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.
7
,
No
.
5
,
Octo
b
e
r
2
0
1
7
,
p
p
.
2
7
7
3
~2
781
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v7
i
5
.
pp
2
7
7
3
-
2
781
2773
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
JE
C
E
H
y
brid Me
thod
H
VS
-
M
RM
R
for
Va
ria
ble Selec
tion
i
n
M
ultilay
er Ar
t
ific
ia
l Neural Ne
tw
o
r
k
Cla
ss
ifier
B
en
-
H
dech
Adil
1
,
G
ha
no
u Y
o
us
s
ef
2
,
E
l
Q
a
di Ab
derr
a
hi
m
3
1,
2
T
IM
T
e
a
m
,
Hig
h
S
c
h
o
o
l
o
f
T
e
c
h
n
o
lo
g
y
,
M
o
u
lay
Is
m
a
il
Un
iv
e
rsit
y
,
M
e
k
n
e
s,
M
o
ro
c
c
o
3
LA
S
T
IM
I,
Hig
h
S
c
h
o
o
l
o
f
T
e
c
h
n
o
l
o
g
y
,
M
o
h
a
m
m
e
d
V
Un
iv
e
rsity
,
Ra
b
a
t,
M
o
ro
c
c
o
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Mar
29
,
2
0
1
7
R
ev
i
s
ed
Ma
y
5
,
2
0
1
7
A
cc
ep
ted
J
u
l
11
,
201
7
T
h
e
v
a
riab
le
se
lec
ti
o
n
is
a
n
im
p
o
rtan
t
tec
h
n
iq
u
e
th
e
re
d
u
c
i
n
g
d
im
e
n
sio
n
a
li
ty
o
f
d
a
ta
f
re
q
u
e
n
tl
y
u
se
d
in
d
a
ta
p
re
p
ro
c
e
ss
in
g
f
o
r
p
e
rf
o
rm
in
g
d
a
ta
m
in
in
g
.
T
h
is
p
a
p
e
r
p
re
se
n
ts
a
n
e
w
v
a
ri
a
b
le
se
lec
ti
o
n
a
lg
o
rit
h
m
u
se
s
th
e
h
e
u
risti
c
v
a
riab
le
se
lec
ti
o
n
(HV
S
)
a
n
d
M
i
n
im
u
m
Re
d
u
n
d
a
n
c
y
M
a
x
i
m
u
m
Re
lev
a
n
c
e
(M
RM
R).
W
e
e
n
h
a
n
c
e
th
e
H
V
S
m
e
th
o
d
f
o
r
v
a
riab
le
s
e
l
e
c
ti
o
n
b
y
in
c
o
rp
o
ra
ti
n
g
(M
RM
R)
f
il
ter.
Ou
r
a
lg
o
rit
h
m
is
b
a
se
d
o
n
w
ra
p
p
e
r
a
p
p
ro
a
c
h
u
sin
g
m
u
lt
i
-
la
y
e
r
p
e
rc
e
p
tro
n
.
W
e
c
a
ll
e
d
th
is
a
l
g
o
rit
h
m
a
H
V
S
-
M
R
M
R
W
ra
p
p
e
r
f
o
r
v
a
riab
les
se
l
e
c
ti
o
n
.
T
h
e
re
lev
a
n
c
e
o
f
a
se
t
o
f
v
a
riab
les
is
m
e
a
su
re
d
b
y
a
c
o
n
v
e
x
c
o
m
b
in
a
ti
o
n
o
f
th
e
re
lev
a
n
c
e
g
iv
e
n
b
y
HV
S
c
rit
e
rio
n
a
n
d
th
e
M
RM
R
c
rit
e
rio
n
.
T
h
is
a
p
p
r
o
a
c
h
se
lec
ts
n
e
w
re
le
v
a
n
t
v
a
r
iab
les
;
w
e
e
v
a
lu
a
te
th
e
p
e
r
f
o
r
m
a
n
c
e
o
f
HV
S
-
M
RM
R
o
n
e
ig
h
t
b
e
n
c
h
m
a
rk
c
l
a
s
sif
i
c
a
ti
o
n
p
ro
b
lem
s.
T
h
e
e
x
p
e
ri
m
e
n
tal
re
su
lt
s
sh
o
w
th
a
t
HV
S
-
M
RM
R
se
lec
ted
a
les
s
n
u
m
b
e
r
o
f
v
a
riab
les
w
it
h
h
ig
h
c
las
sif
ic
a
ti
o
n
a
c
c
u
ra
c
y
c
o
m
p
a
re
d
to
M
RM
R
a
n
d
HV
S
a
n
d
w
it
h
o
u
t
v
a
riab
les
se
lec
ti
o
n
o
n
m
o
st
d
a
tas
e
ts.
HV
S
-
M
RM
R
c
a
n
b
e
a
p
p
li
e
d
to
v
a
rio
u
s
c
las
sif
ica
t
i
o
n
p
ro
b
lem
s
th
a
t
re
q
u
ire
h
ig
h
c
las
sif
i
c
a
ti
o
n
a
c
c
u
ra
c
y
.
K
ey
w
o
r
d
s
:
Heu
r
is
tic
v
ar
iab
le
s
elec
tio
n
Neu
r
o
n
al
n
et
w
o
r
k
Min
i
m
u
m
r
ed
u
n
d
an
c
y
Ma
x
i
m
u
m
r
ele
v
an
ce
Mu
ltil
a
y
er
p
er
ce
p
tr
o
n
Var
iab
le
s
elec
tio
n
Co
p
y
rig
h
t
©
2
0
1
7
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
:
B
en
-
Hd
ec
h
A
d
il
,
Dep
ar
t
m
en
t o
f
E
lectr
ical
an
d
C
o
m
p
u
ter
E
n
g
in
ee
r
i
n
g
,
Natio
n
al
C
h
u
n
g
C
h
e
n
g
Un
i
v
er
s
it
y
,
1
6
8
Un
iv
er
s
it
y
R
o
ad
,
Min
h
s
i
u
n
g
T
o
w
n
s
h
ip
,
C
h
ia
y
i
C
o
u
n
t
y
6
2
1
0
2
,
T
aiw
a
n
,
R
O
C
.
E
m
ail: l
s
n
t
l@
cc
u
.
ed
u
.
t
w
1.
I
NT
RO
D
UCT
I
O
N
R
ed
u
ci
n
g
d
i
m
e
n
s
io
n
alit
y
o
f
d
ataset
h
as
b
ec
o
m
e
i
n
cr
ea
s
i
n
g
l
y
cr
itical
b
ec
a
u
s
e
o
f
th
e
m
u
lt
i
p
licatio
n
o
f
d
ata.
I
n
m
a
n
y
ar
ea
s
,
th
e
s
o
l
u
tio
n
o
f
a
s
y
s
te
m
p
r
o
b
le
m
i
s
b
ased
o
n
a
s
et
o
f
d
atab
ase
(
v
ar
iab
les)
[
1
-
2
]
.
I
n
cr
ea
s
in
g
t
h
e
n
u
m
b
er
o
f
t
h
es
e
v
ar
iab
les
t
h
at
c
h
ar
ac
ter
izes
t
h
e
p
r
o
b
le
m
r
ep
r
esen
t
s
d
if
f
ic
u
l
ties
at
m
a
n
y
le
v
els
s
u
c
h
as
co
m
p
lex
i
t
y
,
co
m
p
u
ti
n
g
ti
m
e,
an
d
d
eter
io
r
atio
n
o
f
th
e
s
y
s
te
m
p
r
o
b
le
m
s
o
lu
t
io
n
in
t
h
e
p
r
esen
ce
o
f
n
o
is
y
d
ata.
A
m
e
th
o
d
o
f
r
ed
u
cin
g
d
i
m
en
s
io
n
ali
t
y
is
to
f
in
d
a
r
ep
r
esen
tatio
n
o
f
t
h
e
o
r
ig
in
al
d
ata
i
n
a
s
m
aller
s
p
ac
e.
Di
m
e
n
s
io
n
alit
y
r
ed
u
cti
o
n
ca
n
r
o
u
g
h
l
y
b
e
d
iv
id
ed
i
n
to
t
w
o
ca
te
g
o
r
ies
[
3
-
4
]
:
f
ea
tu
r
e
ex
tr
ac
tio
n
a
n
d
f
ea
t
u
r
e
s
elec
t
io
n
.
Firstl
y
,
Feat
u
r
e
e
x
tr
ac
tio
n
g
e
n
er
ates
a
s
m
all
s
et
o
f
n
o
v
el
f
ea
t
u
r
es
b
y
m
er
g
in
g
th
e
o
r
ig
i
n
al
f
ea
t
u
r
es
.
Seco
n
d
l
y
,
Featu
r
e
s
e
l
ec
tio
n
p
ick
s
a
s
m
al
l set o
f
t
h
e
o
r
ig
in
al
o
n
e
s
.
Var
iab
les
s
elec
tio
n
o
r
f
ea
tu
r
e
s
s
elec
tio
n
is
a
s
ea
r
ch
p
r
o
ce
s
s
u
s
ed
to
s
elec
t
a
s
u
b
s
et
o
f
v
ar
iab
les
f
o
r
b
u
ild
in
g
r
o
b
u
s
t
lear
n
i
n
g
m
o
d
els
[
5
]
s
u
c
h
as
n
e
u
r
al
n
et
w
o
r
k
s
,
d
ec
is
io
n
tr
ee
s
a
n
d
o
th
er
s
.
So
m
e
ir
r
ele
v
an
t
an
d
/o
r
r
ed
u
n
d
an
t
v
ar
iab
les
ex
i
s
t
in
t
h
e
lear
n
i
n
g
d
ata
th
at
m
a
k
e
lear
n
i
n
g
h
ar
d
er
an
d
d
ec
r
ea
s
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
lear
n
i
n
g
m
o
d
els.
T
h
e
v
ar
i
ab
les
s
elec
tio
n
m
e
th
o
d
s
ca
n
b
e
class
i
f
ied
in
to
th
r
ee
m
ai
n
ca
teg
o
r
ies:
f
ilter
,
w
r
ap
p
er
an
d
e
m
b
ed
d
ed
.
Fil
ter
m
et
h
o
d
s
w
er
e
t
h
e
f
ir
s
t
u
s
ed
f
o
r
th
e
v
ar
iab
les
s
elec
tio
n
.
T
h
is
ca
te
g
o
r
y
allo
w
s
ev
alu
a
tin
g
t
h
e
r
elev
a
n
ce
o
f
a
v
ar
iab
le
ac
co
r
d
in
g
to
m
ea
s
u
r
e
s
th
at
r
el
y
o
n
th
e
p
r
o
p
er
ties
o
f
th
e
lear
n
i
n
g
d
ata.
Fil
ter
tec
h
n
iq
u
es
ar
e
f
a
s
t
f
o
r
h
ig
h
-
d
i
m
e
n
s
io
n
a
l d
ataset
s
,
b
u
t
t
h
e
y
i
g
n
o
r
e
i
n
ter
ac
tio
n
w
it
h
t
h
e
class
i
f
ier
[
6
]
.
T
h
e
w
r
ap
p
er
m
eth
o
d
s
u
s
e
t
h
e
p
r
ed
ictiv
e
ac
cu
r
ac
y
o
f
a
p
r
ed
eter
m
in
ed
lear
n
in
g
al
g
o
r
ith
m
to
d
eter
m
i
n
e
t
h
e
b
est
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
5
,
Octo
b
er
2
0
1
7
:
2
7
7
3
–
2
7
8
1
2774
s
u
b
s
et
s
elec
ted
.
T
h
e
ac
cu
r
ac
y
o
f
th
e
lear
n
i
n
g
al
g
o
r
it
h
m
s
is
u
s
u
all
y
h
i
g
h
[
7
]
.
W
r
ap
p
er
m
eth
o
d
s
t
en
d
to
f
i
n
d
th
e
m
o
s
t
s
u
itab
le
f
ea
t
u
r
e
s
u
b
s
et
f
o
r
th
e
lear
n
in
g
al
g
o
r
ith
m
,
b
u
t
th
e
y
ar
e
v
er
y
co
m
p
u
tatio
n
al
l
y
ex
p
e
n
s
i
v
e.
Un
l
ik
e
th
e
w
r
ap
p
er
an
d
f
ilter
m
et
h
o
d
s
,
e
m
b
ed
d
ed
m
eth
o
d
s
i
n
co
r
p
o
r
ate
th
e
s
elec
t
io
n
o
f
v
ar
iab
les
d
u
r
in
g
th
e
lear
n
i
n
g
p
r
o
ce
s
s
.
T
h
e
e
m
b
ed
d
ed
m
et
h
o
d
s
co
m
b
i
n
e
th
e
ad
v
a
n
ta
g
es
o
f
f
ilt
er
a
n
d
w
r
ap
p
er
tech
n
iq
u
es
[
8
]
.
T
h
e
f
ilter
ap
p
r
o
ac
h
d
eter
m
in
e
s
th
e
r
ele
v
an
t
a
n
d
r
ed
u
n
d
an
t
v
ar
iab
les
in
d
ep
en
d
en
t
o
f
t
h
e
class
i
f
ica
tio
n
,
s
u
c
h
as
u
s
i
n
g
o
n
l
y
M
R
M
R
cr
iter
io
n
,
s
o
it
i
s
n
o
t
r
ec
o
m
m
e
n
d
ed
to
u
s
e
it
alo
n
e
[
9
]
.
T
h
e
m
et
h
o
d
f
i
lter
m
i
g
h
t
i
m
p
r
o
v
e
th
e
s
elec
tio
n
o
f
v
ar
iab
les
if
it
u
n
d
er
s
tan
d
s
h
o
w
t
h
e
f
ilter
ed
v
a
r
iab
les
ar
e
u
s
ed
b
y
th
e
clas
s
i
f
ier
.
T
h
e
w
r
ap
p
er
m
et
h
o
d
ev
alu
ate
s
a
s
u
b
s
et
o
f
f
ea
t
u
r
es
b
y
its
clas
s
i
f
icatio
n
p
er
f
o
r
m
an
ce
u
s
i
n
g
a
lear
n
in
g
alg
o
r
ith
m
[
1
0
]
,
f
o
r
ex
a
m
p
le
,
h
eu
r
i
s
tic
v
ar
iab
le
s
e
lectio
n
(
HVS)
.
I
n
t
h
i
s
w
o
r
k
,
w
e
p
r
o
p
o
s
e
to
in
co
r
p
o
r
ate
M
R
MR
cr
iter
io
n
in
to
th
e
r
an
k
in
g
s
c
h
e
m
e
o
f
HV
S.
W
e
ar
e
m
ak
in
g
h
y
b
r
id
s
b
y
a
co
n
v
ex
co
m
b
in
atio
n
o
f
t
h
e
r
elev
an
c
y
g
iv
e
n
b
y
HVS
cr
iter
io
n
an
d
t
h
e
MRM
R
cr
iter
io
n
.
T
h
e
r
est
o
f
t
h
e
p
ap
er
is
o
r
g
a
n
ized
as
f
lo
w
s
:
Sec
tio
n
2
w
e
p
r
esen
t
r
elate
d
w
o
r
k
s
.
I
n
s
e
ctio
n
3
w
e
p
r
esen
t
r
esear
ch
m
eth
o
d
.
S
ec
tio
n
4
p
r
esen
ts
t
h
e
r
esu
l
ts
o
f
o
u
r
ex
p
er
i
m
e
n
tal
s
tu
d
ies
in
cl
u
d
in
g
th
e
ex
p
er
i
m
e
n
tal
m
eth
o
d
o
lo
g
y
,
e
x
p
er
i
m
e
n
tal
r
es
u
lts
,
a
n
d
th
e
co
m
p
ar
is
o
n
w
it
h
h
e
u
r
is
tic
v
ar
i
ab
les
s
elec
tio
n
HV
S
an
d
MRM
R
Min
i
m
u
m
R
ed
u
n
d
an
c
y
Ma
x
i
m
u
m
R
e
lev
a
n
ce
.
T
h
e
co
n
clu
s
io
n
s
ar
e
d
r
a
w
n
in
Sectio
n
5
.
2.
RE
L
AT
E
D
WO
RK
S
T
h
e
v
ar
iab
les
s
elec
tio
n
is
g
e
n
er
all
y
d
e
f
i
n
ed
as
a
s
ea
r
c
h
p
r
o
ce
s
s
to
f
in
d
a
s
u
b
s
et
o
f
"
r
elev
an
t"
ch
ar
ac
ter
is
tic
s
f
r
o
m
t
h
o
s
e
o
f
t
h
e
o
r
ig
i
n
al
s
et
[
5
]
,
[
1
1
-
1
3
]
.
T
h
e
co
n
ce
p
t
o
f
r
elev
a
n
ce
o
f
a
s
u
b
s
et
o
f
v
ar
iab
le
s
al
w
a
y
s
d
ep
en
d
s
o
n
t
h
e
o
b
j
ec
tiv
es
an
d
s
y
s
te
m
r
eq
u
ir
e
m
en
ts
.
T
h
e
p
r
o
b
le
m
o
f
v
ar
iab
les
s
elec
tio
n
f
o
r
class
i
f
icatio
n
ta
s
k
ca
n
b
e
d
escr
ib
ed
as
f
o
llo
w
s
:
g
i
v
en
t
h
e
o
r
ig
in
a
l
s
et
G,
o
f
N
f
ea
t
u
r
e
s
,
f
in
d
a
s
u
b
s
et
F
co
n
s
is
tin
g
o
f
N
’
r
elev
a
n
t
f
e
atu
r
es
w
h
er
e
N’
<
N
.
T
h
e
s
elec
tio
n
o
f
a
s
u
b
s
et
F
allo
w
s
m
a
x
i
m
izin
g
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
clas
s
i
f
icati
o
n
b
y
co
n
s
tr
u
cti
n
g
lear
n
i
n
g
m
o
d
els.
2
.
1
T
he
H
euri
s
t
ic
Va
ria
ble Sele
c
t
io
n
L
et
’
s
co
n
s
id
er
t
h
at
a
m
u
lt
ila
y
e
r
p
er
ce
p
tr
o
n
(
ML
P
)
[
1
4
]
is
n
eu
r
al
n
et
w
o
r
k
s
ar
ch
itect
u
r
e
d
e
f
in
ed
b
y
A
(
I
,
H,
0
)
w
h
er
e
I
i
n
p
u
t
la
y
er
,
H
h
id
d
en
la
y
er
s
an
d
O
o
u
tp
u
t
la
y
er
,
an
d
W
w
ei
g
h
t
m
atr
ix
.
T
h
e
v
alu
e
o
f
w
_
ij
co
n
n
ec
tio
n
b
et
w
ee
n
t
w
o
n
e
u
r
o
n
s
j
an
d
i
r
ef
lect
s
t
h
e
i
m
p
o
r
tan
ce
o
f
th
eir
r
ela
tio
n
s
h
ip
.
T
h
is
v
alu
e
ca
n
b
e
p
o
s
itiv
e
o
r
n
e
g
ati
v
e
d
ep
en
d
i
n
g
o
n
if
th
e
co
n
n
ec
tio
n
is
e
x
cit
ato
r
y
(
+)
o
r
i
n
h
ib
ito
r
y
(
-
)
.
Yac
o
u
b
et
al
p
r
o
p
o
s
ed
a
m
et
h
o
d
f
o
r
v
ar
iab
le
s
elec
tio
n
n
a
m
ed
h
eu
r
i
s
ti
c
v
ar
iab
le
s
elec
tio
n
HVS
[
1
5
]
.
T
h
e
HVS
cr
iter
io
n
is
in
ter
e
s
ted
i
n
th
e
s
tr
e
n
g
t
h
o
f
t
h
ese
co
n
n
ec
ti
o
n
s
.
T
h
is
s
tr
e
n
g
th
i
s
q
u
a
n
ti
f
ie
d
b
y
|
w
_
ij
|
.
T
h
e
p
ar
tial
co
n
tr
ib
u
tio
n
π
(
i,
j
)
o
f
th
e
h
id
d
en
n
e
u
r
o
n
j
o
n
th
e
o
u
tp
u
t
i
i
s
g
i
v
en
b
y
th
e
p
r
o
p
o
r
tio
n
o
f
al
l
t
h
e
co
n
n
ec
t
io
n
s
tr
en
g
t
h
a
r
r
iv
in
g
to
n
e
u
r
o
n
i
s
ee
Fig
u
r
e
1
.
Fig
u
r
e
1
.
T
h
e
p
ar
tial c
o
n
tr
ib
u
tio
n
o
f
th
e
n
e
u
r
o
n
j
co
n
n
ec
t
w
it
h
i
π
i
,
j
=
|
w
i
,
j
|
∑
|
w
i
,
j
|
Nc
j
(
1
)
Fo
r
esti
m
ate
t
h
e
r
elati
v
e
co
n
tr
ib
u
tio
n
o
f
u
n
it
j
is
f
i
n
al
d
e
cisi
o
n
o
f
t
h
e
s
y
s
te
m
.
T
h
e
u
n
it
j
s
en
d
s
co
n
n
ec
tio
n
s
to
,
a
s
et
o
f
u
n
i
ts
(
j
)
w
it
h
p
ar
tial c
o
n
tr
ib
u
tio
n
s
π
i
,
j
Fig
u
r
e
2
.
C
j
=
∑
π
ij
δ
i
Nj
i
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Hyb
r
id
Meth
o
d
HV
S
-
MRMR
fo
r
V
a
r
ia
b
le
S
elec
tio
n
in
Mu
ltil
a
ye
r
A
r
ti
ficia
l Neu
r
a
l …
(
B
en
-
Hd
ec
h
A
d
il)
2775
W
h
er
e
δ
i
=
{
1
if
unit
i
ϵ
O
C
i
if
unit
i
∈
H
(
3
)
Fig
u
r
e
2
.
T
h
e
r
elativ
e
co
n
tr
ib
u
tio
n
o
f
t
h
e
n
e
u
r
o
n
j
T
h
e
s
elec
tio
n
alg
o
r
it
h
m
b
ased
o
n
HVS
cr
iter
io
n
as
f
o
llo
w
s
:
̶
L
ea
r
n
i
n
g
th
e
M
L
P
u
n
til
w
e
r
e
ac
h
a
lo
ca
l
m
i
n
i
m
u
m
.
̶
C
alcu
late
th
e
r
ele
v
an
ce
o
f
ea
ch
v
ar
iab
le
ac
co
r
d
in
g
to
2
.
̶
So
r
t v
ar
iab
les in
asce
n
d
i
n
g
o
r
d
er
o
f
r
elev
an
ce
.
̶
R
e
m
o
v
e
th
e
lo
w
er
v
ar
iab
le
i
m
p
o
r
tan
ce
̶
B
ac
k
to
1
)
to
th
e
last
v
ar
iab
le
2
.
2
.
M
ini
m
u
m
Re
du
n
da
ncy
M
a
x
i
m
u
m
Re
lev
a
nce
T
h
e
MRMR
(
Min
i
m
u
m
R
ed
u
n
d
an
c
y
M
a
x
i
m
u
m
R
ele
v
a
n
ce
)
m
et
h
o
d
[
1
6
]
s
elec
ts
v
ar
iab
les
th
at
h
a
v
e
th
e
m
a
x
i
m
all
y
r
elev
a
n
ce
w
it
h
th
e
tar
g
et
clas
s
an
d
w
h
ich
ar
e
also
m
i
n
i
m
all
y
r
ed
u
n
d
an
t.
I
n
t
h
is
w
o
r
k
,
to
f
i
n
d
a
m
ax
i
m
a
ll
y
r
ele
v
an
t
a
n
d
m
in
i
m
all
y
r
ed
u
n
d
an
t
s
et
o
f
v
ar
ia
b
les,
w
e
u
s
e
m
u
tu
al
i
n
f
o
r
m
a
tio
n
b
ased
MRMR
cr
iter
io
n
.
T
h
e
ca
lcu
latio
n
o
f
r
ed
u
n
d
an
c
y
a
n
d
r
elev
an
ce
o
f
a
v
ar
iab
le
is
g
iv
en
b
y
eq
u
atio
n
s
(
4
)
an
d
(
5
)
.
T
h
e
I
(
i
,
Y
)
is
t
h
e
m
u
t
u
al
i
n
f
o
r
m
atio
n
b
et
w
ee
n
cla
s
s
lab
els
y
an
d
v
ar
iab
le
i
.
T
h
is
allo
w
s
u
s
to
q
u
an
t
if
y
t
h
e
r
elev
an
ce
o
f
v
ar
iab
le
i
to
th
e
cl
ass
i
f
icatio
n
.
T
h
e
r
elev
an
ce
o
f
v
ar
iab
le
is
g
i
v
e
n
b
y
:
Rl
i
=
1
|
S
|
2
∑
I
(
i
,
Y
)
y
(
4
)
W
h
er
e
(
,
)
=
∑
∑
(
,
)
l
og
(
(
,
)
(
)
(
)
∈
∈
)
(
5
)
T
h
e
r
ed
u
n
d
an
c
y
o
f
a
v
ar
iab
le
s
u
b
s
et
i
s
d
eter
m
in
ed
b
y
t
h
e
m
u
tu
al
i
n
f
o
r
m
atio
n
a
m
o
n
g
t
h
e
v
ar
iab
les.
T
h
e
r
ed
u
n
d
an
c
y
o
f
v
ar
iab
le
i
w
ith
t
h
e
o
th
er
v
ar
iab
les i
s
g
i
v
e
n
b
y
:
Rd
i
=
1
|
|
2
∑
(
,
)
∈
(
6
)
W
h
er
e
S
an
d
|
S
|
r
esp
ec
ti
v
el
y
d
e
n
o
te
th
e
s
e
t
o
f
v
ar
iab
le
s
an
d
its
s
ize
an
d
I
(
i
,
j
)
is
th
e
m
u
tu
a
l
in
f
o
r
m
atio
n
b
et
w
ee
n
i
an
d
j
.
T
h
e
s
co
r
e
o
f
a
v
ar
iab
le
is
th
e
co
m
b
in
at
io
n
o
f
t
h
ese
t
w
o
f
ac
to
r
s
:
Sc
i
=
Rl
i
Rd
i
(
7
)
T
h
e
m
ea
s
u
r
es
o
f
r
elev
a
n
ce
an
d
r
ed
u
n
d
an
c
y
o
f
v
ar
iab
les
ca
n
b
e
f
o
r
m
ed
in
s
e
v
er
al
w
a
y
s
,
b
u
t
th
e
q
u
o
tien
t
o
f
th
e
r
elev
a
n
ce
b
y
r
ed
u
n
d
an
c
y
s
elec
t
h
i
g
h
l
y
r
ele
v
an
t
v
ar
iab
les
w
it
h
le
s
s
r
ed
u
n
d
an
c
y
[
1
7
]
.
Af
t
er
t
h
i
s
in
d
iv
id
u
al
v
ar
iab
le
e
v
alu
at
io
n
,
a
s
eq
u
en
tial
s
ea
r
ch
tec
h
n
iq
u
e
is
u
s
ed
w
it
h
a
class
i
f
ier
to
s
el
ec
t
th
e
f
i
n
al
s
u
b
s
et
o
f
v
ar
iab
les.
A
clas
s
i
f
ier
is
u
s
ed
to
ev
alu
ate
th
e
s
u
b
s
ets
s
tar
tin
g
w
it
h
t
h
e
v
ar
iab
le
th
a
t
h
a
s
th
e
b
est
s
co
r
e,
th
e
b
est t
w
o
,
u
n
ti
l
w
e
f
i
n
d
th
e
s
u
b
s
et
th
at
m
i
n
i
m
izes t
h
e
clas
s
i
f
ic
atio
n
er
r
o
r
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
5
,
Octo
b
er
2
0
1
7
:
2
7
7
3
–
2
7
8
1
2776
3.
RE
S
E
ARCH
M
E
T
H
O
D
Usi
n
g
t
h
e
f
il
ter
m
et
h
o
d
s
alo
n
e
f
o
r
ex
a
m
p
le
MRM
R
,
m
a
y
n
o
t
g
iv
e
t
h
e
b
est
p
er
f
o
r
m
an
ce
b
ec
au
s
e
it
o
p
er
ates in
d
ep
en
d
en
tl
y
t
h
e
cla
s
s
i
f
ier
an
d
i
s
n
o
t i
n
v
o
lv
ed
i
n
t
h
e
s
elec
t
io
n
o
f
v
ar
iab
les
.
O
n
t
h
e
o
th
er
h
an
d
,
HVS
d
o
es
n
o
t
tak
e
i
n
to
ac
co
u
n
t
t
h
e
r
ed
u
n
d
a
n
c
y
a
m
o
n
g
v
ar
iab
les.
Ou
r
o
b
j
ec
tiv
e
is
to
i
m
p
r
o
v
e
th
e
v
ar
iab
les
s
elec
tio
n
HVS
b
y
i
n
tr
o
d
u
ci
n
g
a
n
M
R
MR
f
i
lter
to
m
i
n
i
m
i
ze
th
e
r
ed
u
n
d
an
c
y
a
m
o
n
g
r
el
ev
an
t
v
ar
iab
le
s
.
A
s
s
ee
n
later
,
th
i
s
i
m
p
r
o
v
es
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
class
if
ier
b
y
co
m
p
r
o
m
is
i
n
g
r
elev
a
n
c
y
a
n
d
r
ed
u
n
d
an
c
y
o
f
v
ar
iab
les.
I
n
o
u
r
ap
p
r
o
ac
h
o
f
HVS
-
M
R
MR
v
ar
iab
les
s
elec
tio
n
,
t
h
e
v
ar
iab
les
ar
e
s
elec
ted
b
y
a
co
n
v
e
x
co
m
b
i
n
atio
n
o
f
th
e
r
elev
a
n
c
y
g
iv
e
n
b
y
HVS
co
n
tr
ib
u
tio
n
s
an
d
th
e
MRM
R
cr
iter
io
n
.
Fo
r
i
th
e
v
ar
iab
le,
th
e
r
an
k
i
n
g
Me
as
u
r
e
R
_
i is
g
iv
e
n
b
y
R
i
=
α
|
C
i
|
+
(
1
−
α
)
Sc
i
(
8
)
W
h
er
e
th
e
p
ar
a
m
eter
α
∈
[
0
,
1
]
d
eter
m
i
n
es t
h
e
co
m
p
r
o
m
i
s
e
b
et
w
ee
n
HVS
an
d
MR
MR c
r
iter
io
n
,
T
h
e
s
ea
r
ch
s
tr
ateg
y
i
s
o
n
e
o
f
th
e
p
r
o
p
er
ties
o
f
t
h
e
v
ar
iab
l
e
s
elec
tio
n
al
g
o
r
ith
m
s
.
T
h
er
e
ar
e
th
r
ee
s
tr
ateg
ie
s
,
f
o
r
w
ar
d
s
elec
tio
n
,
b
ac
k
w
ar
d
eli
m
i
n
atio
n
an
d
s
te
p
w
is
e
s
elec
tio
n
.
I
n
f
o
r
w
ar
d
s
e
lectio
n
,
v
ar
iab
les
ar
e
p
r
o
g
r
ess
iv
el
y
in
co
r
p
o
r
ated
in
t
o
lar
g
er
an
d
lar
g
er
s
u
b
s
et
s
.
I
n
b
ac
k
w
ar
d
eli
m
i
n
atio
n
o
n
e
s
t
ar
ts
w
it
h
t
h
e
s
et
o
f
all
v
ar
iab
les
an
d
p
r
o
g
r
ess
iv
el
y
eli
m
i
n
ate
s
th
e
least
p
r
o
m
is
i
n
g
o
n
e
s
[
3
]
.
I
n
o
u
r
alg
o
r
ith
m
w
e
u
s
e
th
e
s
tr
ate
g
y
b
ac
k
w
ar
d
s
eli
m
i
n
atio
n
.
T
o
b
etter
co
m
p
r
o
m
i
s
e
w
i
th
r
ed
u
n
d
a
n
c
y
an
d
R
elev
a
n
c
y
o
f
v
ar
iab
les,
w
e
u
s
e
S
(
i
)
MRMR
cr
iter
io
n
f
o
r
r
an
k
i
n
g
.
A
ls
o
,
w
e
u
s
e
|
C
i
|
t
h
e
cr
iter
i
o
n
o
f
HVS
as
t
h
e
m
ea
s
u
r
e
o
f
r
elev
an
ce
v
ar
iab
les.
A
l
g
o
r
ith
m
1
ill
u
s
tr
ates
HV
S
-
MRMR
v
ar
iab
les
s
elec
tio
n
m
eth
o
d
.
I
n
ea
ch
iter
atio
n
,
w
e
id
en
ti
f
ied
th
e
least
i
m
p
o
r
ta
n
t
v
ar
iab
le
a
f
ter
r
an
k
i
n
g
th
e
v
ar
iab
les
i
n
t
h
e
s
e
t
G
.
T
h
e
v
ar
iab
le
least
s
i
g
n
if
ica
n
t
ar
e
r
etir
ed
an
d
th
e
r
e
m
ain
in
g
s
u
b
s
et
G
w
il
l
g
o
th
r
o
u
g
h
p
r
o
ce
s
s
iter
ati
v
e,
ea
ch
ti
m
e
w
e
r
e
m
o
v
e
a
v
ar
i
ab
le,
r
elea
r
n
in
g
is
r
eq
u
ir
ed
.
T
h
e
alg
o
r
ith
m
r
e
m
o
v
es t
h
e
v
ar
iab
les o
n
e
b
y
o
n
e
u
n
til t
h
e
last
v
ar
iab
le.
L
et
’
s
co
n
s
id
er
th
at
S
p
is
a
s
u
b
s
et
o
f
v
ar
iab
les an
d
p
is
th
e
n
u
m
b
e
r
o
f
th
ese
v
ar
iab
les.
Alg
o
rit
h
m
1
:
H
VS
-
M
R
M
R
f
o
r
v
a
ri
a
bles
s
elec
t
io
n
B
eg
in
Set
α
G
iv
en
s
et
o
f
v
a
ria
ble,
S
⊂
G
R
ep
ea
t :
T
ra
in t
he
M
L
P
w
it
h t
est
d
a
t
a
s
et
;
F
o
r
e
a
ch
i
∈
do
Co
m
p
ute
t
he
b
y
eq
u
atio
n
(
2
)
Co
m
p
ute
t
he
b
y
eq
u
atio
n
(
7
)
Co
m
p
ute
t
he
b
y
eq
u
atio
n
(
8
)
E
n
d
f
o
r
Select
t
he
v
a
ria
ble
∗
=
{
}
Upda
t
e
−
=
\
{
∗
}
Unt
il
G
=
{
∅
}
E
n
d
Fo
r
s
elec
tin
g
a
s
u
b
s
et
S
p
o
f
v
ar
i
ab
les,
alg
o
r
ith
m
.
1
g
en
er
ate
s
a
s
et
o
f
N
n
eu
r
o
n
al
n
et
w
o
r
k
s
(
N
is
t
h
e
n
u
m
b
er
o
f
v
ar
iab
les)
h
a
v
i
n
g
less
an
d
les
s
v
ar
iab
les.
T
h
e
ch
o
ice
o
f
s
u
b
s
et
in
cl
u
d
es
u
s
i
n
g
a
s
ta
tis
tica
l
tes
t
(
Fis
h
er
tes
t)
an
d
s
ea
r
ch
es
a
m
o
n
g
all
n
et
w
o
r
k
s
MLP
(
p
)
th
o
s
e
th
at
ar
e
s
tat
is
t
icall
y
n
ea
r
MLP
(
p
∗
)
.
T
h
is
p
r
in
cip
le
p
r
o
v
id
es
a
s
et
o
f
n
eu
r
o
n
al
n
et
w
o
r
k
s
s
u
c
h
as
E
(
p
i
)
≈
E
(
p
∗
)
.
T
h
e
s
u
b
s
et
o
f
s
elec
ted
v
ar
iab
le
s
is
t
h
e
s
m
al
lest
s
u
b
s
et
o
f
p
0
s
tati
s
ticall
y
n
ea
r
MLP
(
p
∗
)
v
ar
iab
les.
W
h
er
e
E
(
p
i
)
is
th
e
er
r
o
r
o
f
ML
P
f
o
r
s
u
b
s
e
t
S
p
.
W
h
er
e,
E
(
p
)
is
th
e
er
r
o
r
o
f
ML
P
f
o
r
s
u
b
s
et
S
p
,
MLP
(
p
∗
)
=
a
gr
min
M
L
P
(
p
)
E
(
p
)
(
9
)
p
0
=
min
i
{
i
}
(
1
0
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Hyb
r
id
Meth
o
d
HV
S
-
MRMR
fo
r
V
a
r
ia
b
le
S
elec
tio
n
in
Mu
ltil
a
ye
r
A
r
ti
ficia
l Neu
r
a
l …
(
B
en
-
Hd
ec
h
A
d
il)
2777
4.
RE
SU
L
T
S AN
D
AN
AL
Y
SI
S
T
o
ev
alu
ate
th
e
p
er
f
o
r
m
a
n
c
e
o
f
t
h
e
HVS
-
M
R
M
R
,
w
e
u
s
e
s
o
m
e
d
ata
s
et
s
f
o
r
r
ea
l
w
o
r
ld
class
i
f
icatio
n
.
T
ab
le
1
s
h
o
w
s
t
h
e
d
etailed
in
f
o
r
m
a
tio
n
o
f
ea
ch
d
ataset.
T
h
e
y
w
er
e
p
ar
titi
o
n
ed
in
to
th
r
ee
s
e
ts
:
a
tr
ain
i
n
g
s
et,
a
v
alid
atio
n
s
et
a
n
d
a
te
s
ti
n
g
s
et.
T
h
e
tr
ain
in
g
an
d
test
in
g
s
ets
w
er
e
u
s
ed
to
tr
ain
M
L
P
an
d
to
ev
alu
a
te
t
h
e
cla
s
s
i
f
icat
io
n
ac
cu
r
ac
y
o
f
tr
ain
ed
M
L
P
,
r
esp
ec
tiv
el
y
.
T
h
e
v
alid
atio
n
s
et
i
s
u
s
ed
to
e
s
ti
m
ate
p
r
ed
ictio
n
er
r
o
r
f
o
r
ML
P
.
T
ab
le
1
Descr
ip
tio
n
s
an
d
P
ar
titi
o
n
s
f
o
r
Dif
f
er
en
t
C
las
s
i
f
icati
o
n
s
Data
s
e
ts
D
a
t
e
se
t
N
u
mb
e
r
o
f
V
a
r
i
a
b
l
e
s
T
r
a
i
n
i
n
g
e
x
a
mp
l
e
s
V
a
l
i
d
a
t
i
o
n
e
x
a
mp
l
e
s
T
e
st
i
n
g
e
x
a
mp
l
e
s
C
l
a
sse
s
D
i
a
b
e
t
e
s
8
3
8
4
1
9
2
1
9
2
2
C
a
n
c
e
r
9
3
4
9
1
7
5
1
7
5
2
G
l
a
ss
9
1
0
8
53
53
6
Ve
h
i
c
l
e
18
2
4
2
2
1
1
2
1
1
4
H
e
p
a
t
i
t
i
s
19
77
39
39
2
Wa
v
e
f
o
rm
21
2
5
0
0
1
2
5
0
1
2
5
0
3
H
o
rse
21
1
7
2
86
86
2
I
o
n
o
s
p
h
e
re
34
1
7
5
88
88
2
4
.
1
P
re
pro
ce
s
s
ing
W
e
p
r
ep
r
o
ce
s
s
ed
th
e
d
ataset
s
b
y
r
escali
n
g
i
n
p
u
t
v
ar
iab
l
es
v
al
u
es
b
et
w
ee
n
0
a
n
d
1
th
e
li
n
ea
r
n
o
r
m
aliza
t
io
n
f
u
n
ctio
n
:
Af
ter
n
o
r
m
aliza
tio
n
,
all
f
ea
t
u
r
es
o
f
th
ese
ex
a
m
p
les
ar
e
b
et
w
ee
n
ze
r
o
an
d
o
n
e.
T
h
e
s
tan
d
ar
d
izatio
n
f
o
r
m
u
la
is
as
f
o
llo
w
s
:
x
i
,
n
ew
=
x
i
,
o
l
d
−
x
i
,
m
in
x
i
,
m
ax
−
x
i
,
m
in
(
1
1
)
W
h
er
e
x
i
,
n
ew
an
d
x
i
,
o
l
d
ar
e
th
e
n
e
w
an
d
o
ld
v
alu
e
o
f
attr
ib
u
te,
r
esp
ec
tiv
el
y
.
The
x
i
,
m
ax
an
d
x
i
,
m
i
n
ar
e
th
e
m
a
x
i
m
u
m
a
n
d
m
i
n
i
m
u
m
v
al
u
e,
r
esp
ec
ti
v
el
y
o
f
v
ar
ia
b
le
i
.
T
ab
le
2
.
P
er
f
o
r
m
a
n
ce
o
f
HVS
-
MR
MR
f
o
r
d
if
f
er
e
n
t c
las
s
i
f
ic
atio
n
s
d
ataset
s
,
St.
d
ev
.
g
i
v
e
n
af
ter
th
e
±
s
i
g
n
D
a
t
a
se
t
M
e
a
su
r
e
W
i
t
h
o
u
t
V
S
HVS
M
R
M
R
HVS
-
M
R
M
R
D
i
a
b
e
t
e
s
M
e
a
n
N
o
.
o
f
v
a
r
i
a
b
l
e
8
.
0
0
±
0
.
0
0
5
.
9
0
±
0
.
9
2
6
.
5
0
±
0
.
6
3
5
.
5
5
±
1
.
0
2
A
c
c
%
7
5
.
6
8
±
1
.
0
1
7
6
.
3
2
±
2
.
0
6
7
6
.
2
5
±
2
.
0
1
7
6
.
6
3
±
3
.
2
3
C
a
n
c
e
r
M
e
a
n
N
o
.
o
f
v
a
r
i
a
b
l
e
9
.
0
0
±
0
.
0
0
6
.
8
3
±
1
.
0
1
5
.
6
0
±
0
.
9
2
6
.
1
3
±
0
.
9
8
A
c
c
%
9
7
.
9
6
±
0
.
8
8
9
8
.
4
3
±
1
.
0
6
9
8
.
1
7
±
1
.
5
2
9
8
.
6
8
±
1
.
8
8
G
l
a
ss
M
e
a
n
N
o
.
o
f
v
a
r
i
a
b
l
e
9
.
0
0
±
0
.
0
0
5
.
1
3
±
0
.
5
5
4
.
9
0
±
0
.
4
5
4
.
3
3
±
0
.
6
5
A
c
c
%
7
4
.
2
1
±
5
.
6
2
7
6
.
6
1
±
4
.
6
6
7
6
.
7
3
±
5
.
2
0
7
7
.
0
3
±
4
.
9
5
V
e
h
i
c
l
e
M
e
a
n
N
o
.
o
f
v
a
r
i
a
b
l
e
1
8
.
0
0
±
0
.
0
0
4
.
5
1
±
0
.
6
7
5
.
3
3
±
0
.
6
4
4
.
9
0
±
0
.
7
6
A
c
c
%
7
3
.
3
7
±
4
.
8
7
7
4
.
3
5
±
3
.
5
7
7
4
.
2
1
±
4
.
0
3
7
5
.
1
7
±
3
.
4
7
H
e
p
a
t
i
t
i
s
M
e
a
n
N
o
.
o
f
v
a
r
i
a
b
l
e
1
9
.
0
0
±
0
.
0
0
3
.
7
3
±
0
.
8
7
5
.
1
0
±
0
.
7
6
3
.
5
5
±
0
.
9
2
A
c
c
%
7
0
.
6
3
±
3
.
6
0
7
7
.
6
5
±
2
.
7
9
7
6
.
3
2
±
3
.
4
3
7
8
.
6
7
±
3
.
1
5
W
a
v
e
f
o
r
m
M
e
a
n
N
o
.
o
f
v
a
r
i
a
b
l
e
2
1
.
0
0
±
0
.
0
0
4
.
8
5
±
0
.
9
6
5
.
5
0
±
0
.
8
1
4
.
8
5
±
1
.
0
4
A
c
c
%
8
5
.
3
0
±
1
.
2
3
8
5
.
9
1
±
2
.
5
4
8
5
.
2
7
±
2
.
9
8
8
4
.
9
1
±
3
.
1
3
H
o
r
se
M
e
a
n
N
o
.
o
f
v
a
r
i
a
b
l
e
2
1
.
0
0
±
0
.
0
0
7
.
9
4
±
1
.
8
7
6
.
9
3
±
1
.
6
5
6
.
5
5
±
1
.
4
5
A
c
c
%
8
4
.
5
1
±
2
.
5
2
8
6
.
0
3
±
2
.
1
0
8
5
.
2
7
±
2
.
0
5
8
6
.
2
1
±
1
.
9
5
I
o
n
o
sp
h
e
r
e
M
e
a
n
N
o
.
o
f
v
a
r
i
a
b
l
e
3
4
.
0
0
±
0
.
0
0
6
.
5
2
±
2
.
0
3
7
.
1
1
±
1
.
8
7
6
.
8
7
±
1
.
9
7
A
c
c
%
9
4
.
6
6
±
2
.
0
3
9
5
.
8
1
±
1
.
9
5
9
6
.
2
7
±
2
.
2
5
9
6
.
3
1
±
2
.
9
8
W
h
er
e
Me
an
No
.
o
f
v
ar
iab
le
: th
e
a
v
e
r
ag
e
n
u
m
b
er
s
o
f
v
ar
iab
les s
e
le
cted
A
cc
: T
h
e
clas
s
if
icatio
n
ac
c
u
r
ac
y
W
ith
o
u
t V
S: W
ith
o
u
t
v
ar
iab
le
s
elec
tio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
5
,
Octo
b
er
2
0
1
7
:
2
7
7
3
–
2
7
8
1
2778
4
.
2
P
a
ra
m
et
er
E
s
t
i
m
a
t
io
n
I
n
th
is
p
ar
ag
r
ap
h
,
w
e
h
av
e
s
p
ec
if
ied
s
o
m
e
p
ar
a
m
eter
s
o
f
HVS
-
M
R
MR.
T
h
ese
ar
e
d
e
s
cr
ib
ed
as
f
o
llo
w
s
:
T
h
e
i
n
itia
l
w
ei
g
h
ts
f
o
r
an
M
L
P
w
er
e
r
a
n
d
o
m
l
y
c
h
o
s
en
i
n
t
h
e
v
al
u
e
b
et
w
ee
n
-
1
an
d
1
.
T
h
e
tr
ain
in
g
er
r
o
r
th
r
esh
o
ld
v
al
u
e
f
o
r
ca
n
c
er
,
d
iab
etes,
g
la
s
s
,
h
ep
atiti
s
,
h
o
r
s
e,
io
n
o
s
p
h
er
e,
v
eh
icle,
a
n
d
w
av
e
f
o
r
m
d
ataset
s
w
a
s
s
et
to
0
.
0
0
2
,
0
.
0
0
3
,
0
.
0
4
,
0
.
0
2
,
0
.
0
0
3
,
0
.
0
2
,
0
.
0
1
an
d
0
.
0
3
r
esp
ec
tiv
el
y
.
A
l
s
o
,
th
e
v
alid
atio
n
er
r
o
r
th
r
es
h
o
ld
v
alu
e
w
as
s
et
to
0
.
0
0
1
,
0
.
0
0
2
,
0
.
0
2
5
,
0
.
0
1
8
,
0
.
0
0
1
,
0
.
0
1
4
,
0
.
0
0
7
an
d
0
.
0
2
5
f
o
r
ca
n
ce
r
,
d
iab
etes,
g
las
s
,
h
ep
atit
is
,
h
o
r
s
e
,
io
n
o
s
p
h
er
e,
v
e
h
icle,
a
n
d
w
av
e
f
o
r
m
d
atasets
,
r
esp
ec
ti
v
el
y
.
T
h
e
lear
n
in
g
r
ate
a
n
d
i
n
itial
w
ei
g
h
t
v
alu
e
s
ar
e
t
h
e
p
ar
a
m
eter
s
o
f
t
h
e
w
ell
k
n
o
w
n
b
a
ck
-
p
r
o
p
ag
atio
n
alg
o
r
it
h
m
[
1
8
-
1
9
]
.
Fro
m
to
th
e
s
u
g
g
e
s
tio
n
s
o
f
m
an
y
p
r
ev
io
u
s
w
o
r
k
s
[
2
0
-
2
1
]
an
d
af
ter
s
o
m
e
p
r
elim
i
n
ar
y
te
s
ts
t
h
ese
v
alu
es
w
er
e
s
et.
T
h
e
α
v
alu
e
f
o
r
t
h
e
H
VS
-
MR
MR
v
a
r
iab
le
s
elec
tio
n
w
as t
h
e
n
d
eter
m
i
n
ed
e
m
p
ir
icall
y
f
r
o
m
th
e
s
e
t {
0
.
2
,
0
.
3
,
0
.
4
,
0
.
5
,
0
.
6
,
0
.
7
,
0
.
8
}
b
ased
o
n
th
e
b
est ten
f
o
ld
cr
o
s
s
-
v
alid
atio
n
p
er
f
o
r
m
a
n
ce
.
4
.
3
Resul
t
s
T
ab
le
2
s
h
o
w
s
t
h
e
r
esu
lts
o
f
HVS
-
MRM
R
,
MRM
R
a
n
d
H
VS
o
v
er
2
0
in
d
ep
en
d
en
t
r
u
n
s
o
n
eig
h
t
class
i
f
icatio
n
d
ataset
s
.
T
h
e
cla
s
s
i
f
icatio
n
ac
c
u
r
ac
y
(
A
cc
)
in
T
ab
le
2
r
ef
er
s
to
th
e
p
er
ce
n
ta
g
e
o
f
clas
s
i
f
icatio
n
s
ac
cu
r
ac
y
p
r
o
d
u
ce
d
b
y
tr
ain
ed
ML
P
o
n
t
h
e
te
s
ti
n
g
s
et
o
f
a
cl
ass
i
f
icatio
n
d
ataset
a
n
d
(
Me
an
No
.
o
f
v
ar
iab
le)
p
r
esen
ts
t
h
e
av
er
a
g
e
n
u
m
b
er
s
o
f
v
ar
iab
les
s
elec
ted
.
I
t
ca
n
b
e
o
b
s
er
v
ed
f
r
o
m
tab
le
2
th
at
HVS
-
M
R
M
R
w
a
s
s
elec
ted
a
s
m
aller
n
u
m
b
er
o
f
v
ar
iab
les
f
o
r
d
if
f
er
e
n
t
b
en
c
h
m
ar
k
d
atasets
.
Fo
r
ex
a
m
p
le,
HVS
-
M
R
M
R
s
elec
ted
o
n
av
er
ag
e
6
.
1
3
v
ar
iab
les
f
r
o
m
a
s
et
o
f
9
v
ar
iab
le
s
f
o
r
s
o
l
v
e
th
e
ca
n
ce
r
d
ataset.
I
t
also
s
elec
ted
o
n
av
er
ag
e
6
.
8
7
v
ar
iab
les
f
r
o
m
a
s
et
o
f
3
4
v
ar
iab
les
f
o
r
s
o
lv
e
th
e
I
o
n
o
s
p
h
er
e
d
ataset.
I
n
f
ac
t,
HV
S
-
MR
MR
s
elec
ted
a
s
m
al
l n
u
m
b
er
o
f
v
ar
iab
les f
o
r
ea
ch
d
atasets
w
i
th
m
o
r
e
v
ar
ia
b
les.
Fig
u
r
e
3
s
h
o
w
s
t
h
e
f
r
eq
u
en
c
y
o
f
s
elec
ted
b
y
HVS
-
M
R
MR
f
o
r
C
an
ce
r
d
ataset
Fig
u
r
e
4
s
h
o
w
s
t
h
e
f
r
eq
u
en
c
y
o
f
s
elec
ted
b
y
HVS
-
M
R
MR
f
o
r
Diab
etes d
ataset
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Hyb
r
id
Meth
o
d
HV
S
-
MRMR
fo
r
V
a
r
ia
b
le
S
elec
tio
n
in
Mu
ltil
a
ye
r
A
r
ti
ficia
l Neu
r
a
l …
(
B
en
-
Hd
ec
h
A
d
il)
2779
Fig
u
r
e
5
s
h
o
w
s
t
h
e
f
r
eq
u
en
c
y
o
f
s
elec
ted
b
y
HVS
-
M
R
MR
f
o
r
Glass
d
ataset
I
n
o
r
d
er
to
d
eter
m
in
e
th
e
e
s
s
e
n
ce
o
f
s
elec
ted
v
ar
iab
les,
w
e
m
ea
s
u
r
ed
th
e
f
r
eq
u
e
n
c
y
o
f
v
ar
iab
les.
T
h
e
f
r
eq
u
en
c
y
o
f
a
v
ar
iab
le
i
ca
n
b
e
d
ef
in
ed
as
f
i
=
H
i
T
(
1
2
)
W
h
er
e
H
i
is
th
e
n
u
m
b
er
o
f
ti
m
e
s
a
p
ar
ticu
lar
v
ar
iab
le
is
s
elec
te
d
in
all
test
an
d
T
is
t
h
e
to
tal
n
u
m
b
er
o
f
test
.
Fig
u
r
e
3
,
Fig
u
r
e
4
an
d
Fi
g
u
r
e
5
s
h
o
w
t
h
e
f
r
eq
u
e
n
c
y
o
f
v
ar
i
ab
les
s
elec
ted
f
o
r
d
iab
etes,
ca
n
ce
r
,
an
d
g
las
s
d
ataset
s
,
r
esp
ec
ti
v
el
y
.
I
t
ca
n
b
e
s
ee
n
in
Fi
g
u
r
e
3
t
h
at
HVS
-
MRM
R
s
elec
ted
f
ea
tu
r
e
s
1
,
2
,
6
,
7
,
8
an
d
9
o
f
th
e
C
a
n
ce
r
d
ataset
v
ar
y
f
r
eq
u
en
tl
y
.
T
h
e
f
r
eq
u
e
n
c
y
o
f
s
ele
ctio
n
f
o
r
th
e
s
e
v
ar
iab
les i
s
o
n
e
o
r
n
ea
r
ly
o
n
e.
T
ab
le
3
.
C
o
m
p
ar
is
o
n
a
m
o
n
g
HVS
-
MRM
R
,
HVS,
M
R
MR
A
D
HOC,
GP
SF
SC
D
,
E
I
R
-
M
L
P
FS
an
d
A
N
NI
GM
A
-
W
R
A
P
P
E
R
f
o
r
th
e
ca
n
ce
r
,
d
iab
etes,
g
la
s
s
,
h
ep
atitis
,
h
o
r
s
e,
io
n
o
s
p
h
er
e,
v
e
h
ic
le,
an
d
w
a
v
e
f
o
r
m
d
atasets
D
a
t
a
se
t
HVS
M
R
M
R
A
D
H
O
C
G
P
S
F
S
C
D
E
I
R
-
M
L
P
F
S
A
N
N
I
G
M
A
-
W
R
A
P
P
ER
HVS
-
M
R
M
R
D
i
a
b
e
t
e
s
7
6
.
3
2
7
6
.
2
5
7
1
.
2
_
_
7
7
.
8
7
6
.
6
3
C
a
n
c
e
r
9
8
.
4
3
9
8
.
1
7
_
9
6
.
8
4
8
9
.
4
0
9
6
.
5
9
8
.
6
8
G
l
a
ss
7
6
.
6
1
7
6
.
7
3
7
0
.
5
_
4
4
.
1
0
_
7
7
.
0
3
V
e
h
i
c
l
e
7
4
.
3
5
7
4
.
2
1
6
9
.
6
7
8
.
4
5
7
4
.
6
0
_
7
5
.
1
7
H
e
p
a
t
i
t
i
s
7
7
.
6
5
7
6
.
3
2
_
_
_
_
7
8
.
6
W
a
v
e
f
o
r
m
8
5
.
9
1
8
5
.
2
7
_
_
_
_
8
4
.
9
1
H
o
r
se
8
6
.
0
3
8
5
.
2
7
_
_
_
_
8
6
.
2
1
I
o
n
o
sp
h
e
r
e
9
5
.
8
1
9
6
.
2
7
_
_
9
0
.
6
0
9
0
.
2
9
6
.
3
1
˝
̶
˝
m
ea
n
s
n
o
t a
v
a
ilab
le
W
e
ca
n
b
e
o
b
s
er
v
ed
th
at
o
u
r
m
e
th
o
d
ac
h
ie
v
ed
th
e
b
est
class
i
f
icatio
n
ac
cu
r
ac
y
a
m
o
n
g
all
o
th
er
alg
o
r
ith
m
s
f
o
r
f
i
v
e
o
u
t
(
C
a
n
ce
r
,
Glass
,
o
u
t,
Hep
ati
tis
,
H
o
r
s
e
an
d
I
o
n
o
s
p
h
er
e)
o
f
e
ig
h
t
d
ata
s
ets.
Fo
r
th
e
r
e
m
ain
in
g
th
r
ee
d
atasets
,
HV
S
-
M
R
MR
ac
h
ie
v
ed
as
a
s
ec
o
n
d
b
est.
w
h
ile
HVS
(
W
av
ef
o
r
m
)
,
A
NNI
GM
A
-
W
R
A
P
P
E
R
(
Diab
etes)
an
d
GP
SF
S
C
D(
Veh
i
cle)
ac
h
ie
v
ed
th
e
b
est cla
s
s
i
f
icatio
n
ac
cu
r
ac
y
f
o
r
o
n
e
d
ataset
ea
ch
.
W
e
ca
n
b
e
s
aid
th
at
th
e
v
ar
i
ab
les
s
elec
tio
n
in
cr
ea
s
e
s
th
e
class
i
f
icatio
n
ac
cu
r
ac
y
b
y
ig
n
o
r
in
g
th
e
ir
r
elev
an
t
v
ar
iab
le
s
f
r
o
m
th
e
o
r
ig
in
al
f
ea
t
u
r
e
s
et.
T
h
e
v
ar
iab
les
s
elec
tio
n
is
an
i
m
p
o
r
ta
n
t
tas
k
in
s
u
ch
a
p
r
o
ce
s
s
is
to
s
elec
t
n
ec
es
s
ar
y
in
f
o
r
m
atio
n
(
ir
r
elev
a
n
t
v
ar
ia
b
les).
Oth
er
w
i
s
e,
t
h
e
p
er
f
o
r
m
an
ce
o
f
clas
s
i
f
ier
s
m
i
g
h
t b
e
d
ec
r
ea
s
ed
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
5
,
Octo
b
er
2
0
1
7
:
2
7
7
3
–
2
7
8
1
2780
T
h
e
ef
f
icac
y
o
f
e
m
b
ed
d
i
n
g
o
f
MRM
R
f
i
lter
in
HV
S
w
a
s
ev
id
en
ce
d
b
y
i
m
p
r
o
v
ed
cla
s
s
i
f
icatio
n
p
er
f
o
r
m
a
n
ce
o
n
b
en
ch
m
ar
k
d
atasets
.
I
n
th
i
s
p
ap
er
,
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
o
u
tp
er
f
o
r
m
ed
o
th
er
m
et
h
o
d
s
in
th
e
class
i
f
icatio
n
o
n
t
h
e
m
o
s
t
d
atab
ase
test
ed
,
it
w
as
ab
le
to
s
e
lect
t
h
e
r
elev
a
n
ce
v
ar
iab
les
a
m
o
n
g
d
atase
ts
.
W
e
ca
n
b
e
ch
o
o
s
e
f
r
o
m
th
ese
d
at
a
w
it
h
th
e
an
a
l
y
s
is
o
f
p
er
f
o
r
m
an
ce
o
f
t
h
e
s
u
b
s
et
o
f
v
ar
iab
les
th
at
h
a
v
e
s
tr
o
n
g
r
elatio
n
s
h
ip
w
it
h
th
e
cla
s
s
i
f
ic
atio
n
.
Ho
w
e
v
er
i
n
ter
m
s
o
f
e
x
ec
u
t
io
n
ti
m
e,
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
co
n
s
u
m
e
s
m
o
r
e
t
h
a
n
HVS
a
n
d
MRMR
.
5.
CO
NCLU
SI
O
N
I
t
is
v
er
y
i
m
p
o
r
tan
t
to
r
e
m
o
v
e
th
e
r
ed
u
n
d
an
t
a
n
d
ir
r
elev
a
n
t
v
ar
iab
les
in
d
ata
b
ef
o
r
e
ap
p
ly
in
g
s
o
m
e
d
ata
m
i
n
i
n
g
tec
h
n
iq
u
es
to
a
n
al
y
ze
t
h
e
d
ata
s
et
s
.
I
n
o
u
r
r
e
s
ea
r
ch
,
w
e
s
u
g
g
e
s
t
a
n
e
w
m
eth
o
d
o
f
v
ar
iab
les
s
elec
tio
n
b
ased
o
n
HV
S
cr
iter
io
n
an
d
MRM
R
cr
iter
io
n
,
ca
ll
ed
th
e
HVS
-
MRM
R
,
to
in
te
g
r
ate
th
e
p
r
o
ce
d
u
r
es
o
f
v
ar
iab
le
s
elec
tio
n
f
i
lter
an
d
v
ar
iab
le
s
elec
tio
n
w
r
ap
p
er
to
im
p
r
o
v
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
cl
ass
i
f
icatio
n
.
W
e
ap
p
lied
HVS
-
MRM
R
f
o
r
f
o
u
r
clas
s
i
f
icatio
n
p
r
o
b
le
m
s
.
T
h
e
ex
p
er
i
m
en
t
r
es
u
lt
s
s
h
o
w
th
at
HV
S
-
MRMR
v
ar
iab
les
s
elec
tio
n
s
e
l
ec
ted
a
less
n
u
m
b
er
o
f
v
ar
iab
les
w
it
h
h
i
g
h
c
lass
if
ica
tio
n
ac
c
u
r
ac
y
co
m
p
ar
ed
to
MRMR
,
HV
S.
I
n
a
f
o
r
th
co
m
in
g
r
esear
c
h
w
o
r
k
,
w
e
i
n
ten
d
to
i
m
p
r
o
v
e
th
i
s
ap
p
r
o
ac
h
to
f
i
n
d
b
ette
r
s
u
b
s
et
o
f
v
ar
iab
les s
elec
ted
a
n
d
to
i
m
p
r
o
v
e
class
i
f
icat
io
n
ac
cu
r
ac
y
[
2
6
-
2
7
]
.
ACK
NO
WL
E
D
G
E
M
E
NT
S
W
e
ex
p
r
ess
o
u
r
th
a
n
k
s
to
th
e
a
s
s
o
ciate
ed
ito
r
an
d
an
o
n
y
m
o
u
s
r
ef
er
ee
s
w
h
o
s
e
v
al
u
ab
le
s
u
g
g
esti
o
n
s
h
elp
ed
to
i
m
p
r
o
v
e
th
i
s
w
o
r
k
s
i
g
n
i
f
ica
n
tl
y
a
n
d
also
to
th
eir
co
lleag
u
e.
RE
F
E
R
E
NC
E
S
[1
]
Hin
to
n
G
e
o
ff
re
y
E,
S
a
lak
h
u
td
in
o
v
Ru
sla
n
R.
“
Re
d
u
c
in
g
th
e
Dim
e
n
sio
n
a
li
ty
o
f
Da
ta
w
it
h
Ne
u
ra
l
Ne
tw
o
rk
s
.
”
S
c
i
e
n
c
e
,
2
0
0
6
;
3
1
3
(
5
7
8
6
)
;
p
.
5
0
4
-
5
0
7
.
[2
]
De
m
e
rs
D
a
v
id
,
Co
tt
re
ll
G
W
.
“
n
–
l
in
e
a
r
d
im
e
n
sio
n
a
li
ty
re
d
u
c
ti
o
n
.
”
In
:
Ad
v
a
n
c
e
s
in
n
e
u
ra
l
in
fo
rm
a
ti
o
n
p
ro
c
e
ss
in
g
sy
ste
ms
,
1
9
9
3
;
5
;
5
8
0
-
5
8
7
.
[3
]
F
o
d
o
r
,
Im
o
la
K.
“
A
S
u
rv
e
y
o
f
Dime
n
sio
n
Red
u
c
ti
o
n
T
e
c
h
n
iq
u
e
s
.
”
Ce
n
ter
f
o
r
A
p
p
li
e
d
S
c
ien
ti
f
ic
Co
m
p
u
ti
n
g
,
L
a
w
r
e
n
c
e
L
i
v
e
r
m
o
re
Na
ti
o
n
a
l
L
a
b
o
ra
to
ry
,
2
0
0
2
;
9
;
1
-
1
8
.
[4
]
W
e
i,
Hu
a
-
L
ian
g
Et
Bil
li
n
g
s
,
S
tep
h
e
n
A
.
“
F
e
a
tu
re
S
u
b
se
t
S
e
lec
ti
o
n
a
n
d
Ra
n
k
in
g
f
o
r
Da
ta
Dim
e
n
sio
n
a
li
ty
Re
d
u
c
ti
o
n
.”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Pa
tt
e
rn
A
n
a
lys
is
a
n
d
M
a
c
h
i
n
e
I
n
telli
g
e
n
c
e
,
2
0
0
7
;
2
9
(1
)
;
1
6
2
-
1
6
6
.
[5
]
G
u
y
o
n
Isa
b
e
ll
e
,
El
isse
e
ff
A
n
d
ré
.
“
An
I
n
tro
d
u
c
ti
o
n
t
o
Va
ri
a
b
l
e
a
n
d
Fe
a
tu
re
S
e
lec
ti
o
n
.
”
J
o
u
r
n
a
l
o
f
M
a
c
h
i
n
e
L
e
a
rn
in
g
Res
e
a
rc
h
.
2
0
0
3
;
3
(3
);
1
1
5
7
-
1
1
8
2
.
[6
]
L
iu
h
u
a
n
,
M
o
t
o
d
a
Hir
o
sh
i,
YU
L
e
i.
“
Fea
tu
re
S
e
lec
ti
o
n
wit
h
S
e
lec
t
ive
S
a
mp
li
n
g
.
”
In
:
P
ro
c
e
e
d
i
n
g
s
o
f
th
e
Nin
e
tee
n
th
In
tern
a
ti
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
M
a
c
h
in
e
L
e
a
rn
in
g
,
ICM
L
.
2
0
0
2
;
3
9
5
–
4
0
2
.
[7
]
D
y
Je
n
n
if
e
r
G
,
BR
OD
L
EY
Ca
rl
a
E.
“
F
e
a
tu
re
S
e
lec
ti
o
n
f
o
r
Un
su
p
e
rv
ise
d
L
e
a
rn
in
g
.”
J
o
u
rn
a
l
o
f
ma
c
h
in
e
lea
r
n
in
g
re
se
a
rc
h
.
2
0
0
4
;
5;
8
4
5
-
8
8
9
.
[8
]
Ka
n
n
a
n
S
,
S
e
n
t
h
a
m
a
r
a
i
e
t
R
a
m
a
ra
j
N
A.
”
No
v
e
l
h
y
b
rid
F
e
a
tu
re
S
e
lec
ti
o
n
v
ia
S
y
m
m
e
tri
c
a
l
Un
c
e
rtain
ty
R
a
n
k
in
g
Ba
se
d
L
o
c
a
l
M
e
m
e
ti
c
S
e
a
rc
h
A
lg
o
rit
h
m
.
”
Kn
o
wled
g
e
-
Ba
se
d
S
y
ste
ms
,
2
0
1
0
;
2
3
(6
);
5
8
0
-
5
8
5
.
[9
]
P
e
n
g
Ha
n
c
h
u
a
n
,
L
o
n
F
u
h
u
i
,
D
in
g
Ch
ris.
“
F
e
a
tu
re
S
e
lec
ti
o
n
Ba
se
d
on
M
u
t
u
a
l
I
n
f
o
rm
a
ti
o
n
Crit
e
ri
a
o
f
M
a
x
-
d
e
p
e
n
d
e
n
c
y
,
M
a
x
-
r
e
lev
a
n
c
e
,
a
n
d
M
in
-
re
d
u
n
d
a
n
c
y
.
”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
p
a
tt
e
rn
a
n
a
lys
i
s
a
n
d
m
a
c
h
i
n
e
in
telli
g
e
n
c
e
.
2
0
0
5
;
2
7
(
8
);
1
2
2
6
-
1
2
3
8
.
[1
0
]
Ko
h
a
v
i
Ro
n
,
Jo
h
n
G
e
o
rg
e
H.
“
W
ra
p
p
e
rs
f
o
r
F
e
a
tu
re
S
u
b
se
t
S
e
l
e
c
ti
o
n
.”
Arti
fi
c
i
a
l
in
tell
ig
e
n
c
e
,
1
9
9
7
;
9
7
(1
-
2
),
2
7
3
-
3
2
4
.
[1
1
]
Ch
o
rm
u
n
g
e
S
m
it
a
,
Je
n
a
S
u
d
a
r
so
n
.
“
Eff
icie
n
t
F
e
a
tu
re
S
u
b
se
t
S
e
lec
ti
o
n
A
lg
o
rit
h
m
f
o
r
Hig
h
Dim
e
n
sio
n
a
l
Da
ta
.”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
Co
m
p
u
ter
E
n
g
i
n
e
e
rin
g
.
2
0
1
6
;
6
(4
);
1
8
8
0
.
[1
2
]
L
iu
Ch
u
a
n
,
W
a
n
g
,
W
e
n
y
o
n
g
,
Zh
a
o
Qia
n
g
,
e
t
a
l.
“
A
Ne
w
F
e
a
tu
re
S
e
lec
ti
o
n
M
e
th
o
d
Ba
se
d
o
n
a
Va
li
d
it
y
In
d
e
x
o
f
F
e
a
tu
re
S
u
b
se
t
.”
Pa
t
ter
n
Rec
o
g
n
i
ti
o
n
L
e
tt
e
rs
,
2
0
1
7
.
[1
3
]
S
e
la
En
n
y
Itj
e
,
Ha
rtati
S
ri,
Ha
rjo
k
o
Ag
u
s,
e
t
a
l.
“
F
e
a
tu
re
S
e
lec
ti
o
n
o
f
th
e
Co
m
b
in
a
ti
o
n
o
f
P
o
ro
u
s
T
ra
b
e
c
u
lar
w
it
h
A
n
th
ro
p
o
m
e
tri
c
F
e
a
tu
re
s
f
o
r
Os
teo
p
o
r
o
sis
S
c
re
e
n
in
g
.
”
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
El
e
c
trica
l
a
n
d
Co
m
p
u
ter
En
g
i
n
e
e
rin
g
.
2
0
1
5
;
5
(1
)
7
8
.
[1
4
]
Ro
se
n
b
latt
F
ra
n
k
.
“
Prin
c
i
p
les
o
f
Ne
u
ro
d
y
n
a
mic
s.
Per
c
e
p
tro
n
s
a
n
d
th
e
T
h
e
o
ry
o
f
Bra
i
n
M
e
c
h
a
n
isms
.
”
Co
rn
e
ll
A
e
ro
n
a
u
ti
c
a
l
L
a
b
o
ra
to
ry
In
c
Bu
ffa
lo
n
y
,
1
9
6
1
.
[1
5
]
Ya
c
o
u
b
M
e
z
ian
e
,
Be
n
n
a
n
i
Y
.
“
HV
S
:
A
He
u
risti
c
f
o
r
V
a
riab
le
S
e
lec
ti
o
n
i
n
M
u
lt
il
a
y
e
r
A
rti
f
icia
l
Ne
u
ra
l
Ne
tw
o
rk
Clas
sif
ier
.
”
In
telli
g
e
n
t
E
n
g
i
n
e
e
rin
g
S
y
ste
ms
th
ro
u
g
h
Arti
fi
c
i
a
l
Ne
u
ra
l
Ne
two
rk
s
,
S
t.
L
o
u
is,
M
i
ss
o
u
ri.
1
9
9
7
;
7
;
5
2
7
-
5
3
2
.
[1
6
]
P
e
n
g
Ha
n
c
h
u
a
n
,
L
o
n
g
F
u
h
u
i,
Din
g
Ch
ris
.
“
F
e
a
tu
re
S
e
lec
ti
o
n
Ba
se
d
on
M
u
tu
a
l
In
f
o
rm
a
ti
o
n
Crit
e
ria
o
f
M
a
x
-
d
e
p
e
n
d
e
n
c
y
,
M
a
x
-
r
e
lev
a
n
c
e
,
a
n
d
M
in
-
re
d
u
n
d
a
n
c
y
.
”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
p
a
tt
e
rn
a
n
a
lys
i
s
a
n
d
m
a
c
h
i
n
e
in
telli
g
e
n
c
e
.
2
0
0
5
;
2
7
(
8
);
1
2
2
6
-
1
2
3
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Hyb
r
id
Meth
o
d
HV
S
-
MRMR
fo
r
V
a
r
ia
b
le
S
elec
tio
n
in
Mu
ltil
a
ye
r
A
r
ti
ficia
l Neu
r
a
l …
(
B
en
-
Hd
ec
h
A
d
il)
2781
[1
7
]
Din
g
Ch
ris,
P
e
n
g
Ha
n
c
h
u
a
n
.
“
M
in
im
u
m
Re
d
u
n
d
a
n
c
y
F
e
a
tu
re
S
e
l
e
c
ti
o
n
f
ro
m
M
icro
a
rra
y
Ge
n
e
Ex
p
re
ss
io
n
Da
ta
.”
J
o
u
rn
a
l
o
f
b
i
o
i
n
fo
rm
a
ti
c
s a
n
d
c
o
mp
u
ta
ti
o
n
a
l
b
i
o
l
o
g
y
,
2
0
0
5
;
3
(
0
2
;
1
8
5
-
2
0
5
.
[1
8
]
Ru
m
e
lh
a
rt
D E
,
M
c
Clellan
d
J L
,
P
D
P
Re
se
a
rc
h
G
ro
u
p
.
“
P
a
ra
ll
e
l
D
istr
ib
u
te
d
P
r
o
c
e
ss
in
g
.”
IEE
E
.
9
8
8
;
1
;
4
4
3
-
4
5
3
.
[1
9
]
Ya
n
g
,
Ya
n
g
,
Hu
,
Ju
n
,
e
t
Zh
a
n
g
,
M
u
.
“
P
re
d
ictio
n
s
o
n
t
h
e
De
v
e
lo
p
m
e
n
t
Dim
e
n
sio
n
s
o
f
P
r
o
v
in
c
ial
T
o
u
rism
Disc
ip
li
n
e
Ba
se
d
o
n
t
h
e
A
rti
f
ici
a
l
Ne
u
ra
l
Ne
tw
o
rk
BP
M
o
d
e
l.
”
Bu
l
letin
o
f
El
e
c
trica
l
En
g
i
n
e
e
rin
g
a
n
d
I
n
fo
rm
a
ti
c
s
,
2
0
1
4
;
3
(
2
);
69
-
76.
[2
0
]
Isla
m
,
M
d
M
.
,
Ya
o
,
Xin
,
e
t
M
u
ra
se
,
Ka
z
u
y
u
k
i.
“
A
Co
n
stru
c
ti
v
e
A
l
g
o
rit
h
m
f
o
r
T
r
a
in
in
g
Co
o
p
e
ra
ti
v
e
Ne
u
ra
l
Ne
tw
o
rk
En
se
m
b
les
”
.
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
n
e
u
r
a
l
n
e
two
rk
s
,
2
0
0
3
;
1
4
(4
);
8
2
0
-
8
3
4
.
[2
1
]
P
re
c
h
e
lt
L
u
tz,
e
t
a
l.
“
Pro
b
e
n
1
:
A
S
e
t
o
f
Ne
u
ra
l
Ne
tw
o
rk
Ben
c
h
ma
r
k
Pro
b
lem
s
a
n
d
Ben
c
h
ma
rk
i
n
g
R
u
les
.”
1
9
9
4
.
[2
2
]
Rich
e
ld
i
M
a
rc
o
,
L
a
n
z
i
P
ier
L
u
c
a
.
“
ADHO
C:
A
to
o
l
fo
r
p
e
rfo
r
min
g
e
ff
e
c
ti
v
e
fea
tu
re
se
lec
ti
o
n
.
In
:
T
o
o
ls
wit
h
Arti
fi
c
ia
l
I
n
telli
g
e
n
c
e
.
”
P
r
o
c
e
e
d
in
g
s E
ig
h
th
IEE
E
In
ter
n
a
ti
o
n
a
l
Co
n
f
e
r
e
n
c
e
o
n
.
IE
EE
.
1
9
9
6
;
1
0
2
-
1
0
5
.
[2
3
]
M
u
n
i
Du
rg
a
P
ra
sa
d
,
P
A
L
Nik
h
il
R,
DA
S
J
y
o
ti
r
m
a
y
.
“
G
e
n
e
ti
c
P
ro
g
ra
m
m
in
g
f
o
r
S
im
u
lt
a
n
e
o
u
s
F
e
a
tu
re
S
e
lec
ti
o
n
a
n
d
Clas
sif
ier
De
sig
n
.
”
IEE
E
T
r
a
n
sa
c
ti
o
n
s
o
n
S
y
ste
ms
,
M
a
n
,
a
n
d
Cy
b
e
rn
e
ti
c
s,
P
a
rt
B
(Cy
b
e
rn
e
ti
c
s).
2
0
0
6
;
3
6
(1
);
1
0
6
-
1
1
7
.
[2
4
]
G
a
sc
a
E,
S
á
n
c
h
e
z
J
S,
A
lo
n
so
R.
“
El
im
in
a
ti
n
g
R
e
d
u
n
d
a
n
c
y
a
n
d
Irre
lev
a
n
c
e
Us
in
g
a
N
e
w
M
LP
-
b
a
se
d
F
e
a
tu
re
S
e
lec
ti
o
n
M
e
t
h
o
d
.”
P
a
tt
e
rn
Rec
o
g
n
it
i
o
n
.
2
0
0
6
;
3
9
(
2
),
3
1
3
-
3
1
5
.
[2
5
]
Hsu
Ch
u
n
-
Na
n
,
Hu
a
n
g
H
u
n
g
-
Ju
,
Die
tri
c
h
S
.
“
T
h
e
A
n
n
ig
m
a
-
w
ra
p
p
e
r
A
p
p
ro
a
c
h
to
F
a
st
F
e
a
tu
re
S
e
le
c
ti
o
n
f
o
r
Ne
u
ra
l
Ne
ts
.”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
S
y
ste
ms
,
M
a
n
,
a
n
d
Cy
b
e
rn
e
ti
c
s,
P
a
rt
B
(Cy
b
e
rn
e
ti
c
s).
2002
;
3
2
(
2
)
;
2
0
7
-
2
1
2
.
[2
6
]
Be
n
-
Hd
e
c
h
A
,
G
h
a
n
o
u
Y,
A
b
d
e
rra
h
im
E.
L
.
“
Ro
b
u
st
M
u
lt
i
-
c
o
mb
in
a
ti
o
n
Fea
t
u
re
S
e
lec
ti
o
n
fo
r
M
icr
o
a
rr
a
y
Da
ta
.
”
A
d
v
a
n
c
e
s in
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
lo
g
y
:
T
h
e
o
ry
a
n
d
A
p
p
li
c
a
t
io
n
,
2
0
1
6
.
[2
7
]
G
h
a
n
o
u
,
Y.,
&
Be
n
c
h
e
ik
h
,
G
.
“
A
rc
h
it
e
c
tu
re
Op
ti
m
iza
ti
o
n
a
n
d
Train
in
g
f
o
r
th
e
M
u
lt
il
a
y
e
r
P
e
rc
e
p
tro
n
Us
in
g
A
n
t
S
y
st
e
m
.
”
IAE
NG In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
C
o
mp
u
ter
S
c
ien
c
e
.
2
0
1
6
;
2
8
;
1
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.