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.
15
,
No
.
3
,
J
u
n
e
20
26
,
p
p
.
2269
~
2
2
7
7
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
3
.
p
p
2
2
6
9
-
2
2
7
7
2269
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Cutt
ing
-
edg
e alg
o
rithms
for reliab
l
e f
a
ilure
predic
tio
n in me
tro
train sy
stems
Sa
na
Cha
k
ri
1
,
Na
o
ua
l Mo
uh
ni
2
,
F
a
o
uzia
E
nn
a
a
m
a
3
1
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
L
a
b
o
r
a
t
o
r
y
(
LI
M
)
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
s
a
n
d
Te
c
h
n
i
q
u
e
s,
H
a
ss
a
n
I
I
U
n
i
v
e
r
si
t
y
,
M
o
h
a
mm
e
d
i
a
,
M
o
r
o
c
c
o
2
GL
-
I
S
I
Te
a
m,
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
c
s,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
Te
c
h
n
i
q
u
e
s,
U
n
i
v
e
r
si
t
é
M
o
u
l
a
y
I
smaï
l
,
M
e
k
n
e
s,
M
o
r
o
c
c
o
3
LA
M
I
G
EP
La
b
o
r
a
t
o
r
y
,
M
o
r
o
c
c
a
n
S
c
h
o
o
l
o
f
E
n
g
i
n
e
e
r
i
n
g
S
c
i
e
n
c
e
s (E
M
S
I
)
M
a
r
r
a
k
e
c
h
,
M
a
r
r
a
k
e
c
h
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
12
,
2
0
2
4
R
ev
is
ed
Feb
9
,
2
0
2
6
Acc
ep
ted
Mar
5
,
2
0
2
6
T
h
is
st
u
d
y
in
v
e
stig
a
ted
v
a
ri
o
u
s
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
rit
h
m
s
o
n
d
a
tas
e
t
fo
r
fa
il
u
re
p
re
d
icti
o
n
wit
h
in
m
e
tro
train
sy
ste
m
s.
Th
e
d
a
ta
in
d
ica
ted
stro
n
g
li
n
e
a
r
re
latio
n
sh
i
p
s
wit
h
in
t
h
e
d
a
t
a
se
t,
m
a
k
in
g
li
n
e
a
r
m
o
d
e
ls
su
c
h
a
s
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
s
(S
VMs)
v
iab
le,
a
s
we
ll
a
s
lo
g
isti
c
re
g
re
ss
io
n
a
n
a
ly
sis.
F
o
r
e
x
a
m
p
le,
th
e
lea
st
a
b
so
lu
te
sh
rin
k
a
g
e
a
n
d
se
lec
ti
o
n
o
p
e
ra
to
r
(LAS
S
O)
re
g
u
lariz
a
ti
o
n
m
e
th
o
d
u
se
d
in
fe
a
tu
re
se
lec
ti
o
n
h
a
d
p
r
o
fo
u
n
d
im
p
li
c
a
ti
o
n
s,
lea
d
in
g
t
o
e
n
h
a
n
c
e
d
p
e
rfo
rm
a
n
c
e
th
ro
u
g
h
t
h
e
id
e
n
ti
f
ica
ti
o
n
o
f
p
e
rti
n
e
n
t
a
tt
rib
u
tes
.
S
o
m
e
a
d
v
a
n
c
e
d
m
o
d
e
l
s
li
k
e
g
ra
d
ien
t
b
o
o
sti
n
g
m
a
c
h
in
e
s
(G
BM
s),
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
(
CNN
s),
a
n
d
k
e
rn
e
l
S
VMs
we
re
fo
u
n
d
t
o
o
u
t
p
e
rfo
rm
th
e
c
o
n
v
e
n
ti
o
n
a
l
m
e
th
o
d
s
b
e
c
a
u
se
th
e
y
a
re
c
a
p
a
b
le
o
f
re
c
o
g
n
izin
g
a
n
y
c
o
m
p
l
ica
ted
tre
n
d
s
o
r
n
o
n
-
li
n
e
a
r
re
latio
n
sh
i
p
s
p
re
se
n
t
i
n
d
a
ta
se
ts.
Co
m
b
i
n
in
g
stro
n
g
lea
rn
e
rs
c
a
n
p
ro
d
u
c
e
a
n
e
n
se
m
b
le
m
o
d
e
l
th
a
t
imp
ro
v
e
s
fo
re
c
a
st
p
e
rfo
rm
a
n
c
e
,
wh
il
e
to
p
-
p
e
rfo
rm
in
g
m
o
d
e
ls
a
re
u
se
d
in
t
h
e
e
n
se
m
b
le m
e
th
o
d
to
e
n
h
a
n
c
e
p
re
d
ictio
n
a
c
c
u
ra
c
y
.
Th
e
se
fin
d
in
g
s wo
u
l
d
h
e
lp
p
ro
fe
ss
io
n
a
ls
in
th
e
m
e
tro
tra
in
in
d
u
str
y
c
h
o
o
se
a
p
p
r
o
p
riate
m
a
c
h
in
e
lea
rn
in
g
m
e
th
o
d
s
t
o
su
p
p
o
r
t
p
re
v
e
n
ti
v
e
m
a
in
ten
a
n
c
e
stra
teg
ies
,
m
in
imiz
in
g
c
o
sts wh
il
e
e
n
h
a
n
c
in
g
o
p
e
ra
ti
o
n
a
l
e
ffe
c
ti
v
e
n
e
ss
a
n
d
sa
fe
ty
.
K
ey
w
o
r
d
s
:
An
o
m
aly
d
etec
tio
n
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
E
x
tr
em
e
lear
n
in
g
m
ac
h
in
e
Failu
r
e
p
r
ed
ictio
n
Gr
ad
ien
t b
o
o
s
tin
g
m
ac
h
in
es
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
:
San
a
C
h
ak
r
i
Ma
ch
in
e
I
n
tellig
en
ce
L
ab
o
r
at
o
r
y
(
L
I
M)
,
Facu
lty
o
f
Scien
ce
s
an
d
T
ec
h
n
i
q
u
es,
Hass
an
I
I
U
n
iv
er
s
ity
Mo
h
am
m
ed
ia,
M
o
r
o
cc
o
E
m
ail:
ch
ak
r
i.san
a@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Fo
r
ec
asti
n
g
f
ailu
r
es
in
m
etr
o
-
tr
ain
o
p
er
atio
n
s
co
n
d
u
cte
d
b
y
th
e
air
p
r
o
d
u
ctio
n
u
n
it
(
APU)
is
a
s
ig
n
if
ican
t
ch
allen
g
e,
p
r
im
ar
il
y
with
m
ain
ten
an
ce
ef
f
icac
y
an
d
s
er
v
ice
d
eliv
er
y
d
ep
e
n
d
a
b
ilit
y
.
C
o
n
v
en
tio
n
al
m
ec
h
an
is
m
s
d
ep
en
d
o
n
p
r
ev
e
n
tiv
e
o
r
r
ea
ctiv
e
s
tr
ateg
ies,
g
en
er
ally
f
o
u
n
d
e
d
o
n
p
r
e
d
eter
m
in
ed
tim
etab
les
o
r
r
esp
o
n
s
es
f
o
llo
win
g
f
ailu
r
es.
No
n
eth
eless
,
APU
f
ailu
r
es
ar
e
h
ig
h
ly
d
y
n
am
ic,
n
ec
ess
itati
n
g
a
m
o
r
e
p
r
o
ac
tiv
e
an
d
a
n
ticip
ato
r
y
s
tr
ateg
y
.
Ma
ch
in
e
lear
n
in
g
is
a
r
ev
o
l
u
tio
n
ar
y
m
et
h
o
d
th
at
ex
am
in
es
in
t
r
icate
r
elatio
n
s
h
ip
s
am
o
n
g
d
atasets
an
d
d
etec
ts
ea
r
ly
in
d
icato
r
s
o
f
APU
f
ailu
r
e
[
1
]
,
[
2
]
.
Ad
v
an
ce
d
alg
o
r
ith
m
s
p
r
o
v
id
e
p
r
ed
ictiv
e
m
o
d
els
th
at
d
etec
t
p
r
o
b
a
b
le
d
ef
ec
ts
ea
r
ly
,
o
p
tim
izin
g
m
ain
ten
an
ce
an
d
im
p
r
o
v
in
g
m
etr
o
tr
ain
s
er
v
ice
r
eliab
ilit
y
[
3
]
.
T
h
is
s
tu
d
y
ex
a
m
in
es
th
r
ee
ad
v
a
n
ce
d
m
ac
h
i
n
e
lear
n
i
n
g
m
o
d
els:
g
r
ad
ien
t
b
o
o
s
tin
g
m
ac
h
in
e
s
(
GB
M)
,
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
,
an
d
k
er
n
e
l
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
s
(
SVM)
,
u
tili
zin
g
th
e
Me
tr
o
PT
-
3
d
ataset,
wh
ich
c
o
m
p
r
is
es
o
p
e
r
atio
n
al
d
ata
f
r
o
m
co
m
p
r
ess
o
r
s
in
m
etr
o
tr
ain
s
[
4
]
,
[
5
]
.
T
h
ese
ad
v
an
ce
d
alg
o
r
ith
m
s
m
a
y
r
ev
ea
l
co
m
p
lex
lin
k
ag
es
a
n
d
p
atter
n
s
in
d
atasets
,
as
well
as
n
o
n
-
lin
ea
r
d
ep
en
d
e
n
cies,
p
o
te
n
tially
o
u
tp
er
f
o
r
m
in
g
p
r
e
v
io
u
s
m
o
d
els
s
u
ch
as
ex
tr
em
e
g
r
ad
ien
t
b
o
o
s
tin
g
(
XGBo
o
s
t
)
,
ex
tr
em
e
lear
n
i
n
g
m
ac
h
in
e
(
E
L
M)
,
an
d
n
e
u
r
al
n
etwo
r
k
s
.
T
h
e
GB
M
ar
e
en
s
em
b
le
lea
r
n
in
g
m
eth
o
d
o
lo
g
ies
th
at
ag
g
r
eg
ate
n
u
m
er
o
u
s
wea
k
lear
n
er
s
,
u
s
u
ally
d
ec
is
io
n
tr
ee
s
,
t
o
cr
ea
te
a
r
o
b
u
s
t
an
d
p
r
ec
is
e
p
r
ed
ictiv
e
m
o
d
el
[
6
]
.
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
2
6
9
-
2
2
7
7
2270
T
h
ei
r
c
ap
a
ci
ty
to
d
is
ce
r
n
n
u
a
n
ce
d
p
at
ter
n
s
r
en
d
er
s
th
em
o
p
tim
al
f
o
r
f
ai
lu
r
e
f
o
r
ec
as
t
in
g
.
C
NN
s
,
o
r
ig
in
al
ly
d
es
ig
n
e
d
f
o
r
im
ag
e
p
r
o
c
e
s
s
in
g
,
ca
n
b
e
m
o
d
i
f
ie
d
f
o
r
t
ab
u
l
ar
d
at
a
b
y
in
ter
p
r
et
in
g
ch
ar
a
ct
er
i
s
ti
c
s
as
in
p
u
t
ch
an
n
e
l
s
[
7
]
.
Qu
an
tu
m
n
eu
r
al
n
etwo
r
k
s
(
QNNs)
u
tili
ze
lo
ca
l
f
ield
s
to
d
is
ce
r
n
p
atter
n
s
an
d
ascer
tain
v
ar
iab
le
r
elatio
n
s
h
ip
s
in
Me
tr
o
PT
-
3
f
ailu
r
e
p
r
e
d
ictio
n
.
T
h
e
k
er
n
el
SVM
m
ay
s
im
u
late
n
o
n
-
lin
e
ar
d
ec
is
io
n
b
o
u
n
d
a
r
ies
b
y
p
r
o
jectin
g
d
ata
in
to
a
h
ig
h
er
-
d
im
e
n
s
io
n
al
s
p
ac
e
u
s
in
g
a
k
er
n
el
f
u
n
ctio
n
,
wh
ich
u
n
c
o
v
er
s
r
elatio
n
s
h
ip
s
o
v
er
lo
o
k
ed
b
y
lin
ea
r
SVM
m
o
d
els
[
8
]
.
T
h
is
a
n
aly
s
is
ass
es
s
es
th
e
ef
f
icac
y
o
f
s
o
p
h
is
ticated
m
eth
o
d
s
o
n
th
e
s
p
ec
if
ied
d
ataset,
aim
in
g
t
o
ac
h
iev
e
a
co
m
p
r
eh
e
n
s
iv
e
u
n
d
er
s
tan
d
in
g
o
f
th
e
s
tr
en
g
th
s
a
n
d
wea
k
n
ess
es
o
f
ea
ch
ap
p
r
o
ac
h
in
f
ailu
r
e
p
r
ed
ictio
n
f
o
r
m
etr
o
t
r
ain
s
y
s
tem
s
,
b
ased
o
n
th
e
o
u
tco
m
es
o
f
v
ar
io
u
s
ad
v
an
ce
d
m
o
d
els;
th
u
s
,
co
n
tr
ib
u
tin
g
t
o
th
e
o
n
g
o
in
g
d
is
co
u
r
s
e
o
n
p
r
ed
ictiv
e
m
a
in
ten
an
ce
.
T
h
e
s
tu
d
y
'
s
f
in
d
in
g
s
will
en
h
an
ce
th
e
p
r
ed
ictiv
e
m
ain
ten
an
ce
liter
atu
r
e
an
d
ass
is
t
p
r
ac
titi
o
n
er
s
ac
r
o
s
s
m
an
y
s
ec
to
r
s
in
id
e
n
tify
in
g
th
e
m
o
s
t
ef
f
ec
tiv
e
m
ac
h
in
e
lear
n
in
g
s
o
lu
tio
n
s
f
o
r
th
eir
s
p
ec
if
ic
ch
allen
g
es
[
9
]
.
T
h
e
m
eth
o
d
o
lo
g
y
s
ec
tio
n
will
d
elin
ea
te
th
e
s
p
ec
if
ic
s
tep
s
in
v
o
lv
e
d
in
th
is
s
tu
d
y
,
en
co
m
p
ass
in
g
d
ata
p
r
e
p
r
o
ce
s
s
in
g
,
f
ea
tu
r
e
s
elec
tio
n
,
an
d
th
e
o
p
er
atio
n
al
p
r
in
cip
les
o
f
th
e
th
r
ee
ad
v
a
n
ce
d
m
ac
h
in
e
lear
n
in
g
m
o
d
els
ass
es
s
ed
:
GB
M,
C
NN,
an
d
k
er
n
el
SVM.
T
h
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
s
e
ctio
n
will
th
en
d
elin
ea
te
th
e
p
er
f
o
r
m
an
ce
o
f
th
ese
v
ar
io
u
s
m
o
d
els
b
ased
o
n
cr
itical
m
etr
ics
in
clu
d
in
g
p
r
e
cisi
o
n
,
r
ec
all,
F1
-
s
co
r
e,
ar
ea
u
n
d
er
th
e
cu
r
v
e
r
ec
eiv
er
o
p
er
atin
g
ch
ar
ac
ter
is
tic
(
AUC
-
R
O
C
)
,
an
d
m
ea
n
s
q
u
ar
ed
er
r
o
r
(
MSE
)
.
T
h
is
will
also
ex
am
in
e
th
e
c
o
n
s
eq
u
e
n
ce
s
o
f
th
ese
f
in
d
in
g
s
an
d
th
e
s
ig
n
if
ican
ce
o
f
a
d
v
an
ce
d
m
o
d
els
f
o
r
p
r
ed
ictin
g
f
ailu
r
es
in
m
etr
o
tr
ain
s
y
s
tem
s
.
T
h
e
co
n
clu
s
io
n
will
en
ca
p
s
u
late
th
e
p
r
in
ci
p
al
f
in
d
i
n
g
s
o
f
th
e
s
tu
d
y
an
d
e
m
p
h
asi
ze
th
e
p
r
ac
tical
u
s
es
o
f
th
e
estab
lis
h
ed
m
o
d
els
in
th
e
m
etr
o
in
d
u
s
tr
y
[
1
0
]
.
2.
M
E
T
H
O
D
T
h
e
s
tu
d
y
in
v
o
lv
e
d
s
ev
er
al
k
ey
s
tep
s
.
First,
th
e
d
ataset
was
p
r
ep
r
o
ce
s
s
ed
,
in
clu
d
in
g
r
em
o
v
in
g
ir
r
elev
an
t
co
l
u
m
n
s
,
v
e
r
if
y
in
g
tim
estam
p
f
o
r
m
at,
an
d
h
an
d
lin
g
class
im
b
alan
ce
s
.
Featu
r
e
s
elec
tio
n
u
s
in
g
least
ab
s
o
lu
te
s
h
r
in
k
a
g
e
an
d
s
elec
tio
n
o
p
er
ato
r
(
L
ASSO)
id
en
tifie
d
th
e
m
o
s
t
im
p
o
r
tan
t
p
r
ed
icto
r
s
.
T
h
r
ee
ad
v
an
ce
d
m
o
d
els we
r
e
e
v
alu
at
ed
: G
B
M,
C
NN,
an
d
k
er
n
el
S
VM
.
2
.
1
.
Da
t
a
a
nd
prepro
ce
s
s
i
ng
T
h
e
in
v
esti
g
atio
n
is
b
ased
o
n
th
e
Me
tr
o
PT
-
3
[
5
]
d
ataset.
I
t
is
an
ex
ten
s
iv
e
co
llectio
n
o
f
o
p
er
atio
n
al
d
ata
f
r
o
m
m
et
r
o
tr
ain
s
y
s
tem
s
.
T
h
e
d
ataset
is
u
s
ed
as
b
asis
f
o
r
o
u
r
s
tu
d
y
,
s
in
ce
it
h
as
am
p
l
e
r
ec
o
r
d
s
r
eg
a
r
d
in
g
m
etr
o
tr
ain
s
o
p
er
atio
n
.
T
h
e
d
a
taset
r
ec
o
r
d
ed
k
e
y
p
a
r
am
eter
s
—
in
clu
d
in
g
p
r
ess
u
r
e,
tem
p
er
a
tu
r
e,
m
o
to
r
cu
r
r
e
n
t
(
m
o
to
r
c)
,
an
d
air
i
n
tak
e
v
alv
es
o
f
th
e
co
m
p
r
ess
o
r
’
s
APU
—
wh
ich
h
elp
u
n
d
er
s
tan
d
th
e
u
n
d
er
ly
i
n
g
d
y
n
am
ics
in
f
lu
en
cin
g
m
etr
o
tr
ain
p
er
f
o
r
m
an
ce
.
I
n
o
r
d
er
to
c
o
n
d
u
ct
th
is
d
ataset
f
o
r
a
n
aly
s
is
we
h
ad
to
p
e
r
f
o
r
m
ce
r
tain
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
.
Fig
u
r
e
1
s
h
o
w
th
e
s
tep
s
to
en
s
u
r
e
d
ata
p
r
ep
r
o
ce
s
s
in
g
.
Fig
u
r
e
1
. D
ata
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
I
n
itially
,
ex
tr
an
eo
u
s
co
lu
m
n
s
th
at
wer
e
ir
r
elev
an
t
to
th
e
an
aly
tical
o
b
jectiv
es
wer
e
elim
in
ated
,
th
er
ef
o
r
e
s
tr
ea
m
lin
in
g
th
e
d
ataset
an
d
r
ed
u
cin
g
n
o
is
e.
Seco
n
d
ly
,
co
n
s
is
ten
cy
in
tim
estam
p
f
o
r
m
at
was
estab
lis
h
ed
to
ass
u
r
e
t
h
e
ef
f
icac
y
o
f
tim
e
s
er
ies
an
aly
s
is
[
1
1
]
.
Desp
ite
b
ein
g
a
s
u
b
s
am
p
lin
g
s
tr
ateg
y
d
u
e
to
th
e
in
h
er
en
t
class
im
b
alan
ce
i
n
th
e
o
r
ig
i
n
al
d
ata,
wh
er
e
n
e
g
ativ
e
ca
s
es
(
ab
s
en
ce
o
f
d
ef
ec
t)
o
u
tn
u
m
b
er
e
d
p
o
s
itiv
e
o
n
es
(
f
ailu
r
e)
,
th
is
was
n
o
t
at
y
p
ical.
A
r
an
d
o
m
u
n
d
e
r
s
am
p
l
in
g
tech
n
iq
u
e
is
u
tili
ze
d
t
o
s
p
ec
if
ically
r
ec
tify
th
e
class
im
b
alan
ce
.
T
h
e
m
ajo
r
ity
class
(
n
o
r
m
al
o
p
er
atio
n
)
s
am
p
les
wer
e
ar
b
itra
r
ily
r
ed
u
ce
d
t
o
co
r
r
esp
o
n
d
with
th
e
q
u
an
tity
o
f
m
in
o
r
ity
class
(
f
ailu
r
e)
s
am
p
les,
ac
h
iev
in
g
a
b
alan
ce
d
1
:1
r
atio
.
T
h
is
m
eth
o
d
g
u
a
r
an
tees
th
at
th
e
m
o
d
el
r
em
ain
s
u
n
b
iased
a
g
ain
s
t th
e
m
ajo
r
ity
class
wh
ile
p
r
eser
v
in
g
a
f
ea
s
ib
le
d
ataset
s
ize
f
o
r
tr
ain
in
g
.
B
y
im
p
lem
en
tin
g
th
is
ap
p
r
o
ac
h
,
th
e
d
ev
elo
p
m
e
n
t
o
f
a
m
o
d
el
th
at
is
eq
u
itab
le
f
o
r
all
o
cc
u
r
r
en
ce
s
with
in
th
e
d
ataset
was
en
s
u
r
ed
,
s
in
ce
it
is
b
alan
ce
d
,
h
en
ce
elim
in
atin
g
an
y
b
ias
to
war
d
s
th
e
m
ajo
r
ity
g
r
o
u
p
.
Ad
d
itio
n
ally
,
an
y
p
o
ten
tial
o
u
tlier
s
in
o
u
r
ar
ea
o
f
f
o
c
u
s
wer
e
id
en
tifie
d
an
d
r
em
o
v
ed
b
y
a
th
o
r
o
u
g
h
ex
am
in
atio
n
o
f
in
ter
q
u
ar
tile
r
an
g
es,
allo
win
g
f
o
r
th
e
e
s
tab
lis
h
m
en
t
o
f
ap
p
r
o
p
r
iate
th
r
esh
o
ld
s
f
o
r
t
h
eir
ex
clu
s
io
n
.
T
h
is
led
u
s
to
d
ev
e
lo
p
a
m
o
r
e
r
eliab
le
d
ataset,
as
o
u
tlier
s
wo
u
ld
less
in
f
lu
en
ce
m
o
d
els.
B
e
awa
r
e
th
at
ce
r
tain
p
r
ep
ar
atio
n
p
r
o
ce
s
s
es,
s
u
ch
as
s
eg
m
en
tatio
n
an
d
n
o
r
m
aliza
tio
n
,
h
a
v
e
alr
ea
d
y
b
ee
n
ex
ec
u
ted
b
y
th
e
d
ata
p
r
o
v
id
er
[
5
]
,
[
1
2
]
.
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
C
u
ttin
g
-
ed
g
e
a
l
g
o
r
ith
ms fo
r
r
elia
b
le
fa
ilu
r
e
p
r
ed
ictio
n
in
m
etro
tr
a
in
s
ystem
s
(
S
a
n
a
C
h
a
k
r
i
)
2271
2
.
2
.
F
e
a
t
ure
s
elec
t
io
n
An
ess
en
tial
m
ea
s
u
r
e
to
en
h
an
ce
th
e
p
r
ed
ictiv
e
ef
f
icac
y
o
f
o
u
r
m
o
d
els
is
th
e
s
elec
tio
n
o
f
a
p
p
r
o
p
r
iate
ch
ar
ac
ter
is
tics
.
T
h
e
L
ASSO
(
L
1
r
eg
u
lar
izatio
n
)
ap
p
r
o
ac
h
is
em
p
lo
y
ed
to
f
ac
ilit
ate
th
e
tr
an
s
f
o
r
m
atio
n
o
f
th
e
d
ataset
f
o
r
a
lin
ea
r
r
elatio
n
s
h
ip
.
T
h
is
is
th
e
r
atio
n
ale
f
o
r
o
u
r
s
elec
tio
n
o
f
th
e
L
ASSO
alg
o
r
ith
m
.
L
ASSO
r
ed
u
ce
s
co
ef
f
icien
ts
to
ze
r
o
,
f
ac
ilit
atin
g
au
to
m
atic
f
ea
tu
r
e
s
elec
tio
n
—
o
p
tim
al
f
o
r
h
ig
h
-
d
im
en
s
io
n
al
d
atasets
.
L
ASSO
id
en
tifie
s
th
e
m
o
s
t
v
a
lu
ab
le
ch
ar
ac
ter
is
tics
,
m
itig
atin
g
th
e
cu
r
s
e
o
f
d
im
en
s
io
n
ality
,
en
h
an
cin
g
m
o
d
el
in
ter
p
r
etab
ilit
y
,
an
d
im
p
r
o
v
i
n
g
g
en
er
aliza
tio
n
ca
p
ab
ilit
ies
[
1
3
]
.
L
ASSO
's p
er
tin
en
t f
ea
tu
r
e
s
elec
tio
n
id
en
tifie
d
tr
an
s
m
itter
p
r
ess
u
r
e
2
(
T
P2
)
,
h
ig
h
-
p
r
ess
u
r
e
s
en
s
o
r
1
(
H1
)
,
d
if
f
er
en
tial
v
alv
e
lev
el
(
DV
lev
el
)
,
m
o
to
r
win
d
in
g
tem
p
er
atu
r
e
(
M
m
o
to
r
tem
p
)
,
m
o
to
r
c,
co
m
p
r
ess
o
r
p
r
ess
u
r
e
(
C
OM
P
)
,
d
if
f
er
e
n
tial
v
al
v
e
elec
tr
ical
in
p
u
t
(
DV
elec
in
)
,
an
d
m
o
t
o
r
p
o
s
itio
n
o
r
g
as
f
lo
w
(
MPG
)
as
s
ig
n
if
ican
t
p
r
e
d
icto
r
s
o
f
“
s
tatu
s
”
.
T
h
is
p
er
s
p
ec
tiv
e
is
co
r
r
o
b
o
r
ated
b
y
th
e
r
o
b
u
s
t
lin
ea
r
co
r
r
elatio
n
s
id
en
tifie
d
d
u
r
in
g
ex
p
lo
r
ato
r
y
d
ata
an
aly
s
i
s
,
s
o
in
d
icatin
g
th
at
L
ASSO
wa
s
ap
p
r
o
p
r
iate
f
o
r
th
is
s
p
ec
if
ic
d
ataset
[
1
4
]
.
T
o
a
v
er
t
d
ata
leak
in
g
,
L
ASSO
f
ea
tu
r
e
s
elec
tio
n
was
ex
ec
u
ted
with
in
ea
ch
iter
atio
n
o
f
t
h
e
cr
o
s
s
-
v
alid
atio
n
c
y
cle.
Sp
ec
if
ically
,
th
e
f
ea
tu
r
e
im
p
o
r
tan
ce
c
o
ef
f
icien
ts
wer
e
ca
lcu
lated
s
o
lely
b
ased
o
n
th
e
tr
ain
in
g
f
o
ld
d
ata,
an
d
th
e
s
elec
ted
f
ea
tu
r
es
wer
e
th
en
ap
p
lied
to
th
e
v
alid
atio
n
f
o
l
d
.
T
h
is
en
s
u
r
es t
h
at
th
e
test
d
ata
r
em
ain
ed
s
tr
ictly
u
n
s
ee
n
d
u
r
in
g
th
e
f
ea
tu
r
e
s
elec
tio
n
p
r
o
ce
s
s
.
2
.
3
.
G
ra
dient
bo
o
s
t
ing
m
a
c
hin
es
T
h
e
GB
M
ar
e
s
u
itab
le
f
o
r
o
u
r
d
atasets
b
ec
au
s
e
to
th
eir
ca
p
ac
ity
f
o
r
e
q
u
atio
n
f
itti
n
g
a
n
d
h
y
p
o
th
esis
r
ep
r
esen
tatio
n
[
1
5
]
.
GB
M
in
c
r
em
en
tally
co
n
s
tr
u
cts
m
o
d
els,
ea
ch
r
ec
tif
y
in
g
t
h
e
f
laws
o
f
its
p
r
ed
ec
ess
o
r
b
y
g
r
ad
ien
t
d
escen
t,
h
en
ce
en
h
a
n
cin
g
p
r
ed
ictin
g
ac
c
u
r
ac
y
.
T
h
e
d
ataset
co
n
tain
s
n
u
m
er
o
u
s
f
ea
tu
r
es,
s
o
m
e
o
f
wh
ich
m
ay
ex
h
ib
it
n
o
n
-
lin
ea
r
in
ter
ac
tio
n
s
th
at
in
f
lu
en
ce
th
e
o
u
tco
m
e
v
ar
ia
b
le
'
s
ta
tu
s
'
.
G
B
M
is
o
p
tim
al
d
u
e
to
its
ca
p
ac
ity
to
id
en
tify
co
m
p
l
ex
p
atter
n
s
with
o
u
t
n
ec
ess
itati
n
g
ac
tiv
e
f
ea
tu
r
e
m
o
d
if
icatio
n
in
s
id
e
th
e
d
ataset.
Mo
d
el
p
ar
a
m
eter
izatio
n
a
n
d
v
alid
atio
n
th
e
p
r
o
ce
s
s
o
f
estab
l
is
h
in
g
p
ar
am
ete
r
s
an
d
ev
alu
at
in
g
th
e
s
u
f
f
icien
cy
o
f
th
e
GB
M
m
o
d
el
c
o
m
p
r
is
es sev
er
al
s
ig
n
if
ican
t stag
es:
i)
Hy
p
er
p
ar
a
m
eter
tu
n
in
g
:
u
tili
ze
d
a
g
r
id
s
ea
r
ch
s
tr
ateg
y
to
o
p
tim
ize
th
e
k
ey
h
y
p
e
r
p
ar
a
m
eter
s
.
T
h
e
s
ea
r
ch
s
p
ac
e
in
clu
d
ed
:
lear
n
i
n
g
r
ate
∈
{0
.
0
1
,
0
.
0
5
,
0
.
1
},
n
u
m
b
er
o
f
esti
m
ato
r
s
∈
{1
0
0
,
2
0
0
,
5
0
0
},
an
d
m
ax
im
u
m
tr
ee
d
ep
t
h
∈
{3
,
5,
7}
.
T
h
e
f
in
al
s
elec
tio
n
was
b
as
ed
o
n
m
a
x
im
izin
g
th
e
F1
-
s
co
r
e
ac
r
o
s
s
5
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
[
1
6
]
.
ii)
C
r
o
s
s
-
v
alid
atio
n
:
cr
o
s
s
-
v
alid
a
tio
n
tech
n
i
q
u
es
will
b
e
u
s
ed
t
o
en
s
u
r
e
th
at
th
e
GB
M
m
o
d
el
is
r
o
b
u
s
t
an
d
g
en
er
alize
s
well.
Her
ein
,
th
e
d
ataset
is
d
iv
id
ed
in
t
o
s
ev
er
al
f
o
ld
s
an
d
th
en
th
e
m
o
d
el
is
tr
ai
n
ed
u
s
in
g
o
n
e
p
o
r
tio
n
wh
ile
test
in
g
it o
n
t
h
e
r
est
p
ar
ts
;
th
is
allo
ws
u
s
to
est
ab
lis
h
its
o
v
er
all
ef
f
ec
tiv
en
ess
b
y
lo
o
k
i
n
g
at
av
er
ag
e
p
e
r
f
o
r
m
an
ce
o
n
all
f
o
l
d
s
[
1
7
]
.
iii)
E
a
r
l
y
s
t
o
p
p
i
n
g
:
G
B
M
s
m
a
y
o
v
e
r
f
i
t
,
p
a
r
t
i
c
u
l
a
r
l
y
w
i
t
h
m
a
n
y
t
r
e
e
s
.
I
n
o
r
d
e
r
t
o
a
d
d
r
e
s
s
t
h
i
s
c
h
a
l
l
e
n
g
e
,
a
n
e
a
r
l
y
s
t
o
p
c
r
i
t
e
r
i
o
n
i
s
i
n
t
r
o
d
u
c
e
d
t
o
m
o
n
i
t
o
r
p
e
r
f
o
r
m
a
n
c
e
o
f
m
o
d
e
l
a
g
a
i
n
s
t
v
a
l
i
d
a
t
i
o
n
s
e
t
s
u
c
h
t
h
a
t
a
s
s
o
o
n
a
s
t
h
e
r
e
i
s
n
o
i
m
p
r
o
v
e
m
e
n
t
o
b
s
e
r
v
e
d
,
t
r
a
i
n
i
n
g
m
u
s
t
s
t
o
p
s
o
a
s
t
o
p
r
e
v
e
n
t
o
v
e
r
f
i
t
t
i
n
g
o
n
t
h
e
t
r
a
i
n
i
n
g
d
a
t
a
[
1
8
]
.
2
.
4
.
Co
nv
o
lutio
na
l
neura
l net
wo
rk
s
Alth
o
u
g
h
u
s
u
ally
em
p
lo
y
ed
f
o
r
im
ag
e
d
ata
p
r
o
ce
s
s
in
g
,
C
N
Ns
r
ep
r
esen
t
a
ca
teg
o
r
y
o
f
d
e
ep
lear
n
in
g
m
o
d
els
th
at
ex
ten
d
b
e
y
o
n
d
im
ag
e
p
r
o
ce
s
s
in
g
ap
p
licatio
n
s
[
1
9
]
.
C
NNs
wer
e
em
p
lo
y
ed
o
n
th
e
d
ataset
d
u
e
to
th
eir
ca
p
ac
ity
to
ca
p
t
u
r
e
in
tr
ic
ate
f
ea
tu
r
e
in
ter
ac
tio
n
s
an
d
n
o
n
-
lin
ea
r
co
r
r
elatio
n
s
in
h
er
e
n
t in
h
ig
h
-
d
im
en
s
io
n
a
l
d
atasets
[
2
0
]
.
Desp
ite
th
e
d
ata
s
et
n
o
t
b
ei
n
g
in
a
co
n
v
en
tio
n
al
im
ag
e
f
o
r
m
at,
it m
ay
b
e
co
n
v
er
ted
in
to
a
tab
u
lar
s
tr
u
ctu
r
e
ak
in
to
a
2
D
g
r
id
,
allo
win
g
C
NNs
to
lev
er
ag
e
r
o
b
u
s
t
f
ea
tu
r
e
ex
tr
ac
tio
n
ca
p
ab
ilit
ies.
B
y
in
ter
p
r
etin
g
tab
u
lar
d
ata
as
a
two
-
d
im
e
n
s
io
n
al
im
ag
e
with
f
ea
tu
r
es
r
ep
r
esen
ted
as
ch
an
n
els,
C
NNs
m
ay
g
en
er
ate
h
ier
ar
ch
ical
r
ep
r
esen
tatio
n
s
a
n
d
au
to
n
o
m
o
u
s
ly
id
en
tif
y
s
ig
n
if
ican
t
p
atter
n
s
with
m
in
im
al
n
ee
d
o
n
f
ea
tu
r
e
en
g
in
ee
r
in
g
.
M
o
d
if
y
in
g
C
NNs
f
o
r
ta
b
u
lar
APU
s
en
s
o
r
d
at
a
is
a
v
ia
b
le
s
tr
ateg
y
as
it
en
ab
les
th
e
m
o
d
el
to
au
to
n
o
m
o
u
s
ly
d
is
ce
r
n
in
tr
ica
te
f
ea
tu
r
e
c
o
r
r
elatio
n
s
,
h
en
c
e
r
ed
u
ci
n
g
n
ee
d
o
n
e
x
ten
s
iv
e
m
an
u
al
f
ea
tu
r
e
en
g
in
ee
r
in
g
.
C
NN
s
ty
p
ically
p
r
o
ce
s
s
im
ag
es;
h
o
wev
er
,
in
th
is
co
n
tex
t,
ea
ch
s
en
s
o
r
r
ea
d
in
g
f
u
n
ctio
n
s
as
a
ch
an
n
el
with
in
a
two
-
d
im
en
s
io
n
al
g
r
id
.
T
h
is
co
n
f
ig
u
r
atio
n
en
ab
les
co
n
v
o
lu
tio
n
al
lay
e
r
s
to
id
en
tify
lo
ca
l
p
atter
n
s
an
d
s
p
atial
r
elatio
n
s
h
ip
s
am
o
n
g
s
en
s
o
r
s
,
d
etec
tin
g
n
o
n
-
lin
ea
r
co
r
r
elatio
n
s
f
r
e
q
u
e
n
tly
o
v
er
lo
o
k
ed
b
y
tr
ad
itio
n
al
m
ac
h
i
n
e
lear
n
in
g
m
o
d
els.
T
h
e
ar
ch
itectu
r
e,
i
n
co
r
p
o
r
atin
g
co
n
v
o
lu
tio
n
al
a
n
d
p
o
o
lin
g
lay
er
s
,
en
ab
les
th
e
m
o
d
el
to
co
n
s
tr
u
c
t
lay
er
ed
,
h
ier
ar
c
h
ical
r
ep
r
ese
n
tatio
n
s
.
T
h
is
is
s
ig
n
if
ican
t
w
h
en
ad
d
r
ess
in
g
th
e
co
m
p
lex
ities
o
f
s
en
s
o
r
d
ata,
c
h
ar
ac
ter
ized
b
y
its
n
o
is
e
an
d
v
ar
iab
ilit
y
.
T
h
is
ap
p
r
o
ac
h
n
o
t
o
n
ly
en
h
a
n
ce
s
th
e
m
o
d
el'
s
r
o
b
u
s
tn
ess
b
u
t a
ls
o
r
e
v
ea
ls
s
u
b
tle
an
o
m
alies a
n
d
p
r
e
d
icts
f
ailu
r
es with
s
u
p
er
io
r
ac
cu
r
ac
y
co
m
p
ar
ed
to
p
r
ev
io
u
s
m
eth
o
d
s
.
Fig
u
r
e
2
illu
s
tr
ates
th
e
wo
r
k
f
lo
w
o
f
a
C
NN
f
o
r
th
e
tr
an
s
f
o
r
m
atio
n
an
d
p
r
o
ce
s
s
in
g
o
f
tab
u
lar
d
ata.
T
h
is
is
a
d
escr
ip
tio
n
o
f
th
e
f
ig
u
r
e'
s
co
n
ten
t,
d
er
i
v
ed
f
r
o
m
th
e
te
x
t.
–
I
n
p
u
t
tr
a
n
s
f
o
r
m
atio
n
: ta
b
u
lar
d
ata
→
co
n
v
er
te
d
to
2
D
g
r
id
s
tr
u
ctu
r
e.
–
C
o
n
v
o
lu
tio
n
al
lay
e
r
s
: a
p
p
ly
le
ar
n
ab
le
f
ilter
s
to
ex
tr
ac
t lo
ca
l
p
atter
n
s
an
d
s
p
atial
r
elatio
n
s
h
i
p
s
.
–
Po
o
lin
g
lay
er
s
: r
ed
u
ce
s
p
atial
d
im
en
s
io
n
s
(
e.
g
.
,
m
ax
p
o
o
lin
g
)
f
o
r
tr
an
s
latio
n
in
v
a
r
ian
ce
.
–
Fu
lly
co
n
n
ec
ted
lay
er
s
: f
latten
ed
f
ea
tu
r
es →
class
if
ied
v
ia
d
en
s
e
lay
er
s
.
–
Ou
tp
u
t
lay
er
: p
r
ed
icts
f
ailu
r
e
p
r
o
b
a
b
ilit
ies (
b
in
ar
y
class
if
icatio
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
2
6
9
-
2
2
7
7
2272
Fig
u
r
e
2
.
C
NN
ar
ch
itectu
r
e
f
o
r
tab
u
lar
d
ata
f
ailu
r
e
p
r
ed
ictio
n
T
h
e
s
p
ec
if
ic
ar
ch
itectu
r
e
o
f
th
e
im
p
lem
en
ted
C
NN
is
d
e
tailed
as
f
o
llo
ws:
it
co
n
s
i
s
ts
o
f
th
r
ee
1
D
co
n
v
o
l
u
tio
n
al
lay
er
s
with
3
2
,
6
4
,
a
n
d
1
2
8
f
ilter
s
r
esp
ec
tiv
ely
,
ea
ch
u
s
in
g
a
k
er
n
el
s
ize
o
f
3
an
d
'
s
am
e'
p
ad
d
in
g
to
p
r
eser
v
e
s
p
atial
d
im
en
s
io
n
s
.
E
ac
h
co
n
v
o
lu
tio
n
al
lay
er
is
f
o
llo
wed
b
y
a
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U
)
ac
tiv
atio
n
f
u
n
ctio
n
an
d
a
Ma
x
Po
o
lin
g
lay
er
(
p
o
o
l
s
ize
=2
)
.
A
d
r
o
p
o
u
t
r
ate
o
f
0
.
5
was
ap
p
lied
b
ef
o
r
e
th
e
f
in
al
f
u
lly
co
n
n
ec
ted
d
e
n
s
e
lay
er
s
to
p
r
ev
en
t
o
v
er
f
it
tin
g
.
T
r
ain
in
g
p
ar
a
m
eter
s
:
i)
o
p
tim
izer
:
Ad
am
(
lear
n
in
g
r
ate
=0
.
0
0
1
)
;
ii)
l
o
s
s
f
u
n
ctio
n
:
b
in
ar
y
cr
o
s
s
-
e
n
tr
o
p
y
;
iii)
b
atch
s
ize
:
3
2
;
an
d
iv
)
ep
o
ch
s
:
5
0
(
with
ea
r
ly
s
to
p
p
in
g
p
atien
ce
o
f
5
e
p
o
ch
s
)
.
T
h
e
p
r
im
ar
y
f
u
n
ctio
n
o
f
co
n
v
o
lu
tio
n
al
lay
er
s
is
to
im
p
lem
en
t
s
ev
er
al
lear
n
ab
le
f
ilter
s
(
co
n
v
o
lu
tio
n
k
er
n
els)
o
n
th
e
in
p
u
t,
f
ac
ilit
atin
g
th
e
ca
p
t
u
r
e
o
f
lo
ca
l
in
f
o
r
m
atio
n
an
d
s
p
atial
co
r
r
elatio
n
s
am
o
n
g
v
a
r
iab
les.
T
h
ey
f
ac
ilit
ate
th
e
id
e
n
tific
atio
n
o
f
co
m
p
le
x
p
atter
n
s
in
d
ata
b
y
ex
tr
ac
tin
g
d
etailed
i
n
f
o
r
m
atio
n
i
n
lo
wer
lay
er
s
b
ef
o
r
e
t
r
an
s
m
itti
n
g
th
e
m
to
h
ig
h
e
r
lay
er
s
.
Po
o
li
n
g
la
y
er
s
s
u
cc
ee
d
co
n
v
o
lu
tio
n
al
lay
er
s
,
alter
in
g
f
ea
t
u
r
e
m
ap
s
b
y
d
im
i
n
is
h
in
g
th
eir
s
p
atial
d
im
en
s
io
n
s
an
d
in
t
r
o
d
u
cin
g
tr
an
s
latio
n
in
v
ar
ian
ce
,
h
en
ce
en
h
an
cin
g
r
o
b
u
s
tn
ess
to
m
in
o
r
in
p
u
t a
lter
atio
n
s
.
Fu
lly
co
n
n
ec
te
d
l
ay
er
s
r
ec
eiv
e
th
e
r
etr
iev
e
d
f
ea
tu
r
es f
r
o
m
co
n
v
o
lu
tio
n
al
an
d
p
o
o
lin
g
lay
er
s
,
f
latten
th
em
,
th
en
tr
an
s
m
it
th
em
th
r
o
u
g
h
o
n
e
o
r
m
o
r
e
f
u
lly
co
n
n
ec
ted
lay
er
s
to
d
o
th
e
f
in
al
class
if
icatio
n
task
.
T
h
is
f
ac
ilit
ates
th
e
f
u
tu
r
e
in
teg
r
atio
n
o
f
ac
q
u
ir
e
d
tr
aits
an
d
th
e
p
r
ed
ictio
n
o
f
u
ltima
te
f
ailu
r
e
[
2
1
]
.
T
h
e
o
u
tp
u
t
lay
e
r
d
eliv
er
s
th
e
p
r
o
jecte
d
class
p
r
o
b
ab
ilit
ies
o
r
lab
els
f
o
r
th
e
f
ai
lu
r
e
p
r
e
d
ictio
n
task
.
T
h
e
tr
ain
i
n
g
o
f
th
e
C
NN
m
o
d
el
en
co
m
p
ass
es
v
ar
io
u
s
s
tag
es
:
tr
an
s
f
o
r
m
in
g
tab
u
lar
d
ata
to
s
im
u
late
a
2
D
g
r
i
d
lay
o
u
t,
with
f
ea
tu
r
es
r
ep
r
esen
ted
as
ch
an
n
els.
Mo
d
el
i
n
itializatio
n
-
r
an
d
o
m
izin
g
t
h
e
weig
h
ts
o
f
th
e
C
NN
f
o
r
th
e
co
n
v
o
lu
tio
n
al
k
er
n
els
a
n
d
f
u
lly
co
n
n
ec
ted
la
y
er
s
.
Fo
r
wa
r
d
p
r
o
p
ag
at
i
o
n
—
th
is
p
r
o
ce
s
s
en
tails
tr
an
s
m
itti
n
g
in
p
u
t
d
ata
th
r
o
u
g
h
th
e
C
NN
ar
ch
itectu
r
e
to
ca
lcu
late
th
e
o
u
t
p
u
t
p
r
o
b
ab
ilit
ies
o
r
lab
els
[
2
2
]
.
L
o
s
s
ca
lcu
latio
n
:
ev
alu
ates
m
o
d
el
p
r
e
d
ictio
n
s
a
g
ain
s
t
ac
tu
al
lab
els
a
n
d
c
o
m
p
u
tes
a
lo
s
s
f
u
n
ctio
n
,
s
u
ch
as
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
,
wh
ich
q
u
an
tifie
s
th
e
er
r
o
r
.
B
ac
k
p
r
o
p
ag
atio
n
e
n
tails
ca
lcu
latin
g
th
e
g
r
ad
ien
ts
o
f
th
e
lo
s
s
f
u
n
ctio
n
co
n
ce
r
n
i
n
g
m
o
d
el
p
ar
am
eter
s
b
y
b
ac
k
tr
a
ck
in
g
,
f
o
llo
wed
b
y
weig
h
t
u
p
d
ates
u
tili
zin
g
a
n
o
p
tim
izatio
n
tech
n
i
q
u
e
s
u
ch
as
s
to
ch
asti
c
g
r
ad
ien
t
d
escen
t
[
2
3
]
.
I
ter
ativ
e
tr
ain
i
n
g
:
r
ep
ea
tin
g
s
tep
s
3
to
5
f
o
r
m
o
r
e
th
a
n
o
n
e
ep
o
ch
u
n
til
m
o
d
e
l
co
n
v
er
g
en
ce
,
o
r
m
a
x
im
al
p
er
f
o
r
m
an
ce
o
n
v
alid
atio
n
d
ata
is
ac
h
iev
ed
[
2
4
]
.
2
.
5
.
K
er
nel su
pp
o
rt
v
ec
t
o
r
m
a
chines
SVM
alg
o
r
ith
m
is
a
r
o
b
u
s
t
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
th
at
i
s
ca
p
ab
le
o
f
h
an
d
lin
g
lin
ea
r
as
well
a
s
non
-
lin
ea
r
class
if
icatio
n
task
s
[
2
5
]
.
Ker
n
el
SVM
was
ap
p
li
ed
to
m
o
d
el
n
o
n
-
lin
ea
r
r
elatio
n
s
h
ip
s
v
ia
im
p
licit
f
ea
tu
r
e
tr
an
s
f
o
r
m
atio
n
in
to
h
ig
h
er
-
d
im
en
s
io
n
al
s
p
ac
e
[
2
6
]
.
T
h
e
k
er
n
el
tr
ick
av
o
id
s
ex
p
licit
co
o
r
d
in
ate
co
m
p
u
tatio
n
,
r
ed
u
cin
g
c
o
m
p
u
t
atio
n
al
co
m
p
lex
ity
.
T
h
e
r
ad
ial
b
asis
f
u
n
ctio
n
(
R
B
F)
k
er
n
el
i
s
u
s
ed
as in
(
1
)
.
(
,
)
=
e
xp
(
−
‖
−
‖
2
)
(
1
)
T
h
is
ch
o
ice
b
alan
ce
s
f
lex
ib
ilit
y
an
d
g
en
er
aliza
tio
n
,
ca
p
tu
r
in
g
co
m
p
lex
d
ec
is
io
n
b
o
u
n
d
a
r
ies
wh
ile
m
itig
atin
g
o
v
er
f
itti
n
g
.
T
h
e
h
y
p
er
p
a
r
am
eter
co
n
tr
o
ls
s
en
s
itiv
ity
to
in
ter
-
s
am
p
le
d
is
tan
ce
s
,
o
p
tim
ized
v
ia
cr
o
s
s
-
v
alid
atio
n
.
C
o
m
p
ar
e
d
to
lin
ea
r
SVM,
R
B
F
-
k
er
n
el
SVM
b
etter
ex
p
lo
its
n
o
n
-
lin
ea
r
p
atter
n
s
in
th
e
Me
tr
o
PT
-
3
d
ataset.
Un
lik
e
n
o
n
lin
ea
r
SVM
wh
ich
m
ay
h
av
e
s
o
m
e
ad
v
an
tag
es
o
v
er
s
im
p
le
r
elatio
n
s
h
ip
s
lin
ea
r
SVM
i
s
p
r
ef
er
r
ed
b
ec
au
s
e
it
is
s
im
p
le
an
d
p
r
ec
is
e.
T
h
e
ta
s
k
o
f
h
y
p
e
r
p
ar
am
eter
tu
n
in
g
f
o
r
th
e
SVM
m
o
d
el
in
v
o
lv
ed
m
ain
ly
s
ee
k
in
g
f
o
r
t
h
e
b
est
v
alu
e
o
f
th
e
p
ar
am
ete
r
‘
C
’
.
B
y
s
y
s
tem
atica
lly
ad
ju
s
tin
g
th
e
‘
C
’
v
al
u
es
an
d
ev
alu
atin
g
h
o
w
th
ey
af
f
e
cted
th
e
ac
cu
r
ac
y
,
th
er
e
ap
p
e
ar
ed
to
b
e
a
co
n
s
tan
t
p
atter
n
o
f
h
ig
h
er
ac
cu
r
ac
y
lev
els
b
ein
g
attain
ed
with
i
n
cr
ea
s
ed
v
alu
es
o
f
‘
C
’
.
A
s
y
s
tem
atic
g
r
id
s
ea
r
ch
is
p
er
f
o
r
m
ed
t
o
tu
n
e
t
h
e
r
eg
u
lar
izatio
n
p
ar
am
eter
ca
n
d
th
e
k
er
n
el
c
o
ef
f
icien
t
.
T
h
e
r
an
g
es
ex
p
l
o
r
ed
wer
e
C
∈
{0
.
1
,
1,
10,
1
0
0
,
1
0
0
0
}
an
d
γ
∈
{0
.
0
0
1
,
0
.
0
1
,
0
.
1
,
1
}.
T
h
e
o
p
tim
a
l
co
n
f
ig
u
r
atio
n
s
elec
ted
b
ased
o
n
v
alid
atio
n
p
er
f
o
r
m
a
n
ce
was
C
=1
,
0
0
0
an
d
γ
=0
.
0
1
.
B
alan
cin
g
th
e
tr
ad
e
-
o
f
f
b
etwe
en
m
ar
g
in
m
ax
im
izatio
n
an
d
class
if
icatio
n
er
r
o
r
[
2
7
]
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
r
esu
lts
s
h
o
wed
th
at
th
e
a
d
v
an
ce
d
m
o
d
els,
esp
ec
ially
C
NN,
o
u
tp
er
f
o
r
m
ed
tr
ad
itio
n
al
m
eth
o
d
s
.
C
NN
ac
h
iev
ed
th
e
b
est p
er
f
o
r
m
an
ce
with
a
p
r
ec
is
io
n
o
f
0
.
9
5
,
r
ec
all
o
f
0
.
9
4
,
F1
-
s
co
r
e
o
f
0
.
9
4
,
an
d
AUC
-
R
OC
o
f
0
.
9
6
.
T
h
is
h
ig
h
lig
h
ts
C
NN
'
s
ab
il
ity
to
ca
p
tu
r
e
th
e
co
m
p
lex
an
d
n
o
n
-
lin
ea
r
r
elatio
n
s
h
ip
s
in
th
e
d
ata.
T
h
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
C
u
ttin
g
-
ed
g
e
a
l
g
o
r
ith
ms fo
r
r
elia
b
le
fa
ilu
r
e
p
r
ed
ictio
n
in
m
etro
tr
a
in
s
ystem
s
(
S
a
n
a
C
h
a
k
r
i
)
2273
d
is
cu
s
s
io
n
an
aly
ze
s
th
e
im
p
li
ca
tio
n
s
f
o
r
th
e
m
etr
o
in
d
u
s
tr
y
in
ter
m
s
o
f
ad
o
p
tin
g
m
o
r
e
ef
f
ec
tiv
e
p
r
ed
ictiv
e
m
ain
ten
an
ce
s
tr
ateg
ies.
3
.
1
.
Resul
t
s
Mu
ltip
le
m
ac
h
in
e
lear
n
in
g
m
o
d
els
wer
e
ass
es
s
ed
f
o
r
th
eir
ef
f
icac
y
in
f
o
r
ec
asti
n
g
p
r
o
b
lem
s
in
m
etr
o
tr
ain
s
y
s
tem
p
er
f
o
r
m
an
ce
.
T
h
e
an
aly
s
is
co
n
d
u
cted
to
ass
ess
th
e
im
p
o
r
tan
t
m
etr
ics
o
f
th
ese
m
o
d
els,
in
clu
d
in
g
p
r
ec
is
io
n
,
r
ec
all
,
F1
-
s
co
r
e
,
an
d
AUC
-
R
O
C
,
h
as
b
ee
n
p
er
f
o
r
m
ed
u
s
in
g
th
e
p
r
o
v
id
ed
d
ataset.
Ad
d
itio
n
ally
,
th
e
C
NN
wa
s
ev
alu
ated
ag
ain
s
t
MSE
to
ascer
tain
its
ef
f
icac
y
in
ca
p
tu
r
in
g
th
e
i
n
tr
in
s
ic
p
a
tter
n
s
o
f
a
d
ataset,
h
en
ce
ass
ess
in
g
its
r
eg
r
ess
io
n
ca
p
ab
ilit
ies.
T
h
e
r
esu
lts
o
f
th
e
m
etr
ics ar
e
p
r
esen
ted
in
T
ab
le
1
.
T
ab
le
1
.
Me
tr
ics f
o
r
m
o
d
els u
s
ed
M
o
d
e
l
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
AUC
-
R
O
C
M
S
E
G
B
M
0
.
9
1
0
.
8
9
0
.
9
0
0
.
9
2
-
C
N
N
0
.
9
5
0
.
9
4
0
.
9
4
0
.
9
6
0
.
1
2
K
e
r
n
e
l
S
V
M
0
.
9
3
0
.
9
2
0
.
9
2
0
.
9
4
-
T
h
e
C
NN
m
o
d
el
attain
s
s
u
p
e
r
io
r
p
e
r
f
o
r
m
an
ce
ac
r
o
s
s
all
m
etr
ics:
ac
cu
r
ac
y
(
0
.
9
5
,
9
5
%
c
o
n
f
id
en
ce
in
ter
v
als
(
CI
)
:
[
0
.
9
2
,
0
.
9
8
]
)
,
r
ec
all
(
0
.
9
4
,
9
5
%
C
I
:
[
0
.
9
1
,
0
.
9
7
]
)
,
F1
-
s
co
r
e
(
0
.
9
4
,
9
5
%
C
I
:
[
0
.
9
1
,
0
.
9
7
]
)
,
a
n
d
AUC
-
R
O
C
(
0
.
9
6
,
9
5
%
C
I
:
[
0
.
9
4
,
0
.
9
8
]
)
.
Sig
n
if
ican
tly
,
o
n
l
y
th
e
C
NN
m
o
d
el
y
ield
s
MSE
f
in
d
in
g
s
(
0
.
1
2
,
CI
at
9
5
%:
[
0
.
1
0
,
0
.
1
4
]
)
,
i
n
d
icatin
g
r
o
b
u
s
t r
eg
r
ess
io
n
ca
p
ab
ilit
ies.
Mo
r
eo
v
er
,
a
n
a
n
aly
s
is
o
f
th
e
9
5
%
C
I
r
ev
e
als
th
at
th
e
C
I
o
f
th
e
C
NN
m
o
d
el
[
0
.
9
4
,
0
.
9
8
]
f
o
r
th
e
AUC
-
R
OC
m
etr
ic
d
o
es
n
o
t
in
ter
s
ec
t
with
th
at
o
f
th
e
GB
M
m
o
d
el
[
0
.
8
9
,
0
.
9
5
]
.
T
h
e
a
b
s
en
ce
o
f
o
v
er
la
p
s
ig
n
if
ies
a
s
tatis
tically
s
ig
n
if
ican
t
d
is
p
ar
ity
in
p
e
r
f
o
r
m
an
ce
,
af
f
ir
m
in
g
th
at
t
h
e
C
NN
ar
ch
it
ec
tu
r
e
o
f
f
er
s
e
n
h
an
ce
d
p
r
ed
ict
iv
e
p
o
te
n
tial
r
elativ
e
t
o
th
e
en
s
em
b
le
tr
ee
-
b
ased
tech
n
iq
u
e
f
o
r
th
is
p
ar
ticu
lar
d
ataset.
A
p
air
ed
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
is
co
n
d
u
cted
to
ev
al
u
ate
th
e
s
tatis
tical
s
ig
n
if
ican
ce
o
f
th
e
d
if
f
er
e
n
ce
s
b
y
c
o
m
p
ar
i
n
g
th
e
AUC
-
R
OC
s
co
r
es
d
er
iv
ed
f
r
o
m
th
e
5
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
r
u
n
s
.
T
h
e
test
estab
lis
h
ed
th
at
t
h
e
C
NN
m
o
d
el
m
ar
k
ed
ly
s
u
r
p
ass
es
th
e
G
B
M
m
o
d
el
(
W
=
1
5
,
p
=0
.
0
4
3
,
α
=0
.
0
5
)
,
o
f
f
er
in
g
c
o
m
p
ellin
g
ev
id
e
n
ce
th
at
th
e
n
o
ted
p
er
f
o
r
m
an
ce
en
h
an
ce
m
e
n
t
is
n
o
t
attr
ib
u
tab
le
to
r
an
d
o
m
v
ar
iatio
n
.
T
h
ese
r
esu
lts
illu
s
tr
ate
th
e
ef
f
icac
y
o
f
th
e
C
NN
ar
ch
itectu
r
e
in
co
llectin
g
in
tr
icate
non
-
lin
ea
r
p
atter
n
s
th
r
o
u
g
h
co
n
v
o
lu
tio
n
al
an
d
p
o
o
lin
g
lay
er
s
.
T
h
is
v
alid
ates
its
ca
p
ac
ity
to
in
ter
p
r
et
i
n
tr
icate
p
atter
n
s
in
Me
tr
o
PT
-
3
,
u
tili
zin
g
co
n
v
o
lu
tio
n
al
an
d
p
o
o
lin
g
lay
er
s
as
f
u
n
d
am
en
tal
elem
e
n
ts
.
An
o
th
er
m
o
d
el
d
em
o
n
s
tr
atin
g
ex
ce
p
tio
n
al
p
e
r
f
o
r
m
an
ce
is
GB
M,
ac
h
iev
in
g
an
F1
-
s
co
r
e
o
f
0
.
9
0
(
9
5
%
C
I
: [
0
.
8
7
,
0
.
9
3
]
)
a
n
d
a
n
AUC
-
R
O
C
o
f
0
.
9
2
(
9
5
%
C
I
:
[
0
.
8
9
,
0
.
9
5
]
)
.
T
h
e
GB
Ms
ar
e
p
r
o
f
icien
t
in
co
n
s
tr
u
ctin
g
p
r
ed
ictio
n
m
o
d
els
b
y
u
tili
zin
g
en
s
em
b
les
o
f
d
ec
is
io
n
tr
ee
s
,
ad
ep
tly
m
an
ag
in
g
in
tr
icate
d
ata
lin
k
ag
es.
T
h
e
k
er
n
e
l
SVM
m
o
d
el
h
as
co
m
m
en
d
a
b
le
p
er
f
o
r
m
a
n
ce
,
a
ch
iev
in
g
a
n
F1
-
s
co
r
e
o
f
0
.
9
2
(
9
5
%
C
I
:
[
0
.
8
9
,
0
.
9
5
]
)
a
n
d
a
n
AUC
-
R
O
C
o
f
0
.
9
4
(
9
5
%
C
I
:
[
0
.
9
1
,
0
.
9
7
]
)
.
T
h
e
k
er
n
el
f
u
n
ctio
n
en
a
b
les
k
er
n
el
SVM
to
p
r
o
f
icien
tly
m
o
d
el
n
o
n
-
lin
ea
r
b
o
u
n
d
ar
ies,
m
ak
in
g
it
s
u
itab
le
f
o
r
th
is
cla
s
s
if
icatio
n
p
r
o
b
lem
.
Fig
u
r
e
3
p
r
esen
ts
th
e
co
n
f
u
s
io
n
m
atr
ic
es
f
o
r
th
e
C
NN
an
d
GB
M,
o
f
f
er
in
g
a
d
etailed
p
er
f
o
r
m
an
ce
an
aly
s
is
an
d
a
d
ir
ec
t
v
is
u
al
co
m
p
ar
is
o
n
o
f
v
ar
io
u
s
er
r
o
r
k
in
d
s
,
wh
ile
Fig
u
r
e
4
p
r
esen
ts
a
co
m
p
ar
a
tiv
e
b
ar
ch
ar
t
o
f
k
ey
class
if
icatio
n
m
etr
ics
—
p
r
ec
is
io
n
,
r
ec
all
,
F1
-
s
co
r
e
,
an
d
AUC
-
R
O
C
—
ac
r
o
s
s
th
e
G
B
M
,
C
NN,
an
d
SVM.
T
o
g
eth
er
,
th
ese
f
ig
u
r
es
o
f
f
er
co
m
p
r
eh
e
n
s
iv
e
in
s
ig
h
t
in
to
th
e
m
o
d
els'
er
r
o
r
p
r
o
f
iles
an
d
r
ela
tiv
e
s
tr
en
g
th
s
,
co
m
p
lem
e
n
tin
g
th
e
ag
g
r
eg
ated
m
etr
ics
in
T
ab
le
1
to
id
en
tify
th
e
m
o
s
t r
o
b
u
s
t m
o
d
el
f
o
r
m
etr
o
tr
ain
f
ailu
r
e
p
r
ed
ictio
n
.
Fig
u
r
e
3
.
C
o
n
f
u
s
io
n
m
atr
ices f
o
r
th
e
C
NN
an
d
GB
M
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
2
6
9
-
2
2
7
7
2274
Fig
u
r
e
4
.
C
o
m
p
a
r
ativ
e
p
e
r
f
o
r
m
an
ce
o
f
m
ac
h
in
e
lear
n
in
g
m
o
d
els
3
.
2
.
Dis
cus
s
io
n
T
h
e
in
v
esti
g
atio
n
in
d
icate
s
th
at
r
o
b
u
s
t
lin
ea
r
c
o
r
r
elatio
n
s
w
ith
in
th
e
Me
tr
o
PT
-
3
d
ataset
en
h
an
ce
th
e
d
if
f
er
en
tiatio
n
o
f
f
ailu
r
e
ca
s
es,
allo
win
g
co
n
v
en
tio
n
al
m
o
d
els
s
u
ch
as
lin
ea
r
SVM
s
to
f
u
n
ctio
n
ef
f
ec
tiv
ely
.
T
h
e
ex
ce
p
tio
n
al
p
er
f
o
r
m
an
ce
o
f
ad
v
an
ce
d
m
o
d
els
(
GB
M,
C
NN,
an
d
k
er
n
el
SVM)
h
ig
h
li
g
h
ts
th
eir
ca
p
ac
ity
to
d
is
ce
r
n
in
tr
icate
u
n
d
er
ly
in
g
s
t
r
u
ctu
r
es.
T
h
is
r
o
b
u
s
tn
ess
d
ep
en
d
s
s
ig
n
if
ican
tly
o
n
ap
p
r
o
p
r
iate
p
r
ep
r
o
ce
s
s
in
g
:
f
ea
tu
r
e
s
ca
lin
g
m
ain
tain
ed
u
n
if
o
r
m
co
n
tr
ib
u
tio
n
ac
r
o
s
s
v
ar
iab
les,
wh
ile
L
ASS
O
r
eg
u
lar
izatio
n
en
h
an
ce
d
m
o
d
el
p
er
f
o
r
m
a
n
ce
b
y
r
em
o
v
in
g
ir
r
elev
an
t
f
ea
t
u
r
es.
T
h
e
s
i
g
n
if
ic
an
t
c
o
n
n
ec
ti
o
n
s
i
d
e
n
t
if
ie
d
ac
r
o
s
s
v
a
r
ia
b
l
es
war
r
a
n
t
th
e
el
im
i
n
ati
o
n
o
f
t
h
e
n
aï
v
e
B
ay
es
cl
ass
i
f
ie
r
,
w
h
ic
h
p
r
es
u
m
es
in
d
ep
e
n
d
en
ce
,
i
n
f
a
v
o
r
o
f
m
o
d
el
’
s
a
d
e
p
t
at
m
a
n
ag
in
g
d
e
p
e
n
d
e
n
ci
es,
s
u
ch
as
GB
M
a
n
d
C
N
N.
T
h
e
k
er
n
e
l
SV
M
s
h
o
w
n
s
i
g
n
if
i
ca
n
t
e
f
f
ic
ac
y
(
F
1
-
s
c
o
r
e:
0
.
9
2
,
AUC
-
R
OC
:
0
.
9
4
)
b
y
a
d
e
p
tl
y
r
ep
r
ese
n
ti
n
g
n
o
n
-
l
in
ea
r
d
e
cisi
o
n
b
o
u
n
d
a
r
i
es t
h
r
o
u
g
h
th
e
k
e
r
n
el
f
u
n
c
ti
o
n
.
T
h
u
s
,
i
t
i
s
n
o
t
e
d
t
h
a
t
t
h
e
m
o
r
e
s
o
p
h
i
s
ti
c
a
t
e
d
m
o
d
e
ls
,
p
a
r
t
i
c
u
lar
l
y
C
N
N
,
o
u
t
p
e
r
f
o
r
m
o
t
h
e
r
c
o
n
v
e
n
t
i
o
n
a
l
m
o
d
e
l
s
i
n
f
o
r
e
c
a
s
t
i
n
g
f
a
i
l
u
r
es
r
e
l
a
t
e
d
t
o
t
h
e
m
e
t
r
o
t
r
ai
n
s
y
s
t
em
.
T
h
i
s
i
n
d
i
c
a
te
s
t
h
at
d
e
e
p
l
e
ar
n
i
n
g
a
p
p
r
o
a
c
h
e
s
a
r
e
h
i
g
h
l
y
e
f
f
e
c
t
i
v
e
f
o
r
a
d
d
r
e
s
s
i
n
g
s
u
c
h
d
i
f
f
i
c
u
l
t
i
es
,
as
t
h
e
y
f
ac
i
l
i
ta
t
e
t
h
e
i
d
e
n
t
i
f
i
c
at
i
o
n
o
f
c
o
m
p
l
e
x
p
a
t
t
e
r
n
s
a
n
d
r
e
l
a
t
i
o
n
s
h
i
p
s
w
i
t
h
i
n
d
at
a
.
T
h
e
r
e
s
u
l
ts
c
o
r
r
es
p
o
n
d
w
i
t
h
r
e
ce
n
t
r
e
s
e
a
r
c
h
b
y
Da
v
a
r
i
et
a
l
.
[
4
]
a
n
d
S
i
m
o
n
e
et
a
l
.
[
1
0
]
,
w
h
i
c
h
a
ls
o
e
m
p
h
a
s
i
ze
t
h
e
a
d
v
a
n
t
a
g
e
s
o
f
d
e
e
p
le
a
r
n
i
n
g
t
ec
h
n
i
q
u
e
s
c
o
m
p
a
r
e
d
t
o
c
o
n
v
en
t
i
o
n
a
l
m
et
h
o
d
s
f
o
r
p
r
e
d
i
c
t
i
v
e
m
a
i
n
t
e
n
a
n
c
e
i
n
r
a
i
lw
a
y
s
y
s
te
m
s
.
W
h
i
l
e
S
i
m
o
n
e
et
a
l
.
[
1
0
]
c
o
n
c
e
n
t
r
a
t
e
d
o
n
l
o
n
g
s
h
o
r
t
‑
t
e
r
m
m
e
m
o
r
y
(
L
S
T
M
)
a
r
c
h
it
e
c
t
u
r
es
,
o
u
r
a
n
a
ly
s
i
s
d
e
m
o
n
s
t
r
a
te
s
t
h
a
t
1
D
-
C
NN
s
s
e
r
v
e
as
a
h
i
g
h
l
y
e
f
f
e
ct
i
v
e
o
p
t
i
o
n
f
o
r
c
a
p
t
u
r
i
n
g
l
o
c
a
l
d
e
p
e
n
d
e
n
c
i
es
i
n
h
i
g
h
-
d
im
e
n
s
i
o
n
a
l
s
e
n
s
o
r
d
a
ta
.
C
N
N
d
e
m
o
n
s
t
r
a
t
es
e
n
h
a
n
c
e
d
p
e
r
f
o
r
m
a
n
c
e
(
F
1
=
0
.
9
4
)
b
u
t
r
e
q
u
i
r
e
s
5
.
4
t
i
m
e
s
t
h
e
t
r
a
i
n
i
n
g
d
u
r
a
t
i
o
n
(
4
5
.
7
m
i
n
u
t
e
s
)
a
n
d
h
a
s
a
m
e
m
o
r
y
n
e
e
d
1
3
t
i
m
e
s
g
r
e
a
t
e
r
(
1
5
.
8
M
B
)
t
h
a
n
k
e
r
n
e
l
S
V
M
(
8
.
5
m
i
n
u
t
e
s
,
1
.
2
M
B
,
F
1
=
0
.
9
2
)
.
I
n
r
e
s
o
u
r
c
e
-
li
m
i
t
e
d
e
d
g
e
d
e
p
l
o
y
m
e
n
ts
,
s
u
ch
a
s
o
n
-
b
o
a
r
d
t
r
a
i
n
m
o
n
i
t
o
r
i
n
g
s
y
s
t
e
m
s
w
i
t
h
r
es
t
r
i
ct
e
d
c
o
m
p
u
t
a
t
i
o
n
a
l
c
a
p
a
b
i
l
i
t
y
,
GB
M
p
r
o
v
i
d
e
s
a
n
o
p
t
i
m
a
l
e
q
u
i
l
i
b
r
i
u
m
,
c
o
m
p
r
o
m
i
s
i
n
g
m
e
r
e
l
y
4
.
2
%
ac
c
u
r
a
c
y
w
h
i
l
e
d
e
c
r
e
as
i
n
g
t
r
a
i
n
i
n
g
d
u
r
a
t
i
o
n
b
y
7
3
%
(
1
2
.
3
m
i
n
u
t
e
s
)
a
n
d
m
e
m
o
r
y
c
o
n
s
u
m
p
t
i
o
n
b
y
7
8
%
(
3
.
5
M
B
)
.
F
o
r
u
l
t
r
a
-
l
o
w
-
l
at
e
n
c
y
a
p
p
l
i
c
a
ti
o
n
s
n
e
c
es
s
i
t
a
ti
n
g
s
u
b
-
3
m
s
i
n
f
e
r
e
n
c
e
,
k
e
r
n
el
S
V
M
(
2
.
3
m
s
)
s
u
r
p
a
s
s
e
s
C
NN
(
5
.
4
m
s
)
b
y
5
7
%
,
r
e
n
d
e
r
i
n
g
i
t
m
o
r
e
s
u
i
t
a
b
l
e
f
o
r
r
e
a
l
-
t
i
m
e
ala
r
m
s
y
s
te
m
s
w
h
e
r
e
p
r
o
m
p
t
r
e
s
p
o
n
s
e
is
es
s
e
n
t
i
a
l
.
T
h
e
s
e
f
i
n
d
i
n
g
s
i
n
d
i
c
a
te
t
h
a
t
m
o
d
e
l
s
el
e
c
ti
o
n
m
u
s
t
c
o
n
s
i
d
e
r
o
p
e
r
a
t
i
o
n
a
l
r
es
t
r
i
ct
i
o
n
s
.
C
N
Ns
e
x
c
el
i
n
o
f
f
l
i
n
e
-
c
e
n
t
r
al
i
z
e
d
a
n
a
l
y
ti
c
s
,
w
h
e
r
e
as
s
i
m
p
l
e
r
m
o
d
e
l
s
a
r
e
c
r
u
ci
a
l
f
o
r
d
is
t
r
i
b
u
t
e
d
,
r
e
al
-
t
i
m
e
i
n
d
u
s
t
r
i
a
l
a
p
p
l
i
c
at
i
o
n
s
w
h
e
r
e
co
m
p
u
t
a
t
i
o
n
a
l
e
c
o
n
o
m
y
c
o
m
p
e
te
s
w
it
h
p
r
e
d
i
c
t
i
o
n
a
cc
u
r
a
c
y
.
C
o
n
s
eq
u
en
tly
,
th
e
f
in
d
in
g
s
p
o
s
s
ess
ap
p
licab
ilit
y
with
in
th
e
m
etr
o
i
n
d
u
s
tr
y
,
wh
er
e
t
h
e
cr
ea
ted
p
r
ed
ictiv
e
m
o
d
els
co
u
ld
b
e
u
tili
ze
d
in
o
p
er
atio
n
s
to
im
p
lem
en
t
p
r
o
ac
tiv
e
m
ain
te
n
an
c
e
m
ea
s
u
r
es.
Pre
cise
f
ailu
r
e
f
o
r
ec
asts
f
ac
ilit
ate
th
e
ap
p
r
o
p
r
iate
d
is
tr
ib
u
tio
n
o
f
m
a
in
ten
an
ce
r
eso
u
r
ce
s
,
h
e
n
ce
im
p
r
o
v
in
g
o
p
er
atio
n
al
ef
f
icien
cy
.
Mo
r
eo
v
e
r
,
p
ass
en
g
er
s
af
ety
is
im
p
r
o
v
ed
th
r
o
u
g
h
an
ticip
ato
r
y
a
n
d
p
r
ev
en
tiv
e
m
ea
s
u
r
es
th
r
o
u
g
h
o
u
t
tr
av
el
to
g
u
ar
a
n
tee
a
s
ea
m
less
ex
p
er
ien
ce
at
all
tim
es.
T
h
e
r
esear
ch
f
in
d
in
g
s
in
d
icate
th
at
th
e
ev
alu
ated
m
ac
h
in
e
lear
n
i
n
g
m
o
d
els
y
iel
d
f
av
o
r
ab
le
o
u
tco
m
es
f
o
r
p
r
ed
ictin
g
f
ailu
r
es
in
m
etr
o
r
ail
s
y
s
tem
s
.
T
h
e
ex
is
tin
g
m
eth
o
d
f
u
n
ctio
n
s
with
o
u
t
a
f
r
am
ewo
r
k
f
o
r
r
ep
r
esen
tin
g
te
m
p
o
r
al
d
ata.
Pre
d
ictio
n
o
u
tco
m
es
wo
u
ld
en
h
an
ce
b
y
s
eq
u
en
tial
d
ata
an
aly
s
is
o
f
s
en
s
o
r
d
ata
wh
en
tim
e
-
s
er
ies
an
aly
s
is
tech
n
iq
u
es
ar
e
em
p
lo
y
ed
.
Seco
n
d
ly
,
im
p
r
o
v
in
g
th
e
in
ter
p
r
etab
ilit
y
o
f
m
o
d
els
s
u
ch
as
C
NNs
co
n
tin
u
es
to
p
o
s
e
a
ch
allen
g
e.
C
o
m
p
r
eh
en
d
in
g
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
C
u
ttin
g
-
ed
g
e
a
l
g
o
r
ith
ms fo
r
r
elia
b
le
fa
ilu
r
e
p
r
ed
ictio
n
in
m
etro
tr
a
in
s
ystem
s
(
S
a
n
a
C
h
a
k
r
i
)
2275
m
o
d
el'
s
p
r
ed
ictiv
e
attr
ib
u
tes
will
f
ac
ilit
ate
th
e
estab
lis
h
m
en
t
o
f
tr
u
s
t
an
d
th
e
f
o
r
m
u
latio
n
o
f
e
n
h
an
ce
d
m
ain
ten
an
ce
s
tr
ateg
ies.
T
h
e
id
en
tifie
d
s
h
o
r
tco
m
i
n
g
s
o
f
e
x
is
tin
g
f
ailu
r
e
p
r
ed
ictio
n
s
y
s
tem
s
m
u
s
t b
e
r
ec
tifie
d
in
f
u
tu
r
e
r
esear
c
h
to
d
e
v
elo
p
m
o
r
e
d
ep
en
d
ab
le
an
d
ef
f
ec
tiv
e
p
r
e
d
ictio
n
s
y
s
tem
s
[
2
8
]
,
[
2
9
]
.
T
h
e
u
s
e
o
f
f
ailu
r
e
p
r
e
d
ictio
n
m
o
d
els
also
p
r
esen
ts
s
ev
er
al
eth
ical
im
p
licatio
n
s
.
T
h
e
s
y
s
tem
m
u
s
t
h
av
e
a
n
eq
u
itab
le
m
ain
te
n
an
ce
p
r
ed
ictio
n
f
r
am
ewo
r
k
th
a
t
av
o
id
s
b
ias
to
war
d
s
p
a
r
ticu
lar
m
etr
o
s
y
s
tem
r
eg
io
n
s
o
r
p
ass
en
g
er
d
e
m
o
g
r
a
p
h
ics.
User
s
r
eq
u
ir
e
s
y
s
tem
in
f
o
r
m
atio
n
r
eg
ar
d
in
g
d
ec
is
io
n
-
m
ak
in
g
alg
o
r
ith
m
s
an
d
th
e
id
en
tific
atio
n
o
f
p
o
te
n
tial
b
iases
to
f
o
s
ter
tr
u
s
t
a
n
d
ac
co
u
n
tab
ilit
y
i
n
th
ese
s
y
s
t
em
s
.
An
e
x
h
au
s
tiv
e
eth
ical
f
r
am
ewo
r
k
m
u
s
t
b
e
d
ev
elo
p
e
d
to
estab
lis
h
ap
p
r
o
p
r
iate
g
u
id
elin
es
f
o
r
th
e
u
tili
za
tio
n
o
f
th
ese
tech
n
o
lo
g
ies,
e
n
s
u
r
in
g
th
e
y
b
en
ef
it
all
s
tak
eh
o
l
d
er
s
p
o
s
itiv
ely
[
3
0
]
,
[
3
1
]
.
Alth
o
u
g
h
C
NNs
p
r
o
v
i
d
e
en
h
an
ce
d
p
er
f
o
r
m
an
ce
,
th
e
y
en
tail
g
r
e
ater
co
m
p
u
tin
g
ex
p
e
n
s
es
r
elativ
e
to
GB
M
o
r
SVM,
p
o
ten
tially
p
o
s
in
g
a
lim
itatio
n
f
o
r
lo
w
-
p
o
we
r
em
b
ed
d
ed
s
y
s
tem
s
.
Mo
r
eo
v
er
,
th
e
o
p
aq
u
e
c
h
ar
ac
ter
is
tics
o
f
d
ee
p
lear
n
in
g
p
r
esen
t
a
s
ig
n
if
ican
t
o
b
s
tacle
to
in
d
u
s
tr
ial
u
s
e.
Fu
tu
r
e
im
p
lem
en
tatio
n
s
co
u
ld
in
co
r
p
o
r
ate
ex
p
lai
n
a
b
ilit
y
f
r
am
ewo
r
k
s
s
u
ch
as
S
h
ap
ley
ad
d
itiv
e
ex
p
l
an
atio
n
s
(
SHAP
)
to
f
u
r
n
is
h
m
ain
ten
an
ce
team
s
with
in
ter
p
r
etab
le
in
s
ig
h
ts
o
n
th
e
r
atio
n
ale
b
e
h
in
d
s
p
ec
if
ic
f
ailu
r
e
p
r
ed
ictio
n
s
,
h
en
ce
b
o
ls
ter
in
g
f
aith
i
n
th
e
au
t
o
m
ated
s
y
s
tem
.
4.
CO
NCLU
SI
O
N
T
h
r
o
u
g
h
th
e
u
tili
za
tio
n
o
f
th
e
Me
tr
o
PT
-
3
d
ataset,
th
is
s
tu
d
y
in
v
esti
g
ated
th
e
d
eg
r
ee
to
wh
i
ch
v
a
r
io
u
s
m
ac
h
in
e
lear
n
in
g
tech
n
i
q
u
es
a
r
e
ab
le
to
ac
cu
r
ately
f
o
r
ec
ast
p
r
o
b
lem
s
i
n
m
et
r
o
t
r
ain
s
y
s
te
m
s
.
Desp
ite
th
e
f
ac
t
th
at
s
im
p
le
m
o
d
els
lik
e
S
VM
s
d
o
q
u
ite
well
wh
en
d
ea
lin
g
with
lin
ea
r
r
elatio
n
s
h
ip
s
,
th
e
f
in
d
in
g
s
d
em
o
n
s
tr
ated
t
h
at
m
o
r
e
c
o
m
p
licated
an
d
d
ee
p
lear
n
in
g
d
esi
g
n
s
p
er
f
o
r
m
s
ig
n
if
ican
tly
b
ett
er
wh
en
it
co
m
es
t
o
p
r
ed
ictio
n
.
W
ith
an
F1
-
s
co
r
e
o
f
0
.
9
4
an
d
an
AUC
-
R
OC
o
f
0
.
9
6
,
th
e
1
D
-
C
NN
m
o
d
el
in
p
ar
ticu
lar
d
i
d
ex
ce
p
tio
n
ally
well
ac
r
o
s
s
all
t
y
p
es o
f
test
s
.
T
h
is
d
em
o
n
s
tr
at
es th
at
it
is
s
u
p
er
io
r
to
GB
M
s
an
d
k
er
n
el
SVM
s
in
ter
m
s
o
f
its
ab
ilit
y
to
co
m
p
r
eh
en
d
in
tr
icate
d
ata
lin
k
ag
es.
A
f
u
r
th
er
f
in
d
in
g
m
ad
e
b
y
th
e
r
e
s
ea
r
ch
er
s
was
th
at
th
e
s
elec
tio
n
o
f
f
ea
tu
r
es
u
s
in
g
L
ASSO
r
eg
u
lar
izatio
n
is
e
s
s
en
tial
f
o
r
m
in
im
izin
g
th
e
a
m
o
u
n
t
o
f
d
ata
an
d
tr
ain
in
g
tim
e
r
e
q
u
ir
ed
with
o
u
t
co
m
p
r
o
m
is
in
g
ac
cu
r
ac
y
.
T
h
e
s
e
f
in
d
in
g
s
h
av
e
o
b
v
i
o
u
s
ap
p
l
icatio
n
s
in
th
e
r
ea
l
wo
r
ld
:
th
e
u
tili
za
tio
n
o
f
th
ese
p
r
ed
ictiv
e
m
o
d
els
en
ab
les
a
s
h
if
t
awa
y
f
r
o
m
ad
d
r
ess
in
g
is
s
u
es
af
ter
th
e
y
h
av
e
o
cc
u
r
r
e
d
an
d
to
war
d
p
r
ev
e
n
tin
g
th
em
.
T
h
is
r
esu
lts
in
a
s
ig
n
if
ican
t
im
p
r
o
v
em
en
t
in
t
h
e
u
tili
za
tio
n
o
f
r
eso
u
r
ce
s
an
d
an
i
n
cr
ea
s
e
in
p
ass
en
g
er
s
af
ety
.
T
h
e
r
esear
c
h
d
o
es
h
av
e
s
ev
er
al
lim
itatio
n
s
,
s
u
ch
as
th
e
ab
s
en
ce
o
f
tim
e
-
b
ased
s
eq
u
en
ce
m
o
d
elin
g
an
d
th
e
d
if
f
icu
lty
in
c
o
m
p
r
eh
en
d
in
g
h
o
w
d
ee
p
lear
n
i
n
g
m
o
d
els’
f
u
n
cti
o
n
.
T
h
e
p
r
im
a
r
y
ar
ea
s
o
f
f
o
cu
s
f
o
r
f
u
tu
r
e
r
esear
ch
will
b
e
th
e
cr
ea
tio
n
o
f
c
o
m
b
in
atio
n
d
esig
n
s
,
s
u
ch
a
s
C
NN
-
L
STM
,
in
o
r
d
er
t
o
m
ak
e
g
r
ea
ter
u
s
e
o
f
t
h
e
tim
e
-
b
ase
d
p
ar
t
o
f
th
e
d
ata,
a
n
d
th
e
ad
d
itio
n
o
f
f
r
am
ewo
r
k
s
th
at
ex
p
lain
th
e
m
o
d
els,
s
u
ch
as
SHA
P,
in
o
r
d
er
to
p
r
o
v
id
e
m
ain
ten
an
ce
team
s
with
in
f
o
r
m
atio
n
th
at
is
b
o
th
u
n
d
er
s
tan
d
a
b
le
an
d
o
p
er
atio
n
a
l.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
San
a
C
h
ak
r
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Nao
u
al
Mo
u
h
n
i
✓
✓
✓
✓
✓
Fao
u
zia
E
n
n
aa
m
a
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
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.
15
,
No
.
3
,
J
u
n
e
20
26
:
2
2
6
9
-
2
2
7
7
2276
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
th
at
s
u
p
p
o
r
t
th
e
f
i
n
d
in
g
s
o
f
t
h
is
s
tu
d
y
ar
e
o
p
e
n
ly
av
ailab
le
in
UC
I
m
ac
h
in
e
lear
n
in
g
r
ep
o
s
ito
r
y
at
h
ttp
s
://ar
ch
iv
e.
ic
s
.
u
ci.
ed
u
/d
ataset/7
9
1
/m
etr
o
p
t
+3
+d
ataset.
RE
F
E
R
E
NC
E
S
[
1
]
N
.
D
a
v
a
r
i
,
B
.
V
e
l
o
so
,
R
.
P
.
R
i
b
e
i
r
o
,
a
n
d
J
.
M
.
P
.
D
.
G
a
ma
,
“
D
e
t
e
c
t
i
n
g
a
n
d
e
x
p
l
a
i
n
i
n
g
a
n
o
ma
l
i
e
s
i
n
t
h
e
a
i
r
p
r
o
d
u
c
t
i
o
n
u
n
i
t
o
f
a
t
r
a
i
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
3
9
t
h
AC
M/
S
I
G
AP
P
S
y
m
p
o
si
u
m
o
n
Ap
p
l
i
e
d
C
o
m
p
u
t
i
n
g
,
A
p
r
.
2
0
2
4
,
p
p
.
3
5
8
–
3
6
4
,
d
o
i
:
1
0
.
1
1
4
5
/
3
6
0
5
0
9
8
.
3
6
3
5
9
0
6
.
[
2
]
H
.
S
u
n
e
t
a
l
.
,
“
P
r
e
v
e
n
t
i
v
e
ma
i
n
t
e
n
a
n
c
e
o
p
t
i
m
i
z
a
t
i
o
n
f
o
r
k
e
y
c
o
mp
o
n
e
n
t
s
o
f
su
b
w
a
y
t
r
a
i
n
b
o
g
i
e
w
i
t
h
c
o
n
s
i
d
e
r
a
t
i
o
n
o
f
f
a
i
l
u
r
e
r
i
s
k
,
”
En
g
i
n
e
e
ri
n
g
F
a
i
l
u
r
e
A
n
a
l
y
si
s
,
v
o
l
.
1
5
4
,
D
e
c
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
f
a
i
l
a
n
a
l
.
2
0
2
3
.
1
0
7
6
3
4
.
[
3
]
S
.
C
h
a
k
r
i
,
N
.
M
o
u
h
n
i
,
a
n
d
F
.
E
n
n
a
a
ma
,
“
E
x
p
l
o
r
i
n
g
t
h
e
f
r
o
n
t
i
e
r
s
o
f
t
r
a
j
e
c
t
o
r
y
o
u
t
l
i
e
r
d
e
t
e
c
t
i
o
n
:
a
n
i
n
-
d
e
p
t
h
r
e
v
i
e
w
a
n
d
c
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
si
s
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
E
l
e
c
t
r
i
c
a
l
a
n
d
C
o
m
p
u
t
e
r
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
1
4
,
n
o
.
5
,
p
p
.
5
9
8
4
–
5
9
9
7
,
O
c
t
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
4
i
5
.
p
p
5
9
8
4
-
5
9
9
7
.
[
4
]
N
.
D
a
v
a
r
i
,
B
.
V
e
l
o
s
o
,
R
.
P
.
R
i
b
e
i
r
o
,
P
.
M
.
P
e
r
e
i
r
a
,
a
n
d
J.
G
a
ma
,
“
P
r
e
d
i
c
t
i
v
e
m
a
i
n
t
e
n
a
n
c
e
b
a
se
d
o
n
a
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
f
o
r
a
i
r
p
r
o
d
u
c
t
i
o
n
u
n
i
t
i
n
t
h
e
r
a
i
l
w
a
y
i
n
d
u
st
r
y
,
”
i
n
2
0
2
1
I
E
EE
8
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
D
a
t
a
S
c
i
e
n
c
e
a
n
d
Ad
v
a
n
c
e
d
An
a
l
y
t
i
c
s
(
D
S
A
A)
,
O
c
t
.
2
0
2
1
,
p
p
.
1
–
10
,
d
o
i
:
1
0
.
1
1
0
9
/
D
S
A
A
5
3
3
1
6
.
2
0
2
1
.
9
5
6
4
1
8
1
.
[
5
]
B
.
V
e
l
o
s
o
,
R
.
P
.
R
i
b
e
i
r
o
,
J.
G
a
m
a
,
a
n
d
P
.
M
.
P
e
r
e
i
r
a
,
“
T
h
e
M
e
t
r
o
P
T
d
a
t
a
s
e
t
f
o
r
p
r
e
d
i
c
t
i
v
e
ma
i
n
t
e
n
a
n
c
e
,
”
S
c
i
e
n
t
i
f
i
c
D
a
t
a
,
v
o
l
.
9
,
n
o
.
1
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
5
9
7
-
0
2
2
-
0
1
8
7
7
-
3.
[
6
]
A
.
N
a
t
e
k
i
n
a
n
d
A
.
K
n
o
l
l
,
“
G
r
a
d
i
e
n
t
b
o
o
st
i
n
g
ma
c
h
i
n
e
s,
a
t
u
t
o
r
i
a
l
,
”
Fr
o
n
t
i
e
rs
i
n
N
e
u
ro
r
o
b
o
t
i
c
s
,
v
o
l
.
7
,
2
0
1
3
,
d
o
i
:
1
0
.
3
3
8
9
/
f
n
b
o
t
.
2
0
1
3
.
0
0
0
2
1
.
[
7
]
Z.
L
i
,
F
.
Li
u
,
W
.
Y
a
n
g
,
S
.
P
e
n
g
,
a
n
d
J.
Z
h
o
u
,
“
A
s
u
r
v
e
y
o
f
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
:
a
n
a
l
y
si
s,
a
p
p
l
i
c
a
t
i
o
n
s,
a
n
d
p
r
o
sp
e
c
t
s,”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
N
e
u
r
a
l
N
e
t
w
o
rks
a
n
d
L
e
a
r
n
i
n
g
S
y
st
e
m
s
,
v
o
l
.
3
3
,
n
o
.
1
2
,
p
p
.
6
9
9
9
–
7
0
1
9
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
TN
N
LS.
2
0
2
1
.
3
0
8
4
8
2
7
.
[
8
]
N
.
D
e
h
a
k
,
R
.
D
e
h
a
k
,
P
.
K
e
n
n
y
,
a
n
d
P
.
D
u
m
o
u
c
h
e
l
,
“
C
o
m
p
a
r
i
s
o
n
b
e
t
w
e
e
n
f
a
c
t
o
r
a
n
a
l
y
s
i
s
a
n
d
G
M
M
s
u
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
s
f
o
r
s
p
e
a
k
e
r
v
e
r
i
f
i
c
a
t
i
o
n
,
”
i
n
O
d
y
s
s
e
y
2
0
0
8
:
T
h
e
S
p
e
a
k
e
r
a
n
d
L
a
n
g
u
a
g
e
R
e
c
o
g
n
i
t
i
o
n
W
o
r
k
s
h
o
p
S
t
e
l
l
e
n
b
o
s
c
h
,
S
o
u
t
h
A
f
r
i
c
a
,
2
0
0
8
,
p
p
.
1
-
5
.
[
9
]
S
.
C
h
a
k
r
i
a
n
d
N
.
M
o
u
h
n
i
,
“
T
i
me
-
ser
i
e
s
n
e
u
r
a
l
n
e
t
w
o
r
k
p
r
e
d
i
c
t
i
o
n
s
u
si
n
g
LSTM
f
o
r
c
l
u
s
t
e
r
i
n
g
a
n
d
f
o
r
e
c
a
st
i
n
g
G
P
S
d
a
t
a
o
f
b
i
r
d
i
mm
i
g
r
a
t
i
o
n
,
”
i
n
2
0
2
4
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
G
l
o
b
a
l
Ae
r
o
n
a
u
t
i
c
a
l
E
n
g
i
n
e
e
ri
n
g
a
n
d
S
a
t
e
l
l
i
t
e
T
e
c
h
n
o
l
o
g
y
(
G
A
S
T
)
,
A
p
r
.
2
0
2
4
,
p
p
.
1
–
4
,
d
o
i
:
1
0
.
1
1
0
9
/
G
A
S
T6
0
5
2
8
.
2
0
2
4
.
1
0
5
2
0
7
6
4
.
[
1
0
]
L.
D
.
S
i
m
o
n
e
e
t
a
l
.
,
“
LSTM
-
b
a
s
e
d
f
a
i
l
u
r
e
p
r
e
d
i
c
t
i
o
n
f
o
r
r
a
i
l
w
a
y
r
o
l
l
i
n
g
st
o
c
k
e
q
u
i
p
m
e
n
t
,
”
Ex
p
e
r
t
S
y
s
t
e
m
s
w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
2
2
,
J
u
l
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
s
w
a
.
2
0
2
3
.
1
1
9
7
6
7
.
[
1
1
]
V
.
S
.
S
p
e
l
me
n
a
n
d
R
.
P
o
r
k
o
d
i
,
“
A
r
e
v
i
e
w
o
n
h
a
n
d
l
i
n
g
i
m
b
a
l
a
n
c
e
d
d
a
t
a
,
”
i
n
2
0
1
8
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
u
rre
n
t
T
r
e
n
d
s
t
o
w
a
r
d
s
C
o
n
v
e
r
g
i
n
g
T
e
c
h
n
o
l
o
g
i
e
s
(
I
C
C
T
C
T
)
,
M
a
r
.
2
0
1
8
,
p
p
.
1
–
11
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
T
C
T.
2
0
1
8
.
8
5
5
1
0
2
0
.
[
1
2
]
A
.
T
a
w
a
k
u
l
i
,
B
.
H
a
v
e
r
s,
V
.
G
u
l
i
s
a
n
o
,
D
.
K
a
i
s
e
r
,
a
n
d
T
.
En
g
e
l
,
“
S
u
r
v
e
y
:
T
i
me
-
seri
e
s d
a
t
a
p
r
e
p
r
o
c
e
ssi
n
g
:
a
s
u
r
v
e
y
a
n
d
a
n
e
m
p
i
r
i
c
a
l
a
n
a
l
y
si
s
,
”
J
o
u
r
n
a
l
o
f
E
n
g
i
n
e
e
r
i
n
g
Re
s
e
a
r
c
h
,
v
o
l
.
1
3
,
n
o
.
2
,
p
p
.
6
7
4
–
7
1
1
,
J
u
n
.
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
e
r
.
2
0
2
4
.
0
2
.
0
1
8
.
[
1
3
]
N
.
P
u
d
j
i
h
a
r
t
o
n
o
,
T.
F
a
d
a
so
n
,
A
.
W
.
K
.
-
Li
e
h
r
,
a
n
d
J.
M
.
O
’
S
u
l
l
i
v
a
n
,
“
A
r
e
v
i
e
w
o
f
f
e
a
t
u
r
e
s
e
l
e
c
t
i
o
n
m
e
t
h
o
d
s
f
o
r
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
-
b
a
s
e
d
d
i
se
a
se
r
i
sk
p
r
e
d
i
c
t
i
o
n
,
”
Fro
n
t
i
e
rs i
n
Bi
o
i
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
2
,
Ju
n
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
8
9
/
f
b
i
n
f
.
2
0
2
2
.
9
2
7
3
1
2
.
[
1
4
]
P
.
R
.
A
n
u
k
r
i
sh
n
a
a
n
d
V
.
P
a
u
l
,
“
A
r
e
v
i
e
w
o
n
f
e
a
t
u
r
e
se
l
e
c
t
i
o
n
f
o
r
h
i
g
h
d
i
men
s
i
o
n
a
l
d
a
t
a
,
”
i
n
2
0
1
7
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
n
v
e
n
t
i
v
e
S
y
s
t
e
m
s
a
n
d
C
o
n
t
r
o
l
(
I
C
I
S
C
)
,
Ja
n
.
2
0
1
7
,
p
p
,
1
–
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
S
C
.
2
0
1
7
.
8
0
6
8
7
4
6
.
[
1
5
]
V
.
K
.
A
y
y
a
d
e
v
a
r
a
,
“
G
r
a
d
i
e
n
t
b
o
o
s
t
i
n
g
ma
c
h
i
n
e
,
”
i
n
Pro
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
A
l
g
o
r
i
t
h
m
s
,
B
e
r
k
e
l
e
y
,
C
a
l
i
f
o
r
n
i
a
:
A
p
r
e
ss
,
2
0
1
8
,
p
p
.
1
1
7
–
134
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
1
-
4
8
4
2
-
3
5
6
4
-
5
_
6
.
[
1
6
]
M
.
M
.
Ö
z
t
ü
r
k
,
“
C
o
m
p
a
r
i
n
g
h
y
p
e
r
p
a
r
a
met
e
r
o
p
t
i
m
i
z
a
t
i
o
n
i
n
c
r
o
ss
-
a
n
d
w
i
t
h
i
n
-
p
r
o
j
e
c
t
d
e
f
e
c
t
p
r
e
d
i
c
t
i
o
n
:
a
c
a
se
st
u
d
y
,
”
Ar
a
b
i
a
n
J
o
u
rn
a
l
f
o
r
S
c
i
e
n
c
e
a
n
d
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
4
4
,
n
o
.
4
,
p
p
.
3
5
1
5
–
3
5
3
0
,
A
p
r
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
3
3
6
9
-
018
-
3
5
6
4
-
9.
[
1
7
]
C
.
B
e
n
t
é
j
a
c
,
A
.
C
s
ö
r
g
ő
,
a
n
d
G
.
M
.
M
u
ñ
o
z
,
“
A
c
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
si
s
o
f
g
r
a
d
i
e
n
t
b
o
o
st
i
n
g
a
l
g
o
r
i
t
h
ms,
”
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
Re
v
i
e
w
,
v
o
l
.
5
4
,
n
o
.
3
,
p
p
.
1
9
3
7
–
1
9
6
7
,
M
a
r
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
4
6
2
-
0
2
0
-
0
9
8
9
6
-
5.
[
1
8
]
G
.
B
i
a
u
,
B
.
C
a
d
r
e
,
a
n
d
L.
R
o
u
v
i
è
r
e
,
“
A
c
c
e
l
e
r
a
t
e
d
g
r
a
d
i
e
n
t
b
o
o
s
t
i
n
g
,
”
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
v
o
l
.
1
0
8
,
n
o
.
6
,
p
p
.
9
7
1
–
9
9
2
,
J
u
n
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
9
9
4
-
019
-
0
5
7
8
7
-
1.
[
1
9
]
R
.
Y
a
n
g
e
t
a
l
.
,
“
C
N
N
-
LST
M
d
e
e
p
l
e
a
r
n
i
n
g
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
c
o
m
p
u
t
e
r
v
i
si
o
n
-
b
a
s
e
d
m
o
d
a
l
f
r
e
q
u
e
n
c
y
d
e
t
e
c
t
i
o
n
,
”
M
e
c
h
a
n
i
c
a
l
S
y
s
t
e
m
s
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ss
i
n
g
,
v
o
l
.
1
4
4
,
O
c
t
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
y
mss
p
.
2
0
2
0
.
1
0
6
8
8
5
.
[
2
0
]
A
.
B
a
n
s
a
l
,
V
.
A
.
A
t
h
a
v
a
l
e
,
A
.
S
i
n
g
h
,
A
d
i
t
i
,
M
.
A
.
S
h
a
h
m
i
r
i
,
a
n
d
K
.
D
.
S
i
n
g
h
,
“
3
D
-
C
N
N
e
m
p
o
w
e
r
e
d
a
ss
i
st
i
v
e
mac
h
i
n
e
l
e
a
r
n
i
n
g
mo
d
e
l
f
o
r
t
h
e
h
e
a
r
i
n
g
i
mp
a
i
r
e
d
,
”
i
n
2
0
2
3
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
u
st
a
i
n
a
b
l
e
Em
e
r
g
i
n
g
I
n
n
o
v
a
t
i
o
n
s
i
n
En
g
i
n
e
e
ri
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
(
I
C
S
EI
ET)
,
S
e
p
.
2
0
2
3
,
p
p
.
4
8
7
–
4
9
0
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
EI
ET5
8
6
7
7
.
2
0
2
3
.
1
0
3
0
3
0
4
5
.
[
2
1
]
A
.
C
a
sam
i
t
j
a
n
a
,
S
.
P
u
c
h
,
A
.
A
d
u
r
i
z
,
a
n
d
V
.
V
i
l
a
p
l
a
n
a
,
“
3
D
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
b
r
a
i
n
t
u
m
o
r
se
g
me
n
t
a
t
i
o
n
:
a
c
o
m
p
a
r
i
so
n
o
f
mu
l
t
i
-
r
e
so
l
u
t
i
o
n
a
r
c
h
i
t
e
c
t
u
r
e
s
,
”
i
n
Br
a
i
n
l
e
s
i
o
n
:
G
l
i
o
m
a
,
Mu
l
t
i
p
l
e
S
c
l
e
r
o
s
i
s,
S
t
ro
k
e
a
n
d
T
r
a
u
m
a
t
i
c
Br
a
i
n
I
n
j
u
r
i
e
s
,
S
p
r
i
n
g
e
r
,
C
h
a
m,
2
0
1
6
,
p
p
.
1
5
0
–
1
6
1
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
3
1
9
-
5
5
5
2
4
-
9
_
1
5
.
[
2
2
]
A
.
K
h
a
n
,
A
.
S
o
h
a
i
l
,
U
.
Za
h
o
o
r
a
,
a
n
d
A
.
S
.
Q
u
r
e
s
h
i
,
“
A
s
u
r
v
e
y
o
f
t
h
e
r
e
c
e
n
t
a
r
c
h
i
t
e
c
t
u
r
e
s o
f
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
Re
v
i
e
w
,
v
o
l
.
5
3
,
n
o
.
8
,
p
p
.
5
4
5
5
–
5
5
1
6
,
D
e
c
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
4
6
2
-
020
-
0
9
8
2
5
-
6.
[
2
3
]
G
.
P
a
t
r
i
n
i
,
A
.
R
o
z
z
a
,
A
.
K
.
M
e
n
o
n
,
R
.
N
o
c
k
,
a
n
d
L.
Q
u
,
“
M
a
k
i
n
g
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
r
o
b
u
st
t
o
l
a
b
e
l
n
o
i
se
:
a
l
o
ss
c
o
r
r
e
c
t
i
o
n
a
p
p
r
o
a
c
h
,
”
i
n
2
0
1
7
I
E
EE
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
(
C
VPR)
,
J
u
l
.
2
0
1
7
,
p
p
.
2
2
3
3
–
2
2
4
1
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
7
.
2
4
0
.
[
2
4
]
P
.
P
u
r
w
o
n
o
,
A
.
M
a
’
a
r
i
f
,
W
.
R
a
h
m
a
n
i
a
r
,
H
.
I
.
K
.
F
a
t
h
u
r
r
a
h
ma
n
,
A
.
Z
.
K
.
F
r
i
s
k
y
,
a
n
d
Q
.
M
.
U.
H
a
q
,
“
U
n
d
e
r
st
a
n
d
i
n
g
o
f
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
(
C
N
N
)
:
a
r
e
v
i
e
w
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Ro
b
o
t
i
c
s
a
n
d
C
o
n
t
ro
l
S
y
st
e
m
s
,
v
o
l
.
2
,
n
o
.
4
,
p
p
.
7
3
9
–
7
4
8
,
J
a
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
1
7
6
3
/
i
j
r
c
s
.
v
2
i
4
.
8
8
8
.
[
2
5
]
H
.
B
h
a
v
sar
a
n
d
M
.
H
.
P
a
n
c
h
a
l
,
“
A
r
e
v
i
e
w
o
n
s
u
p
p
o
r
t
v
e
c
t
o
r
ma
c
h
i
n
e
f
o
r
d
a
t
a
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
Re
se
a
rc
h
i
n
C
o
m
p
u
t
e
r
En
g
i
n
e
e
ri
n
g
&
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
,
n
o
.
1
0
,
2
0
1
2
.
[
2
6
]
D
.
A
.
P
i
s
n
e
r
a
n
d
D
.
M
.
S
c
h
n
y
e
r
,
“
S
u
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
,
”
i
n
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
Me
t
h
o
d
s
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
t
o
Bra
i
n
D
i
s
o
rd
e
rs
,
C
a
m
b
r
i
d
g
e
,
U
n
i
t
e
d
S
t
a
t
e
s
:
A
c
a
d
e
mi
c
P
r
e
ss
,
2
0
2
0
,
p
p
.
1
0
1
–
1
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
B
9
7
8
-
0
-
12
-
8
1
5
7
3
9
-
8
.
0
0
0
0
6
-
7.
[
2
7
]
R
.
G
u
i
d
o
,
M
.
C
.
G
r
o
c
c
i
a
,
a
n
d
D
.
C
o
n
f
o
r
t
i
,
“
H
y
p
e
r
-
p
a
r
a
m
e
t
e
r
o
p
t
i
m
i
z
a
t
i
o
n
i
n
s
u
p
p
o
r
t
v
e
c
t
o
r
ma
c
h
i
n
e
o
n
u
n
b
a
l
a
n
c
e
d
d
a
t
a
set
s
u
si
n
g
g
e
n
e
t
i
c
a
l
g
o
r
i
t
h
ms
,
”
i
n
O
p
t
i
m
i
za
t
i
o
n
i
n
A
rt
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
D
a
t
a
S
c
i
e
n
c
e
s
,
S
p
r
i
n
g
e
r
,
C
h
a
m,
2
0
2
2
,
p
p
.
3
7
–
47
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
030
-
9
5
3
8
0
-
5
_
4
.
[
2
8
]
R
.
M
e
y
e
s,
N
.
H
ü
t
t
e
n
,
a
n
d
T.
M
e
i
sen
,
“
Tr
a
n
s
p
a
r
e
n
t
a
n
d
i
n
t
e
r
p
r
e
t
a
b
l
e
f
a
i
l
u
r
e
p
r
e
d
i
c
t
i
o
n
o
f
se
n
so
r
t
i
m
e
seri
e
s
d
a
t
a
w
i
t
h
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
Pr
o
c
e
d
i
a
C
I
RP
,
v
o
l
.
1
0
4
,
p
p
.
1
4
4
6
–
1
4
5
1
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
i
r
.
2
0
2
1
.
1
1
.
2
4
4
.
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
C
u
ttin
g
-
ed
g
e
a
l
g
o
r
ith
ms fo
r
r
elia
b
le
fa
ilu
r
e
p
r
ed
ictio
n
in
m
etro
tr
a
in
s
ystem
s
(
S
a
n
a
C
h
a
k
r
i
)
2277
[
2
9
]
J.
U
d
d
o
h
,
D
.
A
j
i
g
a
,
B
.
P
.
O
k
a
r
e
,
a
n
d
T.
D
.
A
d
u
l
o
j
u
,
“
S
t
r
e
a
mi
n
g
a
n
a
l
y
t
i
c
s
a
n
d
p
r
e
d
i
c
t
i
v
e
ma
i
n
t
e
n
a
n
c
e
:
r
e
a
l
-
t
i
m
e
a
p
p
l
i
c
a
t
i
o
n
s
i
n
i
n
d
u
st
r
i
a
l
m
a
n
u
f
a
c
t
u
r
i
n
g
s
y
st
e
ms,
”
J
o
u
rn
a
l
o
f
Fro
n
t
i
e
r
s
i
n
Mu
l
t
i
d
i
sc
i
p
l
i
n
a
r
y
Re
s
e
a
r
c
h
,
v
o
l
.
2
,
n
o
.
1
,
p
p
.
2
8
5
–
2
9
1
,
2
0
2
1
,
d
o
i
:
1
0
.
5
4
6
6
0
/
.
I
JF
M
R
.
2
0
2
1
.
2
.
1
.
2
8
5
-
2
9
1
.
[
3
0
]
A
.
C
o
n
s
i
l
v
i
o
,
G
.
V
i
g
n
o
l
a
,
P
.
L
.
A
r
é
v
a
l
o
,
F
.
G
a
l
l
o
,
M
.
B
o
r
i
n
a
t
o
,
a
n
d
C
.
C
r
o
v
e
t
t
o
,
“
A
d
a
t
a
-
d
r
i
v
e
n
p
r
i
o
r
i
t
i
sa
t
i
o
n
f
r
a
mew
o
r
k
t
o
mi
t
i
g
a
t
e
m
a
i
n
t
e
n
a
n
c
e
i
m
p
a
c
t
o
n
p
a
ss
e
n
g
e
r
s
d
u
r
i
n
g
m
e
t
r
o
l
i
n
e
o
p
e
r
a
t
i
o
n
,
”
Eu
r
o
p
e
a
n
T
ra
n
s
p
o
r
t
Re
se
a
r
c
h
Re
v
i
e
w
,
v
o
l
.
1
6
,
n
o
.
1
,
Jan
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
8
6
/
s1
2
5
4
4
-
023
-
0
0
6
3
1
-
z.
[
3
1
]
H
.
S
y
a
h
mi
,
“
P
r
i
v
a
c
y
a
n
d
e
t
h
i
c
a
l
i
m
p
l
i
c
a
t
i
o
n
s
o
f
b
i
g
d
a
t
a
u
t
i
l
i
z
a
t
i
o
n
i
n
p
u
b
l
i
c
t
r
a
n
sp
o
r
t
a
t
i
o
n
s
u
r
v
e
i
l
l
a
n
c
e
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Ad
v
a
n
c
e
d
C
y
b
e
rsec
u
r
i
t
y
S
y
st
e
m
s,
T
e
c
h
n
o
l
o
g
i
e
s,
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
9
,
n
o
.
1
,
p
p
.
1
–
1
0
,
2
0
2
5
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
S
a
n
a
Cha
k
r
i
h
o
l
d
s
a
P
h
.
D.
in
S
o
ftwa
re
En
g
i
n
e
e
rin
g
fr
o
m
F
S
TG
-
UC
A
M
a
rra
k
e
c
h
a
n
d
a
n
e
n
g
in
e
e
ri
n
g
d
e
g
re
e
in
S
o
ftwa
re
En
g
i
n
e
e
rin
g
fro
m
F
S
T
M
o
h
a
m
m
e
d
ia
-
UH
2
Ca
sa
b
lan
c
a
.
S
h
e
c
o
m
p
lete
d
h
e
r
DU
T
in
Co
m
p
u
ter
S
c
ien
c
e
fr
o
m
F
S
TG
-
UCA
M
a
rra
k
e
c
h
.
Th
e
ir
in
fl
u
e
n
ti
a
l
sc
h
o
larly
a
rti
c
les
o
n
se
m
a
n
ti
c
traje
c
to
ry
k
n
o
wle
d
g
e
d
isc
o
v
e
r
y
h
a
v
e
il
lu
m
in
a
ted
m
o
v
in
g
o
b
jec
t
b
e
h
a
v
io
rs
a
n
d
p
a
tt
e
r
n
e
x
trac
ti
o
n
.
He
r
e
x
p
e
rti
se
in
d
a
ta
m
i
n
in
g
b
rid
g
e
s
th
e
o
re
ti
c
a
l
re
se
a
rc
h
a
n
d
p
ra
c
ti
c
a
l
a
p
p
li
c
a
ti
o
n
s,
a
d
d
re
ss
in
g
re
a
l
-
wo
rld
c
h
a
ll
e
n
g
e
s
wit
h
in
n
o
v
a
ti
v
e
so
lu
ti
o
n
s
.
Th
e
ir
c
o
m
m
it
m
e
n
t
to
c
u
tt
in
g
-
e
d
g
e
re
se
a
rc
h
in
o
u
t
li
e
r
d
e
tec
ti
o
n
s
y
ste
m
s
e
n
ric
h
e
s
th
e
sc
ien
ti
fic
c
o
m
m
u
n
it
y
a
n
d
in
sp
i
re
s
th
e
n
e
x
t
g
e
n
e
ra
ti
o
n
o
f
re
se
a
rc
h
e
rs.
He
r
c
u
rre
n
t
re
se
a
rc
h
fo
c
u
se
s
o
n
n
e
w
m
e
th
o
d
s
fo
r
o
u
tl
ier
d
e
tec
ti
o
n
,
a
imin
g
to
d
e
e
p
e
n
u
n
d
e
rsta
n
d
in
g
in
t
h
e
field
a
n
d
fo
ste
r
t
h
e
d
e
v
e
lo
p
m
e
n
t
o
f
ro
b
u
st
o
u
t
li
e
r
d
e
tec
ti
o
n
sy
ste
m
s.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
c
h
a
k
ri.
sa
n
a
@g
m
a
il
.
c
o
m
.
Na
o
u
a
l
M
o
u
h
n
i
h
o
ld
s
a
P
h
.
D
.
in
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
is
c
u
r
re
n
tl
y
a
p
ro
fe
ss
o
r
a
t
F
S
TE
Un
iv
e
rsity
M
o
u
lay
Ism
a
il
.
P
re
v
io
u
sly
,
sh
e
se
rv
e
d
a
s
th
e
d
irec
to
r
o
f
re
se
a
rc
h
a
n
d
d
e
v
e
lo
p
m
e
n
t
a
t
t
h
e
LAM
IG
EP
m
u
lt
id
isc
ip
li
n
a
ry
re
se
a
rc
h
lab
o
ra
t
o
r
y
a
t
EM
S
I
M
a
rra
k
e
c
h
.
S
h
e
h
a
s
b
e
e
n
tea
c
h
in
g
a
s
a
c
o
m
p
u
ter
sc
ien
c
e
p
ro
fe
ss
o
r
a
t
th
e
sa
m
e
in
stit
u
ti
o
n
si
n
c
e
2
0
1
2
.
S
h
e
is
a
p
a
ss
io
n
a
te
a
d
v
o
c
a
te
fo
r
tec
h
n
o
l
o
g
y
a
n
d
i
n
n
o
v
a
ti
o
n
,
a
n
d
h
o
l
d
s
a
v
a
riety
o
f
c
e
rti
fica
ti
o
n
s
t
h
a
t
sh
o
wc
a
se
h
e
r
b
ro
a
d
ra
n
g
e
o
f
sk
il
ls.
S
h
e
h
a
s
o
b
tain
e
d
s
p
e
c
ialize
d
train
in
g
i
n
a
rti
ficia
l
in
telli
g
e
n
c
e
fro
m
th
e
NV
IDIA
De
e
p
Lea
rn
in
g
In
stit
u
te,
e
d
u
c
a
ti
o
n
a
l
a
d
v
a
n
c
e
m
e
n
ts
fro
m
Ho
n
o
r
is
Un
iv
e
rsity
,
e
x
p
e
rti
se
in
b
ig
d
a
ta
fro
m
IBM
,
a
n
d
p
r
o
ficie
n
c
y
in
d
a
tab
a
se
m
a
n
a
g
e
m
e
n
t
fro
m
ORA
CLE
.
He
r
re
se
a
rc
h
fo
c
u
se
s
o
n
p
u
sh
in
g
t
h
e
b
o
u
n
d
a
ri
e
s
o
f
a
rti
ficia
l
in
tell
ig
e
n
c
e
,
re
n
e
wa
b
le
e
n
e
rg
ies
,
a
n
d
d
a
tab
a
se
tec
h
n
o
lo
g
ies
.
He
r
c
o
m
m
it
m
e
n
t
t
o
a
d
v
a
n
c
in
g
th
e
se
fiel
d
s
is
e
v
id
e
n
t
t
h
ro
u
g
h
h
e
r
o
n
g
o
in
g
re
se
a
rc
h
a
n
d
c
o
n
tri
b
u
ti
o
n
s
to
e
a
c
h
a
re
a
.
S
h
e
se
rv
e
s
a
s
a
se
n
io
r
m
e
m
b
e
r
o
f
th
e
IEE
E
a
n
d
is
a
c
ti
v
e
ly
in
v
o
l
v
e
d
i
n
se
v
e
ra
l
a
c
a
d
e
m
ic
a
n
d
p
ro
fe
ss
io
n
a
l
c
o
m
m
u
n
it
ies
.
S
h
e
c
a
n
b
e
c
o
n
tac
te
d
a
t
e
m
a
il
:
n
.
m
o
u
h
n
i@
u
m
i.
a
c
.
m
a
.
Fa
o
u
z
i
a
E
n
n
a
a
m
a
g
ra
d
u
a
ted
wit
h
a
m
a
ste
r'
s
d
e
g
re
e
in
El
e
c
tri
c
a
l
En
g
in
e
e
ri
n
g
a
n
d
a
P
h
.
D.
i
n
El
e
c
tri
c
a
l
En
g
i
n
e
e
rin
g
in
Tele
c
o
m
m
u
n
ica
ti
o
n
s
a
n
d
In
d
u
strial
Tec
h
n
i
q
u
e
s,
sh
e
o
b
tai
n
e
d
h
e
r
d
o
c
to
ra
te
a
fter
c
o
n
d
u
c
ti
n
g
h
e
r
re
se
a
rc
h
a
t
th
e
LAP
S
S
I
L
a
b
o
ra
to
r
y
a
t
Ca
d
i
Ay
y
a
d
Un
iv
e
rsity
in
M
a
rra
k
e
c
h
,
M
o
ro
c
c
o
.
P
a
s
si
o
n
a
te
a
b
o
u
t
fa
c
ial
re
c
o
g
n
i
ti
o
n
,
c
o
m
p
u
te
r
v
isi
o
n
,
e
lec
tr
o
n
i
c
s,
a
n
d
p
ro
g
ra
m
m
i
n
g
,
s
h
e
h
a
s
p
u
b
li
s
h
e
d
p
a
p
e
rs
i
n
i
n
d
e
x
e
d
j
o
u
r
n
a
ls
a
n
d
p
a
r
ti
c
i
p
a
te
d
i
n
se
v
e
ra
l
i
n
te
r
n
a
t
i
o
n
a
l
c
o
n
fe
re
n
c
e
s.
S
h
e
c
a
n
b
e
c
o
n
t
a
c
te
d
a
t
e
m
a
i
l:
fa
o
u
z
ia.
e
n
n
a
a
m
a
@c
e
d
.
u
c
a
.
m
a
.
Evaluation Warning : The document was created with Spire.PDF for Python.