I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
,
p
p
.
140
~
153
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
1
.
p
p
1
4
0
-
1
5
3
140
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
M
a
lwa
re det
ec
tion usin
g
conv
o
luti
o
na
l neural net
wo
rk
-
di
stra
tegy
pola
r f
o
x
optimiza
tion a
lg
o
rithm
P
a
rv
a
t
hi Sa
t
hena
ha
lli J
a
y
a
p
ra
k
a
s
h
1,
2
,
Yo
g
ee
s
h Am
ba
la
g
er
e
Cha
nd
ra
s
hek
a
ra
ia
h
1
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
G
o
v
e
r
n
me
n
t
E
n
g
i
n
e
e
r
i
n
g
C
o
l
l
e
g
e
C
h
a
m
a
r
a
j
a
n
a
g
a
r
,
A
f
f
i
l
i
a
t
e
d
t
o
V
i
s
v
e
s
v
a
r
a
y
a
T
e
c
h
n
o
l
o
g
i
c
a
l
U
n
i
v
e
r
si
t
y
,
B
e
l
a
g
a
v
i
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
JS
S
S
c
i
e
n
c
e
a
n
d
Te
c
h
n
o
l
o
g
y
U
n
i
v
e
r
si
t
y
,
M
y
s
u
r
u
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
r
2
0
,
2
0
2
5
R
ev
is
ed
No
v
1
3
,
2
0
2
5
Acc
ep
ted
J
an
1
0
,
2
0
2
6
M
a
lwa
re
a
tt
a
c
k
s
h
a
v
e
e
sc
a
late
d
sig
n
ifi
c
a
n
tl
y
wi
th
a
n
in
c
re
a
se
o
f
i
n
tern
e
t
u
se
rs
a
n
d
c
o
n
n
e
c
ted
d
e
v
ice
s.
W
it
h
th
e
rise
o
f
v
a
ri
o
u
s
t
y
p
e
s
o
f
m
a
lwa
re
s
re
lea
se
d
b
y
th
e
h
a
c
k
e
rs,
c
o
n
stru
c
t
in
g
n
e
w co
m
p
e
ti
ti
v
e
m
e
th
o
d
s a
re
n
e
c
e
ss
a
ry
to
i
d
e
n
ti
f
y
th
e
a
d
v
a
n
c
e
d
m
a
lwa
re
.
Ho
we
v
e
r,
c
o
n
v
e
n
ti
o
n
a
l
m
a
lwa
re
d
e
tec
ti
o
n
stru
g
g
les
to
id
e
n
ti
fy
n
e
w
a
n
d
e
v
o
lv
i
n
g
m
a
lwa
re
v
a
rian
ts
a
c
c
u
ra
tely
b
e
c
a
u
se
o
f
it
s
d
e
p
e
n
d
e
n
c
e
o
n
h
a
n
d
c
ra
fted
fe
a
tu
re
s
a
n
d
sta
ti
c
-
sig
n
a
tu
re
b
a
se
d
m
e
th
o
d
s.
T
o
a
d
d
re
ss
th
is
p
ro
b
l
e
m
,
th
is
re
se
a
rc
h
p
ro
p
o
se
s
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
(CNN
)
b
a
se
d
d
i
stra
teg
y
p
o
lar
f
o
x
o
p
ti
m
iza
ti
o
n
a
lg
o
r
it
h
m
(DSP
F
OA
)
fo
r
m
a
lwa
re
d
e
tec
ti
o
n
t
o
fin
e
-
t
u
n
e
th
e
CNN
p
a
ra
m
e
ter
s
e
ffe
c
ti
v
e
ly
wh
ic
h
late
r
a
ss
ists
t
o
o
v
e
rc
o
m
e
t
h
e
li
m
it
a
ti
o
n
s
o
f
CNN
.
Th
e
m
o
d
e
l
in
te
g
ra
tes
th
e
sin
e
c
h
a
o
t
ic
m
a
p
p
in
g
a
n
d
Ca
u
c
h
y
o
p
e
ra
to
r
m
u
tatio
n
a
s
DSP
F
OA
p
re
v
e
n
ts
th
e
m
o
d
e
l
fr
o
m
lo
c
a
l
o
p
ti
m
a
issu
e
,
a
n
d
e
x
ten
d
s
se
a
rc
h
sp
a
c
e
so
lu
ti
o
n
,
a
lso
e
n
h
a
n
c
e
c
o
n
v
e
rg
e
n
c
e
.
T
h
is
e
n
s
u
re
s
th
a
t
t
h
e
C
NN
lea
rn
s
h
ig
h
ly
d
isc
rimin
a
ti
v
e
fe
a
tu
re
s
wh
ich
m
a
k
e
s
th
e
sy
ste
m
m
o
re
a
c
c
u
ra
te
a
n
d
ro
b
u
st
i
n
d
e
te
c
ti
n
g
b
o
th
k
n
o
w
n
a
n
d
e
v
o
l
v
in
g
m
a
lwa
re
v
a
rian
ts.
Th
e
CNN
-
DSP
F
OA
a
c
h
iev
e
s
a
h
ig
h
a
c
c
u
ra
c
y
o
f
9
9
.
6
5
a
n
d
9
9
.
7
6
%
b
y
u
ti
l
izin
g
BIG
2
0
1
5
a
n
d
M
a
li
m
g
d
a
tas
e
t
r
e
sp
e
c
ti
v
e
ly
c
o
m
p
a
re
d
t
o
e
x
isti
n
g
m
e
th
o
d
s
li
k
e
m
a
sk
e
d
se
lf
-
su
p
e
r
v
ise
d
m
o
d
e
l
with
sw
i
n
tran
sf
o
rm
e
r
(M
a
lS
o
rt)
.
K
ey
w
o
r
d
s
:
C
au
ch
y
o
p
e
r
ato
r
m
u
tatio
n
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
D
i
s
t
r
a
t
e
g
y
p
o
l
a
r
f
o
x
o
p
t
i
m
i
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m
Ma
lwar
e
d
etec
tio
n
Sin
e
ch
ao
tic
m
ap
p
i
n
g
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
:
Par
v
ath
i Sath
en
ah
alli Jay
ap
r
a
k
ash
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
G
o
v
er
n
m
en
t E
n
g
in
ee
r
in
g
C
o
lleg
e
C
h
am
ar
ajan
ag
ar
Af
f
iliated
to
Vis
v
esv
ar
ay
a
T
e
ch
n
o
lo
g
ical
Un
iv
er
s
ity
B
elag
av
i,
I
n
d
ia
E
m
ail:
s
jp
ar
v
ath
i2
5
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Ma
lwar
e
is
a
m
alicio
u
s
p
r
o
g
r
am
th
at
h
ar
m
s
a
n
d
in
v
ad
e
s
th
e
co
m
p
u
tin
g
d
ev
ices
ty
p
ically
b
y
co
m
p
r
o
m
is
in
g
th
e
p
r
o
to
co
ls
wh
ich
i
n
clu
d
e
th
r
ea
ts
s
u
c
h
as
T
r
o
jan
h
o
r
s
es,
v
ir
u
s
es,
ad
war
e,
wo
r
m
s
,
r
an
s
o
m
war
e,
r
o
o
t
k
its
,
b
o
t
n
ets,
an
d
s
p
y
war
e
[
1
]
.
Ma
lwar
e
is
o
f
ten
u
s
ed
to
s
teal
in
f
o
r
m
atio
n
,
u
tili
ze
h
ar
d
wa
r
e
to
d
is
r
u
p
t
f
o
r
r
e
p
u
tatio
n
al
o
r
f
in
a
n
cial
g
ain
,
o
r
o
th
e
r
u
n
au
th
o
r
ize
d
ac
tiv
ity
.
Ma
lwar
e
d
ef
en
s
e
i
n
v
o
l
v
es
n
u
m
er
o
u
s
lay
er
s
,
in
clu
d
in
g
al
ter
in
g
u
s
er
s
wh
en
a
s
y
s
tem
is
co
m
p
r
o
m
is
ed
,
p
r
e
v
en
tin
g
m
alwa
r
e
in
f
iltra
tio
n
,
an
d
r
em
o
v
in
g
m
alwa
r
e
f
r
o
m
in
f
ec
ted
s
y
s
tem
s
[
2
]
.
I
n
r
ec
en
t
y
ea
r
s
,
th
e
n
u
m
b
er
o
f
cy
b
er
c
r
im
es
an
d
b
r
ea
ch
es
h
av
e
s
ig
n
if
ican
tly
i
n
cr
ea
s
ed
,
e
s
p
ec
ially
in
W
in
d
o
ws an
d
in
te
r
n
et
o
f
t
h
in
g
s
(
I
o
T
)
en
v
ir
o
n
m
e
n
ts
[
3
]
.
L
o
k
iB
o
t
wh
ich
is
a
m
alicio
u
s
s
o
f
twar
e
is
d
esig
n
e
d
to
s
teal
cr
y
p
to
cu
r
r
en
c
y
wallets,
p
ass
wo
r
d
s
,
an
d
o
th
er
s
en
s
itiv
e
in
f
o
r
m
atio
n
b
y
p
lacin
g
m
alicio
u
s
co
d
e
in
to
f
iles
th
at
ar
e
clo
u
d
an
d
h
o
s
ted
-
b
ased
s
er
v
ices
[
4
]
.
Ma
lwar
e
d
etec
tio
n
ap
p
r
o
ac
h
e
s
ar
e
ty
p
ically
b
ased
o
n
d
y
n
a
m
ic
an
aly
s
is
o
r
s
tatic
an
aly
s
is
.
T
h
e
s
tatic
d
o
n
o
t
co
n
tain
th
e
ac
tu
al
e
x
ec
u
tio
n
o
f
a
p
r
o
g
r
am
,
w
h
ich
h
as
co
n
tr
o
l
f
lo
w
g
r
ap
h
s
a
n
d
o
p
co
d
e
s
eq
u
en
ce
s
[
5
]
.
B
u
t
th
e
d
y
n
am
ic
a
n
aly
s
is
ex
ec
u
tes
th
e
p
r
o
g
r
am
b
ased
o
n
v
i
r
tu
al
e
n
v
ir
o
n
m
en
ts
.
I
t
ass
is
ts
in
m
an
ag
in
g
wh
eth
er
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Ma
lw
a
r
e
d
etec
tio
n
u
s
in
g
co
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
etw
o
r
k
-
d
i str
a
teg
y
…
(
P
a
r
va
th
i S
a
th
en
a
h
a
lli Ja
ya
p
r
a
ka
s
h
)
141
ap
p
licatio
n
p
r
o
g
r
am
m
i
n
g
in
te
r
f
ac
e
(
API
)
is
b
ein
g
ca
lled
,
s
y
s
tem
ca
lls
ar
e
ex
ec
u
ted
,
in
s
tr
u
ctio
n
s
ar
e
tr
ac
ed
,
r
eg
is
tr
y
ch
an
g
es
o
cc
u
r
,
o
r
m
e
m
o
r
y
is
m
o
d
if
ied
[
6
]
.
H
o
wev
er
,
b
o
th
d
y
n
am
ic
an
d
s
tatic
an
aly
s
es
f
ailed
d
u
e
to
attac
k
er
s
h
av
in
g
d
eter
m
in
ed
v
ar
io
u
s
m
eth
o
d
s
f
o
r
e
v
ad
in
g
d
etec
tio
n
.
Fo
r
in
s
tan
ce
,
a
s
tatic
m
alwa
r
e
an
aly
s
is
ev
alu
ates
th
e
ex
ec
u
tab
le
f
ile
t
h
at
is
b
y
p
ass
ed
b
y
a
well
-
d
is
g
u
is
ed
m
alwa
r
e.
I
n
ad
d
itio
n
,
m
alwa
r
e
ca
n
b
y
p
ass
th
e
d
etec
tio
n
in
d
y
n
am
ic
an
aly
s
is
b
y
s
im
p
ly
alter
in
g
its
b
eh
av
io
r
.
All
th
ese
cir
cu
m
s
tan
ce
s
led
to
th
e
d
eter
m
in
atio
n
o
f
n
ew
e
f
f
ec
tiv
e
m
eth
o
d
s
f
o
r
m
alwa
r
e
d
etec
tio
n
[
7
]
.
T
o
s
o
lv
e
m
alwa
r
e
attac
k
s
,
a
s
ig
n
atu
r
e
-
b
ased
m
alwa
r
e
d
etec
ti
o
n
s
y
s
tem
a
p
p
licatio
n
,
lik
e
an
an
ti
-
v
ir
u
s
p
r
o
g
r
a
m
,
is
estab
lis
h
ed
b
ased
o
n
th
e
s
ig
n
atu
r
e
d
ata
b
ase
wh
ich
is
ex
tr
ac
ted
f
r
o
m
ea
r
ly
d
eter
m
in
ed
m
alwa
r
e
s
am
p
les
th
at
b
ec
o
m
e
p
o
p
u
lar
.
I
t
co
m
p
ar
es
f
ile
ch
ar
ac
te
r
is
tics
to
k
n
o
wn
s
ig
n
atu
r
es
an
d
ev
alu
ates
o
n
ly
if
a
m
atch
is
d
eter
m
in
ed
to
id
en
tif
y
an
d
p
r
e
v
en
t m
alwa
r
e
th
r
ea
ts
[
8
]
,
[
9
]
.
Mo
r
e
o
v
er
,
s
ig
n
at
u
r
e
-
b
ased
m
alwa
r
e
also
h
as
a
lim
itatio
n
,
wh
er
e
it
d
o
es
n
o
t
m
o
d
if
y
o
r
id
en
tify
n
ew
th
r
ea
ts
an
d
is
co
m
p
u
tatio
n
ally
in
ten
s
iv
e
b
ec
au
s
e
o
f
p
r
o
ce
s
s
in
g
a
h
u
g
e
am
o
u
n
t
o
f
d
ata.
A
m
alwa
r
e
f
am
ily
is
a
co
llectio
n
o
f
m
alwa
r
e
s
am
p
les
with
th
e
b
ase
o
f
a
s
am
e
co
d
e
[
1
0
]
,
[
1
1
]
.
Acc
o
r
d
i
n
g
ly
,
ea
c
h
m
alwa
r
e
f
am
ily
co
n
tain
s
its
v
is
u
al
s
im
ilar
ities
an
d
p
r
o
p
er
ties
,
wh
ic
h
ar
e
d
if
f
e
r
en
t
f
r
o
m
o
th
er
m
alwa
r
e
f
am
ilies
.
A
m
alwa
r
e
wr
ite
r
tr
ies
to
co
n
tam
in
ate
th
e
tr
ain
in
g
d
ata;
h
en
ce
,
t
h
e
r
esu
ltin
g
ap
p
r
o
ac
h
is
less
ef
f
icien
t.
A
p
o
s
s
ib
le
m
eth
o
d
f
o
r
s
u
ch
an
ad
v
er
s
ar
ial
attac
k
o
n
an
im
ag
e
-
b
ased
m
alwa
r
e
d
etec
tio
n
is
f
o
r
g
en
e
r
atin
g
“d
ee
p
f
ak
e”
im
a
g
es
to
p
o
llu
te
th
e
tr
ain
in
g
d
ata.
T
h
er
ef
o
r
e,
d
ee
p
lear
n
i
n
g
(
DL
)
p
r
o
v
i
d
es
n
u
m
er
o
u
s
b
en
ef
its
o
v
er
co
n
v
en
tio
n
al
m
ac
h
in
e
lear
n
in
g
(
ML
)
[
1
2
]
,
[
1
3
]
,
s
u
ch
as
au
to
m
atic
g
en
er
atio
n
o
f
h
ig
h
-
q
u
ality
f
ea
t
u
r
es a
n
d
th
e
ab
ilit
y
to
m
a
n
ag
e
lar
g
e
d
ata
ef
f
ec
ti
v
ely
[
1
4
]
.
W
an
g
et
a
l
.
[
1
5
]
p
r
esen
ted
a
m
ask
ed
s
elf
-
s
u
p
er
v
is
ed
m
o
d
el
with
s
win
tr
an
s
f
o
r
m
er
(
M
alSo
r
t)
to
class
if
y
m
alwa
r
e
ef
f
ec
tiv
el
y
.
I
n
itially
,
ea
ch
m
alwa
r
e
i
n
s
tan
ce
was
co
n
v
er
ted
in
to
co
lo
r
im
ag
e,
an
d
t
h
en
s
win
tr
an
s
f
o
r
m
er
was
ap
p
lied
to
ex
tr
ac
t
th
e
m
u
lti
-
s
ca
le
k
ey
f
ea
tu
r
e
v
ec
to
r
s
.
At
last
,
a
n
e
n
co
d
er
was
f
in
e
-
tu
n
ed
t
o
p
er
f
o
r
m
m
alwa
r
e
class
if
icatio
n
ef
f
icien
tly
u
s
in
g
Ma
lSo
r
t.
H
o
wev
er
,
th
e
Ma
lSo
r
t
s
tr
u
g
g
le
d
with
f
in
e
-
g
r
ain
ed
m
alwa
r
e
class
if
icatio
n
as
it
r
elied
o
n
th
e
s
elf
-
s
u
p
er
v
is
ed
lear
n
in
g
th
at
o
v
er
lo
o
k
ed
s
u
b
tle
v
ar
iatio
n
s
in
m
alwa
r
e
p
atter
n
s
.
L
iu
et
a
l
.
[
1
6
]
estab
lis
h
ed
a
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
f
o
r
m
u
lti
-
class
m
alwa
r
e
d
etec
tio
n
.
T
h
e
b
alan
ce
d
s
am
p
lin
g
an
d
au
g
m
en
tatio
n
m
eth
o
d
s
wer
e
em
p
lo
y
ed
t
o
s
o
lv
e
th
e
im
b
alan
ce
is
s
u
e,
wh
ich
en
s
u
r
e
d
g
e
n
er
aliza
tio
n
ab
ilit
y
.
T
h
en
,
th
e
g
r
a
y
lev
el
c
o
-
o
cc
u
r
r
en
ce
m
atr
ix
(
GL
C
M)
was
u
s
ed
to
ex
t
r
ac
t
th
e
f
ea
tu
r
es
f
r
o
m
m
alicio
u
s
c
o
d
e
f
ea
tu
r
e
m
ap
s
.
T
h
e
o
u
tco
m
es
f
o
cu
s
ed
o
n
th
e
m
o
d
el
ac
cu
r
ac
y
f
o
r
ea
ch
d
ata
r
ec
o
g
n
itio
n
an
d
also
c
o
n
s
id
er
e
d
th
e
r
ec
o
g
n
itio
n
e
f
f
ec
t
o
f
ea
c
h
m
alwa
r
e
class
es.
Nev
er
th
eless
,
C
N
N
s
tr
u
g
g
led
to
d
etec
t
m
alwa
r
e
ef
f
ec
tiv
ely
b
ec
au
s
e
o
f
its
lim
ited
ab
ilit
y
t
o
ca
p
tu
r
e
f
in
e
-
g
r
ain
ed
d
if
f
er
e
n
ce
s
am
o
n
g
s
u
b
tle
v
ar
ian
ts
o
f
m
alwa
r
e.
Mo
s
l
eh
a
n
d
S
h
a
r
i
f
i
an
[
1
7
]
s
u
g
g
est
ed
a
h
ie
r
a
r
c
h
i
ca
l
cl
o
u
d
d
ee
p
n
eu
r
al
n
etw
o
r
k
(
D
NN)
f
o
r
m
alw
ar
e
class
i
f
i
ca
t
io
n
i
n
I
o
T
.
T
h
e
s
u
g
g
est
e
d
m
o
d
el
was
e
f
f
ec
ti
v
el
y
s
ca
le
d
f
r
o
m
I
o
T
d
e
v
i
ce
s
t
o
clo
u
d
a
n
d
e
d
g
e
f
o
r
en
h
a
n
ci
n
g
t
h
e
m
a
lwa
r
e
d
e
tec
t
io
n
a
n
d
s
i
m
u
lt
a
n
e
o
u
s
ly
m
a
n
a
g
i
n
g
th
e
ac
c
u
r
ac
y
le
v
e
l
a
n
d
m
i
n
im
izi
n
g
r
es
o
u
r
c
e
u
ti
liz
ati
o
n
.
T
h
e
h
i
er
ar
c
h
ic
al
c
lo
u
d
DNN
e
f
f
ec
t
iv
el
y
r
e
d
u
ce
d
b
o
th
r
eso
u
r
c
e
c
o
n
s
u
m
p
ti
o
n
an
d
r
u
n
-
t
im
e
als
o
m
a
n
a
g
e
d
t
h
e
p
e
r
f
o
r
m
a
n
c
e
ch
ar
ac
t
er
is
tics
.
H
o
we
v
er
,
h
ie
r
a
r
ch
i
ca
l
cl
o
u
d
DN
N
f
a
ce
d
c
h
a
l
len
g
es
in
s
c
al
ab
ilit
y
is
s
u
es
b
e
ca
u
s
e
o
f
m
u
lti
-
l
ay
er
ed
p
r
o
ce
s
s
i
n
g
ac
r
o
s
s
cl
o
u
d
n
o
d
es
t
h
a
t
le
d
to
d
e
la
y
e
d
th
r
ea
t
d
et
ec
ti
o
n
.
Sh
a
u
k
at
et
a
l
.
[
1
8
]
d
ev
el
o
p
e
d
a
h
y
b
r
i
d
m
o
d
el
b
y
i
n
t
eg
r
at
in
g
ML
a
n
d
d
ee
p
t
r
a
n
s
f
e
r
le
ar
n
i
n
g
f
o
r
m
al
wa
r
e
d
et
ec
ti
o
n
.
I
n
iti
all
y
,
th
e
d
ee
p
t
r
an
s
f
e
r
l
ea
r
n
in
g
m
et
h
o
d
was
e
m
p
lo
y
ed
f
o
r
ex
tr
ac
ti
n
g
ea
ch
d
ee
p
f
ea
t
u
r
e
f
r
o
m
t
h
e
f
u
ll
y
co
n
n
e
cte
d
la
y
er
o
f
t
h
e
D
L
m
o
d
el.
T
h
en
,
t
h
e
ML
w
as
u
t
iliz
ed
as
a
f
i
n
al
d
et
ec
t
o
r
,
w
h
i
c
h
f
u
ll
y
e
m
p
lo
y
ed
t
h
e
in
h
e
r
e
n
t
ass
o
cia
ti
o
n
s
am
o
n
g
i
n
p
u
t
a
n
d
o
u
t
p
u
t
.
T
h
e
d
e
v
el
o
p
e
d
a
p
p
r
o
ac
h
e
lim
in
ate
d
t
h
e
n
ec
es
s
ity
f
o
r
k
n
o
wle
d
g
e
f
r
o
m
d
o
m
ai
n
ex
p
e
r
ts
f
o
r
in
v
e
r
s
e
e
n
g
i
n
e
er
i
n
g
p
e
r
f
o
r
m
a
n
c
e.
T
h
e
d
e
v
el
o
p
e
d
h
y
b
r
i
d
m
o
d
e
l
was
c
o
s
t
-
e
f
f
ec
ti
v
e
,
s
ca
l
ab
le
,
a
n
d
e
f
f
i
ci
en
t
i
n
m
al
war
e
d
et
ec
t
io
n
.
Ne
v
e
r
t
h
eless
,
t
h
e
d
ev
el
o
p
e
d
a
p
p
r
o
a
c
h
s
u
f
f
e
r
e
d
f
r
o
m
f
ea
t
u
r
e
r
e
d
u
n
d
an
cy
b
e
ca
u
s
e
o
f
o
v
er
la
p
p
i
n
g
ex
tr
ac
t
ed
f
e
at
u
r
es,
w
h
ic
h
r
es
u
lt
e
d
i
n
m
in
im
i
ze
d
e
f
f
ec
t
iv
en
ess
o
f
t
h
e
m
o
d
el
.
Sin
g
h
et
a
l
.
[
1
9
]
in
tr
o
d
u
ce
d
a
m
u
lti
-
le
v
el
f
ea
t
u
r
e
e
x
t
r
a
cti
o
n
t
o
ca
te
g
o
r
i
ze
m
alw
a
r
e
f
am
ili
es
.
I
n
iti
all
y
,
s
ig
n
i
f
ic
a
n
t
f
ea
t
u
r
es
f
r
o
m
m
al
war
e
im
ag
es
we
r
e
e
x
t
r
a
ct
ed
b
y
u
s
i
n
g
g
ate
d
r
ec
u
r
r
e
n
t
u
n
i
t
(
GR
U)
,
w
h
i
c
h
we
r
e
p
ass
ed
t
h
r
o
u
g
h
C
NN
f
o
r
th
e
f
i
n
al
f
e
at
u
r
e
v
ec
t
o
r
e
x
t
r
a
cti
o
n
.
At
last
,
n
u
m
e
r
o
u
s
m
al
wa
r
e
f
a
m
ili
es
we
r
e
class
i
f
i
e
d
b
y
e
m
p
l
o
y
i
n
g
c
o
s
t
-
s
e
n
s
it
iv
e
b
o
o
t
s
t
r
ap
p
e
d
we
ig
h
t
e
d
r
a
n
d
o
m
f
o
r
est
(
C
SB
W
-
R
F
)
,
w
h
ic
h
en
h
a
n
ce
d
t
h
e
m
o
d
el
p
e
r
f
o
r
m
a
n
c
e.
H
o
w
ev
er
,
t
h
e
m
u
lt
i
-
l
e
v
el
f
e
at
u
r
e
e
x
t
r
a
cti
o
n
s
tr
u
g
g
le
d
i
n
ca
p
t
u
r
i
n
g
s
u
b
tl
e
a
n
d
e
v
o
l
v
i
n
g
f
e
at
u
r
es
ac
r
o
s
s
n
ew
m
a
lwa
r
e
v
ar
ia
n
ts
t
h
at
r
es
u
lt
ed
i
n
o
v
e
r
f
i
tti
n
g
o
r
p
o
o
r
g
en
er
ali
za
ti
o
n
.
K
im
et
a
l
.
[
2
0
]
est
ab
lis
h
e
d
a
cr
o
s
s
-
m
o
d
al
at
te
n
ti
o
n
m
e
ch
an
is
m
w
it
h
C
NN
t
o
c
lass
i
f
y
t
h
e
m
alw
ar
e
ef
f
e
cti
v
e
ly
.
T
h
e
m
alwa
r
e
i
m
a
g
es
a
n
d
s
tr
u
ct
u
r
al
e
n
t
r
o
p
i
es
we
r
e
th
e
two
m
o
d
ali
ties
w
h
i
c
h
we
r
e
t
r
an
s
f
o
r
m
e
d
a
n
d
e
x
t
r
a
cte
d
f
r
o
m
b
in
ar
y
f
il
es.
B
o
t
h
m
o
d
alit
ies
co
n
t
ai
n
e
d
d
i
v
e
r
s
e
g
r
a
n
u
la
r
it
ies
o
f
ch
u
n
k
s
a
n
d
b
y
t
es
w
h
ic
h
c
o
m
p
l
em
en
te
d
e
ac
h
o
t
h
e
r
.
T
o
i
n
t
e
g
r
ate
two
m
o
d
ali
ties
’
f
e
at
u
r
es,
c
r
o
s
s
-
m
o
d
al
at
te
n
ti
o
n
m
o
d
e
l
wa
s
u
s
e
d
w
h
i
c
h
e
n
h
a
n
c
ed
t
h
e
m
o
d
el
p
e
r
f
o
r
m
a
n
ce
.
T
a
b
le
1
s
h
o
ws
t
h
e
d
esc
r
i
p
ti
o
n
o
f
l
ite
r
a
tu
r
e
s
u
r
v
ey
wi
th
s
t
r
e
n
g
th
s
,
l
im
itat
i
o
n
s
,
a
n
d
m
et
r
i
cs.
Ho
wev
er
,
c
o
n
v
e
n
tio
n
al
m
alwa
r
e
d
etec
tio
n
s
tr
u
g
g
le
d
in
id
en
tify
in
g
n
ew
an
d
ev
o
l
v
in
g
m
alwa
r
e
v
ar
ian
ts
ac
cu
r
ately
d
u
e
to
it
s
d
ep
en
d
e
n
ce
o
n
h
a
n
d
cr
af
te
d
f
ea
tu
r
es
a
n
d
s
tatic
-
s
ig
n
atu
r
e
b
ased
m
et
h
o
d
s
.
Mo
r
eo
v
er
,
m
alwa
r
e
d
etec
tio
n
is
an
ar
tific
ial
in
tellig
en
ce
(
AI
)
-
d
r
iv
e
n
p
atter
n
r
ec
o
g
n
itio
n
an
d
g
en
er
aliza
tio
n
task
wh
er
e
m
o
d
els
ar
e
n
ec
ess
itated
f
o
r
lear
n
in
g
d
is
cr
im
in
ati
v
e
f
ea
tu
r
es
t
h
at
d
is
tin
g
u
is
h
ed
m
alicio
u
s
b
eh
av
i
o
r
f
r
o
m
b
en
ig
n
.
T
o
s
o
lv
e
th
is
li
m
itatio
n
,
C
NN
b
ased
d
i
s
tr
ate
g
y
p
o
lar
f
o
x
o
p
tim
izatio
n
alg
o
r
ith
m
(
DSPF
OA)
is
p
r
o
p
o
s
ed
to
d
etec
t
th
e
m
alwa
r
e
ef
f
ec
tiv
ely
b
y
r
e
m
o
v
in
g
th
e
d
ep
e
n
d
en
ce
o
f
h
an
d
cr
af
ted
f
ea
tu
r
es
an
d
s
tatic
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
140
-
1
5
3
142
s
ig
n
atu
r
es.
C
NN
em
p
lo
y
ed
a
d
ap
tiv
e
in
tellig
en
ce
m
ec
h
an
i
s
m
f
o
r
a
n
au
to
m
atic
ex
tr
ac
ti
o
n
o
f
h
ier
a
r
ch
ical
f
ea
tu
r
es
an
d
th
e
n
f
in
e
-
tu
n
in
g
th
e
p
ar
am
ete
r
s
u
s
in
g
DSPF
O
A
wh
ich
e
n
s
u
r
ed
b
etter
a
d
ap
t
ab
ilit
y
to
n
ew
an
d
ev
o
lv
in
g
v
a
r
ian
ts
.
T
h
e
n
o
v
el
ty
o
f
th
is
r
esear
c
h
is
C
NN’
s
f
ea
tu
r
e
lea
r
n
in
g
with
a
n
a
d
ap
tiv
e
a
b
ilit
y
o
f
DSPF
O
A
wh
ich
en
ab
les
ac
cu
r
ate
m
alwa
r
e
d
etec
tio
n
.
Ov
er
all,
b
y
in
teg
r
atin
g
DL
with
m
etah
eu
r
is
tic
o
p
tim
izatio
n
,
s
y
s
tem
g
e
n
er
ali
ze
s
ac
r
o
s
s
d
if
f
er
en
t
m
alwa
r
e
f
am
ilies
th
at
s
o
lv
es
th
e
d
y
n
am
i
c
n
atu
r
e
o
f
m
o
d
er
n
cy
b
er
th
r
ea
ts
.
T
h
e
m
ajo
r
c
o
n
tr
ib
u
tio
n
o
f
th
is
r
esear
ch
is
in
tr
ad
itio
n
al
p
o
lar
f
o
x
o
p
ti
m
izatio
n
alg
o
r
ith
m
(
PF
OA
)
,
th
e
s
in
e
ch
ao
tic
m
ap
p
in
g
a
n
d
C
au
ch
y
b
eh
a
v
io
r
m
u
tatio
n
wer
e
u
s
ed
as
d
i
-
s
tr
ateg
ies
f
o
r
in
itial
p
o
p
u
latio
n
an
d
s
o
lv
ed
lo
ca
l
o
p
tim
a
to
f
in
e
-
tu
n
e
th
e
C
NN
p
ar
am
eter
’
s
v
alu
es
e
f
f
ec
tiv
ely
.
C
NN’
s
co
n
v
o
l
u
tio
n
al
lay
e
r
m
in
im
ize
d
th
e
n
u
m
b
er
o
f
p
ar
am
eter
s
,
wh
ich
r
esu
lted
i
n
r
a
p
id
p
r
o
ce
s
s
in
g
an
d
m
e
m
o
r
y
u
s
ag
e
th
at
was a
p
p
r
o
p
r
iate
f
o
r
d
etec
tin
g
m
alwa
r
e
s
y
s
tem
s
.
Min
-
m
ax
n
o
r
m
aliza
tio
n
was u
s
ed
to
en
s
u
r
e
th
at
all
f
ea
tu
r
es a
r
e
s
ca
led
with
a
co
n
s
is
ten
t r
an
g
e
th
at
im
p
r
o
v
ed
m
o
d
el
ac
cu
r
ac
y
a
n
d
e
f
f
ec
tiv
en
ess
d
u
r
in
g
tr
ain
in
g
.
T
h
is
r
esear
ch
p
ap
e
r
is
o
r
g
a
n
iz
ed
as
f
o
llo
ws:
s
ec
tio
n
2
s
h
o
w
s
a
d
etailed
ex
p
lan
atio
n
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
Sectio
n
3
p
r
esen
ts
C
NN
-
DSPF
OA
.
Sectio
n
4
illu
s
tr
ates
r
esu
lts
an
d
d
is
cu
s
s
io
n
.
Fin
ally
,
s
ec
tio
n
5
d
em
o
n
s
tr
ates th
e
co
n
cl
u
s
io
n
o
f
th
e
p
ap
e
r
.
T
ab
le
1
.
Su
m
m
a
r
y
o
f
ex
is
tin
g
m
alwa
r
e
d
etec
tio
n
m
eth
o
d
s
with
s
tr
en
g
th
,
lim
itatio
n
s
,
an
d
ev
alu
atio
n
m
etr
ics
A
u
t
h
o
r
M
e
t
h
o
d
s
S
t
r
e
n
g
t
h
Li
mi
t
a
t
i
o
n
M
e
t
r
i
c
s
W
a
n
g
e
t
a
l
.
[
1
5
]
M
a
l
S
o
r
t
M
a
l
S
o
r
t
e
m
p
l
o
y
s
p
r
e
-
t
r
a
i
n
e
d
e
n
c
o
d
e
r
f
o
r
f
i
n
e
-
t
u
n
i
n
g
w
h
i
c
h
p
e
r
f
o
r
m
mal
w
a
r
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
e
f
f
i
c
i
e
n
t
l
y
M
a
l
S
o
r
t
st
r
u
g
g
l
e
d
w
i
t
h
f
i
n
e
-
g
r
a
i
n
e
d
mal
w
a
r
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
b
e
c
a
u
se
i
t
r
e
l
i
e
d
o
n
se
l
f
-
su
p
e
r
v
i
se
d
l
e
a
r
n
i
n
g
t
h
a
t
o
v
e
r
l
o
o
k
e
d
s
u
b
t
l
e
v
a
r
i
a
t
i
o
n
s i
n
mal
w
a
r
e
p
a
t
t
e
r
n
s
A
c
c
u
r
a
c
y
,
r
e
c
a
l
l
,
F1
-
sc
o
r
e
,
p
r
e
c
i
si
o
n
Li
u
e
t
a
l
.
[
1
6
]
C
o
n
v
N
e
t
Th
e
o
u
t
c
o
mes
n
o
t
o
n
l
y
f
o
c
u
se
d
o
n
mo
d
e
l
a
c
c
u
r
a
c
y
f
o
r
e
a
c
h
d
a
t
a
r
e
c
o
g
n
i
t
i
o
n
b
u
t
a
l
so
c
o
n
si
d
e
r
e
d
t
h
e
r
e
c
o
g
n
i
t
i
o
n
e
f
f
e
c
t
o
f
a
l
l
ma
l
w
a
r
e
c
l
a
ss
e
s.
C
N
N
st
r
u
g
g
l
e
d
w
i
t
h
d
e
t
e
c
t
i
n
g
ma
l
w
a
r
e
e
f
f
e
c
t
i
v
e
l
y
b
e
c
a
u
se
o
f
i
t
s
l
i
mi
t
e
d
a
b
i
l
i
t
y
t
o
c
a
p
t
u
r
e
f
i
n
e
-
g
r
a
i
n
e
d
d
i
f
f
e
r
e
n
c
e
s
a
mo
n
g
s
u
b
t
l
e
v
a
r
i
a
n
t
s
o
f
m
a
l
w
a
r
e
.
A
c
c
u
r
a
c
y
M
o
s
l
e
h
a
n
d
S
h
a
r
i
f
i
a
n
[
1
7
]
H
i
e
r
a
r
c
h
i
c
a
l
C
l
o
u
d
D
N
N
Th
i
s
me
t
h
o
d
e
f
f
e
c
t
i
v
e
l
y
r
e
d
u
c
e
s
b
o
t
h
r
e
so
u
r
c
e
c
o
n
s
u
m
p
t
i
o
n
a
n
d
r
u
n
-
t
i
m
e
w
h
i
l
e
m
a
n
a
g
i
n
g
p
e
r
f
o
r
ma
n
c
e
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
H
i
e
r
a
r
c
h
i
c
a
l
c
l
o
u
d
D
N
N
f
a
c
e
c
h
a
l
l
e
n
g
e
s
i
n
sca
l
a
b
i
l
i
t
y
i
ss
u
e
s b
e
c
a
u
s
e
o
f
m
u
l
t
i
-
l
a
y
e
r
e
d
p
r
o
c
e
ss
i
n
g
a
c
r
o
ss c
l
o
u
d
n
o
d
e
s
t
h
a
t
l
e
a
d
s
t
o
d
e
l
a
y
e
d
t
h
r
e
a
t
d
e
t
e
c
t
i
o
n
A
c
c
u
r
a
c
y
,
r
e
c
a
l
l
,
F1
-
sc
o
r
e
,
p
r
e
c
i
si
o
n
S
h
a
u
k
a
t
e
t
a
l
.
[
1
8
]
H
y
b
r
i
d
mo
d
e
l
H
y
b
r
i
d
mo
d
e
l
w
a
s
c
o
st
-
e
f
f
e
c
t
i
v
e
,
sca
l
a
b
l
e
,
a
n
d
e
f
f
i
c
i
e
n
t
i
n
d
e
t
e
c
t
i
n
g
t
h
e
ma
l
w
a
r
e
.
H
y
b
r
i
d
mo
d
e
l
s
u
f
f
e
r
e
d
f
r
o
m
f
e
a
t
u
r
e
r
e
d
u
n
d
a
n
c
y
b
e
c
a
u
se
o
f
o
v
e
r
l
a
p
p
i
n
g
e
x
t
r
a
c
t
e
d
f
e
a
t
u
r
e
s,
w
h
i
c
h
r
e
s
u
l
t
e
d
i
n
mi
n
i
mi
z
e
d
m
o
d
e
l
e
f
f
e
c
t
i
v
e
n
e
ss.
A
c
c
u
r
a
c
y
,
r
e
c
a
l
l
,
F1
-
sc
o
r
e
,
p
r
e
c
i
si
o
n
S
i
n
g
h
e
t
a
l
.
[
1
9
]
G
R
U
+
C
N
N
+
RF
N
u
mer
o
u
s
ma
l
w
a
r
e
f
a
mi
l
i
e
s wer
e
c
l
a
ss
i
f
i
e
d
b
y
e
mp
l
o
y
i
n
g
C
S
B
W
-
R
F
w
h
i
c
h
e
n
h
a
n
c
e
s
t
h
e
m
o
d
e
l
p
e
r
f
o
r
m
a
n
c
e
.
mu
l
t
i
-
l
e
v
e
l
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
h
a
s
d
i
f
f
i
c
u
l
t
y
i
n
c
a
p
t
u
r
i
n
g
s
u
b
t
l
e
a
n
d
e
v
o
l
v
i
n
g
f
e
a
t
u
r
e
s
a
c
r
o
ss
n
e
w
m
a
l
w
a
r
e
v
a
r
i
a
n
t
s
t
h
a
t
r
e
s
u
l
t
i
n
o
v
e
r
f
i
t
t
i
n
g
o
r
p
o
o
r
g
e
n
e
r
a
l
i
z
a
t
i
o
n
A
c
c
u
r
a
c
y
,
r
e
c
a
l
l
,
F1
-
sc
o
r
e
,
p
r
e
c
i
si
o
n
K
i
m
e
t
a
l
.
[
2
0
]
A
t
t
e
n
t
i
o
n
-
b
a
s
e
d
c
r
o
s
s
mo
d
e
l
C
N
N
To
i
n
t
e
g
r
a
t
e
t
w
o
m
o
d
a
l
i
t
i
e
s’
f
e
a
t
u
r
e
s,
c
r
o
ss
-
mo
d
a
l
a
t
t
e
n
t
i
o
n
mo
d
e
l
w
a
s
u
se
d
w
h
i
c
h
e
n
h
a
n
c
e
mo
d
e
l
p
e
r
f
o
r
ma
n
c
e
Th
i
s
me
t
h
o
d
s
t
r
u
g
g
l
e
s
w
i
t
h
a
c
c
u
r
a
t
e
l
y
i
d
e
n
t
i
f
y
i
n
g
n
e
w
a
n
d
e
v
o
l
v
i
n
g
m
a
l
w
a
r
e
v
a
r
i
a
n
t
s
A
c
c
u
r
a
c
y
,
F1
-
sc
o
r
e
2.
M
E
T
H
O
D
T
h
is
r
esear
ch
p
r
o
p
o
s
es
C
NN
-
DSPF
O
A
f
o
r
an
ef
f
ec
tiv
e
m
a
lwar
e
d
etec
tio
n
.
I
n
itially
,
th
e
B
I
G2
0
1
5
an
d
Ma
lim
g
d
ataset
ar
e
p
e
r
f
o
r
m
ed
f
o
r
e
v
alu
atin
g
t
h
e
m
o
d
e
l
p
er
f
o
r
m
an
ce
.
A
co
lo
r
m
ap
m
o
d
u
le
is
ap
p
lied
to
co
n
v
er
t
b
in
a
r
y
f
iles
in
to
co
lo
r
f
u
l
im
ag
es.
T
h
en
,
m
in
-
m
a
x
n
o
r
m
aliza
tio
n
is
u
s
ed
to
p
r
e
-
p
r
o
ce
s
s
th
e
im
ag
es
an
d
th
en
C
NN
-
DSFOA
is
p
er
f
o
r
m
ed
f
o
r
m
alwa
r
e
d
etec
tio
n
b
y
f
in
e
-
tu
n
in
g
th
e
C
NN
p
ar
am
eter
v
alu
es.
Fig
u
r
e
1
d
em
o
n
s
tr
ates th
e
b
lo
c
k
d
iag
r
a
m
o
f
th
e
p
r
o
p
o
s
ed
C
NN
-
DSP
FOA.
Fig
u
r
e
1
.
B
lo
ck
d
iag
r
am
illu
s
t
r
atin
g
th
e
wo
r
k
f
lo
w
o
f
p
r
o
p
o
s
ed
C
NN
-
DSPF
OA
m
eth
o
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Ma
lw
a
r
e
d
etec
tio
n
u
s
in
g
co
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
etw
o
r
k
-
d
i str
a
teg
y
…
(
P
a
r
va
th
i S
a
th
en
a
h
a
lli Ja
ya
p
r
a
ka
s
h
)
143
2
.
1
.
Da
t
a
s
et
I
n
th
is
r
esear
c
h
,
th
e
B
I
G2
0
1
5
[
2
1
]
an
d
Ma
lim
g
[
2
2
]
d
ataset
ar
e
u
s
ed
to
d
eter
m
in
e
th
e
p
e
r
f
o
r
m
a
n
ce
o
f
th
e
m
o
d
el
in
m
alwa
r
e.
T
h
es
e
d
atasets
co
v
er
wid
e
v
ar
iety
o
f
m
alwa
r
e
f
am
ilies
,
wh
ich
en
h
an
ce
s
th
e
m
o
d
el’
s
ab
ilit
y
to
g
en
er
alize
ac
r
o
s
s
d
if
f
er
en
t ty
p
es.
A
d
etailed
ex
p
lan
atio
n
o
f
th
is
d
ataset
is
d
escr
ib
ed
as f
o
llo
ws:
i)
B
I
G2
0
1
5
:
it
was
p
u
b
lis
h
ed
b
y
Mic
r
o
s
o
f
t
in
2
0
1
5
with
9
m
alwa
r
e
f
am
ilies
an
d
1
0
,
8
6
8
m
alwa
r
e
f
iles
.
T
h
e
f
iles
in
th
is
d
atase
t
ar
e
t
r
an
s
f
o
r
m
ed
in
t
o
ar
r
ay
s
an
d
p
lace
d
in
to
m
em
o
r
y
f
o
r
g
en
e
r
atin
g
co
lo
r
ed
m
alwa
r
e
im
ag
es.
B
y
co
n
v
e
r
t
in
g
m
alwa
r
e
d
ata
i
n
to
s
tr
u
ct
u
r
ed
f
o
r
m
at
en
a
b
les
to
d
eter
m
in
e
m
alwa
r
e
p
atter
n
s
ag
ain
s
t th
r
ea
ts
.
ii)
Ma
lim
g
:
it
h
as
v
ar
io
u
s
im
b
alan
ce
d
W
in
d
o
ws
p
o
r
tab
le
ex
e
cu
tab
le
(
PE)
im
ag
es
wh
ic
h
a
r
e
ad
o
p
te
d
t
o
class
if
y
m
alwa
r
e
f
am
ily
.
I
t
co
n
tain
s
9
,
3
3
9
m
alwa
r
e
s
am
p
les
th
at
b
elo
n
g
to
2
5
m
alwa
r
e
f
a
m
ilies
.
A
p
ix
el
v
alu
e
o
f
g
r
ay
s
ca
le
im
ag
e
is
ev
alu
ated
to
all
th
r
ee
R
GB
ch
an
n
els
o
f
a
co
lo
r
f
u
l
im
ag
e
f
o
r
g
e
n
er
atin
g
m
alwa
r
e
co
lo
r
f
u
l
im
ag
e.
T
h
e
n
th
e
d
ataset
is
s
p
lit
in
to
8
0
%
tr
ain
in
g
an
d
2
0
%
test
in
g
,
r
esp
ec
tiv
ely
an
d
th
ese
o
b
tain
ed
co
d
es a
r
e
f
e
d
i
n
to
co
lo
r
-
m
ap
m
o
d
u
le
to
g
e
n
e
r
ate
co
lo
r
f
u
l im
ag
e.
2
.
2
.
Co
l
o
r
-
ma
p mo
du
le
A
f
t
e
r
o
b
t
a
in
i
n
g
t
h
e
c
o
d
e
,
t
h
e
c
o
l
o
r
-
m
a
p
m
o
d
u
l
e
[
1
5
]
i
s
u
s
e
d
to
c
o
n
v
e
r
t
b
i
n
ar
y
f
i
l
e
s
c
o
n
t
a
in
i
n
g
m
a
l
w
a
r
e
in
t
o
c
o
lo
r
f
u
l
i
m
ag
es
,
w
h
i
c
h
i
s
e
s
s
e
n
t
ia
l
d
u
e
t
o
i
t
s
a
b
i
l
i
t
y
en
h
an
c
em
e
n
t
in
D
L
t
o
d
i
s
t
in
g
u
i
s
h
b
e
t
w
e
en
p
a
t
t
e
r
n
s
a
n
d
s
u
b
t
l
e
f
e
a
t
u
r
e
s
.
An
im
a
g
e
-
b
a
s
e
d
m
a
lw
a
r
e
r
ep
r
e
s
e
n
t
a
t
i
o
n
m
a
k
e
s
C
N
N
t
o
c
ap
t
u
r
e
b
y
te
-
l
e
v
e
l
a
n
d
s
t
r
u
c
tu
r
a
l
p
a
t
t
er
n
s
t
h
a
t
ar
e
c
o
n
s
i
s
t
e
n
t
i
n
m
al
w
a
r
e
f
a
m
i
l
i
e
s
an
d
en
h
a
n
c
e
t
h
e
c
la
s
s
i
f
i
c
a
t
i
o
n
a
c
c
u
r
a
c
y
.
T
h
i
s
m
e
th
o
d
em
p
lo
y
s
v
i
s
u
a
l
s
i
m
i
la
r
i
t
i
e
s
o
v
er
m
a
l
w
a
r
e
v
ar
i
a
n
t
s
wh
i
ch
m
ak
e
e
f
f
e
c
t
i
v
e
f
ea
t
u
r
e
e
x
t
r
a
c
t
io
n
a
n
d
c
la
s
s
i
f
i
c
a
t
i
o
n
.
C
o
l
o
r
m
ap
s
i
n
tr
o
d
u
c
e
d
im
e
n
s
i
o
n
,
wh
i
ch
m
ak
e
s
i
t
e
a
s
i
e
r
to
d
if
f
er
e
n
t
ia
t
e
an
d
h
i
g
h
l
i
g
h
t
th
e
m
a
l
w
a
r
e
f
a
m
i
l
ie
s
b
y
v
a
r
y
i
n
g
in
t
e
n
s
i
t
i
e
s
a
c
r
o
s
s
R
G
B
c
h
an
n
e
l
s
.
T
h
e
g
e
n
e
r
a
t
i
o
n
o
f
c
o
lo
r
f
u
l
m
a
p
p
i
n
g
i
s
a
t
t
a
in
e
d
b
y
u
s
i
n
g
b
in
a
r
y
f
i
l
e
s
o
r
g
r
ay
s
c
a
l
e
im
a
g
e
s
th
a
t
i
s
r
e
l
a
t
ed
to
m
a
lw
a
r
e
.
I
n
i
t
ia
l
l
y
,
t
h
e
m
a
l
w
a
r
e
b
in
a
r
y
f
i
l
e
s
a
r
e
co
n
v
er
t
e
d
i
n
to
b
y
t
e
ar
r
ay
s
to
p
r
o
d
u
c
e
a
co
l
o
r
f
u
l
im
a
g
e
an
d
th
e
n
s
t
o
r
e
d
i
n
a
m
e
m
o
r
y
b
u
f
f
er
.
T
h
i
s
p
r
o
c
e
d
u
r
e
i
s
u
s
e
d
f
o
r
t
r
an
s
f
o
r
m
in
g
B
I
G
2
0
1
5
d
a
t
a
s
e
t
in
t
o
c
o
lo
r
f
u
l
i
m
ag
e
s
w
h
er
e
as
M
a
l
i
m
g
d
a
t
a
s
e
t
h
a
s
g
r
ay
s
c
al
e
i
m
a
g
e
s
r
e
s
u
l
ta
n
t
f
r
o
m
b
in
a
r
y
f
i
l
e
s
.
A
co
l
o
r
f
u
l
i
m
ag
e
i
s
e
s
t
a
b
l
i
s
h
ed
b
y
a
s
s
i
g
n
i
n
g
a
g
r
ay
s
c
a
l
e
im
a
g
e
’
s
p
i
x
e
l
v
a
lu
e
t
o
e
a
ch
ch
a
n
n
e
l
o
f
R
G
B
.
F
i
g
u
r
e
2
s
h
o
w
s
th
e
p
r
o
c
e
s
s
o
f
c
o
lo
r
f
u
l
i
m
a
g
e
a
n
d
th
i
s
r
ep
r
e
s
e
n
t
a
t
io
n
a
s
s
i
s
t
s
i
n
v
i
s
u
a
l
an
a
ly
s
i
s
a
n
d
p
r
o
v
i
d
e
s
c
l
e
a
r
e
r
in
s
ig
h
t
s
in
t
o
m
a
l
w
ar
e
s
t
r
u
c
t
u
r
e
.
T
h
en
,
m
in
-
m
a
x
n
o
r
m
a
l
iz
a
t
i
o
n
i
s
u
t
i
l
i
z
ed
f
o
r
n
o
r
m
a
l
i
z
in
g
i
m
ag
e’
s
p
i
x
e
l
v
a
l
u
e
to
e
n
s
u
r
e
c
o
n
s
i
s
t
e
n
c
y
a
n
d
i
m
p
r
o
v
e
t
h
e
tr
a
i
n
in
g
p
r
o
c
e
s
s
.
Fig
u
r
e
2
.
Vis
u
aliza
tio
n
o
f
co
lo
r
-
m
ap
m
o
d
u
le
u
tili
ze
d
in
m
al
war
e
im
ag
e
r
ep
r
esen
tatio
n
2
.
3
.
P
re
-
pro
ce
s
s
ing
Af
ter
th
e
co
lo
r
co
d
e
m
o
d
u
le,
th
e
m
in
-
m
ax
n
o
r
m
aliza
tio
n
s
c
ales
th
e
p
ix
el
v
alu
es
o
f
th
e
tr
an
s
f
o
r
m
ed
m
alwa
r
e
im
ag
es to
a
f
ix
e
d
r
an
g
e
[
0
,
1
]
.
T
h
is
ass
is
t
s
in
s
o
lv
in
g
th
e
is
s
u
es wh
ich
ar
e
ca
u
s
ed
b
y
a
v
a
r
y
in
g
i
m
ag
e
in
ten
s
ity
an
d
en
s
u
r
es
t
h
at
th
e
m
o
d
els
t
r
ea
t
ev
e
r
y
f
ea
tu
r
e
s
eq
u
ally
.
Min
-
m
ax
[
2
3
]
p
r
o
v
id
es
r
elatio
n
s
h
ip
s
am
o
n
g
o
r
i
g
in
al
d
ata
v
al
u
es
t
o
m
an
a
g
e
t
h
e
im
a
g
e’
s
o
r
ig
in
al
s
h
ap
e
wh
ic
h
m
i
n
im
ize
th
e
o
u
tlier
im
p
ac
t.
T
h
e
m
ath
em
atica
l f
o
r
m
u
la
f
o
r
m
in
-
m
ax
n
o
r
m
aliza
tio
n
is
ex
p
r
ess
ed
in
(
1
)
an
d
(
2
)
.
=
−
−
(
1
)
=
×
(
−
)
+
(
2
)
W
h
er
e
r
ep
r
esen
ts
th
e
m
in
im
u
m
f
ea
tu
r
e
r
an
g
e
o
f
a
n
in
p
u
t
d
a
ta
an
d
d
en
o
tes
th
e
m
ax
im
u
m
f
ea
tu
r
e
r
an
g
e
r
esp
ec
tiv
ely
.
Min
-
m
a
x
en
h
an
ce
th
e
tr
ain
in
g
p
r
o
ce
s
s
an
d
m
o
d
el’
s
s
tab
ilit
y
b
y
m
in
im
izin
g
th
e
s
en
s
itiv
ity
to
in
p
u
t
v
ar
iatio
n
s
.
T
h
e
p
r
e
-
p
r
o
ce
s
s
ed
im
ag
es a
r
e
th
en
p
ass
ed
th
r
o
u
g
h
t
h
e
C
NN
f
o
r
m
alwa
r
e
d
etec
tio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
140
-
1
5
3
144
3.
CNN
WI
T
H
DI
-
S
T
RA
T
E
G
Y
P
O
L
AR
F
O
X
O
P
T
I
M
I
Z
A
T
I
O
N
A
L
G
O
RIT
H
M
Af
ter
p
r
e
-
p
r
o
ce
s
s
in
g
,
C
NN
is
ap
p
lied
f
o
r
m
alwa
r
e
d
etec
tio
n
d
u
e
t
o
its
ab
ilit
y
to
ex
tr
ac
t
s
p
atial
an
d
h
ier
ar
ch
ical
p
atter
n
s
f
r
o
m
m
al
war
e
b
in
ar
ies
wh
ich
ar
e
r
ep
r
e
s
en
ted
as
im
ag
es.
C
NN
[
2
4
]
au
to
m
atica
lly
lear
n
f
ea
tu
r
es
with
o
u
t
th
e
r
eq
u
ir
em
en
t
o
f
m
an
u
al
f
ea
tu
r
e
e
n
g
in
e
er
in
g
th
at
en
h
an
ce
s
th
e
d
etec
tio
n
ac
cu
r
ac
y
.
T
h
e
co
n
v
o
l
u
tio
n
al
lay
er
o
f
C
NN
ef
f
ec
tiv
ely
ca
p
tu
r
es
th
e
s
tr
u
c
tu
r
al
s
im
ilar
ities
an
d
a
n
o
m
ali
es
with
in
m
alwa
r
e
f
am
ilies
,
wh
ich
m
a
k
es
r
o
b
u
s
t
class
if
icatio
n
.
A
d
etailed
ex
p
lan
atio
n
o
f
C
NN
is
p
r
o
v
id
ed
an
d
its
ar
c
h
itectu
r
e
f
o
r
m
alwa
r
e
d
etec
tio
n
an
d
class
if
icatio
n
p
r
o
ce
s
s
is
s
h
o
wn
in
Fig
u
r
e
3
.
i)
C
o
n
v
o
lu
tio
n
al
lay
er
:
c
o
n
v
o
lu
t
io
n
is
g
e
n
er
alize
d
in
t
o
n
u
m
er
o
u
s
d
im
en
s
io
n
s
w
h
er
e
t
h
e
f
ea
tu
r
e
m
atr
ix
o
f
th
e
f
ilter
,
an
d
im
ag
e
is
d
ef
in
ed
in
an
i
n
teg
er
s
et
with
two
in
p
u
t
d
im
en
s
io
n
s
(
,
)
an
d
r
ep
r
esen
ts
th
e
wid
th
an
d
h
eig
h
t
co
o
r
d
i
n
ates.
T
h
e
k
er
n
el
s
ize
r
ef
er
s
to
th
e
f
ilter
d
im
en
s
io
n
wh
ich
is
em
p
lo
y
ed
in
a
co
n
v
o
lu
tio
n
lay
er
f
o
r
ex
tr
ac
tin
g
lo
ca
l
f
ea
tu
r
es
f
r
o
m
p
r
e
-
p
r
o
ce
s
s
ed
in
p
u
t.
T
h
e
m
ath
em
atica
l
f
o
r
m
u
la
f
o
r
c
o
n
v
o
lu
tio
n
o
u
tp
u
t
is
ex
p
r
ess
ed
in
(
3
)
an
d
(
4
)
.
T
h
e
ze
r
o
p
a
d
d
in
g
s
ty
p
ically
f
i
t
m
ag
n
itu
d
e
a
n
d
h
en
ce
,
th
at
s
p
a
tial o
u
tp
u
t d
im
e
n
s
io
n
h
as a
s
im
ilar
s
p
atial
in
p
u
t size.
ℎ
(
,
)
=
∫
∫
(
,
)
(
−
,
−
)
∞
−
∞
∞
−
∞
(
3
)
ℎ
(
,
)
=
∑
∑
[
,
]
[
−
,
−
]
(
4
)
ii)
R
ec
tifie
d
lin
ea
r
u
n
it (
R
eL
U
)
ac
tiv
atio
n
: it
ass
is
t
s
in
s
o
lv
in
g
th
e
v
an
is
h
in
g
g
r
ad
ie
n
t iss
u
e
th
at
allo
ws d
ee
p
n
etwo
r
k
s
to
lear
n
ef
f
e
ctiv
ely
wh
ich
is
f
o
r
m
u
lated
in
(
5
)
.
I
f
th
e
in
p
u
t is less
th
an
ze
r
o
,
th
en
R
eL
U
o
b
tain
s
an
o
u
t
p
u
t
v
al
u
e
o
f
ze
r
o
;
h
o
we
v
er
,
if
o
th
er
th
an
ze
r
o
,
th
en
o
u
tp
u
t
s
h
o
ws
r
aw
d
ata.
I
f
th
e
d
ata
is
g
r
ea
ter
th
an
ze
r
o
,
th
en
th
e
p
r
o
d
u
ctio
n
is
s
im
ilar
as in
p
u
t.
(
ℎ
)
=
{
0
ℎ
<
0
ℎ
≥
0
(
5
)
iii)
Po
o
lin
g
lay
er
:
it
is
u
s
ed
to
en
s
u
r
e
th
e
o
u
tp
u
t
v
ar
ian
ce
with
th
e
in
teg
r
atio
n
o
f
th
e
co
n
v
o
l
u
tio
n
al
lay
er
.
Af
ter
th
e
p
o
o
lin
g
lay
er
,
th
e
n
etwo
r
k
o
u
tp
u
t
is
d
en
o
ted
u
s
in
g
(
6
)
.
T
h
e
r
ep
r
esen
ts
th
e
len
g
th
o
r
th
e
h
eig
h
t
o
f
an
o
u
tp
u
t
an
d
in
d
icate
s
s
tr
id
e.
T
h
e
p
o
o
lin
g
p
r
o
c
ess
co
m
p
u
tes
a
s
tatis
tical
s
u
m
m
ar
y
o
f
th
e
n
ea
r
est d
ata
b
y
u
tili
zin
g
an
a
r
ith
m
etic
f
u
n
ctio
n
.
=
+
2
−
+
1
(
6
)
iv
)
Fu
lly
co
n
n
ec
ted
:
a
f
ter
co
n
v
o
l
u
tio
n
al
an
d
p
o
o
lin
g
lay
e
r
s
,
th
e
f
u
lly
c
o
n
n
ec
t
ed
lay
er
is
a
p
p
lied
,
wh
ich
co
n
tain
s
n
u
m
er
o
u
s
n
eu
r
o
n
s
wh
er
e
ea
ch
n
eu
r
o
n
is
ass
o
ciate
d
with
all
th
e
n
e
u
r
o
n
s
in
a
d
jace
n
t
lay
er
s
.
T
h
o
s
e
lay
er
s
at
th
e
en
d
o
f
t
h
e
n
etwo
r
k
ar
e
em
p
lo
y
ed
to
m
ak
e
d
etec
tio
n
a
n
d
g
en
e
r
ate
a
f
in
al
f
ea
tu
r
e
non
-
lin
ea
r
co
m
b
in
atio
n
.
T
h
e
c
ateg
o
r
ical
cr
o
s
s
en
tr
o
p
y
lo
s
s
f
u
n
ctio
n
is
u
s
ed
in
C
NN
m
eth
o
d
to
m
in
im
ize
er
r
o
r
.
T
h
e
tech
n
iq
u
es
u
tili
ze
d
to
m
in
im
ize
th
e
e
r
r
o
r
s
a
r
e
co
l
lectiv
ely
k
n
o
wn
as
r
eg
u
la
r
izatio
n
,
wh
ich
is
u
s
ed
as we
ig
h
t d
ec
ay
a
n
d
d
r
o
p
o
u
t,
r
esp
ec
tiv
ely
.
Fig
u
r
e
3
.
Ar
c
h
itectu
r
al
o
v
e
r
v
i
ew
o
f
d
esig
n
ed
C
NN
s
tr
u
ctu
r
e
f
o
r
m
alwa
r
e
d
etec
tio
n
an
d
cla
s
s
if
icatio
n
p
r
o
ce
s
s
3
.
1
.
H
y
perpa
ra
m
e
t
er
t
un
ing
I
n
C
NN,
h
y
p
er
p
ar
am
eter
t
u
n
i
n
g
is
ap
p
lied
u
s
in
g
DSPF
OA
,
wh
ich
is
ess
en
tial
to
o
p
tim
ize
m
o
d
el
p
er
f
o
r
m
an
ce
th
at
en
s
u
r
es
h
i
g
h
r
o
b
u
s
tn
ess
an
d
ac
cu
r
ac
y
a
g
ain
s
t
th
e
ev
o
lv
in
g
th
r
ea
ts
.
E
f
f
ec
tiv
e
tu
n
in
g
o
f
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Ma
lw
a
r
e
d
etec
tio
n
u
s
in
g
co
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
etw
o
r
k
-
d
i str
a
teg
y
…
(
P
a
r
va
th
i S
a
th
en
a
h
a
lli Ja
ya
p
r
a
ka
s
h
)
145
p
ar
am
eter
s
s
u
ch
as
ep
o
ch
s
,
le
ar
n
in
g
r
ate,
o
p
tim
izer
,
ac
tiv
ati
o
n
f
u
n
ctio
n
,
b
atc
h
s
ize,
an
d
d
r
o
p
o
u
t
en
h
an
ce
t
h
e
co
n
v
er
g
en
ce
s
p
ee
d
.
W
ith
o
u
t
tu
n
in
g
,
th
e
m
o
d
el
u
n
d
er
p
er
f
o
r
m
s
o
r
o
v
er
f
its
,
wh
ic
h
r
esu
lts
in
p
o
o
r
g
en
er
aliza
tio
n
o
n
n
ew
m
alwa
r
e
s
am
p
les.
Op
tim
ized
h
y
p
er
p
a
r
am
eter
s
en
h
an
ce
t
h
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
an
d
m
in
im
ize
f
alse p
o
s
itiv
e
(
FP
)
an
d
f
alse n
eg
ativ
e
(
FP
)
,
r
esp
ec
tiv
ely
,
in
d
etec
tin
g
th
e
m
al
war
e.
C
o
m
p
ar
ed
t
o
ex
is
tin
g
m
eth
o
d
s
lik
e
r
e
d
f
o
x
o
p
tim
izatio
n
(
R
FO)
,
f
en
n
ec
f
o
x
o
p
tim
izatio
n
(
FF
O)
,
a
n
d
ar
tic
f
o
x
o
p
tim
izatio
n
(
AFO)
,
th
e
PF
OA
p
r
o
v
id
es
s
tr
o
n
g
ad
a
p
tab
ilit
y
in
h
ar
s
h
en
v
ir
o
n
m
e
n
ts
,
wh
ich
en
s
u
r
es
an
ef
f
ec
tiv
e
b
alan
c
e
b
etwe
en
ex
p
lo
r
atio
n
an
d
ex
p
lo
itatio
n
.
I
ts
d
y
n
a
m
ic
m
o
v
e
m
en
t
s
tr
ateg
ies
m
ak
e
it
escap
e
lo
ca
l
o
p
tim
a
a
n
d
en
h
an
ce
g
l
o
b
al
s
ea
r
ch
a
b
ilit
ies.
I
ts
g
r
o
u
p
c
o
o
r
d
in
atio
n
e
n
h
an
ce
s
d
ec
is
io
n
-
m
ak
in
g
in
h
ig
h
d
im
en
s
io
n
al
a
n
d
co
m
p
lex
is
s
u
es wh
ich
m
ak
es
PF
OA
a
r
o
b
u
s
t so
lu
tio
n
f
o
r
a
wid
e
r
an
g
e
o
f
ap
p
licatio
n
s
.
i)
I
n
itializatio
n
:
PF
OA
is
a
n
atu
r
e
-
in
s
p
ir
ed
m
etah
eu
r
is
tic
th
at
m
im
ics
th
e
ad
a
p
tiv
e
h
u
n
tin
g
an
d
s
u
r
v
i
v
al
s
tr
ateg
ies
o
f
p
o
la
r
f
o
x
es
to
o
p
tim
ize
in
tr
icate
is
s
u
es.
T
h
e
p
o
lar
f
o
x
’
s
i
n
itial
p
o
p
u
latio
n
is
g
en
er
ate
d
r
an
d
o
m
l
y
,
wh
ich
m
a
d
e
th
e
s
o
lu
tio
n
s
p
ac
e
u
s
in
g
(
7
)
a
n
d
(
8
)
.
T
h
e
=
[
1
2
⋯
]
an
d
=
[
1
2
⋯
]
,
in
d
icate
s
p
o
lar
f
o
x
’
s
p
o
s
itio
n
m
atr
ix
,
r
ep
r
esen
ts
ℎ
p
o
lar
f
o
x
with
ℎ
v
alu
e,
d
eter
m
in
es
th
e
n
u
m
b
er
o
f
p
o
la
r
f
o
x
es,
an
d
illu
s
tr
ates
d
im
en
s
io
n
,
1
⃗
⃗
⃗
d
en
o
tes
a
r
a
n
d
o
m
v
ec
to
r
in
[
0
,
1
]
r
an
g
e
,
an
d
d
em
o
n
s
tr
ates
u
p
p
er
an
d
lo
wer
b
o
u
n
d
r
esp
ec
tiv
ely
.
T
h
e
m
ea
n
s
q
u
ar
e
er
r
o
r
(
MSE
)
is
u
s
ed
as
a
f
itn
ess
f
u
n
ctio
n
in
DSPF
OA
b
ec
au
s
e
it
ef
f
ec
tiv
el
y
m
ea
s
u
r
es
t
h
e
d
if
f
er
en
ce
am
o
n
g
p
r
ed
icted
a
n
d
ac
tu
al
o
u
tp
u
ts
th
at
g
u
id
es th
e
o
p
tim
izer
to
r
ed
u
ce
p
r
e
d
ictio
n
er
r
o
r
s
.
=
[
1
1
2
1
⋯
1
1
2
2
1
⋯
2
⋮
1
⋮
2
⋯
⋮
⋯
]
(
7
)
=
+
1
⃗
⃗
⃗
.
(
−
)
(
8
)
ii)
Gr
o
u
p
in
g
p
o
lar
f
o
x
:
i
n
ea
ch
g
r
o
u
p
,
th
e
n
u
m
b
e
r
o
f
m
e
m
b
er
s
is
s
im
ilar
;
h
o
wev
er
,
b
as
ed
o
n
weig
h
t
ca
lcu
lated
f
o
r
ea
ch
g
r
o
u
p
,
th
e
r
esu
lt is
th
e
f
ailu
r
e
o
r
s
u
cc
ess
o
f
th
at
g
r
o
u
p
.
E
ac
h
g
r
o
u
p
weig
h
t is u
p
d
ated
u
s
in
g
(
9
)
.
T
h
e
d
ep
icts
ℎ
g
r
o
u
p
weig
h
t,
th
e
n
u
m
b
er
o
f
p
o
lar
f
o
x
es
in
ℎ
g
r
o
u
p
,
an
d
in
d
icate
s
th
e
p
r
esen
t iter
atio
n
.
T
h
e
g
r
o
u
p
weig
h
t h
as a
n
i
n
itial v
alu
e
th
at
m
in
im
izes th
e
r
ate
o
f
c
h
an
g
e.
=
+
2
(
9
)
iii)
E
x
p
e
r
i
e
n
ce
-
b
as
ed
p
h
as
e:
t
h
e
p
o
la
r
f
o
x
d
o
es
n
o
t
h
i
b
e
r
n
at
e
d
u
r
i
n
g
wi
n
t
e
r
;
it
e
x
h
ib
its
a
n
in
t
eg
r
ati
o
n
o
f
co
m
m
u
n
a
l
an
d
n
o
m
a
d
i
c
b
eh
a
v
io
r
f
o
r
s
ea
r
c
h
i
n
g
f
o
o
d
.
In
(
1
0
)
a
n
d
(
1
1
)
is
a
p
p
li
ed
f
o
r
s
i
m
u
la
tin
g
t
h
e
p
o
la
r
f
o
x
j
u
m
p
in
g
b
e
h
a
v
i
o
r
d
u
r
i
n
g
h
u
n
ti
n
g
wit
h
t
h
e
p
o
we
r
o
f
j
u
m
p
a
n
d
d
i
r
e
cti
o
n
b
y
c
h
a
n
g
i
n
g
th
e
p
o
s
iti
o
n
(
)
to
n
ew
o
n
e
(
+
1
)
u
s
i
n
g
(
1
0
)
.
T
h
is
p
r
o
ce
s
s
is
r
e
p
ea
te
d
t
ill
th
e
s
u
b
s
eq
u
e
n
t
c
o
n
d
iti
o
n
is
tr
u
e
:
o
n
e
is
w
h
e
n
th
e
en
e
r
g
y
p
o
l
ar
f
o
x
is
less
t
h
a
n
s
et
le
v
el,
a
n
d
a
n
o
th
e
r
is
wh
en
a
b
et
te
r
f
it
n
ess
i
s
ac
q
u
i
r
e
d
,
as
s
h
o
wn
i
n
(
1
1
)
.
T
h
e
in
d
i
ca
tes
ℎ
p
o
la
r
f
o
x
ex
p
er
i
m
e
n
t
al
p
o
w
e
r
f
a
ct
o
r
a
n
d
r
e
p
r
ese
n
ts
f
it
n
ess
.
(
+
1
)
=
(
)
+
.
(
1
0
)
{
<
.
(
)
<
(
−
1
)
(
1
1
)
i
v
)
L
e
a
d
e
r
b
a
s
e
d
p
h
a
s
e
:
t
h
e
p
o
l
a
r
f
o
x
e
a
c
h
l
e
a
s
h
h
a
s
a
l
e
a
d
e
r
w
h
i
c
h
l
e
a
d
s
t
h
e
l
e
as
h
t
o
o
b
t
a
in
i
t
s
g
o
a
l
.
T
h
e
p
o
s
i
t
i
o
n
o
f
l
e
a
d
e
r
i
s
b
a
s
e
d
o
n
o
p
t
i
m
a
l
f
i
t
te
s
t
i
n
l
ea
s
h
a
n
d
t
h
e
p
o
l
a
r
f
o
x
c
h
a
n
g
e
s
i
t
s
p
o
s
i
ti
o
n
(
)
t
o
n
e
w
p
o
s
i
t
i
o
n
(
+
1
)
u
s
i
n
g
(
1
2
)
.
T
h
e
r
e
p
r
e
s
e
n
ts
t
h
e
l
e
a
d
e
r
p
o
we
r
f
a
c
t
o
r
an
d
4
⃗
⃗
⃗
i
n
d
i
c
a
t
e
s
r
a
n
d
o
m
v
e
c
t
o
r
.
(
+
1
)
=
(
)
+
4
⃗
⃗
⃗
(
(
)
−
)
.
(
1
2
)
v)
L
ea
d
er
m
o
tiv
atio
n
p
h
ase,
m
u
t
atio
n
an
d
f
atig
u
e
s
im
u
latio
n
:
a
lead
er
m
o
tiv
ates
an
d
m
em
b
er
s
u
p
d
ate
t
h
eir
p
o
s
itio
n
s
r
an
d
o
m
l
y
wh
ile
th
e
p
o
lar
f
o
x
es
s
tr
u
g
g
le
to
d
e
ter
m
in
e
th
e
p
r
e
y
co
n
s
ec
u
tiv
ely
b
y
s
ettin
g
1
,
2
,
3
,
an
d
4
as
m
atr
ix
b
eh
a
v
io
r
.
T
h
is
r
esu
lts
in
an
in
cr
ea
s
e
in
th
e
p
r
o
ce
s
s
f
o
r
a
lim
i
t
e
d
n
u
m
b
er
o
f
s
tep
s
af
ter
t
h
e
p
o
s
itio
n
is
ch
an
g
ed
.
T
h
e
m
ath
e
m
atica
l
eq
u
atio
n
f
o
r
th
e
lea
d
er
m
o
tiv
atio
n
p
h
ase
is
ex
p
r
ess
ed
in
(
1
3
)
.
Par
en
ts
s
o
m
etim
es
ab
an
d
o
n
litt
er
o
f
p
u
p
s
an
d
d
o
m
in
a
n
t
k
its
ex
h
ib
it
ag
g
r
ess
io
n
to
war
d
t
h
eir
s
ib
lin
g
s
.
Af
ter
lead
er
m
o
tiv
atio
n
,
th
e
p
o
lar
f
o
x
g
ets
tire
d
at
ea
c
h
it
er
atio
n
a
n
d
its
ef
f
o
r
ts
ar
e
m
in
im
ize
d
b
y
1
,
2
,
3
,
a
n
d
4
[
2
5
]
.
=
(
>
)
(
>
0
.
8
×
)
(
1
3
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
140
-
1
5
3
146
3
.
1
.
1
.
I
m
pro
v
ed
s
t
ra
t
e
g
ies
I
n
tr
ad
itio
n
al
PF
OA,
th
e
s
in
ch
ao
tic
m
ap
p
in
g
is
in
clu
d
ed
i
n
th
e
p
o
p
u
latio
n
in
itializatio
n
s
tag
e
f
o
r
im
p
r
o
v
in
g
ev
e
n
d
is
tr
ib
u
tio
n
.
T
h
is
ass
is
t
s
in
en
h
an
cin
g
t
h
e
ex
p
lo
r
atio
n
a
b
ilit
ies
an
d
av
o
id
in
g
p
r
em
at
u
r
e
co
n
v
er
g
en
ce
.
T
h
e
n
,
th
e
C
au
ch
y
o
p
er
atio
n
is
in
co
r
p
o
r
ated
t
o
av
o
id
lo
ca
l
o
p
tim
a
is
s
u
es
e
ar
ly
o
n
an
d
in
cr
ea
s
e
th
e
p
o
p
u
latio
n
s
ea
r
ch
s
p
ac
e.
A
d
etailed
ex
p
lan
atio
n
f
o
r
th
ese
im
p
r
o
v
e
d
s
tr
ateg
ies is
ex
p
lain
ed
as f
o
llo
ws.
i)
Sin
ch
ao
tic
m
ap
p
in
g
:
i
t
is
u
s
ed
to
in
itialize
th
e
p
o
p
u
latio
n
,
wh
ich
is
h
ig
h
er
lev
el
o
f
c
h
ao
tic
b
eh
av
io
r
co
m
p
ar
ed
to
lo
g
is
tic
m
ap
p
in
g
.
I
n
co
r
p
o
r
atin
g
a
s
in
e
ch
ao
tic
m
ap
d
u
r
in
g
th
e
PF
OA
in
itializatio
n
p
h
ase
en
s
u
r
es
m
o
r
e
u
n
if
o
r
m
d
is
tr
ib
u
tio
n
o
f
p
o
p
u
latio
n
u
s
in
g
(
1
4
)
.
B
y
u
s
in
g
th
is
s
tr
ateg
y
,
th
e
p
o
p
u
latio
n
is
in
itialized
,
wh
ich
r
esu
lts
in
m
o
r
e
ev
en
ly
d
is
tr
ib
u
ted
PF
OA
wh
ich
en
h
an
ce
s
th
e
m
o
d
el
’
s
p
er
f
o
r
m
a
n
ce
an
d
co
n
v
er
g
e
n
ce
s
p
ee
d
.
{
+
1
=
2
,
=
0
,
1
,
…
,
−
1
<
<
1
,
≠
0
(
1
4
)
ii)
C
au
ch
y
o
p
er
ato
r
m
u
tatio
n
:
C
au
ch
y
d
is
tr
ib
u
tio
n
m
in
im
ized
s
lo
wly
o
n
b
o
th
s
id
es
o
f
t
h
e
p
ea
k
v
alu
e,
an
d
p
o
lar
f
o
x
m
in
im
izes
th
e
l
o
ca
l
o
p
tim
a
is
s
u
e
af
ter
m
u
tatio
n
.
T
o
in
cr
ea
s
e
th
e
s
ea
r
ch
p
r
o
ce
s
s
o
f
PF
OA,
th
e
C
au
ch
y
o
p
er
ato
r
m
u
tatio
n
is
ap
p
lied
an
d
th
e
m
at
h
em
atica
l e
q
u
atio
n
o
f
1
d
im
en
s
io
n
al
C
au
ch
y
f
u
n
ctio
n
is
ex
p
r
ess
ed
in
(
1
5
)
.
W
h
ile
=
1
,
=
0
th
en
,
th
e
f
o
r
m
u
la
is
r
e
p
r
esen
ted
in
(
1
6
)
an
d
m
ath
em
atica
l
f
o
r
m
u
la
f
o
r
c
o
n
v
e
n
tio
n
al
C
au
ch
y
d
is
tr
ib
u
tio
n
is
in
d
icate
d
in
(
1
7
)
.
(
,
,
)
=
1
2
+
(
−
)
2
,
−
∞
<
<
∞
(
1
5
)
(
,
,
)
=
1
1
2
+
1
,
−
∞
<
<
∞
(
1
6
)
ℎ
(
0
,
1
)
=
[
(
−
0
.
5
)
]
,
[
0
,
1
]
(
1
7
)
B
y
in
teg
r
atin
g
th
e
p
o
s
itio
n
u
p
d
ate
o
f
PF
OA
an
d
C
au
ch
y
o
p
e
r
atio
n
v
ar
iatio
n
,
t
h
e
m
u
tan
t in
d
iv
id
u
al
is
g
en
er
ated
u
s
in
g
(
1
8
)
.
T
h
e
u
p
d
ated
in
d
iv
id
u
al
is
m
u
tated
b
y
in
clu
d
in
g
a
C
au
ch
y
o
p
er
ato
r
r
an
d
o
m
l
y
in
ea
ch
d
im
en
s
io
n
to
s
o
lv
e
th
e
lo
ca
l
o
p
tim
a
is
s
u
e.
W
h
er
e
r
ep
r
esen
ts
th
e
d
is
tu
r
b
an
ce
f
ac
to
r
.
B
y
in
co
r
p
o
r
atin
g
th
ese
two
s
tr
ateg
ies,
th
e
p
o
p
u
latio
n
d
is
tr
ib
u
tio
n
is
p
r
esen
ted
as
m
o
r
e
u
n
if
o
r
m
in
th
e
in
itiali
za
tio
n
s
tag
e,
wh
ich
in
cr
ea
s
es c
o
n
v
er
g
en
ce
s
p
ee
d
.
T
h
e
C
au
ch
y
o
p
er
at
o
r
s
o
lv
es th
e
lo
ca
l o
p
tim
a
is
s
u
e
an
d
ex
ten
d
s
th
e
s
ea
r
ch
s
p
ac
e
o
f
th
e
p
o
p
u
latio
n
ef
f
ec
tiv
el
y
i
n
PF
OA.
(
)
=
+
.
ℎ
(
0
,
1
)
(
1
8
)
T
ab
le
2
s
h
o
ws
th
e
h
y
p
er
p
ar
a
m
eter
v
alu
es
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
to
en
s
u
r
e
r
ep
r
o
d
u
ci
b
ilit
y
.
T
h
e
v
alu
es
ar
e
ch
o
s
en
b
ased
o
n
g
r
id
s
ea
r
ch
with
5
0
e
p
o
ch
f
o
r
co
n
v
er
g
en
ce
,
a
b
atch
s
ize
o
f
2
5
6
an
d
5
0
ep
o
c
h
s
p
r
o
v
id
e
s
tab
le
g
r
a
d
ien
t
u
p
d
ate
s
an
d
s
u
f
f
icien
t
lear
n
in
g
iter
at
io
n
s
o
f
1
0
0
.
A
lear
n
in
g
r
ate
o
f
0
.
0
0
1
with
Ad
am
o
p
tim
izer
en
s
u
r
es
s
tead
y
a
n
d
ad
ap
tiv
e
co
n
v
er
g
en
ce
,
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
en
h
an
ce
le
ar
n
in
g
with
s
o
lv
i
n
g
v
an
is
h
in
g
g
r
ad
ien
ts
,
an
d
d
r
o
p
o
u
t
r
ate
o
f
0
.
5
av
o
id
s
o
v
e
r
f
it
tin
g
b
y
d
ea
ctiv
atin
g
h
alf
o
f
t
h
e
n
eu
r
o
n
s
d
u
r
in
g
tr
ain
in
g
wh
ich
p
r
e
v
en
ts
n
etwo
r
k
f
r
o
m
r
ely
in
g
to
o
h
ea
v
ily
o
n
s
p
ec
if
ic
f
ea
tu
r
es.
Alg
o
r
ith
m
1
s
h
o
ws
a
p
s
eu
d
o
co
d
e
f
o
r
p
r
o
p
o
s
ed
m
eth
o
d
to
en
s
u
r
e
r
e
p
r
o
d
u
cib
ilit
y
.
Alg
o
r
ith
m
1
.
Ps
eu
d
o
co
d
e
o
f
p
r
o
p
o
s
ed
C
NN
-
DSPF
OA
f
o
r
m
alwa
r
e
d
etec
tio
n
I
n
p
u
t: Da
taset {
B
I
G2
0
1
5
,
Ma
l
im
g
}
Ou
tp
u
t: Pr
ed
icted
lab
els
Step
1
:
Data
p
r
e
-
p
r
o
ce
s
s
in
g
i)
Fo
r
ea
ch
s
am
p
le
x
in
X
:
–
C
o
n
v
er
t m
alwa
r
e
b
in
a
r
y
in
t
o
i
m
ag
e
–
Ap
p
ly
m
in
-
m
ax
n
o
r
m
aliza
tio
n
ii)
Sp
lit d
ataset
in
to
tr
ain
in
g
,
test
in
g
,
an
d
v
alid
atio
n
Step
2
:
I
n
itialize
C
NN
m
eth
o
d
:
i)
Def
in
e
co
n
v
o
lu
tio
n
al
lay
e
r
s
with
s
tr
id
es,
k
er
n
el
s
izes,
an
d
ac
tiv
atio
n
f
u
n
ctio
n
ii)
Def
in
e
p
o
o
lin
g
an
d
FC
lay
er
s
Step
3
:
I
n
itialize
DSPF
O
A
i)
I
n
itialize
p
o
p
u
latio
n
o
f
ca
n
d
id
ate
C
NN
h
y
p
er
p
ar
am
eter
s
ii)
Ap
p
ly
m
ax
iter
atio
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Ma
lw
a
r
e
d
etec
tio
n
u
s
in
g
co
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
etw
o
r
k
-
d
i str
a
teg
y
…
(
P
a
r
va
th
i S
a
th
en
a
h
a
lli Ja
ya
p
r
a
ka
s
h
)
147
Step
4
:
Op
tim
ize
C
NN
h
y
p
er
p
ar
am
ete
r
s
b
y
u
tili
zin
g
DSPF
OA
Fo
r
iter
atio
n
=1
to
m
ax
iter
atio
n
s
:
Fo
r
ea
ch
ca
n
d
id
ate
s
o
lu
tio
n
i
n
p
o
p
u
latio
n
:
i)
T
r
ain
C
NN
o
n
tr
ain
in
g
d
ata
b
y
u
tili
zin
g
ca
n
d
id
ate
h
y
p
er
p
a
r
am
eter
s
ii)
C
alcu
late
f
itn
ess
=M
SE
iii)
Up
d
ate
ca
n
d
id
ate
s
o
lu
tio
n
b
y
em
p
lo
y
in
g
s
in
e
ch
ao
tic
m
ap
f
o
r
in
itializatio
n
an
d
C
au
ch
y
m
u
tatio
n
to
s
o
lv
e
l
o
ca
l o
p
tim
a
iv
)
C
h
o
o
s
e
th
e
o
p
tim
al
ca
n
d
id
ate
s
o
lu
tio
n
d
ep
e
n
d
in
g
o
n
l
o
west M
SE
Step
5
:
T
r
ain
f
in
al
C
NN
f
o
r
class
if
icatio
n
:
i)
Ap
p
ly
C
NN
h
y
p
er
p
ar
am
eter
s
to
b
e
o
p
tim
al
s
o
lu
tio
n
d
eter
m
in
ed
b
y
DSPF
OA
ii)
T
r
ain
C
NN
o
n
f
u
ll tr
ain
i
n
g
s
et
iii)
Pre
d
ict
lab
els
E
n
d
T
ab
le
2
.
Su
m
m
a
r
y
o
f
o
p
tim
ized
h
y
p
e
r
p
ar
a
m
eter
v
alu
es e
m
p
lo
y
ed
in
p
r
o
p
o
s
ed
C
NN
-
DSPF
OA
P
a
r
a
me
t
e
r
s
V
a
l
u
e
s
Ep
o
c
h
s
50
B
a
t
c
h
si
z
e
2
5
6
Le
a
r
n
i
n
g
r
a
t
e
0
.
0
0
1
O
p
t
i
mi
z
e
r
A
d
a
m
A
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
R
e
LU
D
r
o
p
o
u
t
0
.
5
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
C
NN
-
DSPF
OA
is
s
im
u
lated
u
s
in
g
a
Py
th
o
n
3
.
4
en
v
ir
o
n
m
e
n
t
with
an
in
tel
i7
p
r
o
ce
s
s
o
r
,
W
in
d
o
ws
1
0
o
p
er
atin
g
s
y
s
te
m
,
an
d
6
4
GB
R
AM
.
F1
-
s
co
r
e,
r
ec
all,
ac
cu
r
ac
y
,
a
n
d
p
r
ec
is
io
n
ar
e
em
p
lo
y
ed
to
d
eter
m
in
e
th
e
m
o
d
el
p
e
r
f
o
r
m
an
ce
u
s
in
g
(
1
9
)
to
(
2
2
)
.
T
h
e
d
em
o
n
s
tr
ates
tr
u
e
p
o
s
itiv
e
a
n
d
in
d
icate
s
tr
u
e
n
eg
ativ
e.
=
+
+
+
+
(
1
9
)
=
+
(
2
0
)
=
+
(
2
1
)
1
−
=
2
2
+
+
(
2
2
)
4
.
1
.
P
er
f
o
r
m
a
nce
a
na
ly
s
is
T
ab
le
3
d
em
o
n
s
tr
ates
th
e
ev
al
u
atio
n
o
f
d
if
f
er
e
n
t
d
etec
tio
n
a
n
d
class
if
icatio
n
tech
n
iq
u
es
o
n
B
I
G2
0
1
5
an
d
Ma
lim
g
d
atasets
.
T
h
e
e
x
is
tin
g
tech
n
iq
u
es
s
u
c
h
as
R
esNet,
v
is
io
n
tr
an
s
f
o
r
m
er
(
ViT
)
,
s
win
tr
an
s
f
o
r
m
er
,
an
d
g
r
a
p
h
tr
an
s
f
o
r
m
er
ar
e
co
m
p
ar
ed
with
C
NN.
W
h
en
co
m
p
ar
ed
to
th
ese
m
eth
o
d
s
,
th
e
C
NN
o
b
tain
s
a
h
ig
h
ac
cu
r
ac
y
o
f
9
9
.
6
5
±
0
.
0
9
6
an
d
9
9
.
7
6
±
0
.
0
5
6
u
s
in
g
B
I
G2
0
1
5
a
n
d
Ma
lim
g
d
ataset
d
u
e
to
its
ab
ilit
y
to
e
x
tr
ac
t
h
ier
ar
ch
ical
s
p
atial
f
ea
tu
r
es f
r
o
m
m
alwa
r
e
im
ag
es.
T
ab
le
3
.
Per
f
o
r
m
an
ce
ev
alu
ati
o
n
r
esu
lts
o
f
d
i
f
f
er
en
t
m
alwa
r
e
d
etec
tio
n
an
d
class
if
icatio
n
tech
n
iq
u
es o
n
B
I
G2
0
1
5
an
d
Ma
lim
g
d
atasets
M
e
t
h
o
d
s
D
a
t
a
s
e
t
A
c
c
u
r
a
c
y
(
%)
R
e
c
a
l
l
(
%)
P
r
e
c
i
s
i
o
n
(
%)
F1
-
sc
o
r
e
(
%)
t
-
t
e
s
t
f
r
o
m
p
-
v
a
l
u
e
C
o
n
f
i
d
e
n
c
e
i
n
t
e
r
v
a
l
(
%)
R
e
sN
e
t
B
I
G
2
0
1
5
8
5
.
6
9
±
0
.
1
5
6
8
1
.
2
9
±
2
.
3
6
5
8
5
.
2
9
±
2
.
0
1
5
8
3
.
2
4
±
1
.
8
5
7
0
.
0
3
6
8
6
.
1
2
V
i
T
8
9
.
3
2
±
1
.
2
5
8
7
9
.
4
3
±
1
.
2
8
7
8
8
.
1
1
±
3
.
1
5
8
8
3
.
5
4
±
2
.
3
9
0
0
.
0
3
2
8
8
.
0
6
S
w
i
n
t
r
a
n
sf
o
r
mer
9
2
.
9
1
±
1
.
3
5
6
7
6
.
3
1
±
0
.
2
3
6
8
7
.
4
3
±
2
.
7
8
4
8
1
.
4
9
±
1
.
2
5
8
0
.
0
2
9
9
0
.
7
8
G
r
a
p
h
t
r
a
n
sf
o
r
mer
9
5
.
3
6
±
2
.
8
7
5
9
4
.
0
8
±
0
.
4
5
8
9
3
.
7
8
±
1
.
2
3
6
9
3
.
9
2
±
2
.
3
6
4
0
.
0
2
7
9
1
.
5
0
C
N
N
9
9
.
6
5
±
0
.
0
9
6
9
8
.
0
9
±
0
.
1
4
8
9
8
.
9
7
±
0
.
0
5
8
9
8
.
5
2
±
0
.
1
0
2
3
0
.
0
2
5
9
3
.
2
6
R
e
sN
e
t
M
a
l
i
m
g
8
9
.
4
3
±
0
.
3
6
5
9
2
.
3
8
±
2
.
3
5
4
9
2
.
1
9
±
1
.
4
8
7
9
2
.
2
8
±
2
.
1
5
8
0
.
0
3
8
8
7
.
6
3
V
i
T
9
3
.
2
9
±
0
.
2
5
4
9
4
.
5
0
±
1
.
0
8
7
9
1
.
2
8
±
1
.
4
5
8
9
2
.
8
6
±
3
.
7
8
4
0
.
0
3
4
8
9
.
4
8
S
w
i
n
t
r
a
n
sf
o
r
mer
9
5
.
3
9
±
0
.
3
5
7
9
5
.
3
8
±
0
.
2
3
5
8
9
.
3
l
±
0
.
1
4
8
9
0
.
4
3
±
2
.
7
9
1
0
.
0
3
1
9
1
.
3
8
G
r
a
p
h
t
r
a
n
sf
o
r
mer
9
6
.
2
8
±
1
.
3
6
9
9
5
.
8
9
±
0
.
1
8
5
9
6
.
0
2
±
0
.
3
5
8
9
5
.
9
5
±
3
.
4
9
6
0
.
0
3
0
9
1
.
3
5
C
N
N
9
9
.
7
6
±
0
.
0
5
6
9
8
.
8
7
±
0
.
0
2
5
9
9
.
8
6
±
0
.
0
3
6
9
9
.
3
6
±
0
.
2
3
4
0
.
0
2
8
9
5
.
0
5
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
140
-
1
5
3
148
ViT
,
s
wim
tr
an
s
f
o
r
m
er
,
an
d
g
r
ap
h
tr
a
n
s
f
o
r
m
e
r
r
eq
u
ir
e
l
ar
g
e
an
d
e
x
ten
s
iv
e
tr
ain
in
g
to
ca
p
tu
r
e
p
atter
n
s
ef
f
ec
tiv
ely
wh
ich
m
ak
e
p
r
o
n
e
to
o
v
er
f
itti
n
g
.
Ho
wev
er
,
C
NN
ar
e
b
etter
th
a
n
r
ec
en
t
tr
a
n
s
f
o
r
m
e
r
m
o
d
els
d
u
e
to
ca
p
t
u
r
in
g
lo
c
al
s
p
atial
p
atter
n
s
an
d
h
ier
a
r
ch
ical
s
tr
u
ctu
r
es
in
m
alwa
r
e
im
ag
e
with
less
p
ar
am
eter
s
th
at
en
ab
le
less
p
r
o
n
e
to
o
v
er
f
itti
n
g
.
Mo
r
e
o
v
er
,
th
e
co
n
v
o
l
u
tio
n
al
lay
er
s
ex
tr
ac
t
f
in
e
-
g
r
ain
e
d
tex
tu
r
es
an
d
s
tr
u
ctu
r
al
s
im
ilar
ities
wh
ich
en
s
u
r
es
s
tab
le
lear
n
in
g
ac
r
o
s
s
d
iv
er
s
e
m
alwa
r
e
f
am
ilies
.
T
h
is
ef
f
icien
cy
e
n
s
u
r
es
C
NN
to
p
r
o
v
id
e
r
a
p
id
tr
ai
n
in
g
,
less
co
m
p
u
tatio
n
al
co
m
p
lex
ity
,
a
n
d
o
b
tain
s
u
p
er
io
r
d
etec
tio
n
ac
cu
r
ac
y
.
Hen
ce
,
th
e
C
NN
co
n
tr
ib
u
te
h
ig
h
ac
c
u
r
ac
y
with
tr
ain
in
g
ef
f
icien
cy
b
y
m
ak
in
g
f
ea
tu
r
e
d
if
f
er
en
tiatio
n
a
cc
u
r
ately
c
o
m
p
ar
ed
to
e
x
is
tin
g
m
eth
o
d
s
.
T
h
e
t
-
test
is
em
p
l
o
y
ed
to
id
en
tify
wh
et
h
er
t
h
e
d
if
f
er
e
n
ce
am
o
n
g
ex
p
e
r
im
en
tal
o
u
tco
m
es
is
s
tatis
t
ically
s
ig
n
if
ican
t
wh
ich
en
s
u
r
es
th
at
t
h
e
o
b
s
er
v
ed
en
h
an
ce
m
en
ts
ar
e
n
o
t
b
ec
a
u
s
e
o
f
r
an
d
o
m
ch
an
ce
.
A
p
-
v
alu
e
r
e
p
r
esen
ts
th
e
s
ig
n
if
ican
ce
with
lo
wer
v
alu
es
wh
ich
p
r
o
v
id
es
b
etter
p
er
f
o
r
m
an
ce
ag
ain
s
t
n
u
ll
h
y
p
o
th
esis
.
C
o
n
f
id
en
ce
i
n
ter
v
al
is
a
s
tatis
tical
r
an
g
e
wh
ich
is
ca
lcu
lated
f
r
o
m
s
am
p
le
d
ata
with
in
tr
u
e
p
o
p
u
latio
n
p
ar
am
ete
r
with
s
p
e
cif
ied
p
r
o
b
ab
ilit
y
.
T
a
b
le
4
r
ep
r
es
e
n
ts
a
n
e
v
al
u
a
ti
o
n
o
f
v
a
r
i
o
u
s
h
y
p
e
r
p
a
r
a
m
et
e
r
tu
n
i
n
g
.
C
o
m
p
a
r
e
d
t
o
e
x
is
ti
n
g
t
ec
h
n
iq
u
es
s
u
c
h
as
R
F
O,
AF
O,
PF
OA
,
n
e
u
r
o
e
v
o
lu
ti
o
n
,
a
n
d
B
a
y
es
i
an
o
p
ti
m
i
za
ti
o
n
(
B
O)
-
DSP
F
OA,
t
h
e
p
r
o
p
o
s
ed
DSPF
OA
o
b
ta
in
s
a
h
ig
h
ac
cu
r
a
cy
o
f
9
9
.
6
5
an
d
9
9
.
7
6
%
u
s
in
g
B
I
G
2
0
1
5
an
d
Ma
l
im
g
d
at
aset
b
y
ef
f
e
cti
v
el
y
f
i
n
e
-
t
u
n
i
n
g
t
h
e
C
NN
p
ar
am
ete
r
s
.
B
u
t,
e
x
is
t
in
g
m
et
h
o
d
s
s
u
ch
as
R
F
O,
AFO
,
an
d
B
O
-
DSPF
OA
g
et
t
r
a
p
p
e
d
in
lo
c
al
o
p
t
im
a
b
e
ca
u
s
e
o
f
th
e
r
e
l
ian
ce
o
n
s
u
r
r
o
g
ate
m
o
d
els
a
n
d
c
o
n
s
tr
ai
n
ts
g
l
o
b
a
l
s
e
ar
c
h
a
b
i
li
ty
.
N
eu
r
o
e
v
o
l
u
ti
o
n
o
p
ti
m
iz
es
b
o
t
h
C
NN
a
n
d
w
e
ig
h
ts
b
y
u
t
ili
zi
n
g
e
v
o
l
u
ti
o
n
a
r
y
s
t
r
ate
g
i
es.
H
o
we
v
e
r
,
n
e
u
r
o
ev
o
l
u
ti
o
n
e
v
al
u
a
tes
m
a
n
y
c
a
n
d
id
at
e
n
e
two
r
k
s
o
v
e
r
m
u
lt
ip
le
g
en
e
r
ati
o
n
s
wh
ich
e
n
ab
le
h
ig
h
l
y
r
eso
u
r
c
e
an
d
ti
m
e
-
i
n
te
n
s
i
v
e.
DSPF
OA
e
n
h
a
n
ce
s
C
NN
’
s
f
ea
t
u
r
e
e
x
t
r
a
cti
o
n
ab
ilit
y
b
y
ch
o
o
s
i
n
g
t
h
e
b
es
t
co
n
f
i
g
u
r
ati
o
n
s
t
h
at
r
esu
lts
in
im
p
r
o
v
e
d
d
e
tec
ti
o
n
p
e
r
f
o
r
m
a
n
c
e
i
n
m
al
wa
r
e
im
ag
es.
I
ts
ex
p
l
o
r
at
io
n
an
d
e
x
p
lo
ita
ti
o
n
s
tag
e
m
a
k
es
b
ett
er
co
n
v
e
r
g
e
n
c
e,
s
o
l
v
es
t
h
e
l
o
ca
l
o
p
ti
m
a
is
s
u
e
,
a
n
d
e
n
h
a
n
c
e
s
f
ea
tu
r
e
lea
r
n
i
n
g
e
f
f
ec
t
iv
en
ess
th
at
i
m
p
r
o
v
es
in
t
er
p
r
et
a
b
ili
ty
,
r
o
b
u
s
tn
ess
a
n
d
g
e
n
er
ali
za
t
io
n
a
g
a
in
s
t
d
i
f
f
er
en
t
m
alw
a
r
e
f
am
i
lies
.
T
h
e
r
e
f
o
r
e
,
t
h
is
f
i
n
e
-
t
u
n
i
n
g
lea
d
s
to
s
u
p
e
r
i
o
r
c
lass
i
f
i
ca
t
io
n
ac
c
u
r
a
cy
t
h
a
t
m
a
k
es
C
NN
p
e
r
f
o
r
m
m
o
r
e
e
f
f
ec
t
iv
el
y
i
n
d
e
tec
ti
n
g
m
al
wa
r
e
.
T
ab
le
4
.
Per
f
o
r
m
an
ce
an
aly
s
is
o
f
d
if
f
er
en
t
h
y
p
er
p
ar
am
eter
t
u
n
in
g
m
eth
o
d
s
o
n
d
if
f
er
en
t
d
a
tasets
M
e
t
h
o
d
s
D
a
t
a
s
e
t
A
c
c
u
r
a
c
y
(
%)
R
e
c
a
l
l
(
%)
P
r
e
c
i
s
i
o
n
(
%)
F1
-
sc
o
r
e
(
%)
R
F
O
B
I
G
2
0
1
5
8
2
.
3
9
8
7
.
3
9
8
9
.
3
0
8
8
.
3
3
A
F
O
8
6
.
4
9
8
5
.
3
1
9
2
.
3
6
8
8
.
6
9
P
F
O
A
9
3
.
2
9
9
6
.
4
0
9
4
.
8
7
9
5
.
6
2
N
e
u
r
o
e
v
o
l
u
t
i
o
n
9
4
.
1
2
9
2
.
1
8
9
3
.
4
8
9
2
.
8
2
BO
-
D
S
P
F
O
A
9
5
.
4
6
9
4
.
0
3
9
5
.
6
5
9
4
.
8
3
D
S
P
F
O
A
9
9
.
6
5
9
8
.
0
9
9
8
.
9
7
9
8
.
5
2
R
F
O
M
a
l
i
m
g
8
5
.
3
9
8
9
.
1
9
9
2
.
0
9
9
0
.
6
1
A
F
O
8
7
.
3
2
9
0
.
3
8
8
9
.
4
3
8
9
.
9
0
P
F
O
A
9
2
.
1
0
9
2
.
1
0
8
6
.
9
2
8
9
.
4
3
N
e
u
r
o
e
v
o
l
u
t
i
o
n
9
3
.
2
5
9
1
.
0
8
8
8
.
4
9
8
9
.
7
6
BO
-
D
S
P
F
O
A
9
4
.
8
5
9
4
.
7
8
9
0
.
7
1
9
2
.
7
0
D
S
P
F
O
A
9
9
.
7
6
9
8
.
8
7
9
9
.
8
6
9
9
.
3
6
T
ab
le
5
p
r
o
v
id
es
a
p
e
r
f
o
r
m
an
ce
an
aly
s
is
o
f
k
-
f
o
ld
v
alid
atio
n
f
o
r
p
r
o
p
o
s
ed
C
NN
-
DSPF
O
A.
C
o
m
p
ar
ed
t
o
k
=3
,
7
,
an
d
9
,
k
=5
o
b
tain
s
a
h
ig
h
ac
cu
r
ac
y
f
o
r
b
o
th
d
ataset
d
u
e
t
o
its
b
alan
ce
d
tr
a
d
e
-
o
f
f
b
etwe
en
v
ar
ian
ce
an
d
b
ias.
A
lo
wer
v
alu
e
lik
e
k
=3
m
ak
es
th
e
m
o
d
el
m
o
r
e
s
en
s
itiv
e
to
o
u
tlier
s
an
d
n
o
is
e,
wh
ich
r
esu
lts
in
o
v
er
f
itti
n
g
.
Similar
ly
,
h
ig
h
v
alu
es
s
u
ch
as
k
=7
an
d
9
en
h
a
n
ce
b
ias,
wh
ich
m
ak
es
th
e
d
ec
is
io
n
b
o
u
n
d
ar
y
t
o
o
s
m
o
o
th
an
d
m
i
n
im
izes
s
en
s
itiv
ity
to
lo
ca
l
p
a
tter
n
s
.
W
h
ile
k
=5
,
th
e
m
o
d
el
ef
f
ec
tiv
ely
ca
p
tu
r
es
lo
ca
l
s
tr
u
ctu
r
es
b
y
m
an
ag
i
n
g
n
o
is
e
i
n
f
lu
en
ce
th
at
en
s
u
r
e
s
b
etter
g
e
n
er
aliza
tio
n
.
T
h
e
r
ef
o
r
e,
th
is
o
p
tim
al
n
eig
h
b
o
r
h
o
o
d
s
ize
im
p
r
o
v
es
c
lass
if
icatio
n
p
er
f
o
r
m
a
n
ce
a
n
d
in
ter
p
r
etab
ilit
y
b
y
m
an
ag
i
n
g
a
r
eliab
le
a
n
d
s
tab
le
d
ec
is
io
n
-
m
ak
in
g
p
r
o
ce
s
s
.
T
ab
le
5
.
k
-
f
o
ld
v
alid
atio
n
an
al
y
s
is
o
f
p
r
o
p
o
s
ed
C
NN
-
DSPF
OA
f
o
r
s
tab
ilit
y
ass
ess
m
en
t
D
a
t
a
s
e
t
k
-
f
o
l
d
A
c
c
u
r
a
c
y
(
%)
R
e
c
a
l
l
(
%)
P
r
e
c
i
s
i
o
n
(
%)
F1
-
sc
o
r
e
(
%)
B
I
G
2
0
1
5
k
=
3
8
4
.
3
0
8
2
.
1
9
9
2
.
1
3
8
6
.
8
7
k
=
5
9
9
.
6
5
9
8
.
0
9
9
8
.
9
7
9
8
.
5
2
k
=
7
8
6
.
4
8
8
4
.
2
9
9
3
.
8
1
8
8
.
7
9
k
=
9
9
2
.
1
0
8
9
.
3
0
8
9
.
0
2
8
9
.
1
5
M
a
l
i
m
g
k
=
3
8
7
.
0
3
8
7
.
3
1
8
3
.
9
1
8
5
.
5
7
k
=
5
9
9
.
7
6
9
8
.
8
7
9
9
.
8
6
9
9
.
3
6
k
=
7
8
9
.
0
2
8
9
.
3
0
9
0
.
4
3
8
9
.
8
6
k
=
9
9
2
.
3
8
9
2
.
1
0
9
3
.
4
9
9
2
.
7
8
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Ma
lw
a
r
e
d
etec
tio
n
u
s
in
g
co
n
v
o
lu
tio
n
a
l
n
eu
r
a
l n
etw
o
r
k
-
d
i str
a
teg
y
…
(
P
a
r
va
th
i S
a
th
en
a
h
a
lli Ja
ya
p
r
a
ka
s
h
)
149
T
a
b
l
e
6
d
e
p
i
c
t
s
t
h
e
a
n
a
l
y
s
is
o
f
c
o
m
p
u
t
a
t
i
o
n
a
l
t
i
m
e
a
n
d
m
e
m
o
r
y
c
o
n
s
u
m
p
t
i
o
n
f
o
r
d
i
f
f
e
r
e
n
t
d
e
t
e
c
t
i
o
n
m
e
t
h
o
d
s
.
T
h
e
C
N
N
o
b
t
a
i
n
s
l
es
s
c
o
m
p
u
t
a
t
i
o
n
a
l
t
i
m
e
a
n
d
m
em
o
r
y
c
o
n
s
u
m
p
t
i
o
n
o
f
7
6
,
8
3
s
a
n
d
8
7
,
9
8
KB
f
o
r
B
I
G2
0
1
5
an
d
Ma
lim
g
d
ataset
co
m
p
ar
ed
t
o
ex
is
tin
g
m
eth
o
d
s
lik
e
R
es
Net,
ViT
,
an
d
s
wim
tr
an
s
f
o
r
m
er
d
u
e
to
co
n
v
o
l
u
tio
n
al
lay
er
w
h
ich
m
i
n
im
ize
th
e
n
u
m
b
er
o
f
p
a
r
am
et
er
s
.
I
n
s
tead
o
f
p
r
o
ce
s
s
in
g
th
e
wh
o
le
im
ag
e,
C
NN
em
p
lo
y
s
s
m
all
k
er
n
els
th
at
s
i
g
n
if
ican
tly
m
in
im
ize
co
m
p
u
ta
tio
n
al
co
m
p
lex
ity
.
T
h
e
g
r
ap
h
i
c
p
r
o
ce
s
s
in
g
u
n
i
ts
(
GPUs
)
en
h
an
ce
ef
f
icien
c
y
an
d
m
ak
e
C
NN
r
ap
id
b
u
t le
s
s
m
em
o
r
y
e
f
f
icien
t.
T
a
b
l
e
6
.
C
o
m
p
u
t
a
t
i
o
n
a
l
t
i
m
e
a
n
d
m
e
m
o
r
y
c
o
n
s
u
m
p
t
i
o
n
a
n
a
l
y
s
i
s
o
f
d
i
f
f
e
r
e
n
t
d
e
t
e
c
t
i
o
n
a
n
d
c
l
a
s
s
i
f
i
c
a
t
i
o
n
m
e
t
h
o
d
s
D
a
t
a
s
e
t
M
e
t
h
o
d
s
C
o
m
p
u
t
a
t
i
o
n
a
l
t
i
m
e
(
s)
M
e
m
o
r
y
c
o
n
su
m
p
t
i
o
n
(
K
B
)
B
I
G
2
0
1
5
R
e
sN
e
t
1
0
2
1
7
6
V
i
T
89
1
6
5
S
w
i
n
t
r
a
n
sf
o
r
mer
82
1
2
5
C
N
N
76
87
M
a
l
i
m
g
R
e
sN
e
t
98
1
7
6
V
i
T
92
1
6
5
S
w
i
n
t
r
a
n
sf
o
r
mer
94
1
5
4
C
N
N
83
98
T
ab
le
7
d
em
o
n
s
tr
ates
th
e
ev
al
u
atio
n
o
f
ab
latio
n
s
tu
d
y
f
o
r
in
d
iv
id
u
al
c
o
m
p
o
n
en
t
in
p
r
o
p
o
s
ed
m
eth
o
d
.
ac
r
o
s
s
B
I
G2
0
1
5
a
n
d
Ma
lim
g
d
ataset.
W
h
ile
co
m
p
ar
e
d
to
in
d
iv
id
u
al
c
o
m
p
o
n
en
ts
s
u
ch
a
s
FOA,
PF
OA,
s
in
ch
ao
tic
m
ap
p
in
g
FOA,
C
au
ch
y
o
p
er
ato
r
m
u
tatio
n
FOA,
DSFOA,
s
in
ch
ao
tic
m
ap
p
in
g
PF
OA,
an
d
C
au
ch
y
o
p
er
ato
r
m
u
tatio
n
PF
OA,
p
r
o
p
o
s
ed
DSPF
OA
o
b
tain
s
ac
cu
r
ac
y
o
f
9
9
.
6
5
an
d
9
9
.
7
6
%
d
u
e
to
in
teg
r
atin
g
th
e
b
en
ef
its
o
f
ea
ch
co
m
p
o
n
en
t.
T
r
ad
itio
n
al
FOA
an
d
PF
O
A
o
f
f
er
s
g
lo
b
al
ex
p
lo
r
atio
n
h
o
wev
er
co
n
v
er
g
es
p
r
em
atu
r
ely
.
B
y
ad
d
in
g
s
in
e
ch
ao
tic
m
ap
p
in
g
im
p
r
o
v
es
p
o
p
u
latio
n
d
iv
e
r
s
ity
an
d
av
o
id
s
s
tag
n
atio
n
wh
er
ea
s
C
au
ch
y
m
u
tatio
n
p
r
o
ce
s
s
en
h
an
ce
s
lo
ca
l
s
ea
r
ch
th
r
o
u
g
h
es
ca
p
in
g
lo
ca
l
o
p
tim
a
an
d
PF
OA
in
cr
ea
s
e
s
tab
ilit
y
.
B
y
in
teg
r
atin
g
th
ese
co
m
p
o
n
en
ts
,
DSPF
OA
b
alan
ce
s
ex
p
lo
itatio
n
an
d
ex
p
l
o
r
atio
n
m
o
r
e
ef
f
ec
tiv
ely
th
at
p
r
ev
en
ts
p
r
em
at
u
r
e
co
n
v
er
g
en
ce
an
d
m
a
n
ag
e
d
if
f
er
en
t
p
o
p
u
latio
n
th
at
r
esu
lts
in
h
ig
h
ac
c
u
r
ac
y
,
in
ter
p
r
etab
ilit
y
,
an
d
r
o
b
u
s
tn
ess
.
T
ab
le
7
.
Ab
latio
n
s
tu
d
y
r
esu
lt
s
an
aly
zin
g
th
e
c
o
n
tr
ib
u
tio
n
o
f
in
d
iv
id
u
al
c
o
m
p
o
n
en
ts
in
p
r
o
p
o
s
ed
m
eth
o
d
M
e
t
h
o
d
s
D
a
t
a
s
e
t
A
c
c
u
r
a
c
y
(
%)
R
e
c
a
l
l
(
%)
P
r
e
c
i
s
i
o
n
(
%)
F1
-
sc
o
r
e
(
%)
F
O
A
B
I
G
2
0
1
5
9
2
.
6
9
9
0
.
4
8
9
0
.
4
8
9
0
.
4
8
P
F
O
A
9
3
.
2
9
9
6
.
4
0
9
4
.
8
7
9
5
.
6
2
S
i
n
e
c
h
a
o
t
i
c
ma
p
p
i
n
g
F
O
A
9
3
.
4
7
9
6
.
4
8
9
3
.
5
9
9
5
.
0
1
C
a
u
c
h
y
o
p
e
r
a
t
o
r
m
u
t
a
t
i
o
n
F
O
A
9
5
.
2
6
9
6
.
5
4
9
3
.
9
6
9
5
.
2
3
D
S
F
O
A
9
5
.
7
8
9
6
.
5
9
9
4
.
6
5
9
5
.
6
1
S
i
n
e
c
h
a
o
t
i
c
ma
p
p
i
n
g
P
F
O
A
9
6
.
7
8
9
6
.
7
8
9
4
.
9
0
9
6
.
7
8
C
a
u
c
h
y
o
p
e
r
a
t
o
r
m
u
t
a
t
i
o
n
P
F
O
A
9
7
.
4
8
9
7
.
1
2
9
5
.
3
6
9
7
.
1
2
D
S
P
F
O
A
9
9
.
6
5
9
8
.
0
9
9
8
.
9
7
9
8
.
5
2
F
O
A
M
a
l
i
m
g
9
0
.
7
8
8
4
.
2
6
9
0
.
4
5
8
7
.
2
4
P
F
O
A
9
2
.
1
0
9
2
.
1
0
8
6
.
9
2
8
9
.
4
3
S
i
n
e
c
h
a
o
t
i
c
ma
p
p
i
n
g
F
O
A
9
2
.
4
8
9
2
.
2
5
8
8
.
4
5
9
0
.
3
1
C
a
u
c
h
y
o
p
e
r
a
t
o
r
m
u
t
a
t
i
o
n
F
O
A
9
2
.
8
7
9
2
.
2
9
9
0
.
7
4
9
1
.
5
0
D
S
F
O
A
9
2
.
9
8
9
2
.
4
5
9
1
.
0
2
9
1
.
7
2
S
i
n
e
c
h
a
o
t
i
c
ma
p
p
i
n
g
P
F
O
A
9
3
.
6
0
9
2
.
7
8
9
1
.
6
9
9
2
.
7
8
C
a
u
c
h
y
o
p
e
r
a
t
o
r
m
u
t
a
t
i
o
n
P
F
O
A
9
4
.
8
5
9
3
.
1
9
9
3
.
2
6
9
3
.
1
9
D
S
P
F
O
A
9
9
.
7
6
9
8
.
8
7
9
9
.
8
6
9
9
.
3
6
Fig
u
r
e
4
s
h
o
ws
co
n
f
u
s
io
n
m
a
tr
ix
f
o
r
p
r
o
p
o
s
ed
m
eth
o
d
t
o
c
lass
if
y
th
e
d
if
f
er
e
n
t
class
es
:
Fig
u
r
e
4
(
a)
f
o
r
th
e
B
I
G2
0
1
5
an
d
Fig
u
r
e
4
(
b
)
f
o
r
th
e
Ma
lim
g
d
ataset.
I
n
B
I
G2
0
1
5
,
s
am
p
les
ar
e
class
if
ied
ac
cu
r
ately
with
less
m
is
clas
s
if
icatio
n
.
L
ik
ewise,
m
o
d
el
o
b
tain
s
h
ig
h
p
er
f
o
r
m
an
ce
in
Ma
lim
g
d
ataset
ac
r
o
s
s
d
if
f
er
e
n
t
m
alwa
r
e
f
am
ilies
.
T
h
e
o
u
tco
m
es
r
ep
r
esen
ts
th
at
m
o
d
el
ef
f
icien
tly
d
if
f
er
e
n
tiate
am
o
n
g
d
if
f
e
r
en
t
m
alwa
r
e
f
am
ilies
wh
ich
s
h
o
ws b
etter
p
er
f
o
r
m
a
n
ce
o
n
b
o
th
d
atasets
.
Fig
u
r
e
5
d
em
o
n
s
tr
ates
th
e
ev
a
lu
atio
n
o
f
r
ec
eiv
er
o
p
er
atin
g
c
h
ar
ac
ter
is
tics
(
R
O
C
)
cu
r
v
e
f
o
r
p
r
o
p
o
s
ed
m
eth
o
d
:
Fig
u
r
e
5
(
a)
f
o
r
th
e
B
I
G2
0
1
5
a
n
d
Fig
u
r
e
5
(
b
)
f
o
r
th
e
Ma
lim
g
d
ataset.
I
n
B
I
G2
0
1
5
,
all
m
alwa
r
e
f
am
ilies
o
b
tain
h
ig
h
ar
ea
u
n
d
e
r
th
e
cu
r
v
e
(
AUC)
th
at
r
ep
r
ese
n
ts
tr
ad
e
-
o
f
f
am
o
n
g
T
P a
n
d
F
P.
Similar
ly
,
m
o
d
el
attain
s
b
etter
AUC
o
n
all
c
lass
es
wh
ich
s
h
o
ws
s
u
p
er
io
r
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
Ov
er
all,
o
u
tco
m
es
v
alid
ate
th
e
m
o
d
el
r
eliab
ilit
y
f
o
r
m
alwa
r
e
class
if
icatio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.