lar
ly
ef
f
ec
tiv
e
f
o
r
c
o
m
p
le
x
i
m
ag
e
p
r
o
ce
s
s
in
g
task
s
[
1
7
]
.
Mo
d
e
r
n
f
ea
t
u
r
e
ex
tr
ac
ti
o
n
tec
h
n
i
q
u
es
c
o
m
b
i
n
e
c
o
m
p
l
em
e
n
t
ar
y
d
esc
r
i
p
t
o
r
s
t
o
m
a
x
im
iz
e
p
er
f
o
r
m
a
n
c
e.
Fo
r
e
x
a
m
p
le
,
i
n
t
eg
r
ati
n
g
QC
F
r
HM
s
wit
h
Z
e
r
n
ik
e
m
o
m
e
n
ts
ca
p
t
u
r
es
b
o
th
g
lo
b
a
l
a
n
d
l
o
c
a
l
c
h
a
r
a
cte
r
is
t
ics
o
f
g
e
o
m
et
r
i
c
p
atte
r
n
s
.
L
it
er
at
u
r
e
s
u
g
g
es
ts
th
at
s
u
c
h
c
o
m
b
i
n
a
tio
n
s
a
r
e
p
i
v
o
ta
l
in
r
e
al
-
ti
m
e
a
p
p
li
ca
t
io
n
s
li
k
e
au
t
o
m
at
e
d
p
at
te
r
n
r
ec
o
g
n
iti
o
n
.
A
d
d
iti
o
n
al
ly
,
a
d
v
an
ce
m
e
n
t
s
in
m
ac
h
i
n
e
le
a
r
n
in
g
cl
ass
if
ier
s
,
in
cl
u
d
in
g
R
F,
SVM,
a
n
d
e
n
s
e
m
b
le
le
ar
n
i
n
g
m
et
h
o
d
s
,
c
o
m
p
l
em
en
t
t
h
es
e
f
e
atu
r
e
e
x
t
r
a
cti
o
n
t
ec
h
n
i
q
u
es.
E
n
s
em
b
le
a
p
p
r
o
a
c
h
es
,
s
u
c
h
as
v
o
t
in
g
cl
ass
i
f
ie
r
s
,
en
h
a
n
c
e
ac
c
u
r
a
cy
b
y
l
e
v
e
r
a
g
i
n
g
t
h
e
s
tr
e
n
g
th
s
o
f
m
u
lti
p
l
e
m
o
d
els
[
1
8
]
.
QC
FrH
M
s
an
d
r
elate
d
m
eth
o
d
o
lo
g
ies
h
av
e
ca
taly
ze
d
n
e
w
r
esear
ch
d
ir
ec
tio
n
s
in
im
a
g
e
an
aly
s
is
[
1
9
]
–
[
2
1
]
.
Po
ten
tial a
p
p
licatio
n
s
in
clu
d
e:
a.
C
u
ltu
r
al
h
er
itag
e
p
r
eser
v
atio
n
:
d
ig
itizin
g
a
n
d
class
if
y
in
g
h
i
s
to
r
ical
g
eo
m
etr
ic
p
atter
n
s
f
o
r
r
ec
o
n
s
tr
u
ctio
n
an
d
ar
ch
i
v
al
p
u
r
p
o
s
es.
b.
Me
d
ical
im
ag
in
g
: im
p
r
o
v
in
g
d
iag
n
o
s
tic
ac
cu
r
ac
y
t
h
r
o
u
g
h
e
n
h
an
ce
d
p
atter
n
r
ec
o
g
n
itio
n
.
c.
C
o
n
ten
t
-
b
ased
im
ag
e
r
etr
iev
al
: e
n
ab
lin
g
ef
f
icien
t in
d
e
x
in
g
a
n
d
r
etr
iev
al
i
n
m
u
ltime
d
ia
d
atab
ases
.
d.
Hig
h
-
s
ec
u
r
it
y
a
p
p
li
ca
t
io
n
s
:
d
i
g
ital
w
ate
r
m
a
r
k
in
g
a
n
d
f
o
r
g
e
r
y
d
et
ec
ti
o
n
t
h
r
o
u
g
h
co
m
p
a
ct
a
n
d
d
is
c
r
i
m
i
n
at
iv
e
im
a
g
e
r
e
p
r
ese
n
t
ati
o
n
s
.
T
h
e
in
teg
r
atio
n
o
f
d
ee
p
lea
r
n
i
n
g
f
r
am
ewo
r
k
s
with
QC
FrH
Ms
f
o
r
e
n
d
-
to
-
en
d
class
if
icati
o
n
p
i
p
elin
es
s
h
o
u
ld
b
e
th
e
ce
n
tr
al
f
o
cu
s
f
o
r
f
u
tu
r
e
r
esear
ch
.
E
x
p
lo
r
in
g
th
eir
ap
p
licab
ilit
y
in
3
D
o
b
ject
an
aly
s
is
,
r
ea
l
-
tim
e
v
id
eo
p
r
o
ce
s
s
in
g
,
an
d
g
en
e
r
ativ
e
m
o
d
elin
g
f
o
r
g
e
o
m
etr
ic
p
at
ter
n
s
y
n
th
esis
also
h
o
ld
s
s
ig
n
i
f
ican
t
p
r
o
m
is
e.
B
y
co
m
b
in
in
g
tr
ad
itio
n
al
o
r
th
o
g
o
n
al
m
o
m
en
ts
with
m
o
d
er
n
q
u
a
ter
n
io
n
-
b
ased
ap
p
r
o
ac
h
es
an
d
ad
v
an
ce
d
m
ac
h
in
e
lear
n
in
g
tech
n
iq
u
es,
QC
FrH
Ms
s
et
a
n
ew
b
en
ch
m
a
r
k
in
g
eo
m
etr
ic
p
atter
n
class
if
icatio
n
,
o
f
f
er
in
g
r
o
b
u
s
t,
ef
f
icien
t,
an
d
v
er
s
atile
s
o
lu
tio
n
s
[
2
2
]
–
[
2
4
]
.
3.
M
E
T
H
O
DO
L
O
G
Y
3
.
1
.
Da
t
a
s
et
prepa
ra
t
io
n
3
.
1
.
1
.
Da
t
a
s
et
c
o
ns
t
it
utio
n
I
n
th
e
cu
r
r
en
t
in
v
esti
g
atio
n
,
w
e
ass
em
b
led
a
d
ataset
o
f
im
ag
es
d
ep
ictin
g
g
eo
m
et
r
ic
m
o
tifs
with
two
ty
p
es
o
f
s
y
m
m
etr
ies:
f
o
u
r
-
f
o
ld
s
y
m
m
etr
y
(
p
4
m
)
a
n
d
s
ix
-
f
o
ld
s
y
m
m
etr
y
(
p
6
m
)
.
T
h
ese
ca
teg
o
r
ies
wer
e
d
elib
er
ately
ch
o
s
en
d
u
e
to
th
eir
f
r
eq
u
e
n
t
o
cc
u
r
r
en
ce
in
tili
n
g
ar
t,
o
r
n
am
en
tal
d
esig
n
s
,
a
n
d
cr
y
s
tallo
g
r
a
p
h
ic
p
atter
n
s
.
a.
p4m
s
y
m
m
etr
y
:
C
o
m
m
o
n
ly
f
o
u
n
d
in
s
q
u
ar
e
tili
n
g
ar
r
an
g
e
m
en
ts
,
ce
r
am
ic
ar
t,
an
d
ce
r
tain
m
an
d
ala
-
lik
e
p
atter
n
s
.
Fig
u
r
e
1
s
h
o
ws an
ex
am
p
le
f
r
o
m
th
e
d
ata
b
ase
o
f
th
i
s
ty
p
e
o
f
s
y
m
m
et
r
y
.
b.
p6m
s
y
m
m
etr
y
:
C
h
ar
ac
ter
is
tic
o
f
h
ex
a
g
o
n
al
lay
o
u
ts
,
s
u
ch
as
h
o
n
ey
co
m
b
s
tr
u
ct
u
r
es
o
r
I
s
lam
ic
-
in
s
p
ir
ed
m
o
tifs
with
h
ex
ag
o
n
al
s
y
m
m
etr
y
.
Fig
u
r
e
2
s
h
o
ws
an
ex
am
p
le
f
r
o
m
th
e
d
atab
ase
o
f
t
h
is
s
y
m
m
etr
y
ty
p
e.
B
y
ca
p
tu
r
in
g
t
h
e
d
is
tin
ct
q
u
al
ities
o
f
p
4
m
an
d
p
6
m
,
o
u
r
d
a
taset
p
r
o
v
id
es
a
f
er
tile
g
r
o
u
n
d
f
o
r
ex
p
lo
r
in
g
s
y
m
m
etr
y
-
b
ased
class
if
icatio
n
task
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
xp
lo
r
in
g
en
s
emb
le
lea
r
n
in
g
fo
r
cla
s
s
ifyin
g
g
eo
metric p
a
tter
n
s
:
in
s
ig
h
ts
fr
o
m
…
(
Zo
u
h
a
ir
Ou
a
z
en
e
)
4633
Fig
u
r
e
1
.
E
x
am
p
le
f
r
o
m
t
h
e
d
a
tab
ase
illu
s
tr
atin
g
p
4
m
s
y
m
m
e
tr
y
Fig
u
r
e
2
.
E
x
am
p
le
f
r
o
m
t
h
e
d
a
tab
ase
illu
s
tr
atin
g
p
6
m
s
y
m
m
e
tr
y
3
.
1
.
2
.
I
m
a
g
e
prepro
ce
s
s
ing
T
h
e
p
r
e
p
r
o
ce
s
s
in
g
p
ip
elin
e
in
th
is
s
tu
d
y
en
s
u
r
es
h
i
g
h
-
q
u
ality
in
p
u
t
f
o
r
f
ea
tu
r
e
e
x
tr
ac
tio
n
b
y
in
co
r
p
o
r
atin
g
tec
h
n
iq
u
es
s
u
ch
as
Me
d
ian
f
ilter
in
g
an
d
Ots
u
’
s
th
r
esh
o
ld
in
g
,
alo
n
g
with
au
g
m
en
tatio
n
u
s
in
g
Gau
s
s
ian
n
o
is
e.
Me
d
ian
f
ilter
in
g
is
ap
p
lied
to
th
e
r
esized
g
r
ay
s
ca
le
im
ag
es
to
r
ed
u
ce
im
p
u
ls
iv
e
n
o
is
e
wh
ile
p
r
eser
v
in
g
c
r
itical
ed
g
es.
Fo
r
s
eg
m
en
tatio
n
,
Ots
u
’
s
T
h
r
esh
o
ld
in
g
is
em
p
lo
y
ed
to
s
ep
ar
a
te
th
e
f
o
r
eg
r
o
u
n
d
(
g
eo
m
etr
ic
m
o
tifs)
f
r
o
m
th
e
b
ac
k
g
r
o
u
n
d
.
Ots
u
’
s
m
eth
o
d
co
m
p
u
tes
th
e
o
p
tim
al
th
r
esh
o
ld
b
y
m
in
im
izin
g
in
t
r
a
-
class
v
ar
ian
ce
,
d
ef
in
ed
as:
2
=
(
1
−
1
)
2
1
(
1
−
1
)
(
3
)
w
h
er
e
is
th
e
to
tal
m
ea
n
in
ten
s
ity
o
f
th
e
im
ag
e
1
is
th
e
m
ea
n
in
ten
s
ity
o
f
th
e
f
o
r
e
g
r
o
u
n
d
p
ix
els,
an
d
1
is
th
e
p
r
o
p
o
r
tio
n
o
f
p
ix
els
class
if
ied
as
f
o
r
e
g
r
o
u
n
d
.
T
h
is
en
s
u
r
es
p
r
ec
is
e
s
eg
m
en
tatio
n
,
p
ar
t
icu
lar
ly
f
o
r
im
ag
es
with
v
ar
y
in
g
in
ten
s
ity
d
is
tr
ib
u
tio
n
s
.
T
o
f
u
r
th
er
en
h
an
ce
t
h
e
r
o
b
u
s
t
n
ess
o
f
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
p
r
o
ce
s
s
,
Gau
s
s
ian
n
o
is
e
is
ad
d
ed
d
u
r
in
g
d
ataset
au
g
m
en
tatio
n
.
T
h
is
is
m
o
d
eled
m
at
h
em
atica
lly
as:
(
,
)
=
0
(
,
)
+
(
,
)
(
4
)
w
h
er
e
(
,
)
r
ep
r
esen
ts
th
e
n
o
is
y
im
ag
e,
0
(
,
)
is
th
e
o
r
ig
in
al
im
ag
e,
an
d
(
,
)
is
th
e
Gau
s
s
ian
n
o
is
e
with
ze
r
o
m
ea
n
a
n
d
a
s
p
ec
i
f
ied
v
a
r
ian
ce
.
T
h
is
au
g
m
en
t
atio
n
test
s
th
e
s
y
s
tem
'
s
r
esil
ien
ce
to
r
ea
l
-
wo
r
ld
s
ce
n
ar
io
s
wh
er
e
n
o
is
e
is
p
r
e
v
alen
t,
s
u
ch
as
v
ar
iatio
n
s
i
n
lig
h
tin
g
o
r
s
en
s
o
r
im
p
er
f
ec
tio
n
s
.
T
h
e
p
r
ep
r
o
ce
s
s
in
g
p
ip
elin
e
th
u
s
co
m
b
i
n
es
d
en
o
i
s
in
g
,
s
eg
m
en
tatio
n
,
an
d
a
u
g
m
en
tatio
n
to
p
r
e
p
ar
e
im
a
g
es
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
,
en
s
u
r
in
g
b
o
th
r
o
b
u
s
tn
ess
an
d
p
r
ec
is
io
n
.
3
.
1
.
3
.
Seg
m
ent
a
t
io
n
W
ith
th
e
p
r
e
p
r
o
ce
s
s
ed
im
a
g
es
in
h
an
d
,
th
e
n
e
x
t
s
tep
in
v
o
lv
ed
s
eg
m
en
tatio
n
to
is
o
late
th
e
g
eo
m
et
r
ic
m
o
tif
f
r
o
m
its
b
ac
k
g
r
o
u
n
d
.
W
e
u
tili
ze
d
Ots
u
’
s
th
r
esh
o
ld
in
g
,
a
well
-
estab
lis
h
ed
tech
n
iq
u
e
th
at
au
to
m
atica
lly
d
eter
m
in
es th
e
o
p
tim
al
th
r
esh
o
ld
b
y
m
in
im
izin
g
i
n
tr
a
-
class
v
ar
ian
ce
:
a.
T
h
e
r
esu
lt is
a
b
in
ar
y
m
ask
p
a
r
titi
o
n
in
g
th
e
im
a
g
e
in
to
f
o
r
e
g
r
o
u
n
d
(
m
o
tif
)
an
d
b
ac
k
g
r
o
u
n
d
r
eg
io
n
s
.
b.
T
h
is
s
eg
m
en
tatio
n
p
r
o
v
es c
r
u
c
ial
f
o
r
ac
cu
r
ate
f
ea
tu
r
e
ex
tr
ac
t
io
n
,
as it d
ir
ec
ts
atten
tio
n
to
o
n
ly
th
e
s
h
ap
e
o
f
in
ter
est.
3
.
2
.
F
ea
t
ure
ex
t
ra
c
t
io
n
T
h
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
p
r
o
ce
s
s
in
th
is
s
tu
d
y
em
p
lo
y
s
two
p
o
wer
f
u
l
d
escr
ip
to
r
s
:
QC
F
r
HM
s
an
d
Z
er
n
ik
e
m
o
m
en
ts
,
ch
o
s
en
f
o
r
th
eir
ab
ilit
y
to
r
o
b
u
s
tly
en
co
d
e
g
eo
m
etr
ic
an
d
s
y
m
m
etr
y
f
ea
tu
r
es
u
n
d
er
v
ar
io
u
s
tr
an
s
f
o
r
m
atio
n
s
.
QC
FrH
Ms
b
u
ild
u
p
o
n
th
e
Fra
ctio
n
al
Ha
h
n
Po
ly
n
o
m
ials
,
wh
ich
wer
e
in
tr
o
d
u
ce
d
in
th
e
r
elate
d
wo
r
k
s
ec
tio
n
,
to
d
er
iv
e
r
o
b
u
s
t
m
o
m
en
t
co
ef
f
icie
n
ts
.
T
h
ese
p
o
ly
n
o
m
ials
,
d
ef
in
ed
r
ec
u
r
s
iv
ely
,
p
r
o
v
id
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
6
3
0
-
4
6
4
1
4634
in
h
er
en
t
s
tab
ilit
y
u
n
d
er
tr
an
s
f
o
r
m
atio
n
s
s
u
ch
as
r
o
tatio
n
,
s
c
alin
g
,
an
d
m
o
d
er
ate
n
o
is
e
in
t
er
f
er
en
ce
.
B
y
u
s
in
g
th
e
r
ec
u
r
s
io
n
f
o
r
m
u
la
ℎ
(
,
)
(
)
an
d
i
ts
in
itial
co
n
d
itio
n
s
,
as
d
escr
ib
ed
in
r
elate
d
wo
r
k
,
th
e
co
ef
f
icien
ts
r
eq
u
ir
ed
f
o
r
QC
FrH
Ms a
r
e
ef
f
icien
tly
co
m
p
u
te
d
.
QC
FrH
M
s
f
u
r
th
er
e
n
h
an
ce
th
eir
d
escr
ip
tiv
e
p
o
we
r
b
y
en
co
d
in
g
co
lo
r
in
f
o
r
m
ati
o
n
th
r
o
u
g
h
q
u
ater
n
io
n
alg
eb
r
a.
E
ac
h
p
ix
el
o
f
a
co
l
o
r
im
ag
e
is
r
ep
r
esen
te
d
as a
q
u
ater
n
io
n
(
,
)
=
(
,
)
+
(
,
)
+
(
,
)
(
5
)
wh
er
e
(
,
)
,
(
,
)
,
an
d
(
,
)
co
r
r
esp
o
n
d
to
t
h
e
p
ix
el
in
ten
s
ities
in
th
e
r
ed
,
g
r
ee
n
,
an
d
b
l
u
e
ch
a
n
n
els,
r
esp
ec
tiv
ely
.
T
h
is
h
o
lis
tic
r
e
p
r
esen
tatio
n
en
ab
les
QC
FrH
Ms
to
ca
p
tu
r
e
b
o
th
am
p
litu
d
e
an
d
p
h
ase
d
etails
s
im
u
ltan
eo
u
s
ly
,
o
f
f
er
in
g
a
co
m
p
ac
t a
n
d
m
u
ltid
im
en
s
io
n
al
f
ea
tu
r
e
s
et.
Z
er
n
ik
e
m
o
m
e
n
ts
co
m
p
lem
en
t
QC
FrH
M
s
b
y
p
r
o
v
id
in
g
r
o
t
atio
n
al
in
v
ar
ian
ce
,
wh
ic
h
is
p
ar
ticu
lar
ly
ef
f
ec
tiv
e
f
o
r
g
eo
m
etr
ic
p
atter
n
class
if
icatio
n
.
As
d
etailed
in
th
e
r
elate
d
wo
r
k
s
ec
tio
n
,
Z
er
n
ik
e
m
o
m
en
ts
ar
e
co
m
p
u
ted
u
s
in
g
o
r
th
o
g
o
n
al
p
o
ly
n
o
m
ials
(
r
,
θ)
,
wh
er
e
th
e
r
ad
ial
co
m
p
o
n
en
t
(
r
)
ca
p
tu
r
es
v
a
r
iatio
n
s
in
th
e
r
ad
ial
d
ir
ec
tio
n
,
an
d
th
e
an
g
u
lar
co
m
p
o
n
en
t
en
s
u
r
es
in
v
ar
ian
ce
to
r
o
tatio
n
[
2
5
]
.
T
h
e
r
ad
ial
p
o
ly
n
o
m
ial
(
r
)
is
d
ef
in
ed
as:
(
)
=
∑
(
−
1
)
(
−
)
!
!
(
+
|
|
2
−
)
!
(
−
|
|
2
−
)
!
−
|
|
2
=
0
−
2
(
6
)
T
h
is
d
ec
o
m
p
o
s
itio
n
e
n
ab
les
Z
er
n
ik
e
m
o
m
en
ts
to
en
ca
p
s
u
lat
e
b
o
th
g
l
o
b
al
an
d
lo
ca
l
f
ea
tu
r
es
o
f
th
e
g
eo
m
etr
ic
p
atter
n
s
,
m
ak
in
g
th
em
a
n
in
v
a
lu
ab
le
ad
d
itio
n
to
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
p
r
o
ce
s
s
.
T
h
e
co
m
b
in
e
d
u
s
e
o
f
QC
FrH
Ms
an
d
Z
er
n
ik
e
m
o
m
en
ts
r
esu
lts
in
a
u
n
if
ied
f
ea
tu
r
e
v
ec
to
r
th
at
ca
p
tu
r
es
in
tr
icate
g
eo
m
etr
ic
d
etails
wh
ile
m
ain
tain
in
g
r
o
b
u
s
tn
ess
ac
r
o
s
s
tr
an
s
f
o
r
m
atio
n
s
.
T
h
e
f
ea
tu
r
es
f
r
o
m
b
o
th
d
escr
ip
to
r
s
ar
e
n
o
r
m
aliz
ed
to
en
s
u
r
e
co
m
p
a
r
ab
ilit
y
an
d
co
n
ca
ten
ated
i
n
to
a
s
in
g
le
v
ec
to
r
.
T
h
is
v
ec
to
r
,
in
itially
co
m
p
r
is
in
g
3
4
r
aw
f
ea
tu
r
es,
is
f
u
r
th
er
r
ef
in
ed
d
u
r
in
g
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
to
en
h
an
ce
co
m
p
u
tatio
n
al
ef
f
icien
cy
an
d
m
itig
ate
o
v
er
f
itti
n
g
.
T
h
is
d
u
al
-
f
ea
tu
r
e
ap
p
r
o
ac
h
s
ig
n
if
ican
tly
im
p
r
o
v
es
th
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
b
y
lev
er
ag
in
g
t
h
e
s
tr
en
g
th
s
o
f
b
o
t
h
QC
FrH
Ms a
n
d
Z
er
n
ik
e
m
o
m
en
ts
.
3
.
3
.
Dim
ens
io
na
lity
re
du
ct
io
n
3
.
3
.
1
P
rincipa
l
co
m
po
nent
a
na
ly
s
is
T
o
cu
r
tail
d
im
en
s
io
n
ality
with
o
u
t
s
ac
r
if
icin
g
cr
u
cial
v
ar
ia
n
ce
,
we
em
p
lo
y
ed
p
r
i
n
cip
al
co
m
p
o
n
en
t
an
aly
s
is
(
PC
A)
as
a
d
im
en
s
io
n
ality
r
ed
u
ctio
n
tech
n
iq
u
e.
PC
A
id
en
tifie
s
o
r
th
o
g
o
n
al
ax
es,
o
r
p
r
in
cip
al
co
m
p
o
n
en
ts
,
th
at
ca
p
tu
r
e
t
h
e
h
ig
h
est
v
ar
ian
ce
in
t
h
e
f
ea
tu
r
e
s
p
ac
e
b
y
p
er
f
o
r
m
in
g
eig
en
v
alu
e
d
ec
o
m
p
o
s
itio
n
o
n
th
e
c
o
v
ar
ia
n
ce
m
atr
ix
o
f
t
h
e
d
ata.
Ma
th
e
m
atica
lly
,
th
e
co
v
ar
ian
ce
m
atr
ix
is
co
m
p
u
t
ed
as:
=
,
wh
er
e
r
ep
r
esen
ts
th
e
ce
n
ter
e
d
d
ata
m
atr
ix
.
T
h
e
eig
e
n
v
alu
e
d
ec
o
m
p
o
s
itio
n
o
f
y
ield
s
eig
en
v
alu
es
an
d
eig
en
v
ec
to
r
s
,
e
x
p
r
ess
ed
as
=
[
2
6
]
.
Her
e,
th
e
ei
g
en
v
al
u
es
q
u
an
tify
th
e
am
o
u
n
t
o
f
v
ar
ia
n
ce
ex
p
lain
ed
b
y
th
ei
r
co
r
r
esp
o
n
d
i
n
g
eig
en
v
ec
to
r
s
,
wh
ich
d
ef
in
e
th
e
d
ir
ec
tio
n
s
o
f
m
ax
im
u
m
v
ar
ian
ce
in
th
e
d
ata.
Prin
cip
al
co
m
p
o
n
e
n
ts
ar
e
th
en
s
elec
ted
b
ased
o
n
a
v
ar
ian
ce
th
r
esh
o
ld
;
in
th
is
s
tu
d
y
,
we
r
etain
ed
co
m
p
o
n
en
ts
th
at
co
llectiv
ely
ex
p
lain
e
d
9
5
%
o
f
th
e
to
tal
v
ar
ia
n
ce
,
r
e
d
u
c
in
g
th
e
f
ea
tu
r
e
s
p
ac
e
f
r
o
m
3
4
to
1
0
d
im
en
s
io
n
s
.
T
h
is
ap
p
r
o
ac
h
n
o
t
o
n
ly
m
i
n
im
izes
th
e
r
is
k
o
f
o
v
e
r
f
itti
n
g
b
u
t a
ls
o
s
ig
n
if
ican
tly
r
ed
u
ce
s
th
e
co
m
p
u
tatio
n
al
lo
ad
f
o
r
s
u
b
s
eq
u
e
n
t c
lass
if
icatio
n
task
s
wh
ile
p
r
eser
v
in
g
th
e
m
o
s
t in
f
o
r
m
ativ
e
f
ea
tu
r
es.
3
.
3
.
2
.
t
-
SNE
Vis
ua
liza
t
io
n
W
e
lev
er
ag
ed
t
-
d
is
tr
ib
u
ted
s
to
ch
asti
c
n
eig
h
b
o
r
em
b
e
d
d
in
g
(
t
-
SNE
)
to
v
is
u
alize
t
h
e
s
ep
ar
ab
ilit
y
o
f
th
e
d
ata
i
n
a
m
o
r
e
in
tu
itiv
e
2
D
o
r
3
D
s
p
ac
e,
en
ab
lin
g
b
etter
in
ter
p
r
etab
ilit
y
o
f
th
e
ex
tr
ac
ted
f
ea
tu
r
es.
t
-
SNE
wo
r
k
s
b
y
m
o
d
elin
g
h
ig
h
-
d
im
e
n
s
io
n
al
d
ata
p
o
in
ts
an
d
as
p
r
o
b
ab
ilit
ies
,
wh
er
e
th
e
s
im
ilar
ity
b
etwe
en
d
ata
p
o
in
ts
in
th
e
h
ig
h
-
d
im
en
s
io
n
al
s
p
ac
e
is
d
ef
in
ed
u
s
in
g
a
Gau
s
s
ian
d
is
tr
ib
u
tio
n
:
=
ex
p
(
−
|
|
−
|
|
2
2
2
)
∑
ex
p
(
−
|
|
−
|
|
2
2
2
)
≠
(
)
wh
er
e
|
|
−
|
|
2
is
th
e
s
q
u
ar
ed
E
u
clid
e
an
d
is
tan
ce
b
etwe
en
p
o
in
ts
a
n
d
,
an
d
σ
co
n
tr
o
ls
th
e
b
a
n
d
w
id
th
o
f
th
e
Gau
s
s
ian
k
er
n
el
[
2
7
]
.
I
n
th
e
lo
wer
-
d
im
e
n
s
io
n
al
s
p
ac
e,
t
-
SNE
m
in
im
izes
th
e
Ku
l
lb
ac
k
-
L
eib
ler
(
KL
)
d
iv
er
g
en
ce
b
etwe
en
th
e
jo
in
t
p
r
o
b
a
b
ilit
y
d
is
tr
ib
u
tio
n
in
th
e
o
r
ig
i
n
al
s
p
ac
e
an
d
in
th
e
em
b
ed
d
ed
s
p
ac
e,
d
ef
in
ed
u
s
in
g
a
s
tu
d
en
t
’
s
t
-
d
is
tr
ib
u
tio
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:
2088
-
8
7
0
8
E
xp
lo
r
in
g
en
s
emb
le
lea
r
n
in
g
fo
r
cla
s
s
ifyin
g
g
eo
metric p
a
tter
n
s
:
in
s
ig
h
ts
fr
o
m
…
(
Zo
u
h
a
ir
Ou
a
z
en
e
)
4635
=
(
1
+
|
|
−
|
|
2
)
−
1
∑
(
1
+
|
|
−
|
|
2
)
−
1
≠
(
)
T
h
is
o
p
tim
izatio
n
r
esu
lts
in
a
v
is
u
ally
in
ter
p
r
etab
le
em
b
ed
d
i
n
g
,
wh
e
r
e
s
im
ilar
p
o
in
ts
in
th
e
h
ig
h
-
d
im
en
s
io
n
al
s
p
ac
e
ar
e
p
lace
d
clo
s
er
to
g
e
th
er
in
th
e
lo
w
-
d
im
e
n
s
io
n
al
s
p
ac
e.
Usi
n
g
t
-
SNE,
we
o
b
s
er
v
ed
two
d
is
tin
ct
clu
s
ter
s
co
r
r
esp
o
n
d
in
g
to
th
e
p
4
m
an
d
p
6
m
s
y
m
m
etr
ies,
v
alid
atin
g
th
e
ef
f
icac
y
o
f
o
u
r
f
ea
tu
r
e
ex
tr
ac
tio
n
p
r
o
ce
s
s
.
T
h
ese
v
is
u
aliza
tio
n
s
en
h
an
ce
th
e
i
n
ter
p
r
etab
ilit
y
o
f
th
e
d
ataset,
esp
ec
ially
wh
en
clea
r
g
r
o
u
p
in
g
s
em
er
g
e,
d
e
m
o
n
s
tr
atin
g
t
h
e
d
is
cr
im
in
ativ
e
p
o
wer
o
f
th
e
ex
tr
a
cted
f
ea
tu
r
es.
3
.
4
.
M
o
del t
ra
ini
ng
a
nd
ev
a
lua
t
io
n
3
.
4
.
1
.
Cla
s
s
if
ier
s
elec
t
io
n a
nd
t
ra
ini
ng
Ou
r
in
v
esti
g
atio
n
co
m
p
a
r
ed
t
h
r
ee
class
if
icatio
n
m
eth
o
d
s
R
F,
SVM,
an
d
a
v
o
tin
g
class
if
ier
—
ea
ch
co
n
tr
ib
u
tin
g
u
n
iq
u
e
s
tr
en
g
th
s
to
th
e
class
if
icatio
n
task
.
R
F
is
an
en
s
em
b
le
m
eth
o
d
th
at
co
n
s
tr
u
cts
m
u
ltip
le
d
ec
is
io
n
tr
ee
s
th
r
o
u
g
h
a
b
ag
g
in
g
tech
n
iq
u
e,
wh
e
r
e
th
e
f
i
n
al
p
r
ed
ictio
n
is
d
er
iv
ed
b
y
m
ajo
r
ity
v
o
tin
g
o
r
av
er
ag
in
g
.
T
h
is
ap
p
r
o
ac
h
is
h
ig
h
ly
r
esil
ien
t
to
n
o
is
e
an
d
d
e
m
o
n
s
tr
ates
f
lex
ib
ilit
y
ac
r
o
s
s
d
iv
er
s
e
f
ea
tu
r
e
s
ets.
T
o
o
p
tim
ize
its
p
er
f
o
r
m
an
ce
,
h
y
p
er
p
ar
am
eter
s
s
u
ch
as
th
e
n
u
m
b
er
o
f
tr
ee
s
(
n
esti
m
ato
r
s
)
an
d
th
e
m
ax
im
u
m
tr
ee
d
ep
th
(
m
a
x
_
d
e
p
th
)
wer
e
tu
n
ed
u
s
in
g
Gr
i
d
Sear
ch
C
V.
I
n
co
n
tr
ast,
th
e
SVM
class
if
ie
r
,
eq
u
ip
p
ed
with
a
r
ad
ial
b
asis
f
u
n
ctio
n
(
R
B
F)
k
er
n
el,
was
im
p
lem
en
ted
to
m
o
d
el
c
o
m
p
lex
,
n
o
n
lin
ea
r
d
ec
is
io
n
b
o
u
n
d
ar
ies.
Gr
id
Sear
ch
C
V
was
s
im
ilar
ly
ap
p
lied
to
s
elec
t
th
e
o
p
tim
a
l
co
s
t
p
ar
am
eter
(
C
)
an
d
k
er
n
el
co
ef
f
icien
t
(
γ
)
,
allo
win
g
th
e
SVM
to
e
x
ce
l
in
h
ig
h
-
d
im
e
n
s
io
n
al
s
p
ac
es
an
d
with
lim
ited
d
atasets
.
T
o
co
m
b
in
e
th
e
a
d
v
an
tag
es
o
f
th
ese
two
m
o
d
els,
a
v
o
tin
g
class
if
ier
was
co
n
s
tr
u
cted
,
in
teg
r
atin
g
R
F
an
d
SVM
p
r
e
d
ictio
n
s
th
r
o
u
g
h
a
weig
h
ted
en
s
em
b
le
s
tr
ateg
y
.
W
eig
h
ts
wer
e
as
s
ig
n
ed
to
ea
ch
m
o
d
el
b
ased
o
n
th
eir
v
ali
d
atio
n
p
er
f
o
r
m
a
n
ce
,
en
s
u
r
in
g
b
alan
ce
d
co
n
tr
ib
u
tio
n
s
to
th
e
f
in
al
d
ec
is
io
n
.
T
h
is
en
s
em
b
le
ap
p
r
o
ac
h
ef
f
ec
ti
v
ely
lev
er
ag
ed
th
e
v
ar
ian
ce
r
ed
u
ctio
n
ca
p
a
b
ilit
ies
o
f
R
F
an
d
th
e
m
ar
g
in
o
p
tim
izatio
n
s
tr
en
g
th
s
o
f
SVM,
r
esu
ltin
g
in
en
h
an
ce
d
g
en
er
aliza
tio
n
a
n
d
im
p
r
o
v
e
d
c
lass
if
icatio
n
p
er
f
o
r
m
an
ce
.
T
h
e
f
o
llo
win
g
p
s
eu
d
o
c
o
d
e
o
u
tlin
es
th
e
co
m
p
lete
wo
r
k
f
lo
w
o
f
th
e
class
if
ier
s
elec
tio
n
an
d
tr
ain
in
g
m
eth
o
d
o
l
o
g
y
:
Alg
o
r
ith
m
1
.
C
lass
if
ier
s
elec
t
i
o
n
an
d
tr
ain
in
g
wo
r
k
f
lo
w
1.
L
o
ad
d
ataset
o
f
g
eo
m
etr
ic
p
atter
n
s
(
p
4
m
an
d
p
6
m
s
y
m
m
etr
ie
s
)
.
2.
Pre
p
r
o
ce
s
s
ea
ch
im
ag
e:
a.
C
o
n
v
er
t to
g
r
ay
s
ca
le.
b
.
R
esize
to
2
5
6
×
2
5
6
.
c.
Ap
p
ly
m
e
d
ian
f
ilter
in
g
f
o
r
n
o
is
e
r
ed
u
ctio
n
.
d
.
Seg
m
en
t
u
s
in
g
Ots
u
'
s
th
r
es
h
o
ld
in
g
.
3.
C
o
m
p
u
te
f
ea
tu
r
es:
a.
C
alcu
late
QC
FrH
M
s
u
s
in
g
f
r
ac
tio
n
al
Hah
n
p
o
ly
n
o
m
ials
.
b
.
C
o
m
p
u
te
Z
e
r
n
ik
e
M
o
m
en
ts
.
c.
C
o
m
b
in
e
an
d
n
o
r
m
alize
f
ea
tu
r
es in
to
a
u
n
if
ied
v
ec
to
r
.
4.
Ap
p
ly
PC
A
f
o
r
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
(
r
etain
9
5
% v
a
r
ian
ce
)
.
5.
Sp
lit d
ataset
in
to
tr
ain
in
g
(
8
0
%)
an
d
test
(
2
0
%)
s
ets.
6.
T
r
ain
class
if
ier
s
:
a.
Op
tim
ize
R
F h
y
p
er
p
ar
a
m
eter
s
with
Gr
id
Sear
ch
C
V.
b
.
Op
tim
ize
SVM
h
y
p
er
p
ar
am
eter
s
with
Gr
id
Sear
ch
C
V.
c.
C
o
m
b
in
e
R
F a
n
d
SVM
p
r
ed
ictio
n
s
u
s
in
g
a
v
o
tin
g
class
if
ier
.
7.
E
v
alu
ate
p
er
f
o
r
m
a
n
ce
o
n
th
e
t
est s
et:
a.
C
alcu
late
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
Sco
r
e
.
b
.
Vis
u
alize
r
esu
lts
an
d
an
aly
z
e
m
is
class
if
icat
io
n
s
.
3
.
4
.
2
.
T
ra
ini
ng
a
nd
v
a
lid
a
t
io
n pro
t
o
co
ls
T
h
e
d
ataset,
in
th
e
c
o
u
r
s
e
o
f
en
s
u
r
in
g
r
eliab
le
m
o
d
el
tr
ain
i
n
g
an
d
v
alid
atio
n
,
was
d
iv
id
e
d
in
to
two
s
u
b
s
ets:
8
0
%
was
allo
ca
ted
f
o
r
tr
ain
in
g
an
d
h
y
p
e
r
p
ar
a
m
eter
o
p
tim
izatio
n
,
w
h
ile
th
e
r
e
m
ain
in
g
2
0
%
was
r
eser
v
ed
f
o
r
th
e
f
in
al
test
p
h
ase.
T
o
m
itig
ate
o
v
e
r
f
itti
n
g
an
d
en
h
an
ce
th
e
r
o
b
u
s
tn
ess
o
f
h
y
p
er
p
a
r
am
eter
tu
n
in
g
,
a
k
-
f
o
l
d
cr
o
s
s
-
v
alid
atio
n
s
tr
ateg
y
(
c
o
m
m
o
n
ly
=
5
)
was
im
p
lem
en
te
d
,
p
ar
titi
o
n
i
n
g
th
e
tr
ain
in
g
d
ata
in
to
f
iv
e
f
o
ld
s
an
d
iter
ativ
ely
u
s
in
g
f
o
u
r
f
o
ld
s
f
o
r
tr
ain
in
g
an
d
o
n
e
f
o
r
v
alid
atio
n
.
A
d
d
itio
n
ally
,
all
f
ea
tu
r
e
s
wer
e
s
tan
d
ar
d
ized
u
s
in
g
Stan
d
ar
d
Scaler
to
m
ain
tain
co
n
s
is
ten
cy
ac
r
o
s
s
th
e
d
ataset.
T
h
is
p
r
o
ce
s
s
in
v
o
lv
ed
ze
r
o
-
ce
n
ter
in
g
th
e
m
ea
n
a
n
d
s
ca
lin
g
ea
ch
f
ea
tu
r
e
to
u
n
it
v
ar
ian
ce
,
en
s
u
r
in
g
u
n
if
o
r
m
ity
in
f
ea
tu
r
e
m
ag
n
it
u
d
es
an
d
f
ac
ilit
atin
g
m
o
r
e
s
tab
le
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
6
3
0
-
4
6
4
1
4636
3
.
4
.
3
.
E
v
a
lua
t
io
n
m
e
t
rics
Ou
r
m
o
d
el
ev
alu
atio
n
was
b
ased
o
n
f
o
u
r
p
r
im
ar
y
m
etr
i
cs,
ea
ch
o
f
f
er
in
g
d
is
tin
ct
in
s
ig
h
ts
in
to
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
:
A
cc
u
r
ac
y
,
Pre
cisi
o
n
,
R
ec
all,
an
d
th
e
F1
-
Sco
r
e.
Acc
u
r
ac
y
,
a
g
en
er
al
m
ea
s
u
r
e
o
f
o
v
er
all
p
er
f
o
r
m
an
ce
,
ca
lcu
l
ates
th
e
f
r
ac
tio
n
o
f
co
r
r
e
c
tly
class
if
ied
s
am
p
les
o
v
e
r
th
e
t
o
tal
d
ataset.
Ma
th
em
atica
lly
,
it is
d
ef
in
ed
a
s
:
=
+
+
+
+
,
(
9
)
wh
er
e
T
P,
T
N,
FP
,
an
d
FN
r
ep
r
esen
t
tr
u
e
p
o
s
itiv
es,
tr
u
e
n
eg
ativ
es,
f
alse
p
o
s
itiv
es,
a
n
d
f
alse
n
eg
ativ
es,
r
esp
ec
tiv
ely
.
Pre
cisi
o
n
,
also
k
n
o
wn
as
p
o
s
itiv
e
p
r
ed
ictiv
e
v
alu
e,
m
ea
s
u
r
es
th
e
p
r
o
p
o
r
tio
n
o
f
p
r
ed
icted
p
o
s
itiv
es
th
at
ar
e
ac
tu
al
p
o
s
iti
v
es,
r
ef
lectin
g
th
e
m
o
d
el'
s
ab
ilit
y
to
av
o
id
f
alse
p
o
s
itiv
es.
R
ec
all,
o
r
s
en
s
itiv
ity
,
ev
alu
ates
th
e
f
r
ac
tio
n
o
f
ac
t
u
a
l
p
o
s
itiv
es
co
r
r
ec
tly
id
en
tifie
d
,
in
d
icatin
g
th
e
m
o
d
el’
s
ca
p
ab
ilit
y
to
d
etec
t
tr
u
e
p
o
s
itiv
es.
T
o
b
ala
n
ce
p
r
ec
is
i
o
n
an
d
r
ec
all,
esp
ec
ially
in
s
ce
n
ar
io
s
with
im
b
alan
ce
d
cla
s
s
d
is
tr
ib
u
tio
n
s
,
th
e
F1
-
Sco
r
e
is
u
s
ed
.
I
t is d
ef
in
e
d
as th
e
h
ar
m
o
n
ic
m
ea
n
o
f
Pre
ci
s
io
n
an
d
R
ec
all
:
1
=
2
.
.
+
,
(
1
0
)
T
h
is
co
m
b
in
atio
n
o
f
m
etr
ic
s
p
r
o
v
i
d
es
a
c
o
m
p
r
eh
en
s
iv
e
ev
alu
atio
n
f
r
am
ewo
r
k
,
allo
win
g
f
o
r
n
u
an
ce
d
in
ter
p
r
etatio
n
o
f
th
e
m
o
d
el'
s
s
t
r
en
g
th
s
an
d
wea
k
n
ess
es a
cr
o
s
s
d
if
f
er
en
t a
s
p
ec
ts
o
f
class
if
icatio
n
.
4.
RE
SU
L
T
S
4
.
1
.
F
ea
t
ure
a
na
ly
s
is
Af
ter
p
r
e
p
r
o
ce
s
s
in
g
,
th
e
d
at
aset
co
n
s
is
ted
o
f
1
,
2
0
4
im
a
g
es,
wh
ich
u
n
d
er
wen
t
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
u
s
in
g
PC
A.
Fro
m
th
e
in
itial
3
4
ex
tr
ac
ted
f
ea
tu
r
es,
1
0
p
r
in
cip
al
co
m
p
o
n
e
n
ts
wer
e
r
etain
ed
,
p
r
eser
v
in
g
ap
p
r
o
x
im
ately
9
5
%
o
f
t
h
e
to
tal
v
ar
ia
n
ce
.
As
s
h
o
wn
in
Fig
u
r
e
3
,
th
e
f
ir
s
t
f
ew
co
m
p
o
n
en
ts
ca
p
tu
r
e
th
e
m
ajo
r
ity
o
f
th
e
d
ataset’
s
v
ar
ian
ce
,
h
ig
h
lig
h
tin
g
th
e
ef
f
icac
y
o
f
PC
A
in
r
ed
u
cin
g
d
im
en
s
io
n
ality
wh
ile
m
ain
tain
in
g
cr
itical
in
f
o
r
m
ati
o
n
.
T
h
e
v
a
r
ian
ce
cu
r
v
e
clea
r
ly
in
d
icate
s
d
im
in
is
h
in
g
r
etu
r
n
s
b
ey
o
n
d
th
e
1
0
t
h
co
m
p
o
n
en
t,
j
u
s
tify
in
g
th
eir
s
elec
tio
n
f
o
r
f
u
r
th
e
r
a
n
aly
s
is
.
T
o
ev
al
u
ate
th
e
f
ea
tu
r
e
s
ep
ar
ab
ilit
y
,
we
a
p
p
lied
t
-
SNE
to
p
r
o
ject
th
e
d
ata
in
to
a
lo
wer
-
d
im
e
n
s
io
n
al
s
p
ac
e
f
o
r
v
is
u
aliza
tio
n
.
As
d
e
p
icted
in
Fig
u
r
e
4
,
th
e
t
-
SNE
p
lo
t
r
ev
ea
ls
d
is
tin
ct
clu
s
ter
in
g
o
f
th
e
two
s
y
m
m
etr
y
class
es
(
p
4
m
an
d
p
6
m
)
.
T
h
e
clu
s
ter
s
in
d
icate
th
at
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
tec
h
n
iq
u
es,
in
clu
d
i
n
g
q
u
ater
n
i
o
n
ca
r
tesi
an
f
r
ac
tio
n
al
Hah
n
m
o
m
en
t
s
(
QC
FrH
Ms)
an
d
Z
er
n
ik
e
m
o
m
e
n
ts
,
ef
f
ec
tiv
ely
ca
p
tu
r
ed
th
e
u
n
i
q
u
e
ch
ar
ac
te
r
is
tics
o
f
ea
ch
clas
s
.
T
h
e
clea
r
s
ep
ar
atio
n
in
th
e
t
-
SNE
p
lo
t
v
alid
ates
th
e
r
o
b
u
s
tn
ess
o
f
th
e
ex
tr
ac
ted
f
ea
t
u
r
es
an
d
th
eir
s
u
itab
ilit
y
f
o
r
class
if
icatio
n
task
s
.
T
h
e
PC
A
an
d
t
-
SNE
r
e
s
u
lts
co
l
lectiv
ely
d
em
o
n
s
tr
ate
th
at
th
e
d
im
en
s
io
n
ality
r
ed
u
ctio
n
an
d
v
is
u
aliza
tio
n
tech
n
iq
u
es
p
r
o
v
id
ed
m
ea
n
in
g
f
u
l
in
s
ig
h
ts
in
to
th
e
d
ataset’
s
s
tr
u
ctu
r
e.
T
h
ese
f
in
d
in
g
s
u
n
d
er
s
co
r
e
th
e
im
p
o
r
tan
ce
o
f
lev
er
a
g
in
g
ad
v
a
n
ce
d
f
ea
tu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
to
ac
h
iev
e
h
i
g
h
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
Fig
u
r
e
3
.
PC
A
v
ar
ian
ce
e
x
p
lai
n
ed
b
y
ea
ch
c
o
m
p
o
n
en
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
xp
lo
r
in
g
en
s
emb
le
lea
r
n
in
g
fo
r
cla
s
s
ifyin
g
g
eo
metric p
a
tter
n
s
:
in
s
ig
h
ts
fr
o
m
…
(
Zo
u
h
a
ir
Ou
a
z
en
e
)
4637
Fig
u
r
e
4
.
t
-
SNE
v
is
u
aliza
tio
n
o
f
th
e
f
ea
t
u
r
e
s
p
ac
e
f
o
r
p
4
m
a
n
d
p
6
m
class
es
4
.
2
.
Cla
s
s
if
ier
perf
o
rma
nce
T
o
ev
alu
ate
t
h
e
ef
f
ec
tiv
e
n
ess
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
s
,
th
r
ee
class
if
ier
s
R
F,
SVM,
an
d
t
h
e
v
o
tin
g
class
if
ier
,
wer
e
ap
p
lied
to
c
lass
if
y
th
e
p
4
m
a
n
d
p
6
m
s
y
m
m
etr
y
class
es.
T
h
e
r
esu
lts
d
em
o
n
s
tr
ate
th
e
r
o
b
u
s
tn
ess
o
f
th
ese
m
o
d
els
in
h
an
d
lin
g
th
e
d
ataset’
s
co
m
p
lex
ity
,
as
s
u
m
m
ar
ized
in
T
ab
le
1
.
T
a
b
le
1
s
h
o
wca
s
es
th
e
p
er
f
o
r
m
a
n
ce
m
etr
ics
p
er
tain
in
g
to
th
r
ee
class
if
ier
s
R
F,
SVM,
an
d
th
e
v
o
tin
g
class
if
ier
,
ap
p
lied
to
th
e
class
if
icat
io
n
o
f
p
4
m
a
n
d
p
6
m
s
y
m
m
etr
y
class
es.
T
h
e
m
etr
ics
in
clu
d
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
Sco
r
e
f
o
r
ea
ch
class
.
Am
o
n
g
t
h
e
m
o
d
els,
th
e
v
o
tin
g
cl
ass
if
ier
ac
h
iev
ed
th
e
h
ig
h
est
o
v
er
all
ac
cu
r
ac
y
o
f
8
2
.
2
%,
alo
n
g
with
b
ala
n
ce
d
m
etr
ics
ac
r
o
s
s
b
o
th
s
y
m
m
e
tr
y
class
es,
d
em
o
n
s
tr
atin
g
th
e
ef
f
ec
tiv
en
ess
o
f
en
s
em
b
le
lear
n
in
g
in
lev
er
a
g
i
n
g
th
e
s
tr
en
g
th
s
o
f
R
F
an
d
SVM.
Sp
ec
if
ically
,
th
e
v
o
tin
g
class
if
ier
ex
ce
lled
with
a
p
r
ec
is
io
n
o
f
0
.
8
3
an
d
r
ec
all
o
f
0
.
8
5
f
o
r
th
e
p
4
m
cla
s
s
,
wh
ile
m
ain
tain
in
g
s
tr
o
n
g
p
er
f
o
r
m
an
ce
in
th
e
p
6
m
class
with
a
p
r
ec
is
io
n
o
f
0
.
8
1
a
n
d
r
ec
all
o
f
0
.
7
8
.
T
h
e
R
F
class
if
ier
f
o
llo
wed
clo
s
ely
with
an
ac
c
u
r
ac
y
o
f
8
1
.
3
%,
s
h
o
wca
s
in
g
its
r
o
b
u
s
tn
ess
d
u
e
to
its
b
ag
g
in
g
-
b
ased
v
a
r
ian
ce
r
ed
u
c
tio
n
ca
p
ab
ilit
ies.
On
th
e
o
th
er
h
an
d
,
th
e
SVM
class
if
ier
,
alth
o
u
g
h
ac
h
iev
in
g
a
r
esp
ec
tab
le
ac
cu
r
ac
y
o
f
7
8
.
0
%,
h
ig
h
lig
h
te
d
th
e
co
m
p
lex
ity
o
f
t
h
e
d
ataset
a
n
d
t
h
e
s
u
b
tle
d
if
f
er
en
ce
s
b
etwe
en
th
e
p
4
m
a
n
d
p
6
m
s
y
m
m
etr
y
p
atter
n
s
.
Desp
ite
th
ese
d
if
f
er
en
ce
s
,
SVM
d
is
p
lay
ed
b
alan
ce
d
p
er
f
o
r
m
an
ce
with
a
p
r
ec
is
io
n
an
d
r
ec
all
o
f
0
.
8
0
f
o
r
th
e
p
4
m
class
an
d
0
.
7
6
p
r
ec
i
s
io
n
an
d
0
.
7
5
r
ec
all
f
o
r
th
e
p
6
m
class
.
T
h
e
co
n
f
u
s
io
n
m
atr
i
x
o
f
f
er
s
a
clea
r
v
is
u
aliza
tio
n
o
f
th
e
class
if
icat
io
n
o
u
tco
m
es,
s
h
o
win
g
th
e
d
is
tr
ib
u
tio
n
o
f
co
r
r
ec
t
an
d
in
co
r
r
ec
t
p
r
e
d
ictio
n
s
f
o
r
ea
ch
s
y
m
m
etr
y
class
.
T
h
is
lev
el
o
f
d
etail
h
elp
s
in
id
en
tify
in
g
s
tr
en
g
th
s
an
d
ar
ea
s
f
o
r
im
p
r
o
v
em
en
t
in
th
e
v
o
ti
n
g
class
if
ier
'
s
p
er
f
o
r
m
a
n
ce
.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
f
o
r
t
h
e
v
o
tin
g
class
if
ier
,
wh
ic
h
p
r
o
v
id
es
a
d
etailed
v
iew
o
f
th
e
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
Fo
r
th
e
p
4
m
class
,
1
1
4
tr
u
e
p
o
s
itiv
es
an
d
1
9
f
alse
n
eg
ativ
es
wer
e
r
ec
o
r
d
ed
,
wh
ile
f
o
r
th
e
p
6
m
class
,
8
6
tr
u
e
p
o
s
itiv
es
an
d
2
2
f
alse
n
eg
ativ
es
wer
e
o
b
s
er
v
ed
.
T
h
is
m
atr
ix
u
n
d
er
s
co
r
es
th
e
v
o
tin
g
class
if
ier
’
s
ab
ilit
y
to
ef
f
ec
tiv
ely
d
is
tin
g
u
is
h
b
etwe
en
th
e
two
s
y
m
m
etr
y
g
r
o
u
p
s
,
with
s
lig
h
tly
h
ig
h
er
ac
cu
r
ac
y
o
b
s
er
v
e
d
f
o
r
th
e
p
4
m
class
co
m
p
ar
ed
t
o
th
e
p
6
m
class
.
Fig
u
r
es
5
an
d
6
p
r
o
v
id
e
v
is
u
al
in
s
ig
h
ts
in
to
th
e
class
if
ic
atio
n
p
er
f
o
r
m
an
ce
.
Fig
u
r
e
5
p
r
esen
ts
th
e
co
n
f
u
s
io
n
m
atr
i
x
,
h
i
g
h
lig
h
tin
g
th
e
d
is
tr
ib
u
tio
n
o
f
tr
u
e
p
o
s
itiv
es,
f
alse
p
o
s
itiv
es,
an
d
m
is
class
if
icatio
n
s
.
Fig
u
r
e
6
co
m
p
ar
es
t
h
e
ac
c
u
r
ac
y
ac
r
o
s
s
all
th
r
ee
class
if
ier
s
,
s
h
o
wca
s
in
g
th
e
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
o
f
th
e
v
o
tin
g
c
lass
if
ier
.
T
h
ese
r
esu
lts
v
alid
ate
th
e
en
s
em
b
le
ap
p
r
o
ac
h
,
wh
ich
s
u
cc
ess
f
u
lly
co
m
b
in
es
R
F’s
v
ar
ian
ce
r
ed
u
ctio
n
with
SVM’
s
m
ar
g
in
-
b
as
ed
d
is
cr
im
in
atio
n
to
ac
h
iev
e
en
h
an
ce
d
ac
c
u
r
ac
y
an
d
s
tab
ilit
y
.
T
ab
le
1
.
Per
f
o
r
m
an
ce
m
etr
ics o
f
class
if
icatio
n
m
o
d
els
C
l
a
s
si
f
i
e
r
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
(
p
4
m)
R
e
c
a
l
l
(
p
4
m)
F1
-
S
c
o
r
e
(
p
4
m)
P
r
e
c
i
s
i
o
n
(
p
6
m)
R
e
c
a
l
l
(
p
6
m)
F1
-
S
c
o
r
e
(
p
6
m)
RF
0
.
8
1
3
0
.
8
1
0
.
8
6
0
.
8
4
0
.
8
2
0
.
7
5
0
.
8
1
S
V
M
0
.
7
8
0
.
8
0
.
8
0
.
8
0
.
7
6
0
.
7
5
0
.
7
5
V
o
t
i
n
g
c
l
a
s
si
f
i
e
r
0
.
8
2
2
0
.
8
3
0
.
8
5
0
.
8
4
0
.
8
1
0
.
7
8
0
.
8
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
6
3
0
-
4
6
4
1
4638
Fig
u
r
e
5
.
C
o
n
f
u
s
io
n
m
atr
i
x
f
o
r
th
e
v
o
tin
g
class
if
ier
Fig
u
r
e
6
.
Acc
u
r
ac
y
c
o
m
p
ar
is
o
n
ac
r
o
s
s
class
if
ier
s
(
R
F,
S
VM
,
an
d
v
o
tin
g
class
if
ier
)
4
.
3
.
M
is
cla
s
s
if
ica
t
io
n
a
na
ly
s
is
Up
o
n
clo
s
er
ex
am
in
atio
n
,
m
is
class
if
icatio
n
s
p
r
ed
o
m
in
a
n
tly
ar
o
s
e
u
n
d
er
th
r
ee
ch
allen
g
in
g
co
n
d
itio
n
s
.
First,
in
te
n
s
e
Gau
s
s
ian
n
o
is
e
o
f
ten
o
b
s
cu
r
e
d
cr
itical
g
e
o
m
etr
ic
f
ea
tu
r
es
,
co
m
p
licatin
g
th
e
id
en
tific
atio
n
o
f
s
y
m
m
etr
y
lin
es
an
d
p
atter
n
s
.
T
h
is
was
p
ar
ticu
lar
ly
ev
id
e
n
t
in
h
ea
v
ily
au
g
m
en
te
d
im
ag
es,
wh
er
e
n
o
is
e
in
tr
o
d
u
ce
d
ir
r
e
g
u
lar
ities
th
at
m
o
m
en
t
d
escr
ip
to
r
s
s
tr
u
g
g
led
to
p
r
o
ce
s
s
.
Seco
n
d
,
ex
tr
em
e
r
o
tatio
n
s
d
is
to
r
ted
th
e
o
r
ig
in
al
s
y
m
m
etr
y
o
f
ce
r
tain
p
4
m
an
d
p
6
m
m
o
tifs,
cr
ea
tin
g
am
b
ig
u
o
u
s
p
atte
r
n
s
th
at
b
lu
r
r
ed
th
e
d
is
tin
ctio
n
s
b
etwe
en
th
e
two
class
es.
Su
ch
d
is
to
r
tio
n
s
p
o
s
e
d
s
ig
n
if
ican
t
ch
allen
g
es
f
o
r
th
e
r
o
b
u
s
tn
ess
o
f
th
e
d
escr
ip
to
r
s
.
L
astl
y
,
lo
w
co
n
tr
ast
in
th
e
im
ag
es
led
to
in
co
m
p
lete
o
r
im
p
er
f
ec
t
s
eg
m
en
ta
tio
n
,
r
esu
ltin
g
in
th
e
lo
s
s
o
f
k
ey
s
tr
u
ctu
r
al
d
etails r
eq
u
ir
ed
f
o
r
ac
c
u
r
ate
f
ea
tu
r
e
ex
t
r
ac
tio
n
.
T
h
e
an
aly
s
is
d
ep
icted
in
Fig
u
r
e
7
p
r
o
v
id
es
in
s
ig
h
ts
in
to
t
h
e
er
r
o
r
d
is
tr
ib
u
tio
n
,
s
h
o
win
g
th
at
m
o
s
t
m
is
class
if
icatio
n
s
ar
is
e
u
n
d
er
co
n
d
itio
n
s
o
f
n
o
is
e,
ex
tr
em
e
r
o
tatio
n
s
,
o
r
lo
w
co
n
tr
ast,
f
u
r
th
er
v
alid
atin
g
th
e
n
ee
d
f
o
r
r
o
b
u
s
t
p
r
e
p
r
o
ce
s
s
in
g
.
B
y
r
ev
iewin
g
th
ese
o
u
tlier
s
,
we
ca
n
p
in
p
o
in
t
ar
ea
s
f
o
r
p
o
t
en
tial
im
p
r
o
v
em
e
n
t
in
b
o
th
p
r
e
p
r
o
ce
s
s
in
g
an
d
th
e
co
m
p
u
tatio
n
o
f
m
o
m
en
t
d
e
s
cr
ip
to
r
s
.
Ad
d
r
ess
in
g
th
ese
ch
allen
g
es
—
s
u
ch
as
en
h
an
cin
g
n
o
is
e
f
ilter
in
g
,
i
m
p
r
o
v
i
n
g
c
o
n
tr
ast
ad
j
u
s
tm
en
t
tech
n
iq
u
es,
o
r
r
ef
in
in
g
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
p
ip
elin
e
—
co
u
ld
f
u
r
th
er
b
o
ls
te
r
th
e
class
if
ier
'
s
p
er
f
o
r
m
an
ce
,
esp
ec
ially
u
n
d
er
a
d
v
er
s
e
co
n
d
itio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
E
xp
lo
r
in
g
en
s
emb
le
lea
r
n
in
g
fo
r
cla
s
s
ifyin
g
g
eo
metric p
a
tter
n
s
:
in
s
ig
h
ts
fr
o
m
…
(
Zo
u
h
a
ir
Ou
a
z
en
e
)
4639
Fig
u
r
e
7
.
Misclass
if
icatio
n
an
aly
s
is
illu
s
tr
atin
g
th
e
f
r
eq
u
en
c
y
o
f
er
r
o
r
s
in
class
if
y
in
g
p
4
m
an
d
p
6
m
s
y
m
m
etr
ies
5.
DIS
CU
SS
I
O
N
T
h
e
in
teg
r
atio
n
o
f
QC
FrH
Ms
p
r
o
v
id
ed
r
o
b
u
s
t
d
escr
ip
to
r
s
th
at
s
ig
n
if
ican
tly
im
p
r
o
v
ed
th
e
class
if
icatio
n
ac
cu
r
ac
y
o
f
g
eo
m
etr
ic
p
atter
n
s
with
en
h
a
n
ce
d
r
o
b
u
s
tn
ess
ag
ain
s
t
r
o
tatio
n
s
an
d
s
ca
lin
g
.
QC
FrH
M
s
ef
f
ec
tiv
ely
ca
p
tu
r
ed
f
in
e
-
g
r
ain
ed
s
h
ap
e
d
eta
ils
,
esp
ec
ially
wh
en
in
teg
r
ated
with
Z
er
n
ik
e
d
escr
ip
to
r
s
,
f
o
r
m
in
g
a
co
m
p
l
em
en
tar
y
,
d
u
al
-
la
y
er
f
ea
tu
r
e
r
ep
r
esen
tatio
n
t
h
at
en
ca
p
s
u
lated
b
o
th
g
lo
b
al
a
n
d
lo
ca
l
s
tr
u
ctu
r
al
s
y
m
m
etr
ies.
T
h
is
tu
r
n
ed
o
u
t
to
b
e
a
v
er
y
ef
f
ec
tiv
e
co
m
b
in
atio
n
an
d
g
av
e
a
r
em
ar
k
a
b
le
ac
cu
r
ac
y
f
o
r
th
e
d
if
f
er
en
tiatio
n
o
f
th
e
class
es
p
4
m
an
d
p
6
m
,
r
ea
ch
i
n
g
ev
en
8
2
.
2
%
with
th
e
v
o
tin
g
class
if
ier
.
T
h
e
en
s
em
b
le
m
o
d
el
ca
p
italized
o
n
th
e
s
tr
en
g
t
h
s
o
f
R
F
an
d
SVM,
wh
er
ein
R
F
p
r
o
v
id
ed
s
tab
ilit
y
b
y
v
ar
ian
ce
r
ed
u
ctio
n
th
r
o
u
g
h
b
ag
g
in
g
,
wh
er
ea
s
SVM
en
s
u
r
ed
r
o
b
u
s
t
m
ar
g
in
-
b
ased
class
if
icatio
n
i
n
h
ig
h
-
d
im
en
s
io
n
al
s
p
ac
es.
T
h
is
s
y
n
er
g
y
th
u
s
cr
e
ated
a
s
itu
atio
n
wh
er
e
th
e
we
ak
n
ess
es
o
f
in
d
iv
id
u
al
m
o
d
el
s
wer
e
o
f
f
s
et,
th
eir
s
tr
en
g
th
s
co
n
s
o
lid
ated
to
p
r
o
v
id
e
p
er
f
o
r
m
an
ce
u
n
if
o
r
m
ly
ac
r
o
s
s
d
iv
er
s
e
co
n
d
itio
n
s
.
Desp
ite
all
th
ese
ad
v
an
ce
s
,
s
o
m
e
o
f
th
e
m
et
h
o
d
o
lo
g
ical
lim
itatio
n
s
ap
p
ea
r
.
C
o
m
p
u
tin
g
h
i
g
h
er
-
o
r
d
e
r
m
o
m
en
ts
o
f
th
e
f
r
ac
tio
n
al
o
r
d
er
p
r
o
v
es
c
o
m
p
u
tatio
n
ally
e
x
p
en
s
iv
e
f
o
r
v
e
r
y
lar
g
e
d
atasets
,
an
d
th
o
u
g
h
m
ed
ian
f
ilter
in
g
was
p
er
f
o
r
m
e
d
,
p
ar
t
s
o
f
th
e
r
esid
u
al
n
o
is
e
m
ay
co
m
p
r
o
m
is
e
th
e
p
r
ec
is
io
n
o
f
th
e
m
o
m
en
t
-
b
ased
d
escr
ip
to
r
s
.
W
h
ile
Gau
s
s
ian
n
o
is
e,
r
o
tatio
n
,
an
d
s
ca
lin
g
en
h
an
ce
d
th
e
r
o
b
u
s
tn
ess
,
o
th
er
r
ea
l
-
w
o
r
ld
co
m
p
lex
ities
,
s
u
ch
as
v
ar
iab
l
e
lig
h
tin
g
o
r
o
cc
lu
s
io
n
s
,
m
ay
n
o
t
b
e
f
u
lly
ca
p
tu
r
ed
b
y
th
ese
au
g
m
en
tatio
n
s
.
C
o
m
p
ar
in
g
th
ese
r
esu
lts
with
th
e
ex
is
tin
g
liter
atu
r
e
wh
e
r
e
Z
er
n
ik
e
M
o
m
en
ts
h
av
e
b
ee
n
wid
ely
lau
d
ed
f
o
r
s
h
ap
e
class
if
icatio
n
,
o
u
r
ap
p
r
o
ac
h
r
ep
r
esen
ts
a
q
u
an
tu
m
le
ap
b
y
in
tr
o
d
u
cin
g
QC
FrH
Ms
th
at
o
f
f
er
en
h
an
ce
d
r
esil
ien
ce
an
d
f
lex
ib
ilit
y
.
W
h
at
is
m
o
r
e,
th
is
em
b
e
d
d
in
g
o
f
en
s
em
b
le
lear
n
i
n
g
is
o
n
tr
en
d
ac
c
o
r
d
in
g
t
o
m
o
d
er
n
te
n
d
en
cies
in
class
if
icatio
n
;
th
is
o
n
ce
ag
ain
u
n
d
er
lin
e
d
its
ca
p
ab
ilit
y
f
o
r
en
h
an
cin
g
th
e
ac
cu
r
ac
y
a
n
d
s
tab
ilit
y
o
v
er
s
in
g
le
m
o
d
els.
T
h
ese
r
esu
lts
p
r
o
v
ed
th
at
th
e
QC
FrH
M
s
ar
e
r
o
b
u
s
t
an
d
,
ev
en
m
o
r
e
im
p
o
r
ta
n
t,
th
e
en
s
em
b
le
s
tr
ateg
y
ef
f
ec
tiv
ely
wo
r
k
ed
t
o
ad
v
a
n
ce
th
e
s
tate
-
of
-
th
e
-
ar
t
in
t
h
e
g
eo
m
etr
ic
p
atter
n
class
if
icatio
n
,
o
p
en
in
g
a
v
er
y
f
ir
m
g
r
o
u
n
d
f
o
r
th
e
a
p
p
licatio
n
s
in
t
o
cu
ltu
r
al
h
e
r
itag
e
p
r
eser
v
atio
n
an
d
a
u
to
m
atio
n
o
f
d
esig
n
.
6.
CO
NCLU
SI
O
N
T
h
is
r
esear
ch
h
as
u
n
d
er
s
co
r
e
d
th
e
ef
f
ec
tiv
en
ess
o
f
co
m
b
i
n
in
g
en
s
em
b
le
lear
n
in
g
with
ad
v
an
ce
d
m
o
m
en
t
-
b
ased
d
escr
ip
to
r
s
p
ar
ticu
lar
ly
q
u
ater
n
io
n
ca
r
tesi
an
f
r
ac
tio
n
al
Hah
n
m
o
m
e
n
t
s
(
QC
FrH
M
s
)
an
d
Z
er
n
ik
e
m
o
m
en
ts
to
class
if
y
g
eo
m
etr
ic
m
o
tifs
ex
h
ib
itin
g
p
4
m
a
n
d
p
6
m
s
y
m
m
etr
ies.
B
y
in
teg
r
atin
g
b
o
t
h
d
escr
ip
to
r
s
ets,
cr
itical
g
eo
m
e
tr
ic
tr
aits
ar
e
m
o
r
e
co
m
p
r
eh
e
n
s
iv
ely
ca
p
tu
r
e
d
,
wh
ile
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
v
ia
PC
A
m
ain
tain
s
a
b
alan
ce
b
etwe
en
co
m
p
lex
ity
a
n
d
ac
cu
r
ac
y
b
y
r
etain
in
g
co
m
p
o
n
e
n
ts
co
v
er
in
g
9
5
% o
f
th
e
v
ar
ian
ce
.
Am
o
n
g
th
e
class
if
ier
s
test
ed
,
th
e
v
o
tin
g
class
if
ier
co
m
p
o
s
ed
o
f
a
R
F
an
d
an
SVM
o
u
tp
er
f
o
r
m
s
s
tan
d
alo
n
e
m
o
d
els,
h
ig
h
lig
h
ti
n
g
th
e
ad
v
an
tag
e
o
f
m
er
g
in
g
co
m
p
lem
en
tar
y
alg
o
r
ith
m
s
.
Alth
o
u
g
h
th
e
cu
r
r
en
t
r
esu
lts
ar
e
p
r
o
m
is
in
g
,
f
u
tu
r
e
wo
r
k
ca
n
f
o
cu
s
o
n
m
itig
a
tin
g
co
m
p
u
tatio
n
al
o
v
er
h
ea
d
,
en
h
a
n
cin
g
n
o
is
e
r
esil
ien
ce
,
an
d
ex
ten
d
in
g
th
e
f
r
am
ewo
r
k
to
ad
d
itio
n
al
s
y
m
m
etr
y
ty
p
es
o
r
r
ea
l
-
tim
e
ap
p
lica
tio
n
s
.
B
y
b
len
d
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.