I
AE
S
I
n
t
e
r
n
at
ion
al
Jou
r
n
al
of
Ar
t
if
icial
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
Vol.
14
,
No.
1
,
F
e
br
ua
r
y
2025
,
pp.
2
1
3
~
22
1
I
S
S
N:
2252
-
8938
,
DO
I
:
10
.
11591/i
jai
.
v
14
.i
1
.
pp
21
3
-
221
213
Jou
r
n
al
h
omepage
:
ht
tp:
//
ij
ai
.
iaes
c
or
e
.
c
om
M
ic
r
oa
r
r
ay
g
e
n
e
e
x
p
r
e
ssi
on
c
la
ssi
f
ic
at
io
n
:
d
w
ar
f
m
on
goos
e
op
t
imiz
at
io
n
w
it
h
d
e
e
p
le
a
r
n
i
n
g
S
h
yam
ala
Gowr
i
B
alaram
an
1
,
Anu
H
.
Nair
1
,
S
a
n
al
Ku
m
ar
2
1
D
e
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ie
nc
e
a
nd
E
ngi
ne
e
r
in
g,
F
a
c
ul
ty
of
E
ngi
ne
e
r
in
g a
nd T
e
c
hnol
ogy,
A
nna
ma
la
i
U
ni
ve
r
s
it
y,
A
nna
ma
la
i
N
a
ga
r
,
I
ndi
a
2
D
e
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ie
nc
e
,
R
a
je
s
w
a
r
i
V
e
da
c
ha
la
m G
ove
r
nme
nt
A
r
ts
C
ol
le
ge
,
C
he
nga
lp
a
tt
u,
I
ndi
a
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
J
a
n
29
,
2024
R
e
vis
e
d
Aug
6
,
2024
Ac
c
e
pted
Aug
30
,
2024
The
d
e
o
x
y
ri
b
o
n
u
c
l
ei
c
aci
d
(D
N
A
)
mi
cro
arra
y
mo
d
el
h
o
l
d
s
s
i
g
n
i
f
i
can
t
p
ro
m
i
s
e
fo
r
rev
eal
i
n
g
ex
p
re
s
s
i
o
n
d
at
a
fro
m
t
h
o
u
s
a
n
d
s
of
g
en
es
.
It
s
erv
es
as
a
v
al
u
ab
l
e
t
o
o
l
f
o
r
i
n
v
es
t
i
g
at
i
n
g
g
e
n
e
ex
p
res
s
i
o
n
s
in
d
i
v
ers
e
b
i
o
l
o
g
i
ca
l
res
earc
h
fi
el
d
s
.
T
h
i
s
s
t
u
d
y
e
x
p
l
o
re
s
ad
v
an
ceme
n
t
s
in
g
e
n
e
s
el
ect
i
o
n
fo
r
ca
n
cer
d
et
ec
t
i
o
n
t
h
ro
u
g
h
art
i
f
i
c
i
al
i
n
t
e
l
l
i
g
e
n
ce,
w
i
t
h
a
fo
c
u
s
on
t
h
e
ch
al
l
en
g
e
of
ex
t
rac
t
i
n
g
p
er
t
i
n
en
t
i
n
fo
rma
t
i
o
n
fr
o
m
v
as
t
d
a
t
ab
a
s
es
.
The
ap
p
l
i
cat
i
o
n
of
d
eep
l
earn
i
n
g
arch
i
t
ect
u
re
in
d
et
ec
t
i
n
g
ch
r
o
n
i
c
d
i
s
eas
es
an
d
ai
d
i
n
g
med
i
cal
d
eci
s
i
o
n
-
ma
k
i
n
g
h
a
s
p
ro
v
en
effect
i
v
e
acro
s
s
v
ar
i
o
u
s
d
o
mai
n
s
.
T
h
erefo
re,
t
h
i
s
s
t
u
d
y
d
es
i
g
n
s
an
en
h
an
c
ed
mi
cr
o
array
g
e
n
e
ex
p
res
s
i
o
n
cl
a
s
s
i
fi
ca
t
i
o
n
by
u
t
i
l
i
zi
n
g
a
d
w
arf
mo
n
g
o
o
s
e
o
p
t
i
mi
za
t
i
on
w
i
t
h
d
eep
l
earn
i
n
g
(MG
E
X
C
-
D
MO
D
L
)
ap
p
ro
ac
h
.
The
MG
E
X
C
-
D
M
O
D
L
ap
p
r
o
ach
i
n
t
e
n
d
s
to
c
l
as
s
i
f
y
the
mi
cro
arra
y
g
e
n
e
e
x
p
re
s
s
i
o
n
(MG
E
).
Fo
r
t
h
i
s
,
the
MG
E
X
C
-
D
M
O
D
L
t
ech
n
i
q
u
e
i
n
i
t
i
al
l
y
a
p
p
l
i
e
s
t
h
e
w
i
en
er
f
i
l
t
eri
n
g
(W
F)
t
ech
n
i
q
u
e
to
era
d
i
ca
t
e
t
h
e
n
o
i
s
e.
In
ad
d
i
t
i
o
n
,
the
MG
E
X
C
-
D
MO
D
L
t
ech
n
i
q
u
e
emp
l
o
y
s
a
d
eep
res
i
d
u
al
s
h
ri
n
k
ag
e
n
e
t
w
o
rk
(
D
RSN
)
to
l
earn
fea
t
u
re
v
ect
o
rs
.
Mean
w
h
i
l
e,
the
co
n
v
o
l
u
t
i
o
n
a
l
au
t
o
e
n
co
d
er
(CA
E
)
mo
d
el
w
as
ex
ecu
t
ed
fo
r
i
d
e
n
t
i
fy
i
n
g
an
d
cl
as
s
i
f
y
i
n
g
t
h
e
MG
E
d
a
t
a.
Fu
r
t
h
erm
o
re,
t
h
e
d
w
a
rf
mo
n
g
o
o
s
e
o
p
t
i
m
i
za
t
i
o
n
(
D
MO
)
-
b
as
e
d
h
y
p
er
p
aramet
er
t
u
n
i
n
g
is
p
erf
o
rmed
to
e
n
h
a
n
ce
t
h
e
d
e
t
ect
i
o
n
o
u
t
co
me
s
of
the
CA
E
mo
d
el
.
The
i
n
v
e
s
t
i
g
a
t
i
o
n
a
l
ev
a
l
u
a
t
i
o
n
of
t
h
e
MG
E
X
C
-
D
MO
D
L
mo
d
el
is
v
al
i
d
a
t
ed
u
s
i
n
g
a
b
e
n
ch
mar
k
d
a
t
ab
a
s
e
.
The
co
m
p
reh
e
n
s
i
v
e
co
mp
ar
i
s
o
n
o
u
t
c
o
me
h
i
g
h
l
i
g
h
t
ed
t
h
e
b
et
t
ermen
t
of
t
h
e
MG
E
X
C
-
D
MO
D
L
mo
d
e
l
o
v
er
rece
n
t
ap
p
ro
a
ch
e
s
.
K
e
y
w
o
r
d
s
:
C
onvolut
ional
a
utoenc
ode
r
De
e
p
lea
r
ning
De
oxyr
ibonucle
ic
a
c
id
Dw
a
r
f
mongoos
e
opti
mi
z
a
ti
on
M
icr
oa
r
r
a
y
ge
ne
e
xpr
e
s
s
ion
Th
i
s
is
an
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
the
CC
BY
-
SA
l
i
ce
n
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
S
hya
mala
Gow
r
i
B
a
lar
a
man
R
e
s
e
a
r
c
h
S
c
holar
,
De
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ienc
e
a
nd
E
nginee
r
ing
,
Anna
mala
i
Unive
r
s
it
y
Anna
mala
i
N
a
ga
r
,
C
hidamba
r
a
m
608002
,
I
ndia
E
mail:
s
hya
mala
gowr
ibala
r
a
man
@gmail.
c
om
1.
I
NT
RODU
C
T
I
ON
T
he
m
icr
oa
r
r
a
y
ge
ne
e
xp
r
e
s
s
io
n
(
M
G
E
)
da
ta
c
la
s
s
if
ica
t
ion
pr
o
blem
mi
d
poin
ts
a
r
ound
the
tas
k
of
pr
e
c
is
e
ly
c
las
s
i
f
yi
ng
bi
olo
gica
l
s
a
mpl
e
s
de
pe
nde
n
t
up
on
t
he
i
r
pr
of
il
e
s
of
ge
ne
e
xpr
e
s
s
i
on
[
1]
.
T
he
m
i
c
r
oa
r
r
a
y
model
pe
r
mi
ts
s
c
ie
nti
s
ts
to
e
va
luate
ma
ny
ge
ne
e
xp
r
e
s
s
ion
leve
ls
t
oge
t
he
r
,
of
f
e
r
ing
a
we
a
l
th
of
da
t
a
t
h
a
t
can
be
vit
a
l
f
or
r
e
c
ogn
izin
g
dis
e
a
s
e
s
,
f
ind
ing
bi
oma
r
ke
r
s
,
a
nd
incr
e
a
s
ing
ta
r
ge
t
e
d
tr
e
a
tm
e
n
ts
[
2
]
.
How
e
ve
r
,
th
is
we
a
l
th
of
in
f
or
mat
ion
a
ls
o
de
l
iver
s
an
im
por
tant
a
na
lyt
ic
a
l
a
nd
c
o
mpu
tati
ona
l
tas
k
.
T
he
p
r
ob
lema
t
ic
r
e
p
or
t
c
onta
ins
de
ve
lop
ing
s
t
r
on
g
a
nd
e
f
f
e
c
t
ua
l
c
las
s
i
f
ica
ti
on
met
hods
that
can
d
is
ti
n
guis
h
a
mong
s
a
mp
les
li
ke
he
a
lt
hy
a
n
d
unhe
a
l
thy
pe
r
s
o
ns
,
ba
s
e
d
on
t
he
i
r
ge
ne
e
xp
r
e
s
s
io
n
in
f
o
r
mat
ion
[
3
]
.
In
thi
s
s
i
tuat
ion
,
the
main
tas
k
s
c
ontai
n
f
e
a
tu
r
e
s
e
le
c
ti
on;
de
a
li
ng
w
it
h
hig
h
-
dim
e
ns
iona
l
da
ta,
a
nd
a
tt
a
i
ning
g
r
e
a
t
c
las
s
if
ica
t
ion
a
c
c
ur
a
c
y
wh
il
e
s
e
c
ur
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
1
,
F
e
br
ua
r
y
2025
:
21
3
-
221
214
biol
o
gica
l
int
e
r
p
r
e
tab
il
i
ty
.
P
r
e
c
is
e
ge
ne
e
xp
r
e
s
s
io
n
da
ta
c
las
s
if
ica
ti
on
is
v
e
r
y
e
s
s
e
nt
ial
f
or
f
ur
ther
ing
ou
r
c
ons
ide
r
a
ti
on
of
d
if
f
ic
ult
d
is
e
a
s
e
s
,
a
l
lowi
ng
in
it
ia
l
diagnos
is
,
a
nd
d
ir
e
c
ti
n
g
pe
r
s
o
na
li
z
e
d
t
r
e
a
tm
e
nt
pla
ns
[
4]
.
M
icr
oa
r
r
a
y
c
a
nc
e
r
da
ta
s
tudy
is
a
vit
a
l
s
tudy
a
r
e
a
a
c
r
os
s
va
r
ious
f
ields
li
ke
mac
hine
lea
r
ning
(
M
L
)
,
pa
tt
e
r
n
r
e
c
ognit
ion
,
s
tatis
ti
c
s
,
c
omput
a
ti
ona
l
biol
ogy
,
a
nd
other
a
s
s
oc
iate
d
a
r
e
a
s
.
It
plays
a
vi
tal
pa
r
t
in
r
e
c
ognit
ion,
a
na
lys
is
,
a
nd
c
a
nc
e
r
tr
e
a
tm
e
nt
[
5
]
.
T
he
s
tudi
e
s
a
r
e
nowa
da
ys
tar
ge
ti
ng
the
im
pr
ove
ment
of
e
xis
tenc
e
r
a
tes
in
c
a
nc
e
r
pa
ti
e
nts
by
de
ve
lopi
ng
the
pr
oc
e
s
s
a
nd
knowle
dge
of
c
he
c
king
a
nd
t
r
e
a
tm
e
nt
[
6]
.
T
he
f
or
e
mos
t
tr
ouble
with
mi
c
r
oa
r
r
a
y
da
tas
e
t
identif
ic
a
ti
on
a
r
is
e
s
f
r
om
numer
ous
is
s
ue
s
li
ke
s
hor
tage
of
e
nough
s
a
mpl
e
s
,
im
ba
lanc
e
d
c
las
s
,
nois
y
da
ta,
a
nd
h
igh
tr
ouble
of
f
e
a
tu
r
e
dim
e
ns
ionalit
y
that
mana
ge
d
to
be
dif
f
icult
to
diagnos
e
a
nd
ha
ve
outcome
s
of
wr
ong
c
las
s
if
ica
ti
on.
S
e
ve
r
a
l
r
e
s
e
a
r
c
h
wor
ks
a
s
s
oc
iate
d
with
dua
l
-
c
las
s
da
ta
c
las
s
if
ica
ti
on
o
f
mi
c
r
oa
r
r
a
y
c
a
nc
e
r
ha
ve
be
e
n
c
onduc
ted
[
7]
.
C
las
s
if
ying
mul
ti
c
las
s
da
ta
of
mi
c
r
oa
r
r
a
y
is
s
ti
ll
an
ope
n
r
e
s
e
a
r
c
h
a
r
e
a
due
to
an
outcome
of
tas
ks
in
c
las
s
im
ba
lanc
e
.
C
las
s
e
s
with
a
ti
ny
a
mount
of
models
ha
ve
be
e
n
ge
ne
r
a
ll
y
igno
r
e
d
due
to
the
bias
of
ma
ny
methods
ne
a
r
c
las
s
e
s
ha
ving
mor
e
a
mount
of
e
leme
nts
[
8]
.
ML
models
a
r
e
c
omm
only
us
e
d
in
r
e
s
olvi
ng
va
r
ious
d
i
f
f
icult
r
e
a
l
is
s
ue
s
a
nd
ha
ve
be
e
n
ve
r
if
i
e
d
to
be
e
f
f
e
c
ti
ve
in
e
xa
mi
ning
ge
ne
e
xpr
e
s
s
ion
da
ta.
M
GE
da
ta
c
las
s
if
ica
ti
on
with
de
e
p
lea
r
ning
(
DL
)
is
an
in
nova
ti
ve
tec
hnique
that
c
onne
c
ts
the
powe
r
of
ne
ur
a
l
ne
twor
ks
to
c
las
s
if
y
biol
ogica
l
s
a
mpl
e
s
p
r
e
c
is
e
ly
ba
s
e
d
on
their
M
GE
pr
of
i
les
[
9]
.
In
thi
s
pr
oc
e
dur
e
,
h
igh
-
dim
e
ns
ional
ge
ne
e
xpr
e
s
s
ion
da
ta
is
c
ha
nge
d
int
o
a
plan
tha
t
will
be
a
ppr
opr
iate
f
or
DL
,
a
nd
c
onvolut
ional
ne
ur
a
l
ne
twor
k
(
C
NN
)
mol
e
c
ular
ba
s
is
of
il
lnes
s
e
s
a
nd
a
ll
ow
a
ppli
c
a
ti
ons
in
pr
e
c
is
ion
medic
ine,
b
iom
a
r
ke
r
identif
ica
ti
on,
a
nd
dr
ug
dis
c
ove
r
y
[
10]
.
T
his
s
tudy
de
s
igns
an
e
nha
n
c
ed
mi
c
r
oa
r
r
a
y
ge
ne
e
xpr
e
s
s
ion
c
la
s
s
if
ica
ti
on
by
uti
li
z
ing
a
dwa
r
f
mongoos
e
opti
mi
z
a
ti
on
with
de
e
p
lea
r
ning
(
M
GE
XC
-
D
M
OD
L
)
a
ppr
oa
c
h.
T
he
M
GE
XC
-
DM
OD
L
a
ppr
oa
c
h
int
e
nds
to
c
las
s
if
y
the
M
GE
.
F
or
thi
s
,
the
M
GE
X
C
-
D
M
OD
L
tec
hnique
ini
ti
a
ll
y
a
ppli
e
s
the
W
iene
r
f
il
ter
ing
(
W
F
)
tec
hnique
to
e
r
a
dica
te
the
nois
e
.
In
a
ddit
ion,
the
M
GE
XC
-
DM
OD
L
tec
hnique
e
mpl
oys
a
de
e
p
r
e
s
idual
s
hr
inkage
ne
twor
k
(
DR
S
N)
to
lea
r
n
f
e
a
tur
e
ve
c
tor
s
.
M
e
a
nwhile,
the
c
onvolut
ional
a
utoenc
ode
r
(
C
A
E
)
model
wa
s
e
xe
c
uted
f
or
identi
f
ying
a
nd
c
las
s
if
ying
the
M
GE
da
ta.
F
u
r
ther
mor
e
,
the
dwa
r
f
mongoos
e
opt
im
iza
ti
on
(
DM
O
)
-
ba
s
e
d
hype
r
pa
r
a
mete
r
tuni
ng
is
pe
r
f
or
med
to
e
nha
nc
e
the
de
tec
ti
on
outcome
s
of
the
C
AE
mo
de
l.
T
he
inves
ti
ga
ti
ona
l
e
va
luation
of
the
M
GE
XC
-
DM
OD
L
model
is
va
li
da
ted
us
ing
a
be
nc
hmar
k
da
taba
s
e
.
T
he
r
e
maining
s
e
c
ti
ons
of
the
a
r
ti
c
le
a
r
e
a
r
r
a
nge
d
a
s
:
s
e
c
ti
on
2
il
lus
tr
a
tes
the
r
e
late
d
wor
ks
.
S
e
c
ti
on
3
por
tr
a
ys
the
pr
opos
e
d
model
.
T
he
n,
s
e
c
ti
on
4
e
labo
r
a
tes
on
the
e
xpe
r
im
e
ntal
va
li
da
ti
on
a
nd
s
e
c
ti
on
5
c
ompl
e
tes
the
wor
k.
2.
RE
L
AT
E
D
WORKS
S
a
he
e
d
[
11]
int
e
nde
d
to
de
ve
lop
an
ML
–
ba
s
e
d
a
ppr
oa
c
h
to
c
a
tegor
ize
a
c
ute
myeloid
a
nd
a
c
ute
lym
phoblas
ti
c
leuke
mi
a
de
pe
nde
nt
upon
M
GE
p
r
of
il
e
s
.
T
he
a
uthor
s
uti
li
z
e
d
l
inea
r
dis
c
r
im
inant
a
na
lys
is
(
L
DA
)
,
Ada
boos
t,
logi
s
ti
c
r
e
gr
e
s
s
ion
(
L
R
)
,
k
-
ne
i
ghbor
method,
e
xtr
e
me
r
a
ndomi
z
e
d
tr
e
e
s
a
lgor
it
h
m,
r
idge
c
las
s
if
ier
,
gr
a
dient
boos
ti
ng,
a
nd
r
a
ndom
f
o
r
e
s
t
(
R
F
)
.
T
he
pr
inciple
c
omponent
a
na
lys
is
(
P
C
A
)
wa
s
e
mpl
oye
d
f
or
d
im
e
ns
ionalit
y
r
e
duc
ti
on
.
T
he
a
uthor
s
uti
l
ize
2
va
r
ious
c
r
os
s
-
va
li
da
ti
on
pr
oc
e
s
s
e
s
due
to
they
mak
e
higher
-
a
c
c
ur
a
te
a
bil
it
y
e
va
luations
than
pr
ior
a
pp
r
oa
c
he
s
.
Va
iyapur
i
e
t
al.
[
12
]
de
s
igned
an
innovative
r
e
d
f
ox
opti
mi
z
e
r
with
a
de
e
p
lea
r
ning
-
ba
s
e
d
mi
c
r
oa
r
r
a
y
ge
ne
e
xpr
e
s
s
ion
c
la
s
s
if
ica
ti
on
(
R
F
OD
L
-
M
GE
C
)
tec
hn
ique.
T
his
model
tar
ge
ts
incr
e
a
s
ing
c
las
s
if
ica
ti
on
e
f
f
e
c
ti
ve
ne
s
s
by
c
hoos
ing
s
uit
a
ble
f
e
a
tur
e
s
.
T
he
R
F
OD
L
-
M
GE
C
method
e
mpl
oys
an
innovative
r
e
que
s
t
f
o
r
o
f
f
e
r
(
R
F
O
)
-
ba
s
e
d
f
e
a
tur
e
s
e
lec
ti
on
(
F
S
)
tec
hnique
f
o
r
de
t
e
r
mi
ning
opti
mum
f
e
a
tur
e
s
ubs
e
ts
.
Additi
ona
ll
y,
the
R
F
OD
L
-
M
GE
C
method
include
s
a
bi
-
dir
e
c
ti
ona
l
c
a
s
c
a
d
e
d
de
e
p
ne
ur
a
l
ne
twor
k
(
B
C
DN
N)
f
or
c
las
s
if
ying
da
ta.
T
he
c
ons
tr
a
int
s
e
xe
c
uted
in
the
B
C
D
NN
method
c
ould
be
tuned
by
e
mpl
oying
the
c
ha
os
ga
me
opti
mi
z
e
r
(
C
GO
)
tec
hnique.
R
os
tami
e
t
al.
[
13]
int
r
oduc
e
d
an
innovat
ive
s
o
c
ial
ne
twor
k
inves
ti
ga
ti
o
n
-
ba
s
e
d
ge
ne
s
e
lec
ti
on
tec
hnique.
T
he
de
ve
loped
tec
hnique
the
r
e
leva
nc
e
maximi
z
a
ti
on
a
nd
r
e
dunda
nc
y
mi
nim
iza
ti
on
(
mR
M
R
)
model.
He
r
e
,
at
e
ve
r
y
r
ound
,
a
s
upr
e
me
c
omm
un
it
y
wa
s
p
r
e
f
e
r
r
e
d
c
onti
nua
ll
y
.
Ke
et
al
.
[
14
]
c
ons
ider
e
d
a
s
wa
r
m
-
opti
mi
z
e
r
-
a
s
s
i
s
ted
f
il
ter
-
wr
a
ppe
r
ge
ne
s
e
lec
ti
on
c
ompr
i
s
ing
2
s
tage
s
:
T
he
pr
im
a
r
y
s
tage
will
be
the
f
il
ter
s
tep
that
c
hoos
e
s
s
mall
top
-
n
pe
r
c
e
ntag
e
s
of
ge
ne
s
a
nd
a
tt
a
ins
de
c
r
e
a
s
e
d
inf
or
mation;
late
r
,
the
s
e
c
onda
r
y
s
tage
e
xa
mi
ne
s
f
or
the
opti
mu
m
ge
ne
s
ubs
e
ts
de
pe
nd
upon
a
w
r
a
ppe
r
s
ys
tem
in
the
r
e
s
idual
ge
ne
s
by
e
mp
loyi
ng
a
s
wa
r
m
opti
mi
z
a
ti
on
r
e
late
d
tec
hnique.
R
e
s
e
a
r
c
h
by
B
a
c
ha
e
t
al.
[
15
]
,
an
innovative
de
c
r
e
a
s
e
d
c
omput
e
r
-
a
ided
diagnos
is
(
C
AD
)
tec
hnique
wa
s
a
ppli
e
d
with
the
M
AT
L
AB
(
ve
r
s
ion
R
2016a
)
platf
or
m
f
or
c
a
tegor
i
z
ing
the
f
our
c
a
nc
e
r
s
ubc
a
tegor
ies
.
T
he
ou
tcome
s
of
the
e
xpe
r
im
e
nt
ha
ve
be
e
n
pe
r
f
o
r
med
with
4
gr
oups
of
ba
s
e
li
ne
da
ta
unde
r
the
a
ppe
a
r
a
nc
e
of
c
a
nc
e
r
ous
ge
ne
s
.
P
a
ndit
e
t
al.
[
16]
pr
ojec
ted
an
e
f
f
e
c
ti
ve
a
nd
hyb
r
id
DL
method
f
o
r
c
las
s
if
ying
mol
e
c
ular
c
a
nc
e
r
with
the
he
lp
of
e
xpr
e
s
s
ion
da
ta
to
r
e
s
olve
thes
e
bor
de
r
s
.
T
he
input
da
ta
wa
s
pr
e
-
pr
oc
e
s
s
e
d
e
mpl
oying
a
s
c
a
lable
r
a
nge
a
da
pti
ve
bil
a
ter
a
l
f
il
ter
(
B
F
)
.
S
ubs
e
que
ntl
y,
c
lus
ter
ing
ha
s
be
e
n
a
c
c
ompl
is
he
d
by
e
mpl
oying
an
e
nr
iche
d
binom
ial
c
lus
ter
ing
tec
hnique.
F
oll
owe
d
by
the
da
ta
mus
t
be
r
e
moved
thr
ough
the
mul
t
if
r
a
c
tal
B
r
ownia
n
mot
ion
(
M
B
M
)
tec
hnique.
L
a
ter
,
the
s
igni
f
ica
nt
f
e
a
tur
e
s
s
hould
be
c
hos
e
n
by
uti
li
z
ing
an
im
pr
ove
d
c
uc
koo
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
icr
oar
r
ay
ge
ne
e
x
p
r
e
s
s
ion
c
las
s
if
ication:
dw
ar
f
mongoos
e
opti
miz
ati
on
…
(
Shy
amala
Gow
r
i
B
alar
aman
)
215
s
e
a
r
c
h
opti
mi
z
e
r
(
I
C
S
O)
method.
L
a
s
tl
y,
the
da
ta
c
las
s
if
ica
ti
on
wa
s
e
xe
c
uted
e
mpl
oying
a
wa
ve
let
-
ba
s
e
d
de
e
p
c
onvolut
ional
ne
ur
a
l
ne
twor
k
(
DC
NN
)
.
Hi
lal
et
al
.
[
17]
p
r
e
s
e
nted
ne
w
f
e
a
tu
r
e
s
ubs
e
t
s
e
lec
ti
on
(
F
S
S
)
with
opti
mum
a
da
pti
ve
ne
ur
o
f
uz
z
y
inf
e
r
e
nc
e
s
ys
tem
(
OA
NFI
S
)
f
o
r
c
las
s
if
ying
ge
ne
e
xpr
e
s
s
ion.
T
he
ma
in
goa
l
is
to
identif
y
as
we
ll
as
c
a
tegor
ize
the
ge
ne
e
xp
r
e
s
s
ion
inf
or
mation
.
To
a
c
hieve
th
is
,
the
a
ppr
oa
c
h
de
v
e
lops
an
e
nha
nc
ed
im
pr
ove
d
gr
e
y
wolf
op
ti
mi
z
e
r
-
ba
s
e
d
f
e
a
tur
e
s
e
lec
ti
on
(
I
GW
O
-
F
S
)
tec
hnique
f
or
a
c
hieving
opti
mum
f
e
a
tur
e
s
ubs
e
ts
.
F
ur
ther
,
the
OA
NF
I
S
tec
hniqu
e
wa
s
e
xploi
ted
in
the
c
las
s
if
ica
ti
on
of
ge
ne
s
a
nd
the
hype
r
pa
r
a
mete
r
tuni
ng
of
the
a
da
pti
ve
ne
ur
o
-
f
uz
z
y
inf
e
r
e
nc
e
s
ys
tem
(
AN
F
I
S
)
s
ys
tem
can
be
mo
dif
ied
by
a
pplyi
ng
a
c
oyote
opti
mi
z
a
ti
on
a
lgor
it
hm
(
C
OA
)
.
3.
T
HE
P
ROP
OS
E
D
M
E
T
HO
D
In
thi
s
s
tudy,
an
e
nha
nc
e
d
M
GE
XC
-
DM
OD
L
a
ppr
oa
c
h
is
de
s
igned.
T
he
M
GE
XC
-
DM
OD
L
a
ppr
oa
c
h
int
e
nds
to
c
las
s
if
y
the
pr
e
s
e
nc
e
of
the
M
GE
c
las
s
if
ica
ti
on.
To
a
c
c
ompl
is
h
thi
s
,
the
M
GE
XC
-
DM
OD
L
method
e
nc
ompas
s
e
s
pr
e
-
pr
oc
e
s
s
ing,
f
e
a
tur
e
e
xtr
a
c
tor
,
c
las
s
if
ica
ti
on,
a
nd
tuni
ng
pr
oc
e
s
s
e
s
.
F
igur
e
1
de
picts
the
s
tr
uc
tur
e
of
the
M
GE
XC
-
DM
OD
L
method.
F
igur
e
1.
W
or
kf
low
of
M
GE
XC
-
DM
OD
L
tec
hnique
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
1
,
F
e
br
ua
r
y
2025
:
21
3
-
221
216
3.
1.
P
r
e
p
r
oc
e
s
s
in
g
I
nit
ially,
the
M
GE
XC
-
DM
OD
L
tec
hnique
a
ppli
e
s
the
WF
tec
hnique
to
e
r
a
dica
te
the
nois
e
that
e
xis
ts
in
it.
WF
is
an
e
f
f
icie
nt
wa
y
to
e
nha
nc
e
the
a
c
c
ur
a
c
y
a
nd
qua
li
ty
of
mi
c
r
oa
r
r
a
y
im
a
ge
s
[
18]
.
T
he
m
icr
oa
r
r
a
y
tec
hnique
include
s
the
s
im
ult
a
ne
ous
a
na
lys
i
s
of
thous
a
nds
of
biol
ogica
l
s
a
mpl
e
s
,
ge
ne
r
a
ti
ng
lar
ge
im
a
ge
s
with
im
pe
r
f
e
c
ti
ons
a
nd
inher
e
nt
nois
e
.
In
thi
s
c
ontext,
WF
is
us
e
d
f
or
e
nha
nc
ing
the
s
ignal
-
to
-
nois
e
r
a
ti
o,
e
nha
nc
ing
the
c
lar
it
y
a
nd
r
e
duc
ing
a
r
ti
f
a
c
ts
of
ge
ne
e
xpr
e
s
s
ions
.
T
he
WF
e
f
f
icie
ntl
y
s
ha
r
pe
ns
the
im
a
ge
by
s
tatis
ti
c
a
ll
y
modeling
the
c
ha
r
a
c
ter
is
ti
c
s
of
nois
e
a
nd
de
s
ir
e
d
s
ignal
(
mi
c
r
oa
r
r
a
y
s
pots
r
e
pr
e
s
e
nti
ng
ge
ne
e
xpr
e
s
s
ions
)
,
whic
h
f
a
c
il
it
a
tes
qua
nti
f
ica
ti
on
a
nd
mor
e
a
c
c
ur
a
te
de
tec
ti
on
of
ge
ne
e
xpr
e
s
s
ions
.
T
his
tec
hnology
a
c
hieve
d
r
e
mar
ka
ble
s
uc
c
e
s
s
in
ge
nomi
c
s
a
nd
bioi
n
f
or
matic
s
r
e
s
e
a
r
c
h,
a
s
s
is
ti
ng
in
the
e
xtr
a
c
ti
on
of
b
iol
ogica
l
da
ta
f
r
om
mi
c
r
oa
r
r
a
y
im
a
ge
s
a
nd
c
ont
r
ibut
ing
to
a
dva
nc
e
ment
in
unde
r
s
tanding
c
ompl
ica
ted
c
e
ll
ular
p
r
oc
e
s
s
e
s
.
3.
2.
F
e
a
t
u
r
e
e
xt
r
ac
t
ion
T
he
M
GE
XC
-
DM
OD
L
tec
hnique
e
mpl
oys
DR
S
N
to
lea
r
n
f
e
a
tur
e
ve
c
tor
s
.
T
he
c
onc
e
pt
be
hind
the
int
e
gr
a
ti
on
of
the
R
S
N
with
the
de
e
p
r
e
s
idual
ne
twor
k
ha
s
r
e
s
ult
e
d
in
the
f
or
mul
a
ti
on
of
the
DR
S
N
model
[
19]
.
DR
S
N
is
a
c
ompl
e
x
c
a
s
c
a
de
d
de
e
p
ne
ur
a
l
ne
twor
k
(
DNN
)
that
e
mpl
oys
a
s
of
t
thr
e
s
hold
f
unc
ti
on
a
nd
a
tt
e
nti
on
mec
ha
nis
m
to
f
i
lt
e
r
out
nois
y
da
ta.
T
he
DR
S
N
us
e
s
s
e
lf
‐
a
tt
e
nti
on
modul
e
s
to
s
ys
tema
ti
c
a
ll
y
s
e
lec
t
va
luable
f
e
a
tur
e
da
ta
while
r
e
movi
ng
nois
e
a
nd
inef
f
e
c
tual
f
e
a
tur
e
s
,
he
nc
e
boos
ti
ng
the
DNN
c
a
pa
c
it
y
to
e
xtr
a
c
t
va
luable
f
e
a
tur
e
da
ta
f
r
om
the
nois
e
.
T
he
f
unda
menta
l
unit
of
DR
S
N
is
the
r
e
s
idual
s
hr
i
nka
ge
buil
ding
un
it
(
R
S
B
U)
.
T
he
DR
S
N
c
ons
is
ts
of
one
identit
y
mapping
,
two
‐
ba
tch
no
r
maliza
ti
on
(
),
one
s
of
t
thr
e
s
hold
lea
r
n
ing
s
ubne
twor
k
,
2
a
c
ti
v
a
ti
on
f
unc
ti
ons
(
M
is
h)
,
a
nd
2
c
onvolut
ional
laye
r
s
(
)
.
T
h
e
r
e
is
a
s
ubne
twor
k
in
each
s
e
gment
a
nd
its
r
ole
is
to
indepe
nde
ntl
y
lea
r
n
a
gr
oup
of
thr
e
s
holds
.
T
he
r
e
f
or
e
,
the
thr
e
s
hold
is
not
too
lar
ge
a
nd
gua
r
a
nte
e
d
to
be
pos
it
ive.
T
he
f
e
a
tur
e
map
lea
r
ns
va
r
ious
thr
e
s
hol
ds
;
he
nc
e
the
a
bove
s
ubne
twor
k
is
uti
li
z
e
d
as
an
a
tt
e
nti
on
modul
e
,
a
nd
the
s
of
t
thr
e
s
hold
is
us
e
d
to
c
onve
r
t
t
he
obs
e
r
ve
d
invalid
f
e
a
tur
e
in
to
z
e
r
o
,
a
nd
r
e
leva
nt
f
e
a
tur
e
s
a
r
e
r
e
taine
d.
T
he
s
of
t
th
r
e
s
holdi
ng
r
e
moves
the
f
e
a
tur
e
c
los
e
r
to
0
a
nd
r
e
tains
the
ne
ga
ti
ve
a
nd
pos
it
ive
f
e
a
tur
e
s
.
=
{
−
ℎ
>
ℎ
0
−
ℎ
≤
≤
ℎ
+
ℎ
<
−
ℎ
(
1)
In
(
1
)
,
a
nd
a
r
e
the
input
a
nd
output
f
e
a
tur
e
s
a
nd
ℎ
s
hows
the
thr
e
s
holdi
ng
f
unc
ti
on
.
=
{
1
>
ℎ
0
−
ℎ
≤
≤
ℎ
1
<
−
ℎ
(
2)
In
(
2
)
is
de
r
ived
f
r
o
m
(
1
)
,
a
nd
a
f
te
r
de
r
ivation
,
the
s
of
t
thr
e
s
holdi
ng
be
c
omes
0
or
1.
=
(
,
)
=
(
)
∗
m
a
x
{
|
|
−
ℎ
,
0
}
(
3)
T
he
s
of
t
thr
e
s
holdi
ng
is
c
onve
r
ted
int
o
(
3)
,
whe
r
e
(
)
de
notes
the
s
ymbol
ic
f
unc
ti
on
.
T
he
s
of
t
thr
e
s
hold
is
to
indepe
nde
ntl
y
a
tt
a
in
the
thr
e
s
hold
r
a
nge
.
T
he
DNN
ha
s
be
tt
e
r
outcome
s
in
s
e
lf
‐
lea
r
ning,
he
nc
e
the
incor
por
a
ti
on
of
s
of
t
th
r
e
s
hold
a
nd
DNN
can
e
f
f
e
c
ti
ve
ly
dif
f
e
r
e
nti
a
te
f
e
a
tur
e
s
f
r
om
the
i
r
r
e
leva
nt
f
e
a
tur
e
s
.
T
he
p
r
inciple
of
the
a
tt
e
nti
on
modu
le
include
s
f
a
c
i
li
tating
the
NN
model
to
lea
r
n
input
f
a
c
tor
s
indep
e
nde
ntl
y
a
nd
a
ll
oc
a
te
we
ight
s
to
them.
T
his
e
na
bles
to
a
s
s
ignm
e
nt
of
c
omput
a
ti
ona
l
r
e
s
our
c
e
s
to
obtain
e
s
s
e
nti
a
l
f
e
a
tur
e
s
,
whic
h
r
e
s
ult
s
in
be
tt
e
r
pe
r
f
or
manc
e
.
T
he
a
tt
e
nti
on
modul
e
include
s
a
ke
y‐
va
lue
mapping
pr
oc
e
s
s
via
que
r
y
ope
r
a
ti
on
of
mat
r
ix
-
ve
c
tor
.
T
his
invol
ve
s
de
r
ivi
ng
the
c
or
r
e
s
ponding
we
ight
va
lues
a
nd
c
a
lcula
ti
ng
the
s
im
il
a
r
it
y
be
twe
e
n
dim
e
ns
ion
ve
c
tor
s
,
whic
h
unde
r
go
nor
maliza
ti
on
thr
ough
the
S
o
f
tm
a
x
f
unc
ti
on.
C
ons
e
que
ntl
y,
the
we
ight
va
lues
a
r
e
mul
ti
pli
e
d
with
a
matr
ix
dim
e
ns
ional
ve
c
tor
a
nd
their
s
um
mation
is
a
tt
a
ined
f
or
f
or
mul
a
ti
ng
the
las
t
a
tt
e
nti
on
matr
ix
.
If
=
=
,
then
it
is
r
e
pr
e
s
e
nted
as
a
s
e
lf
‐
a
tt
e
nti
on
modul
e
a
nd
is
given
as
(
4)
.
(
,
,
)
=
m
a
x
{
√
}
(
4)
W
he
r
e
=
(
1
,
2
,
)
∈
,
=
(
1
,
2
,
)
∈
,
=
(
1
,
2
,
)
∈
,
r
e
f
e
r
s
to
the
dim
e
ns
ional
of
s
ingl
e
ve
c
tor
,
indi
c
a
tes
the
a
mount
of
input
ve
c
tor
s
that
a
r
e
a
tt
a
ined
by
l
inea
r
c
onve
r
s
ion
of
input
matr
ix
.
indi
c
a
tes
the
matr
ix
with
dim
e
ns
ion
to
a
djus
t
th
e
inner
pr
oduc
ts
.
de
notes
the
tr
a
ns
pos
it
ion
of
,
a
nd
the
f
or
mul
a
of
the
input
s
e
que
nc
e
li
ne
a
r
mappi
ng
pr
oc
e
s
s
is
given
a
s
(
5)
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
icr
oar
r
ay
ge
ne
e
x
p
r
e
s
s
ion
c
las
s
if
ication:
dw
ar
f
mongoos
e
opti
miz
ati
on
…
(
Shy
amala
Gow
r
i
B
alar
aman
)
217
{
=
=
=
(
5)
w
he
r
e
the
li
ne
a
r
mapping
pa
r
a
mete
r
matr
ice
s
,
,
a
nd
a
r
e
s
e
lf
-
lea
r
ne
d
in
the
tr
a
in
ing
model
.
3.
3.
Clas
s
if
icat
ion
u
s
in
g
c
on
volu
t
ion
al
au
t
oe
n
c
o
d
e
r
m
od
e
l
At
thi
s
pha
s
e
,
the
C
AE
model
can
be
e
xe
c
uted
f
or
identif
ying
a
nd
c
las
s
if
ying
the
M
GE
da
ta.
C
AE
c
ombi
ne
s
the
be
ne
f
it
s
of
c
onvolut
ion
f
il
ter
ing
in
C
NN
with
uns
upe
r
vis
e
d
pr
e
tr
a
ini
ng
of
a
utoenc
ode
r
s
[
20]
.
R
a
ther
than
the
f
ull
y
c
onne
c
ted
(
FC
)
laye
r
,
the
e
nc
ode
r
ha
s
a
c
onvolut
ion
laye
r
a
nd
the
de
c
ode
r
ha
s
a
de
c
onvolut
ion
laye
r
in
c
ont
r
a
s
t
to
the
topol
ogy
f
or
a
utoenc
ode
r
s
.
T
he
de
c
onvolut
ion
f
il
ter
is
an
inver
s
e
ve
r
s
ion
of
the
c
onvolut
ion
f
il
ter
.
Additi
ona
ll
y
,
the
de
c
onvo
lut
ion
laye
r
s
hould
be
f
oll
owe
d
by
the
unpooli
ng
la
ye
r
.
T
he
unpooli
ng
pr
oc
e
s
s
can
be
done
by
ke
e
ping
the
loca
t
ion
of
maximum
va
lue
dur
ing
pooli
ng
,
whic
h
pr
e
s
e
r
ve
s
the
va
lue
of
that
loca
ti
on
while
unpoo
li
ng
a
nd
z
e
r
o
ing
the
r
e
s
t.
S
pa
ti
a
l
loca
li
ty
can
be
r
e
taine
d
by
incor
po
r
a
ti
ng
th
e
c
onvolut
ion
f
unc
ti
on
at
a
ll
the
ne
ur
ons
.
T
hus
,
f
o
r
the
input
mat
r
ix
,
the
e
nc
ode
r
c
omput
e
s
.
=
(
∗
+
)
(
6)
In
(
6
),
indi
c
a
tes
the
a
c
ti
va
ti
on
f
unc
ti
on,
is
e
nc
ode
r
bias
,
∗
s
igni
f
ies
2
c
onvolut
ion,
a
nd
r
e
pr
e
s
e
nts
ℎ
2D
c
onvolut
ional
f
il
ter
.
Z
e
r
o
pa
dding
a
ppli
e
s
i
nput
matr
ix
f
o
r
r
e
taining
s
pa
ti
a
l
r
e
s
olut
ion.
N
e
xt,
the
r
e
c
ons
tr
uc
ti
on
is
a
tt
a
ined
by
(
7)
.
=
(
∗
F
̃
n
+
̃
)
(
7)
In
(
7
)
,
F
̃
n
s
hows
ℎ
2
c
onvolut
ion
f
il
ter
s
in
the
de
c
ode
r
,
d
e
s
ignate
s
the
r
e
c
ons
tr
uc
ti
on
of
ℎ
input
a
nd
indi
c
a
tes
bias
of
the
de
c
ode
r
.
Uns
upe
r
vis
e
d
pr
e
-
tr
a
ini
ng
is
us
e
d
in
the
ne
twor
k
that
mi
nim
ize
s
the
(
8)
.
(
)
=
∑
(
=
1
−
)
2
(
8)
T
he
FC
laye
r
a
nd
s
of
tm
a
x
c
las
s
if
ier
a
r
e
a
dde
d
a
nd
the
de
c
ode
r
pa
r
t
is
r
e
moved
at
the
e
nd
of
the
ne
twor
k
a
f
ter
uns
upe
r
vis
e
d
pr
e
tr
a
ini
ng
of
the
unpooli
ng
a
nd
de
c
onvolut
ion
laye
r
s
.
3.
4.
D
war
f
m
on
goos
e
op
t
i
m
izat
ion
-
b
as
e
d
h
yp
e
r
p
ar
am
e
t
e
r
t
u
n
i
n
g
E
ve
ntually,
the
DM
O
-
ba
s
e
d
pa
r
a
mete
r
tuni
ng
m
e
thod
is
e
xe
c
uted
f
or
e
nha
nc
ing
the
r
e
c
ognit
ion
output
s
of
the
C
AE
method
.
C
he
n
et
al
.
[
21
]
d
e
ve
loped
a
DM
O
a
lgor
it
hm
whic
h
is
a
populati
on
-
ba
s
e
d
meta
he
ur
is
ti
c
model
.
T
his
method
s
pli
ts
the
mong
oos
e
populac
e
int
o
3
dis
s
im
il
a
r
gr
oups
s
uc
h
as
ba
bys
it
ter
,
s
c
out,
a
nd
a
lpha
.
B
e
low
the
c
ontr
ol
of
a
f
e
male
lea
de
r
,
the
whole
populac
e
joi
n
tl
y
f
e
e
ds
as
an
a
dhe
s
ive
unit
.
If
the
gr
oup
of
a
lpha
f
lops
to
f
ind
f
ood
,
an
int
e
r
c
ha
nge
ha
ppe
ns
a
mong
f
oll
owe
r
s
of
the
ba
bys
it
ter
a
nd
a
lpha
gr
oups
.
S
o
,
a
s
s
oc
iate
s
of
the
a
lpha
g
r
oup
at
the
s
a
me
time
a
r
e
invol
ve
d
in
hunti
ng
a
c
ti
ons
while
pe
ne
tr
a
ti
ng
f
or
a
s
lee
ping
mound
.
DM
O
ne
e
ds
only
phys
ica
ll
y
or
ga
nize
d
limits
to
de
c
r
e
a
s
e
the
dif
f
iculty
of
the
s
ys
tem
us
e
.
W
he
n
the
a
s
s
oc
iate
s
of
the
a
lpha
gr
oup
ha
ve
i
na
de
qua
te
a
pti
tudes
,
they
will
in
ter
c
ha
nge
f
oll
o
we
r
s
of
ba
bys
it
ter
s
,
a
nd
a
lpha
gr
oups
of
f
e
r
DM
O
the
c
a
pa
bil
it
y
to
uphold
populac
e
va
r
iety.
T
he
s
lee
p
moun
d
de
vice
can
s
top
the
a
lgor
it
hm
f
r
o
m
a
r
r
ivi
ng
at
loca
l
goa
ls
.
−
I
nit
ialize
S
e
t
the
DM
O’
s
mathe
matica
l
method,
as
pr
e
s
e
nted
in
(
9
)
.
=
[
1
,
1
1
,
2
…
1
,
−
1
1
,
2
,
1
2
,
1
…
2
,
−
1
2
,
⋮
⋮
,
⋮
⋮
,
1
,
2
…
,
−
1
,
]
(
9)
W
he
r
e
a
s
,
de
notes
the
loca
ti
on
of
the
ℎ
mongoos
e
in
the
ℎ
dim
e
ns
ion;
s
igni
f
ies
the
populac
e
number
;
s
igni
f
ies
the
s
olut
ion
of
the
c
a
ndidate
a
nd
is
the
s
iz
e
of
the
pr
oblem
.
T
he
mathe
matica
l
method
is
dis
pl
a
ye
d
in
(
10)
.
,
=
(
,
,
)
(
10)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
1
,
F
e
br
ua
r
y
2025
:
21
3
-
221
218
He
r
e
is
e
mpl
oye
d
to
make
e
ve
nly
s
pr
e
a
d
r
a
ndom
number
s
;
a
nd
de
note
the
uppe
r
a
nd
lowe
r
li
mi
ts
,
c
or
r
e
s
pondingl
y;
a
nd
s
igni
f
ies
the
dim
e
ns
i
on.
−
Alpha
gr
oup
T
he
f
or
a
ging
dir
e
c
ti
on
of
the
dwa
r
f
mongoos
e
is
d
e
f
ined
by
the
f
e
male
lea
de
r
,
who
is
f
or
med
in
the
gr
oup
of
a
lpha.
T
he
pos
s
ibi
li
ty
of
e
ve
r
y
f
e
male
indi
vidual
in
the
a
lpha
gr
oup
f
e
tching
a
lea
de
r
ha
s
be
e
n
de
f
ined
by
(
11)
.
=
(
)
∑
−
1
(
)
(
11)
w
he
r
e
(
)
de
notes
the
f
it
ne
s
s
output
of
the
ℎ
indi
vidual;
=
−
;
s
igni
f
ies
the
number
of
ind
ivi
dua
ls
in
the
gr
oup
of
a
lpha;
a
nd
r
e
pr
e
s
e
nts
the
indi
vidua
ls
c
ount
in
the
ba
bys
it
ter
gr
oup
.
T
he
a
lpha
f
e
male
s
ha
ve
pr
e
f
e
r
r
e
d
f
or
a
ging
wa
ys
,
a
nd
their
f
or
mul
a
ti
on
is
as
(
12
)
:
+
1
=
+
×
×
(
−
)
(
12)
W
he
r
e
a
s
s
igni
f
ies
the
pos
it
ion
of
the
ℎ
indi
vidual;
+
1
de
notes
the
nove
l
f
ood
s
our
c
e
plac
e
;
s
igni
f
ies
the
r
a
ndom
a
mount
a
mong
[
−
1
,
1
]
;
is
an
a
r
bit
r
a
r
y
ind
ivi
dua
l
in
the
a
lpha
gr
oup
a
nd
is
s
e
t
to
2.
T
he
s
lee
ping
mound
(
S
M
)
is
the
r
e
laxing
loca
ti
on
of
d
wa
r
f
mongoos
e
s
.
I
ts
e
xp
r
e
s
s
ion
is
as
(
13)
:
=
(
+
1
)
−
(
)
m
a
x
{
|
(
+
1
)
,
(
)
|
}
(
13)
T
he
mathe
matica
l
method
of
the
mea
n
SM
is
as
(
1
4)
:
=
∑
=
1
(
14)
−
S
c
out
gr
oup
T
he
s
e
pa
r
a
te
f
oll
owe
r
s
of
the
gr
oup
of
s
c
outs
will
not
a
r
r
ive
at
their
p
r
e
c
e
ding
S
M
.
T
his
pr
omi
s
e
s
the
a
lgor
it
hm’
s
e
xplor
a
ti
on
c
a
pa
bil
it
y.
T
he
SM
mathe
matica
l
f
or
mul
a
is
as
(
15)
:
+
1
=
{
−
×
×
×
|
−
⃗
⃗
|
+
1
>
+
×
×
×
|
−
⃗
⃗
|
(
15)
H
e
r
e
+
1
de
notes
the
loca
ti
on
of
the
s
ubs
e
que
nt
S
M
;
s
igni
f
ies
the
pa
r
a
mete
r
that
moni
to
r
s
the
f
lexibil
it
y
of
the
mongoos
e
populac
e
.
−
B
a
bys
it
ter
gr
oup
T
he
ba
bys
it
ter
gr
oup
dim
e
ns
ion
na
tur
a
ll
y
c
ons
is
ts
of
s
ub
-
or
dinate
indi
viduals
c
onc
e
r
ne
d
f
o
r
their
of
f
s
pr
ing
whe
r
e
it
is
de
f
ined
as
de
pe
nde
nt
upon
the
d
im
e
ns
ion
of
population
.
T
his
a
f
f
e
c
ts
the
s
ys
tem
by
c
ons
is
tently
de
c
li
ning
the
a
lpha
g
r
oup
f
or
a
ging
pos
s
ibl
e
ove
r
the
pe
r
iod
.
P
a
r
a
mete
r
c
ha
nge
s
the
da
ta
r
e
ga
r
ding
f
o
r
a
ging
plac
e
s
f
o
r
o
ther
f
oll
owe
r
s
.
T
he
ba
bys
it
ter
f
it
ne
s
s
we
ight
is
f
ixed
to
0,
whic
h
s
a
f
e
gua
r
ds
the
a
ve
r
a
ge
we
ight
of
the
a
lpha
g
r
oup
in
the
f
oll
owi
ng
it
e
r
a
ti
on
ha
s
be
e
n
de
c
r
e
a
s
e
d
ther
e
f
o
r
e
mea
ns
t
he
gr
oup
e
f
f
or
t
is
de
laye
d.
T
he
DM
O
model
g
r
ows
a
f
it
ne
s
s
f
unc
ti
on
(
F
F
)
f
or
a
tt
a
ini
ng
s
upe
r
ior
c
las
s
if
ier
r
e
s
ult
s
.
It
e
xpr
e
s
s
e
s
a
pos
it
ive
numer
a
l
to
im
ply
the
be
s
t
output
f
o
r
the
c
a
ndidate
's
e
f
f
icie
nc
y.
He
r
e
,
the
les
s
e
ning
of
the
c
las
s
if
ier
e
r
r
or
s
is
mea
s
ur
e
d
as
F
F
,
as
de
noted
in
(
18
)
.
(
)
=
(
)
=
.
×
100
(
16)
4.
E
XP
E
RI
M
E
NT
AL
VA
L
I
DA
T
I
ON
T
he
outcome
e
va
luation
of
the
M
GE
XC
-
DM
OD
L
method
can
be
e
xa
mi
ne
d
uti
l
izing
thr
e
e
be
nc
hmar
k
da
tas
e
ts
[
22]
,
s
uc
h
as
br
e
a
s
t,
c
olon,
a
nd
ova
r
ian
c
a
nc
e
r
.
A
c
ompr
e
he
ns
ive
c
ompar
is
on
r
e
s
ult
of
the
M
GE
XC
-
DM
OD
L
method
on
the
br
e
a
s
t
c
a
nc
e
r
da
tas
e
t
can
be
highl
ight
e
d
in
F
igu
r
e
2
[
23]
–
[
25]
.
T
he
s
e
outcome
s
point
e
d
out
that
the
C
GR
M
D
-
MR
-
AN
F
I
S
,
gr
id
-
ba
s
e
d,
F
uz
z
y
c
mea
ns
,
a
nd
C
NN
model
ha
s
s
hown
t
he
lea
s
t
pe
r
f
or
manc
e
.
M
e
a
nwhile,
the
RF
model
ga
ins
s
li
ghtl
y
boos
ted
outcome
s
.
How
e
ve
r
,
the
M
GE
XC
-
DM
OD
L
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
icr
oar
r
ay
ge
ne
e
x
p
r
e
s
s
ion
c
las
s
if
ication:
dw
ar
f
mongoos
e
opti
miz
ati
on
…
(
Shy
amala
Gow
r
i
B
alar
aman
)
219
tec
hnique
de
mons
tr
a
tes
maximum
pe
r
f
or
manc
e
w
it
h
of
94.
59%
,
of
94
.
12%
,
of
94
.
59%
,
a
nd
of
94
.
02%
.
F
igur
e
2.
C
ompar
a
ti
ve
r
e
s
ult
of
the
M
GE
XC
-
DM
OD
L
s
ys
tem
unde
r
br
e
a
s
t
c
a
nc
e
r
da
tas
e
t
A
wide
c
ompar
a
ti
ve
a
na
lys
is
of
the
M
GE
XC
-
D
M
OD
L
method
with
the
c
olon
c
a
nc
e
r
da
ta
s
e
t
can
be
e
mphas
ize
d
in
F
igur
e
3.
T
he
s
e
obtaine
d
f
indi
ngs
indi
c
a
te
that
the
ge
ne
ti
c
a
lgor
i
thm
(
GA
)
-
s
uppor
t
ve
c
tor
mac
hine
(
S
VM
)
,
GA
-
K
-
ne
a
r
e
s
t
ne
ighbor
s
(
KNN
)
,
r
a
ndom+SVM
,
P
C
A
-
v
oti
ng,
logi
s
ti
c
boots
tr
a
p
(
L
ogit
B
oot
)
,
a
nd
RF
methods
ha
ve
s
hown
poor
e
r
pe
r
f
or
manc
e
.
M
e
a
nwhile,
the
t
wo
-
wa
y
c
lus
ter
ing
tec
hnique
a
c
hieve
s
moder
a
ted
incr
e
a
s
e
d
outcome
s
.
Ne
ve
r
thele
s
s
,
the
M
GE
XC
-
D
M
OD
L
model
r
e
ve
a
ls
s
upr
e
me
pe
r
f
or
m
a
nc
e
with
an
of
96.
15
%,
of
92.
86
%,
of
96
.
15
%,
a
nd
of
94
.
1
5
%.
F
igur
e
3.
C
ompar
a
ti
ve
outcome
of
the
M
GE
XC
-
D
M
OD
L
method
unde
r
c
olon
c
a
nc
e
r
da
tas
e
t
An
e
xtens
ive
c
ompar
a
ti
ve
r
e
s
ult
of
the
M
GE
XC
-
DM
OD
L
method
at
the
ova
r
ian
c
a
nc
e
r
da
tas
e
t
can
be
unde
r
s
c
or
e
d
in
F
igur
e
4.
T
he
s
e
a
c
c
ompl
is
he
d
f
indi
ngs
de
note
that
the
l
inea
r
S
VM
,
R
F
,
e
ns
e
mb
le
S
VM
,
c
omm
on
f
e
a
tur
e
opti
mi
z
a
ti
on
(
C
F
O)
-
L
DA
,
lapl
a
c
e
a
ppr
oxim
a
ti
on
(
L
APO
)
-
KNN,
a
nd
gr
a
dient
boos
ted
c
las
s
if
ier
(
GB
C
O
)
-
LR
tec
hniques
g
et
poor
e
r
pe
r
f
or
manc
e
.
S
im
i
lar
ly,
the
a
da
pti
ve
a
nt
c
olony
opti
mi
z
a
ti
on
(
AAO
)
-
mul
ti
-
laye
r
pe
r
c
e
ptr
on
(
M
L
P
)
tec
hnique
ob
tains
moder
a
ted
boos
ted
outcome
s
.
How
e
ve
r
,
the
M
GE
XC
-
DM
OD
L
tec
hnique
s
hows
e
xc
e
ll
e
nt
pe
r
f
or
manc
e
with
an
of
95
.
31
%,
of
96.
81
%,
of
95.
31
%,
a
nd
of
95.
89
%.
T
he
s
e
outcome
s
c
onf
ir
me
d
the
boos
ted
pe
r
f
or
manc
e
of
the
M
GE
XC
-
DM
OD
L
method
unde
r
ge
ne
e
xpr
e
s
s
ion
c
las
s
if
ica
ti
on.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
1
,
F
e
br
ua
r
y
2025
:
21
3
-
221
220
F
igur
e
4.
C
ompar
a
ti
ve
outcome
of
the
M
GE
XC
-
D
M
OD
L
model
with
ova
r
ian
c
a
nc
e
r
da
tas
e
t
5.
CONC
L
USI
ON
In
thi
s
s
tudy
,
e
nha
nc
e
d
M
GE
XC
-
DM
OD
L
a
ppr
oa
c
h
is
de
s
igned.
T
he
M
GE
XC
-
DM
OD
L
tec
hnique
int
e
nds
to
c
las
s
if
y
the
pr
e
s
e
nc
e
of
the
M
GE
c
l
a
s
s
if
ica
ti
on.
To
a
c
c
ompl
is
h
thi
s
,
the
M
GE
XC
-
DM
OD
L
tec
hnique
ini
ti
a
ll
y
a
ppli
e
s
the
WF
tec
hnique
to
e
r
a
dica
te
the
nois
e
that
e
xis
ts
in
it
.
In
a
ddit
ion
,
the
M
GE
XC
-
DM
OD
L
te
c
hnique
e
mpl
oys
DR
S
N
to
lea
r
n
f
e
a
tur
e
ve
c
tor
s
.
M
e
a
nwhile,
the
C
AE
tec
hnique
can
be
e
xe
c
uted
f
or
the
identif
ica
ti
on
a
nd
c
las
s
if
ica
ti
on
of
M
GE
da
ta.
F
ur
ther
mo
r
e
,
the
DM
O
-
ba
s
e
d
hype
r
pa
r
a
mete
r
tuni
ng
is
pe
r
f
or
med
to
i
mpr
ove
the
r
e
c
ognit
ion
outcome
s
of
the
C
AE
a
lgo
r
it
hm.
T
he
c
ompar
a
ti
ve
a
na
lys
is
of
the
M
GE
XC
-
D
M
OD
L
methodology
highl
ight
e
d
the
s
u
pe
r
ior
outcome
of
94.
59
%
,
96.
15
%
,
a
nd
95.
31%
ov
e
r
r
e
c
e
nt
s
tate
of
a
r
t
a
ppr
oa
c
he
s
unde
r
be
nc
hmar
k
da
tas
e
ts
.
T
he
M
GE
XC
-
DM
O
DL
methodology
e
nc
ounter
s
r
e
s
tr
ictions
in
ha
ndli
ng
lar
ge
da
tas
e
ts
a
nd
wa
r
r
a
nts
e
xplo
r
a
ti
on
in
r
e
a
l
-
wor
ld
c
li
nica
l
s
e
tt
ings
,
p
r
ompt
in
g
f
utur
e
e
nha
nc
e
ment
in
dis
e
a
s
e
de
tec
ti
on
via
M
GE
c
las
s
if
ica
ti
on.
RE
F
E
RE
NC
E
S
[
1]
S
.
O
s
a
ma
,
H
.
S
ha
ba
n,
a
nd
A
.
A
.
A
li
,
“
G
e
ne
r
e
duc
ti
on
a
nd
ma
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hms
f
or
c
a
nc
e
r
c
la
s
s
if
ic
a
ti
on
ba
s
e
d
on
mi
c
r
oa
r
r
a
y
ge
ne
e
xpr
e
s
s
io
n
da
t
a
:
A
c
ompr
e
he
ns
iv
e
r
e
vi
e
w
,”
E
x
pe
r
t
Sy
s
te
m
s
w
it
h
A
ppl
ic
at
io
ns
,
vol
.
213,
2023,
doi
:
10.1016/j
.e
s
w
a
.2022.1189
46.
[
2]
M
.
A
bd
-
E
ln
a
by,
M
. A
lf
ons
e
,
a
nd
M
.
R
ous
hdy,
“
C
la
s
s
if
ic
a
ti
on of
br
e
a
s
t
c
a
n
c
e
r
us
in
g
mi
c
r
oa
r
r
a
y
ge
ne
e
xpr
e
s
s
io
n
da
ta
:
A
s
ur
v
e
y,”
J
our
nal
of
B
io
m
e
di
c
al
I
nf
or
m
at
ic
s
, vol
. 117, 2021, doi:
10.101
6/
j.
jb
i.
2021.103764.
[
3]
R
.
T
a
ba
r
e
s
-
S
ot
o,
S
.
O
r
oz
c
o
-
A
r
ia
s
,
V
.
R
ome
r
o
-
C
a
no,
V
.
S
.
B
uc
he
li
,
J
.
L
.
R
odr
íg
ue
z
-
S
ot
e
lo
,
a
nd
C
.
F
.
J
im
é
ne
z
-
V
a
r
ón,
“
A
c
ompa
r
a
ti
ve
s
tu
dy of
ma
c
hi
ne
l
e
a
r
ni
ng a
nd de
e
p l
e
a
r
ni
ng a
lg
or
it
hms
t
o c
la
s
s
if
y c
a
nc
e
r
t
ype
s
ba
s
e
d on mi
c
r
oa
r
r
a
y ge
ne
e
xpr
e
s
s
i
on
da
ta
,”
P
e
e
r
J
C
om
put
e
r
S
c
ie
nc
e
, vol
. 2020, no.
4, 2020, doi:
10.
7717/pe
e
r
j
-
c
s
.270.
[
4]
E
. A
lh
e
na
w
i,
R
. A
l
-
S
a
yye
d, A
. H
uda
ib
,
a
nd S
. M
ir
ja
li
li
, “
F
e
a
tu
r
e
s
e
le
c
ti
on me
th
ods
on ge
ne
e
xpr
e
s
s
io
n mi
c
r
oa
r
r
a
y da
ta
f
or
c
a
nc
e
r
c
la
s
s
if
ic
a
ti
on:
A
s
ys
t
e
ma
ti
c
r
e
vi
e
w
,
”
C
om
put
e
r
s
in
B
io
lo
gy
and
M
e
di
c
in
e
,
vol
.
140,
2022
,
doi
:
10.1016/j
.
c
ompbi
ome
d.2021.105051.
[
5]
J
. O
. A
gus
ha
ka
, A
. E
. E
z
ugw
u, O
. N
. O
la
id
e
, O
. A
ki
nol
a
, R
. A
. Z
it
a
r
, a
nd
L
. A
bua
li
ga
h, “
I
mpr
ove
d dw
a
r
f
mongoos
e
opt
im
iz
a
ti
on
f
or
c
ons
tr
a
in
e
d
e
ngi
ne
e
r
in
g
de
s
ig
n
pr
obl
e
ms
,”
J
our
nal
of
B
io
ni
c
E
ngi
ne
e
r
in
g
,
vol
.
20,
no.
3,
pp.
1263
–
1295,
2023,
doi
:
10.1007/s
42235
-
022
-
00316
-
8.
[
6]
S
.
A
r
una
a
nd
L
.
V
.
N
a
nda
ki
s
hor
e
,
“
E
mpi
r
ic
a
l
a
na
ly
s
is
of
th
e
e
f
f
e
c
t
of
r
e
s
a
mpl
in
g
on
s
upe
r
vi
s
e
d
le
a
r
ni
ng
a
lg
or
it
hms
in
pr
e
di
c
t
in
g
th
e
t
ype
s
of
l
ung
c
a
nc
e
r
on mul
ti
c
la
s
s
i
mba
la
nc
e
d mi
c
r
oa
r
r
a
y g
e
ne
e
xpr
e
s
s
io
n da
ta
,”
E
A
I
/S
pr
in
ge
r
I
nnov
at
io
ns
i
n C
om
m
uni
c
at
io
n
and C
om
put
in
g
, pp. 15
–
27, 2022, doi:
10.1007/978
-
3
-
030
-
8616
5
-
0_2.
[
7]
N
. F
e
r
na
nde
z
-
P
oz
o
e
t
al
.
, “
P
E
A
T
mos
s
(
P
hys
c
omi
tr
e
ll
a
E
xpr
e
s
s
io
n A
tl
a
s
T
ool
)
:
a
uni
f
ie
d ge
ne
e
xpr
e
s
s
io
n a
tl
a
s
f
or
t
he
mode
l
pl
a
nt
P
hys
c
omi
tr
e
ll
a
pa
te
ns
,”
P
la
nt
J
our
nal
, vol
. 102, no. 1, pp. 165
–
177, 2020, doi:
10.1111/t
pj
.14607.
[
8]
L
.
C
he
n,
L
.
K
le
ba
nov,
A
.
A
lm
ude
va
r
,
C
.
P
r
os
c
he
l,
a
nd
A
.
Y
a
kovl
e
v,
“
A
s
tu
dy
of
th
e
c
or
r
e
la
ti
on
s
tr
uc
tu
r
e
of
mi
c
r
oa
r
r
a
y
ge
ne
e
xpr
e
s
s
io
n
d
a
ta
ba
s
e
d
on
me
c
h
a
ni
s
ti
c
mod
e
ll
in
g
of
c
e
ll
popul
a
ti
on
ki
ne
ti
c
s
,”
St
at
is
ti
c
al
M
ode
li
ng
fo
r
B
io
lo
gi
c
al
Sy
s
te
m
s
:
I
n
M
e
m
or
y
of
A
ndr
e
i
Y
ak
ov
le
v
, pp. 47
–
61, 2020, doi:
10.1007/978
-
3
-
030
-
3
4675
-
1_3.
[
9]
M
.
L
oe
y,
M
.
W
.
J
a
s
im
,
H
.
M
.
E
L
-
B
a
kr
y,
M
.
H
.
N
. T
a
ha
, a
nd N
.
E
.
M
.
K
ha
li
f
a
,
“
B
r
e
a
s
t
a
nd
c
ol
on
c
a
n
c
e
r
c
la
s
s
if
ic
a
ti
on
f
r
om
ge
ne
e
xpr
e
s
s
io
n pr
of
il
e
s
us
in
g da
ta
mi
ni
ng t
e
c
hni
qu
e
s
,”
S
y
m
m
e
tr
y
, v
ol
. 12, no. 3, 2020, doi
:
10.3390/s
ym12030408.
[
10]
Y
.
W
a
ng,
H
.
W
e
i,
L
.
S
ong,
L
.
X
u,
J
.
B
a
o,
a
nd
J
.
L
iu
,
“
G
e
ne
e
xpr
e
s
s
io
n
mi
c
r
oa
r
r
a
y
da
ta
me
ta
-
a
na
ly
s
i
s
id
e
nt
if
ie
s
c
a
ndi
da
te
g
e
ne
s
a
nd
mol
e
c
ul
a
r
me
c
ha
ni
s
m
a
s
s
oc
ia
te
d
w
it
h
c
le
a
r
c
e
ll
r
e
na
l
c
e
ll
c
a
r
c
in
oma
,”
C
e
ll
J
our
nal
,
vol
.
22,
no.
3,
pp.
386
–
393,
2020,
doi
:
10.22074/ce
ll
j.
2020.6561.
[
11]
Y
.
K
.
S
a
he
e
d,
“
E
f
f
e
c
ti
ve
di
me
ns
io
na
li
ty
r
e
duc
ti
on
mode
l
w
it
h
ma
c
hi
ne
le
a
r
ni
ng
c
la
s
s
if
ic
a
ti
on
f
or
mi
c
r
oa
r
r
a
y
ge
ne
e
xpr
e
s
s
io
n
d
a
ta
,”
D
at
a Sc
ie
nc
e
f
or
G
e
nom
ic
s
, pp. 153
–
164, 2022, doi:
10.1016/B
978
-
0
-
323
-
98352
-
5.00006
-
9.
[
12]
T
.
V
a
iy
a
pur
i,
L
iy
a
ka
th
uni
s
a
,
H
.
A
la
s
ka
r
,
E
.
A
l
jo
ha
ni
,
S
.
S
hr
id
e
vi
,
a
nd
A
.
H
us
s
a
in
,
“
R
e
d
f
ox
o
pt
im
iz
e
r
w
it
h
da
ta
-
s
c
ie
nc
e
-
e
na
bl
e
d
mi
c
r
oa
r
r
a
y ge
ne
e
xpr
e
s
s
io
n c
la
s
s
if
ic
a
ti
on mode
l,
”
A
ppl
ie
d Sc
ie
nc
e
s
, vol
. 12, no. 9, 2022, doi
:
10.3390/a
pp12094172.
[
13]
M
.
R
os
ta
mi
,
S
.
F
or
ouz
a
nde
h,
K
.
B
e
r
a
hma
nd,
M
.
S
ol
ta
ni
,
M
.
S
ha
hs
a
va
r
i,
a
nd
M
.
O
us
s
a
la
h,
“
G
e
ne
s
e
le
c
ti
o
n
f
or
mi
c
r
oa
r
r
a
y
da
ta
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
M
icr
oar
r
ay
ge
ne
e
x
p
r
e
s
s
ion
c
las
s
if
ication:
dw
ar
f
mongoos
e
opti
miz
ati
on
…
(
Shy
amala
Gow
r
i
B
alar
aman
)
221
c
la
s
s
if
ic
a
ti
on
vi
a
mul
ti
-
obj
e
c
ti
ve
gr
a
ph
th
e
or
e
ti
c
-
ba
s
e
d
me
t
hod,”
A
r
ti
fi
c
ia
l
I
nt
e
ll
ig
e
nc
e
in
M
e
di
c
in
e
,
vol
.
123,
2022,
doi
:
10.1016/j
.a
r
tm
e
d.2021.102228.
[
14]
L
.
K
e
,
M
.
L
i,
L
.
W
a
ng,
S
.
D
e
ng,
J
.
Y
e
,
a
nd
X
.
Y
u,
“
I
mpr
ove
d
s
w
a
r
m
-
opt
im
iz
a
t
io
n
-
ba
s
e
d
f
il
te
r
-
w
r
a
ppe
r
ge
ne
s
e
le
c
ti
on
f
r
om
mi
c
r
oa
r
r
a
y
da
ta
f
o
r
ge
ne
e
xpr
e
s
s
io
n
tu
mor
c
la
s
s
if
ic
a
ti
on,”
P
at
te
r
n
A
nal
y
s
is
and
A
ppl
ic
at
io
ns
,
vol
.
26,
no.
2,
pp.
455
–
472,
2
023,
doi
:
10.1007/s
10044
-
022
-
01117
-
9.
[
15]
S
.
B
a
c
ha
,
O
.
T
a
oua
li
,
a
nd
N
.
L
io
ua
ne
,
“
R
e
duc
e
d
C
A
D
s
ys
t
e
m
f
or
c
la
s
s
if
ic
a
ti
ons
of
c
a
nc
e
r
ty
pe
s
ba
s
e
d
on
mi
c
r
oa
r
r
a
y
ge
ne
e
xpr
e
s
s
io
n
d
a
ta
,”
2022
I
E
E
E
9t
h
I
nt
e
r
nat
io
nal
C
onf
e
r
e
nc
e
on
Sc
ie
n
c
e
s
of
E
le
c
tr
oni
c
s
,
T
e
c
hnol
ogi
e
s
of
I
nf
or
m
at
io
n
and
T
e
le
c
om
m
uni
c
at
io
ns
, SE
T
I
T
2022
, pp. 133
–
137, 2022, doi:
10.
1109/
S
E
T
I
T
54465.2022.9875863.
[
16]
D
.
P
a
ndi
t
,
J
.
D
hodi
ya
,
a
nd
Y
.
P
a
te
l,
“
M
ol
e
c
ul
a
r
c
a
nc
e
r
c
la
s
s
if
ic
a
ti
on
on
mi
c
r
oa
r
r
a
ys
ge
n
e
e
xpr
e
s
s
io
n
d
a
ta
us
in
g
w
a
ve
le
t
-
ba
s
e
d
de
e
p
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
k,”
I
nt
e
r
nat
io
nal
J
ou
r
nal
of
I
m
agi
n
g
Sy
s
te
m
s
and
T
e
c
hnol
ogy
,
vol
.
32,
no.
6,
pp.
2262
–
2280,
2022,
doi
:
10.1
002/
im
a
.22780.
[
17]
A
.
M
.
H
il
a
l
e
t
al
.
,
“
F
e
a
tu
r
e
s
ubs
e
t
s
e
le
c
ti
on
w
it
h
opt
im
a
l
a
da
pt
iv
e
ne
ur
o
-
f
uz
z
y
s
ys
te
ms
f
or
bi
oi
nf
or
ma
ti
c
s
ge
ne
e
xpr
e
s
s
io
n
c
la
s
s
if
ic
a
ti
on,”
C
om
put
at
io
nal
I
nt
e
ll
ig
e
nc
e
and N
e
ur
o
s
c
ie
nc
e
,
vol
. 2022, 2022, doi:
10.1155/2022/
1698137.
[
18]
R.
L
iu
,
Y
.
L
i,
H
.
W
a
ng,
a
nd
J
.
L
iu
,
“
A
no
is
y
mul
ti
-
obj
e
c
ti
ve
opt
i
mi
z
a
ti
on
a
lg
or
it
hm
ba
s
e
d
on
me
a
n
a
nd
W
ie
ne
r
f
il
te
r
s
,”
K
now
le
dge
-
B
as
e
d Sy
s
te
m
s
, vol
. 228, 2021, doi:
10.1016/j
.knos
ys
.2021.107
215.
[
19]
T
.
H
a
n,
Z
.
Z
ha
ng,
M
.
R
e
n,
C
.
D
ong,
X
.
J
ia
ng,
a
nd
Q
.
Z
hua
n
g
,
“
S
pe
e
c
h
e
mot
io
n
r
e
c
ogni
ti
on
ba
s
e
d
on
d
e
e
p
r
e
s
id
ua
l
s
hr
in
ka
ge
ne
twor
k,”
E
le
c
tr
oni
c
s
, vol
. 12, no. 11, 2023, doi
:
10.3390/e
le
c
t
r
oni
c
s
12112512.
[
20]
C
.
C
a
mpbe
ll
a
nd
F
.
A
hma
d,
“
S
e
mi
-
s
upe
r
vi
s
e
d
a
tt
e
nt
io
n
-
a
ug
me
nt
e
d
c
onvolut
io
na
l
a
ut
oe
nc
od
e
r
f
or
r
a
da
r
-
ba
s
e
d
hum
a
n
a
c
t
iv
it
y
r
e
c
ogni
ti
on,”
SP
I
E
D
e
fe
ns
e
+ C
om
m
e
r
c
ia
l
Se
ns
in
g
,
2022, doi:
10.1117/12
.2622366.
[
21]
S
.
C
he
n,
Y
.
Z
hou,
a
nd
Q
.
L
uo,
“
H
ybr
id
a
da
pt
iv
e
d
w
a
r
f
mongoos
e
opt
im
iz
a
ti
on
w
it
h
w
ha
le
opt
im
iz
a
ti
on
a
lg
or
it
hm
f
or
e
xt
r
a
c
t
in
g
phot
ovol
ta
ic
pa
r
a
me
te
r
s
,”
A
I
M
S E
ne
r
gy
, vol
. 12, no. 1, pp. 84
–
118, 2024, doi:
10.3934/ene
r
gy.2024005.
[
22]
Z
hu
e
t
al
.
“
M
ic
r
oa
r
r
a
y da
ta
s
e
ts
,”
W
e
k
a A
R
F
F
f
or
m
at
.
[
O
nl
in
e
]
.
A
va
il
a
bl
e
:
ht
tp
:/
/c
s
s
e
.
s
z
u.e
du.c
n/
s
ta
f
f
/z
huz
x/
D
a
ta
s
e
ts
.ht
ml
[
23]
P
.
M
is
hr
a
a
nd
N
.
B
hoi
,
“
C
a
nc
e
r
ge
ne
r
e
c
ogni
ti
on
f
r
om
mi
c
r
oa
r
r
a
y
da
ta
w
it
h
ma
nt
a
r
a
y
ba
s
e
d
e
nha
nc
e
d
A
N
F
I
S
te
c
hni
q
ue
,”
B
io
c
y
be
r
ne
ti
c
s
and B
io
m
e
di
c
al
E
ngi
ne
e
r
in
g
, vol
. 41, no. 3, pp. 916
–
932, 2021, doi:
10.1016/j
.bbe
.2021.06.004.
[
24]
A
.
E
l
-
N
a
ba
w
y,
N
.
E
l
-
B
e
nda
r
y,
a
nd
N
.
A
.
B
e
la
l,
“
E
pi
th
e
li
a
l
o
va
r
ia
n
c
a
nc
e
r
s
ta
ge
s
ubt
ype
c
la
s
s
if
ic
a
ti
on
us
in
g
c
li
ni
c
a
l
a
nd
ge
ne
e
xpr
e
s
s
io
n i
nt
e
gr
a
ti
ve
a
ppr
oa
c
h,
”
P
r
oc
e
di
a C
om
put
e
r
Sc
ie
n
c
e
,
vol
. 131, pp. 23
–
30, 2018, doi:
10.1016/j
.pr
oc
s
.2018.04.181.
[
25]
S
.
K
.
P
r
a
bha
ka
r
a
nd
S
.
W
.
L
e
e
,
“
A
n
in
te
gr
a
te
d
a
ppr
oa
c
h
f
o
r
ova
r
ia
n
c
a
nc
e
r
c
la
s
s
if
ic
a
ti
on
w
it
h
th
e
a
ppl
ic
a
ti
on
of
s
to
c
ha
s
ti
c
opt
im
iz
a
ti
on,”
I
E
E
E
A
c
c
e
s
s
, vol
. 8, pp. 127866
–
127882, 2020, doi:
10.1109/AC
C
E
S
S
.2020.3006154.
B
I
OG
AP
HI
E
S
OF
AU
T
HO
R
Shy
a
m
a
l
a
Go
w
ri
Ba
l
a
r
a
m
a
n
is
a
Res
earch
Sch
o
l
ar
in
the
D
ep
art
me
n
t
of
Co
m
p
u
t
er
Sci
en
ce
a
n
d
E
n
g
i
n
eeri
n
g
,
A
n
n
ama
l
ai
U
n
i
v
er
s
i
t
y
fr
o
m
2
0
2
1
J
a
n
u
ar
y
.
Sh
e
h
as
p
u
b
l
i
s
h
ed
2
res
earch
p
ap
er
s
in
i
n
t
er
n
at
i
o
n
al
j
o
u
rn
a
l
s
a
n
d
c
o
n
fere
n
ces
.
Cu
rre
n
t
l
y
,
h
er
res
earc
h
p
r
o
j
ec
t
s
i
n
c
l
u
d
e
i
ma
g
e
p
r
o
ces
s
i
n
g
,
mac
h
i
n
e
l
ear
n
i
n
g
,
d
eep
l
e
arn
i
n
g
,
an
d
d
at
a
s
ci
e
n
ce.
Sh
e
ca
n
be
co
n
t
act
e
d
at
emai
l
:
s
h
y
ama
l
ag
o
w
ri
b
al
arama
n
@
g
mai
l
.
co
m.
A
nu
H
.
Nair
is
an
A
s
s
i
s
t
a
n
t
Pr
o
fes
s
o
r
in
t
h
e
D
e
p
art
m
en
t
of
C
o
mp
u
t
er
Sci
e
n
ce
an
d
E
n
g
i
n
eeri
n
g
,
A
n
n
ama
l
ai
U
n
i
v
er
s
i
t
y
fr
o
m
2
0
0
5
.
H
er
mai
n
res
earch
area
s
i
n
cl
u
d
e
i
mag
e
p
ro
ce
s
s
i
n
g
i
n
mac
h
i
n
e
l
ear
n
i
n
g
t
ec
h
n
i
q
u
es
.
S
h
e
h
a
s
a
u
t
h
o
red
70
res
earc
h
p
a
p
ers
in
i
n
t
er
n
at
i
o
n
a
l
j
o
u
rn
a
l
s
an
d
co
n
fere
n
ces
,
an
d
t
w
o
b
o
o
k
ch
ap
t
ers
.
H
er
cu
rren
t
res
earc
h
p
ro
j
ect
s
i
n
cl
u
d
e
b
i
g
d
a
t
a
an
al
y
t
i
cs
,
b
i
o
met
r
i
c
p
ers
o
n
i
d
e
n
t
i
fi
cat
i
o
n
,
an
d
med
i
ca
l
i
mag
e
p
ro
ces
s
i
n
g.
Sh
e
can
b
e
co
n
t
act
e
d
at
emai
l
:
an
u
_
j
u
l
@
y
ah
o
o
.
co
.
i
n
.
Sa
na
l
Kum
a
r
is
cu
rre
n
t
l
y
w
o
r
k
i
n
g
as
A
s
s
i
s
t
a
n
t
Pr
o
f
es
s
o
r
in
t
h
e
P.
G
.
D
ep
art
me
n
t
of
Co
mp
u
t
er
Sci
e
n
ce
at
R.
V
.
G
o
v
ern
men
t
A
rt
s
Co
l
l
eg
e,
Ch
en
g
al
p
at
t
u
,
T
am
i
l
n
ad
u
,
In
d
i
a
(o
n
D
ep
u
t
a
t
i
o
n
).
He
h
as
au
t
h
o
re
d
65
p
a
p
ers
in
i
n
t
er
n
at
i
o
n
al
j
o
u
rn
a
l
s
an
d
co
n
feren
ce
s
an
d
h
o
l
d
s
l
i
fe
memb
ers
h
i
p
in
CSI,
IST
E
,
an
d
IA
E
N
G
.
His
res
earch
i
n
t
ere
s
t
s
s
p
an
i
ma
g
e
p
ro
ce
s
s
i
n
g
,
p
at
t
ern
reco
g
n
i
t
i
o
n
,
an
d
w
earab
l
e
co
m
p
u
t
i
n
g
.
He
ca
n
b
e
co
n
t
ac
t
ed
at
emai
l
:
s
a
n
al
p
rab
h
a@
y
a
h
o
o
.
co
.
i
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.