T
E
L
KO
M
N
I
KA
T
e
lec
om
m
u
n
icat
ion
,
Com
p
u
t
i
n
g,
E
lec
t
r
on
ics
an
d
Cont
r
ol
Vol.
18
,
No.
3
,
J
une
2020
,
pp.
1643
~
1649
I
S
S
N:
1693
-
6930,
a
c
c
r
e
dit
e
d
F
ir
s
t
G
r
a
de
by
Ke
me
nr
is
tekdikti
,
De
c
r
e
e
No:
21/E
/KP
T
/2018
DO
I
:
10.
12928/
T
E
L
KO
M
NI
KA
.
v18i3.
14847
1643
Jou
r
n
al
h
omepage
:
ht
tp:
//
jour
nal.
uad
.
ac
.
id/
index
.
php/T
E
L
K
OM
N
I
K
A
T
h
e
c
om
p
ar
is
o
n
st
u
d
y of
k
e
r
n
e
l
KC
-
m
e
an
s
an
d
su
p
p
o
r
t
v
e
c
t
or
m
a
c
h
i
n
e
s
f
or
c
la
ssi
f
yi
n
g s
c
h
iz
op
h
r
e
n
ia
S
r
i
Hart
in
i
,
Z
u
h
e
r
m
an
Rus
t
am
D
ep
ar
t
men
t
o
f
Mat
h
em
at
i
cs
,
U
n
i
v
ers
i
t
y
o
f
In
d
o
n
es
i
a
,
In
d
o
n
es
i
a
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
Aug
15
,
2019
R
e
vis
e
d
J
a
n
4
,
2020
Ac
c
e
pted
F
e
b
23
,
2020
Sch
i
z
o
p
h
ren
i
a
i
s
o
n
e
o
f
men
t
a
l
d
i
s
o
r
d
er
t
h
a
t
affect
s
t
h
e
mi
n
d
,
feel
i
n
g
,
an
d
b
eh
a
v
i
o
r.
It
s
t
reat
me
n
t
i
s
u
s
u
al
l
y
p
erman
e
n
t
an
d
q
u
i
t
e
co
mp
l
i
ca
t
ed
;
t
h
eref
o
re,
earl
y
d
et
ec
t
i
o
n
i
s
i
mp
o
rt
a
n
t
.
K
er
n
el
K
C
-
mea
n
s
an
d
s
u
p
p
o
r
t
v
ec
t
o
r
mach
i
n
es
are
t
h
e
met
h
o
d
s
k
n
o
w
n
a
s
a
g
o
o
d
c
l
as
s
i
fi
er.
T
h
i
s
res
earc
h
,
t
h
eref
o
re,
ai
ms
t
o
co
m
p
are
k
er
n
el
K
C
-
mea
n
s
a
n
d
s
u
p
p
o
rt
v
ect
o
r
mach
i
n
e
s
,
u
s
i
n
g
d
a
t
a
o
b
t
a
i
n
ed
fr
o
m
N
o
r
t
h
w
es
t
ern
U
n
i
v
er
s
i
t
y
,
w
h
i
c
h
c
o
n
s
i
s
t
s
o
f
1
7
1
s
c
h
i
z
o
p
h
ren
i
a
an
d
2
2
1
n
o
n
-
s
c
h
i
z
o
p
h
ren
i
a
s
amp
l
es
.
T
h
e
p
erfo
rman
c
e
accu
racy
,
F1
-
s
co
re,
an
d
ru
n
n
i
n
g
t
i
me
w
ere
ex
am
i
n
e
d
u
s
i
n
g
t
h
e
1
0
-
f
o
l
d
cro
s
s
-
v
a
l
i
d
at
i
o
n
met
h
o
d
.
Fr
o
m
t
h
e
ex
p
er
i
men
t
s
,
k
er
n
el
K
C
-
mea
n
s
w
i
t
h
t
h
e
s
i
x
t
h
-
o
rd
er
p
o
l
y
n
o
mi
a
l
k
ern
e
l
g
i
v
es
8
7
.
1
8
p
e
rcen
t
accu
rac
y
an
d
9
3
.
1
5
p
ercen
t
F1
-
s
c
o
re
at
t
h
e
fas
t
er
ru
n
n
i
n
g
t
i
me
t
h
a
n
s
u
p
p
o
r
t
v
ect
o
r
mach
i
n
es
.
H
o
w
ev
er,
w
i
t
h
t
h
e
s
ame
k
ern
e
l
,
i
t
w
a
s
fu
rt
h
er
d
ed
u
ced
f
ro
m
t
h
e
res
u
l
t
s
t
h
a
t
s
u
p
p
o
rt
v
ect
o
r
mach
i
n
es
p
r
o
v
i
d
e
s
b
et
t
er
p
erfo
rman
ce
w
i
t
h
an
accu
racy
o
f
8
8
.
7
8
p
er
cen
t
an
d
F
1
-
s
c
o
re
o
f
9
4
.
0
5
p
erc
en
t
.
K
e
y
w
o
r
d
s
:
F
a
s
t
f
uz
z
y
c
lus
ter
ing
K
C
-
mea
n
s
Ke
r
ne
l
f
unc
ti
on
S
c
hizophr
e
nia
c
las
s
if
ica
ti
on
S
uppor
t
ve
c
tor
mac
hines
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e
.
C
or
r
e
s
pon
din
g
A
u
th
or
:
S
r
i
Ha
r
ti
n
i
,
De
pa
r
tm
e
nt
of
M
a
thema
ti
c
s
,
Unive
r
s
it
y
of
I
ndone
s
ia,
De
pok
16424
,
I
ndone
s
ia
,
E
mail:
s
ri
.
h
ar
t
i
n
i
@
s
ci
.
u
i
.
ac.
i
d
1.
I
NT
RODU
C
T
I
ON
S
c
hizophr
e
nia
is
a
li
f
e
long
menta
l
dis
or
de
r
tha
t
induce
s
the
mi
nd,
f
e
e
li
ng,
a
nd
be
ha
vio
r
[
1
]
.
I
t
is
c
ha
r
a
c
ter
ize
d
by
unus
ua
l
thought
s
o
r
e
x
pe
r
ienc
e
s
,
dis
or
ga
nize
d
s
pe
e
c
h
pa
tt
e
r
ns
,
a
nd
de
c
r
e
a
s
e
d
pa
r
ti
c
ipation
in
s
oc
ial
li
f
e
[
2]
.
T
he
W
o
r
ld
He
a
lt
h
Or
ga
niza
ti
on
s
tate
d
that
s
c
hizophr
e
nia
is
not
a
c
omm
on
menta
l
dis
or
de
r
;
howe
ve
r
,
it
a
f
f
e
c
ts
mor
e
than
2
3
mi
ll
ion
pe
ople
wor
ldwide
[
3
]
.
A
pp
r
oxim
a
tely
220,
000
pe
ople
i
n
E
ngland
a
nd
W
a
les
a
r
e
s
uf
f
e
r
ing
f
r
om
th
is
a
il
ment
[
4]
.
M
or
e
ove
r
,
f
r
o
m
the
s
tudy
of
one
mi
ll
ion
s
ix
hundr
e
d
thous
a
nd
pe
r
s
on
include
s
568
s
ubjec
ts
with
c
li
nica
ll
y
r
e
leva
nt
ps
yc
hoti
c
s
yndr
omes
,
thi
s
menta
l
dis
or
de
r
is
mor
e
c
omm
on
a
mong
male
s
than
f
e
mal
e
s
[
5]
.
T
r
e
a
tm
e
nt
of
thi
s
menta
l
dis
or
de
r
is
us
ua
ll
y
pe
r
mane
nt
a
nd
of
ten
invol
ve
s
a
c
ombi
na
ti
o
n
of
medic
a
ti
ons
,
ps
yc
hother
a
py,
a
nd
c
oor
dinate
d
pa
r
ti
c
ular
c
a
r
e
s
e
r
vice
s
.
T
he
r
e
f
or
e
,
it
c
ould
b
e
be
tt
e
r
a
s
s
umi
ng
s
omeone
s
uf
f
e
r
ing
thi
s
a
il
ment
is
de
tec
ted
e
a
r
ly
to
pr
ovide
a
de
qua
te
f
a
mi
ly
ti
me
to
d
e
li
be
r
a
te
on
a
bunc
h
of
tr
e
a
tm
e
nts
that
c
o
uld
be
o
f
f
e
r
e
d
by
t
his
p
e
r
s
on.
Va
r
ious
s
tudi
e
s
ha
ve
a
lr
e
a
dy
be
e
n
de
ve
loped
in
many
wa
ys
.
F
or
ins
tanc
e
,
the
li
ne
a
r
s
uppor
t
ve
c
to
r
mac
hine
s
wa
s
im
pleme
nted
to
de
tec
t
the
f
ir
s
t
e
pis
ode
of
s
c
hizophr
e
nia
-
s
pe
c
tr
um
dis
or
de
r
(
F
E
S
)
in
pa
ti
e
nts
a
nd
to
dif
f
e
r
e
nti
a
te
thos
e
that
a
r
e
he
a
lt
hy
w
it
h
a
n
a
c
c
ur
a
c
y
of
62
.
34
pe
r
c
e
nt
[
6]
.
I
n
a
nother
s
tudy
,
the
mul
ti
laye
r
pe
r
c
e
ptr
on
ne
ur
a
l
ne
twor
k
(
M
L
P
NN
)
a
l
gor
it
hm
obtain
ed
82
pe
r
c
e
nt
a
c
c
ur
a
c
y
us
ing
the
publi
c
g
e
ne
e
xpr
e
s
s
ion
omni
bus
(
GE
O)
ge
nome
-
wide
e
x
pr
e
s
s
ion
da
tas
e
t
[
7]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
3
,
J
une
2020:
1643
-
1649
1644
T
he
r
e
we
r
e
a
ls
o
k
-
NN
with
8
5.
7
pe
r
c
e
nt
a
c
c
ur
a
c
y,
de
c
is
ion
tr
e
e
with
62.
50
pe
r
c
e
nt
a
c
c
ur
a
c
y,
a
nd
Na
ive
B
a
ye
s
with
a
n
a
c
c
ur
a
c
y
of
87.
71
f
or
s
c
hizophr
e
nia
da
tas
e
t
whic
h
including
in
the
latter
da
t
a
s
e
t
[
8]
.
M
or
e
ove
r
,
with
a
tot
a
l
of
212
pa
r
ti
c
ipants
whic
h
c
ons
is
t
of
141
s
c
hizophr
e
nia
p
a
ti
e
nts
a
nd
71
he
a
lt
hy
c
ontr
ols
,
the
r
e
gular
ize
d
S
VM
model
gives
a
c
c
ur
a
c
y
of
86.
6%
in
the
tr
a
ini
ng
s
e
t
of
127
indi
viduals
,
mea
nwhile
va
li
da
ti
on
a
c
c
ur
a
c
y
of
83.
5
pe
r
c
e
nt
in
a
n
indepe
nde
nt
s
e
t
of
85
pe
ople
[
9
]
.
M
e
a
nwhile,
S
VM
wa
s
a
ble
to
c
las
s
if
y
the
s
c
hizophr
e
nia
da
tas
e
t
with
a
n
a
c
c
ur
a
c
y
of
90
.
1
pe
r
c
e
nt
a
nd
95.
0
pe
r
c
e
nt
in
a
t
l
e
a
s
t
one
s
im
ulation
us
ing
li
ne
a
r
a
nd
Ga
us
s
ian
ke
r
ne
l
[
10]
.
M
e
a
n
a
c
c
ur
a
c
y
of
mor
e
than
90
pe
r
c
e
nt
wa
s
a
ls
o
obtaine
d
us
ing
T
win
S
VM
with
li
ne
a
r
a
nd
Ga
us
s
ian
ke
r
ne
l
[
11]
.
Ac
c
or
ding
to
thes
e
pr
e
vious
r
e
s
e
a
r
c
h,
it
wa
s
de
duc
e
d
that
s
uppor
t
ve
c
tor
mac
hines
is
one
of
the
we
l
l
c
las
s
if
ier
s
.
As
t
he
other
method,
ther
e
is
a
ls
o
k
e
r
ne
l
KC
-
mea
n
s
that
known
is
be
tt
e
r
than
the
f
or
mer
KC
-
mea
n
s
in
the
c
las
s
if
ica
ti
on
tas
k
us
ing
the
c
lu
s
ter
a
na
lys
is
.
T
his
r
e
s
e
a
r
c
h
,
ther
e
f
or
e
,
a
im
s
to
e
xpa
nd
the
pr
e
vious
r
e
s
e
a
r
c
h
by
c
ompar
ing
the
pe
r
f
o
r
manc
e
o
f
ke
r
ne
l
KC
-
mea
n
s
a
nd
s
uppor
t
ve
c
tor
mac
hines
,
both
wit
h
the
polynom
ial
ke
r
ne
l
f
unc
ti
on
,
us
ing
the
s
a
me
da
tas
e
t
a
s
the
las
t
two
method
s
.
2.
RE
S
E
AR
CH
M
E
T
HO
D
2.
1
.
M
a
t
e
r
ia
l
I
n
thi
s
r
e
s
e
a
r
c
h,
the
Nor
thwe
s
ter
n
Unive
r
s
it
y
S
c
hizophr
e
nia
da
tas
e
t
[
1
2
]
,
c
ons
is
ti
ng
of
171
s
c
hizophr
e
nia
a
nd
221
non
-
s
c
hizophr
e
nia
s
a
mpl
e
s
wa
s
us
e
d.
T
he
ins
tanc
e
s
a
r
e
de
s
c
r
ibed
by
64
f
e
a
tur
e
s
,
whic
h
we
r
e
numer
ica
ll
y
a
nd
c
a
tegor
ica
ll
y
c
ha
r
a
c
ter
ize
d.
De
tailed
inf
or
mation
a
bo
ut
th
e
s
e
f
e
a
tur
e
s
a
r
e
s
hown
in
T
a
ble
1
.
T
a
ble
1.
Nor
thwe
s
ter
n
Unive
r
s
it
y
c
hizoph
r
e
nia
da
t
a
s
e
t
f
e
a
tur
e
s
F
e
a
tu
r
e
numbe
r
F
e
a
tu
r
e
na
me
D
e
s
c
r
ip
ti
on
1
–
34
S
c
a
l
e
f
o
r
t
h
e
A
s
s
e
s
s
m
e
n
t
o
f
P
o
s
i
t
i
v
e
S
y
m
p
t
o
m
s
(
S
A
P
S
)
1
–
34
T
he
a
ns
w
e
r
f
or
que
s
ti
onna
ir
e
S
A
P
S
w
hi
c
h s
c
a
le
f
r
om 0 t
o 5
35
–
59
S
c
a
l
e
f
o
r
t
h
e
A
s
s
e
s
s
m
e
n
t
o
f
N
e
g
a
t
i
v
e
S
y
m
p
t
o
m
s
(
S
A
N
S
)
1
–
25
T
he
a
ns
w
e
r
f
or
que
s
ti
onna
ir
e
S
A
N
S
w
hi
c
h s
c
a
le
f
r
om 0 t
o 5
60
G
e
nde
r
T
he
ge
nde
r
of
t
he
pa
ti
e
nt
(
ma
le
or
f
e
ma
le
)
61
D
omi
na
nt
H
a
nd
L
e
f
t
or
r
ig
ht
ha
nd
62
R
a
c
e
C
a
uc
a
s
i
a
n, A
f
r
ic
a
n, A
me
r
ic
a
n or
ot
he
r
63
E
th
ni
c
C
a
uc
a
s
i
a
n, A
f
r
ic
a
n
-
A
me
r
ic
a
n, or
ot
he
r
64
A
ge
A
ge
of
t
he
pa
ti
e
nt
2.
2.
M
e
t
h
od
s
T
his
r
e
s
e
a
r
c
h
make
s
us
e
of
two
methods
,
na
mely
ke
r
ne
l
KC
-
mea
ns
a
nd
s
uppor
t
ve
c
tor
mac
hines
.
T
he
pe
r
f
or
manc
e
s
of
thes
e
two
methods
a
r
e
e
va
luate
d
us
ing
10
-
f
old
c
r
os
s
-
va
li
da
ti
on.
T
he
da
tas
e
t
wa
s
divi
de
d
r
a
ndoml
y
int
o
ten
mut
ua
ll
y
e
xc
lus
ive
f
ol
ds
with
a
n
a
ppr
oxim
a
tely
e
qua
l
number
of
s
a
mpl
e
s
[
13]
.
E
a
c
h
f
old
then
wa
s
take
n
a
s
the
va
li
da
ti
on
da
ta
f
o
r
tes
ti
ng
the
c
las
s
if
ica
ti
on
model
,
while
the
r
e
s
t
wa
s
take
n
f
or
buil
ding
model
a
nd
ha
s
a
r
ole
a
s
the
tr
a
ini
ng
da
ta.
T
his
pr
oc
e
s
s
wa
s
r
e
pe
a
ted
unti
l
e
a
c
h
f
old
ha
ve
a
lr
e
a
dy
be
e
n
va
li
da
ti
on
da
ta,
whic
h
wa
s
be
ne
f
ici
a
l
to
the
r
e
pe
a
ted
r
a
ndom
s
ub
-
s
a
mpl
ing
[
14]
.
T
h
e
us
e
of
the
c
r
os
s
-
va
li
da
ti
on
a
ppr
oa
c
h
in
e
va
luat
ing
c
las
s
if
ier
s
is
c
ons
ider
e
d
due
to
it
s
a
bil
it
y
to
a
void
pi
tf
a
ll
s
in
c
ompar
ing
thes
e
two
methods
[
15
]
.
2.
2.
1.
Ker
n
e
l
f
u
n
c
t
io
n
T
he
r
e
a
l
-
wor
ld
a
ppli
c
a
ti
ons
of
ten
ne
e
d
mor
e
th
a
n
li
ne
a
r
f
unc
ti
ons
.
T
he
r
e
f
or
e
,
a
s
a
n
a
lt
e
r
na
ti
ve
s
olut
ion,
ke
r
ne
l
of
f
e
r
s
ne
w
r
e
p
r
e
s
e
ntation
by
pr
o
jec
ti
ng
the
da
ta
int
o
f
e
a
tu
r
e
s
pa
c
e
that
ha
s
higher
di
mens
ion
f
or
the
c
omput
a
ti
ona
l
ti
me
o
f
li
ne
a
r
lea
r
ning
mac
hines
to
be
incr
e
a
s
e
d
[
16]
.
T
he
ke
r
ne
l
f
unc
ti
on
f
or
e
ve
r
y
,
∈
ℝ
n
is
de
f
ined
in
(
1
)
[
17
]
.
K
(
,
)
=
(
ϕ
(
)
,
ϕ
(
)
)
(
1)
whe
r
e
ϕ
is
the
mapping
f
unc
ti
on
f
r
om
input
da
ta
to
the
f
e
a
tur
e
s
pa
c
e
.
B
e
s
ides
,
the
d
is
tanc
e
be
twe
e
n
two
mappe
d
point
s
in
ke
r
ne
l
r
e
pr
e
s
e
ntation
is
de
f
ined
i
n
(
2)
[
17]
.
d(
x,
y
)
=
√
K
(
,
)
−
2K
(
,
)
+
K
(
,
)
(
2)
T
his
f
or
mu
la
is
s
ubs
ti
tut
e
d
to
r
e
plac
e
the
E
uc
l
idea
n
dis
tanc
e
c
omm
only
us
e
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
T
he
c
ompar
is
on
s
tudy
of
k
e
r
ne
l
K
C
-
me
ans
and
s
u
ppor
t
v
e
c
tor
mac
hine
s
for
c
las
s
if
y
ing
(
Sr
i
Har
ti
ni
)
1645
T
he
r
e
a
r
e
s
e
ve
r
a
l
types
of
ke
r
ne
l
f
unc
ti
ons
,
inc
ludi
ng
li
ne
a
r
,
polynom
ial
,
G
a
us
s
ian
r
a
dial
b
a
s
is
f
unc
ti
on
(
R
B
F
)
,
a
nd
s
igm
oid
ke
r
ne
l
f
unc
ti
on
,
wit
h
f
or
mul
a
s
a
r
e
s
hown
in
T
a
ble
2.
How
e
ve
r
,
thi
s
r
e
s
e
a
r
c
h
us
e
s
the
polynom
ial
k
e
r
ne
l
f
unc
ti
on
with
the
poly
nomi
a
l
or
de
r
ℎ
f
r
om
1
to
10
.
B
a
s
e
d
on
L
iu
e
t
a
l
.
[
1
8]
,
thi
s
kind
of
ke
r
ne
l
f
unc
ti
on
is
a
ppr
op
r
iate
f
o
r
c
ondit
ion
s
whe
r
e
a
ll
the
t
r
a
ini
ng
da
ta
nor
malize
d.
T
a
ble
2.
Ke
r
ne
l
f
unc
t
ions
N
a
me
F
or
mul
a
1.
L
i
n
e
a
r
k
e
r
n
e
l
f
u
n
c
t
i
o
n
K
(
,
)
=
.
2.
P
o
l
y
n
o
m
i
a
l
k
e
r
n
e
l
f
u
n
c
t
i
o
n
K
(
,
)
=
(
.
+
1
)
ℎ
3.
G
a
us
s
ia
n R
a
di
a
l
B
a
s
is
F
unc
ti
on (
R
B
F
)
ke
r
ne
l
f
unc
ti
on
K
(
,
)
=
e
x
p
(
−
‖
−
‖
2
2
2
)
4.
S
ig
moi
d ke
r
ne
l
f
unc
ti
on
K
(
,
)
=
t
a
n
h
(
.
−
)
2.
2.
2.
Ker
n
e
l
KC
-
m
e
an
s
T
his
method
is
a
c
ombi
na
ti
on
of
K
-
M
e
a
ns
,
F
uz
z
y
C
-
M
e
a
ns
a
lgor
it
hm,
a
nd
ke
r
ne
l
f
unc
ti
on
[
19
]
with
the
pur
pos
e
to
make
s
the
r
unning
ti
me
of
f
u
z
z
y
c
-
mea
n
s
f
a
s
ter
with
the
s
a
me
pe
r
f
or
manc
e
w
he
n
only
us
e
it
s
method.
T
he
ke
r
ne
l
KC
-
mea
ns
is
a
ls
o
known
a
s
f
a
s
t
f
uz
z
y
c
lus
ter
ing
ba
s
e
d
on
ke
r
ne
l.
Ke
r
ne
l
KC
-
mea
n
s
is
a
modi
f
ied
ve
r
s
ion
of
KC
-
mea
ns
[
20
]
pr
opos
e
d
by
Atiyah
e
t
a
l.
M
or
e
ove
r
,
we
ha
ve
to
mi
nim
ize
k
-
mea
ns
a
nd
f
uz
z
y
c
-
mea
n
s
objec
ti
ve
f
unc
ti
on,
wit
h
the
modi
f
ica
ti
on
on
the
dis
tanc
e
de
f
ini
ti
on
.
As
s
ume
is
th
e
-
th
s
a
mpl
e
,
a
nd
is
the
-
th
c
e
ntr
oid.
W
he
n
a
ppl
ying
k
-
mea
ns
us
ing
ke
r
ne
l
f
unc
ti
on,
then
our
goa
l
is
to
opti
mi
z
e
the
objec
ti
ve
f
unc
ti
on
s
hown
in
(
3)
[
17
]
.
J
=
∑
∑
r
(
K
(
,
)
−
2K
(
,
)
+
K
(
,
)
)
=
1
=
1
(3
)
whe
r
e
r
=
{
1
,
if
k
=
a
r
g
m
in
d
2
(
x
,
v
)
0
,
o
th
e
r
w
is
e
(4
)
M
e
a
nwhile,
whe
n
a
pplyi
ng
f
uz
z
y
ke
r
ne
l
c
-
mea
ns
,
our
goa
l
is
to
opti
mi
z
e
the
objec
ti
ve
f
unc
ti
on
(
5
)
[
2
1]
:
J
=
∑
∑
(
u
)
[
K
(
,
)
−
2K
(
,
)
+
K
(
,
)
]
=
1
=
1
(5
)
whe
r
e
u
is
the
membe
r
s
hip
va
lue
of
the
-
th
s
a
mpl
e
in
the
-
th
c
lus
ter
that
s
a
ti
s
f
ies
thes
e
two
c
ondit
ions
:
∑
u
=
1
=
1
,
=
1
,
2
,
…
,
(
6
)
0
<
∑
u
=
1
<
,
=
1
,
2
,
…
,
(7
)
with
<
be
ing
a
pos
it
ive
int
e
ge
r
.
T
he
a
lgor
it
h
m
of
thi
s
method
is
given
in
F
igur
e
1.
2.
2.
3.
S
u
p
p
or
t
ve
c
t
or
m
ac
h
in
e
s
T
his
method
is
a
ppr
opr
iate
to
us
e
s
whe
n
ther
e
a
r
e
pr
e
c
is
e
ly
two
c
las
s
e
s
.
Ac
c
or
ding
to
Va
pnik
e
t
a
l
.
[
22
]
,
da
ta
point
s
a
r
e
c
ons
ider
e
d
a
s
s
uppor
t
ve
c
tor
s
,
a
nd
the
goa
l
,
ther
e
f
o
r
e
,
is
to
f
ind
the
be
s
t
hype
r
plane
that
a
ble
t
s
e
pa
r
a
te
them
int
o
c
las
s
e
s
.
T
he
s
c
h
e
me
of
thi
s
method
is
s
e
e
n
in
F
igur
e
2
[
23]
,
whe
r
e
H
is
the
hype
r
plane
,
W
is
the
nor
mal
ve
c
to
r
to
t
he
hype
r
plane
,
while
m
is
the
mi
n
im
um
dis
tanc
e
be
twe
e
n
pos
it
ive
a
nd
ne
ga
ti
ve
hype
r
plane
s
.
M
or
e
ove
r
,
the
ke
r
ne
l
a
ppr
oa
c
h
is
us
e
d
to
r
e
s
olve
a
s
im
ple
hype
r
plan
e
that
is
us
e
f
ul
in
a
c
las
s
if
ica
ti
on
pr
oblem
with
only
two
c
las
s
e
s
.
I
n
other
wor
ds
,
thi
s
method
br
ings
the
f
o
r
m
of
mappi
ng
input
int
o
a
s
pa
c
e
that
ha
s
higher
dim
e
ns
ion
to
s
uppor
t
nonli
ne
a
r
c
las
s
if
ica
ti
on
pr
oblems
whe
r
e
the
hy
pe
r
plane
c
a
us
e
s
th
e
maximum
s
e
pa
r
a
ti
on
be
twe
e
n
e
a
c
h
c
las
s
[
24]
.
T
he
r
e
f
or
e
,
the
opti
mi
z
a
ti
on
pr
oblem
obs
e
r
ve
d
is
de
f
ined
to
mi
nim
ize
the
f
oll
owing
f
unc
ti
on
[
25]
while
de
ter
mi
ning
the
va
lue
of
we
ight
w
∈
ℝ
n
a
n
d
the
bias
b
∈
ℝ
n
.
1
2
‖
‖
2
(8
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
3
,
J
une
2020:
1643
-
1649
1646
s.
t
y
(
T
⋅
+
b
)
≥
1
,
∀
=
1
,
2
,
…
(
9)
T
he
n
the
de
c
is
ion
f
unc
ti
on
to
opti
mi
z
e
the
mar
gin
is
de
f
ined
a
s
(
10)
[
25]
:
f
(
)
=
s
ig
n
(
⋅
+
b
)
(
10)
whe
r
e
the
va
lue
of
a
nd
b
is
obtaine
d
us
ing
the
f
or
m
ula
in
(
11
)
a
nd
(
12)
,
r
e
s
pe
c
ti
ve
ly.
=
∑
a
y
=
1
(
11)
b
=
1
N
∑
(
y
−
∑
a
y
x
∈
)
∈
(
12
)
w
ith
S
is
the
ve
c
tor
s
e
t
whe
r
e
a
≠
0
f
or
e
ve
r
y
∈
S
a
nd
N
=
|
S
|
.
T
hi
s
method
is
s
ti
ll
c
ons
id
e
r
e
d
due
to
it
s
e
xc
e
ll
e
nt
pe
r
f
or
manc
e
that
doe
s
not
only
maximi
z
e
s
mar
gins
but
a
ls
o
mi
nim
ize
s
e
xis
ti
ng
e
r
r
or
s
[
26
]
.
F
igur
e
1.
Ke
r
ne
l
KC
-
mea
ns
a
lgor
it
hm
F
igur
e
2.
S
uppor
t
ve
c
tor
mac
hines
c
las
s
if
ica
ti
on
s
c
he
me
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
T
he
c
ompar
is
on
s
tudy
of
k
e
r
ne
l
K
C
-
me
ans
and
s
u
ppor
t
v
e
c
tor
mac
hine
s
for
c
las
s
if
y
ing
(
Sr
i
Har
ti
ni
)
1647
2.
2.
4.
P
e
r
f
or
m
an
c
e
m
e
as
u
r
e
T
his
r
e
s
e
a
r
c
h
us
e
s
a
c
c
ur
a
c
y
a
nd
F
1
-
s
c
or
e
t
o
mea
s
ur
e
pe
r
f
or
manc
e
ba
s
e
d
on
the
r
e
s
ult
of
the
c
onf
us
ion
matr
ix.
How
e
ve
r
,
the
pr
ope
r
us
e
of
metr
ics
f
r
om
the
c
onf
us
ion
matr
ix
is
ne
c
e
s
s
a
r
y
be
c
a
us
e
it
a
f
f
e
c
ts
a
ll
s
tatis
ti
c
a
l
c
ompar
is
on
met
r
ics
[
27
]
.
T
he
r
e
f
or
e
,
a
c
c
ur
a
c
y
is
a
s
ubjec
ti
ve
mea
s
ur
e
us
e
d
to
e
va
luate
a
c
las
s
if
ier
on
a
s
e
t
of
tes
t
da
ta
with
the
c
las
s
if
ier
’
s
pr
e
diction
c
or
r
e
c
tl
y
divi
de
d
by
the
tot
a
l
number
of
ins
tanc
e
s
[
28]
.
M
e
a
nwhile,
F
1
-
s
c
or
e
,
whic
h
c
ons
ider
s
both
the
pr
e
c
is
ion
a
nd
s
e
ns
it
ivi
ty,
is
us
e
d
to
dis
ti
nguis
h
the
c
or
r
e
c
t
p
r
e
diction
of
labe
ls
withi
n
dif
f
e
r
e
nt
c
las
s
e
s
[
29]
.
T
he
s
e
two
f
or
mul
a
s
a
r
e
s
hown
in
(
13
)
a
nd
(
14
)
:
A
ccu
r
a
cy
=
TP
+
TN
TP
+
TN
+
FP
+
FN
(
13
)
F1
−
S
co
r
e
=
2
∗
s
e
nsi
t
i
v
i
t
y
∗
pr
e
c
i
s
i
on
s
e
nsi
t
i
v
i
t
y
+
pr
e
c
i
s
i
on
(
14
)
with
s
e
ns
it
ivi
ty
a
nd
pr
e
c
is
ion
is
given
in
(
15)
a
nd
(
16)
S
e
ns
it
ivity
=
TP
TP
+
FN
(
15
)
P
r
e
cis
io
n
=
TP
TP
+
FP
(
16
)
T
P
is
the
c
ount
o
f
s
c
hizophr
e
nia
s
a
mpl
e
s
c
or
r
e
c
tl
y
diagnos
e
d,
F
P
is
the
number
of
non
-
s
c
hizophr
e
nia
s
a
mpl
e
s
incor
r
e
c
tl
y
diagnos
e
d,
T
N
is
the
num
be
r
of
non
-
s
c
hizophr
e
nia
s
a
mpl
e
s
c
or
r
e
c
tl
y
diagnos
e
d,
a
nd
F
N
is
the
tot
a
l
o
f
s
c
hizophr
e
nia
s
a
mpl
e
s
incor
r
e
c
tl
y
di
a
gnos
e
d
[
30]
.
3.
RE
S
UL
T
S
AN
D
AN
AL
YSI
S
T
he
pe
r
f
o
r
manc
e
of
both
ke
r
ne
l
KC
-
mea
ns
a
nd
s
uppor
t
v
e
c
tor
m
a
c
hines
us
ing
polynom
ial
ke
r
ne
l
we
r
e
e
xa
mi
ne
d
with
k
-
f
old
c
r
os
s
-
va
li
da
ti
on
wh
e
r
e
k
=
10
.
I
n
10
-
f
old
c
r
os
s
-
va
li
da
ti
on,
ten
f
olds
with
a
ppr
oxim
a
tely
e
qua
l
s
ize
we
r
e
f
or
med
f
r
om
the
da
tas
e
t
a
f
ter
r
a
ndoml
y
divi
de
d
[
13
]
.
F
ur
the
r
mor
e
,
on
e
f
old
is
the
va
li
da
ti
on
da
ta
,
while
the
r
e
s
t
is
t
r
a
ini
ng
da
ta.
T
his
pr
oc
e
s
s
r
e
pe
a
ts
f
o
r
ten
ti
mes
,
whic
h
gives
a
d
va
ntage
s
of
ove
r
r
e
pe
a
ted
r
a
ndom
s
ub
-
s
a
mpl
ing
[
14
]
.
T
he
pe
r
f
or
manc
e
of
ke
r
ne
l
KC
-
mea
ns
us
ing
polynom
i
a
l
ke
r
ne
l
is
s
hown
in
T
a
ble
3
.
I
n
thi
s
tab
le,
the
r
unning
t
im
e
va
r
ied
a
nd
wa
s
c
onduc
ted
unde
r
or
in
0
.
1
s
e
c
onds
.
T
he
r
e
f
or
e
,
it
is
c
onc
luded
that
the
highes
t
va
lue
of
a
c
c
ur
a
c
y
a
nd
F
1
-
s
c
or
e
i
s
obtaine
d
whe
n
the
s
i
xth
-
or
de
r
polynom
ial
ke
r
ne
l
is
uti
li
z
e
d
.
M
e
a
nwhile,
the
pe
r
f
or
manc
e
o
f
s
uppor
t
ve
c
tor
mac
hines
us
ing
pol
ynomi
a
l
ke
r
ne
l
is
s
hown
in
T
a
ble
4.
W
e
c
a
n
s
e
e
in
thi
s
table
that
the
a
c
c
ur
a
c
y
a
nd
F
1
-
s
c
or
e
is
not
a
s
va
r
ious
a
s
ke
r
ne
l
KC
-
mea
ns
.
T
a
ble
3.
T
he
p
e
r
f
or
manc
e
o
f
ke
r
ne
l
KC
-
mea
ns
with
polynom
ial
ke
r
ne
l
P
ol
ynomi
a
l
D
e
gr
e
e
A
c
c
ur
a
c
y
(%)
F1
-
S
c
o
r
e
(%)
R
unni
ng
T
im
e
(
s
)
1
74.36
85.29
0.10
2
72.50
84.06
0.05
3
72.50
84.06
0.02
4
76.92
86.96
0.03
5
82.05
90.14
0.04
6
87.18
93.15
0.02
7
71.79
83.58
0.03
8
76.92
86.96
0.04
9
76.92
86.96
0.02
10
66.67
80.00
0.03
T
a
ble
4.
T
he
p
e
r
f
or
manc
e
o
f
s
uppor
t
ve
c
tor
mac
hines
with
polynom
ial
ke
r
ne
l
P
ol
ynomi
a
l
D
e
gr
e
e
A
c
c
ur
a
c
y
(%)
F1
-
S
c
or
e
(%)
R
unni
ng
T
im
e
(
s
)
1
88.78
94.05
0.55
2
88.78
94.05
0.15
3
88.78
94.05
0.15
4
88.78
94.05
0.15
5
88.78
94.05
0.41
6
88.78
94.05
0.09
7
88.27
93.77
0.08
8
88.52
93.91
0.08
9
88.52
93.91
0.10
10
88.27
93.77
0.08
T
he
a
c
c
ur
a
c
y
a
nd
F
1
-
s
c
or
e
of
thi
s
method
a
r
e
hi
ghe
r
with
88.
78
a
nd
94.
05
pe
r
c
e
nt,
r
e
s
pe
c
ti
ve
ly,
in
the
s
ixt
h
-
or
de
r
polynom
ial
ke
r
ne
l.
B
e
s
ides
,
the
f
a
s
tes
t
r
unning
ti
me
a
nd
the
be
s
t
pe
r
f
o
r
manc
e
is
yielde
d
whe
n
the
s
ixt
h
-
or
de
r
polynom
ial
ke
r
ne
l.
T
he
r
e
f
or
e
,
the
be
s
t
pe
r
f
or
manc
e
o
f
ke
r
ne
l
KC
-
mea
ns
a
nd
s
uppor
t
v
e
c
tor
m
a
c
hines
whic
h
we
r
e
obtaine
d
whe
n
the
s
ixt
h
-
or
de
r
polynom
ial
ke
r
ne
l
is
c
ompar
e
d
in
T
a
ble
5.
I
n
thi
s
table
,
we
c
a
n
c
onc
lude
that
s
uppor
t
ve
c
tor
mac
hines
give
be
tt
e
r
pe
r
f
or
manc
e
be
c
a
us
e
of
the
a
c
c
ur
a
c
y
a
nd
F
1
-
s
c
or
e
a
r
e
1
.
6%
a
nd
0.
9
%
higher
,
r
e
s
pe
c
ti
ve
ly,
than
the
pe
r
f
o
r
manc
e
o
f
ke
r
ne
l
KC
-
mea
ns
.
H
owe
ve
r
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
3
,
J
une
2020:
1643
-
1649
1648
KC
-
mea
n
s
de
li
ve
r
s
the
pe
r
f
or
manc
e
in
f
a
s
ter
r
unning
ti
me
than
s
uppor
t
ve
c
tor
mac
hines
.
T
he
r
e
f
or
e
,
both
methods
de
li
ve
r
e
xc
e
ll
e
nt
r
e
s
ult
s
with
the
KC
-
mea
ns
pr
oduc
ing
be
tt
e
r
r
unning
ti
me.
T
a
ble
5.
T
he
c
ompa
r
is
on
be
twe
e
n
ke
r
ne
l
KC
-
mea
ns
a
nd
s
upp
or
t
ve
c
tor
mac
hines
with
the
s
ixt
h
-
or
de
r
polynom
ial
ke
r
ne
l
f
unc
ti
on
M
e
th
od
A
c
c
ur
a
c
y (
%
)
F1
-
S
c
or
e
(
%
)
R
unni
ng T
im
e
(
s
)
K
e
r
ne
l
K
C
-
me
a
ns
87.18
93.15
0.02
S
uppor
t
V
e
c
to
r
M
a
c
hi
ne
s
88.78
94.05
0.09
4.
CONC
L
USI
ON
T
r
e
a
tm
e
nt
a
s
s
oc
iate
d
with
a
menta
l
dis
or
de
r
,
e
s
pe
c
ially
S
c
hizophr
e
nia,
s
hould
be
a
dmi
nis
ter
e
d
e
a
r
ly.
T
he
medic
a
ti
on
is
ne
c
e
s
s
a
r
y
be
c
a
us
e
thi
s
menta
l
dis
or
de
r
a
f
f
e
c
ts
be
ha
vio
r
,
mi
nd,
a
nd
f
e
e
li
ng.
T
his
r
e
s
e
a
r
c
h
us
e
s
da
ta
whic
h
c
on
s
is
ts
of
171
s
c
hizophr
e
nia
a
nd
221
non
-
s
c
hizophr
e
nia
s
a
mpl
e
s
f
r
om
Nor
thwe
s
ter
n
Unive
r
s
it
y
to
c
ompar
e
the
pe
r
f
or
manc
e
of
ke
r
ne
l
KC
-
mea
ns
a
nd
s
uppor
t
ve
c
tor
mac
hine
s
in
the
s
c
hi
z
ophr
e
nia
c
las
s
if
ica
ti
on
tas
k.
F
r
om
th
e
e
xpe
r
im
e
nt,
ke
r
ne
l
KC
-
mea
ns
wa
s
f
ound
to
g
i
ve
be
tt
e
r
r
unning
ti
me
with
a
n
a
c
c
ur
a
c
y
o
f
87
.
18
pe
r
c
e
nt
a
nd
F
1
-
s
c
or
e
of
93
.
15
pe
r
c
e
nt
,
whic
h
is
lo
we
r
than
the
s
uppor
t
ve
c
tor
mac
hines
.
T
he
latter
method
s
ti
ll
gives
be
tt
e
r
a
c
c
ur
a
c
y
a
nd
F
1
-
s
c
or
e
with
8
8
.
78
a
nd
94.
05
pe
r
c
e
nt,
r
e
s
pe
c
ti
ve
ly.
I
n
thi
s
s
tudy,
ke
r
ne
l
KC
-
mea
ns
,
a
s
we
ll
a
s
s
uppor
t
ve
c
tor
mac
hines
,
pr
ovided
h
igh
pe
r
f
or
manc
e
.
How
e
ve
r
,
ther
e
is
s
pa
c
e
f
o
r
im
p
r
ove
ment.
F
or
f
utur
e
r
e
s
e
a
r
c
h
e
s
,
it
wa
s
then
e
nc
our
a
ge
d
to
e
xplor
e
the
pos
s
ibi
li
ty
of
c
ons
t
r
uc
ti
ng
ne
w
models
to
obtain
be
tt
e
r
pe
r
f
or
manc
e
,
e
s
pe
c
ially
c
ons
ider
ing
that
polynom
ial
ke
r
ne
l
in
both
methods
did
not
de
li
ve
r
a
bove
90
pe
r
c
e
nt
of
a
c
c
ur
a
c
y.
I
t
is
a
ls
o
pos
s
ibl
e
to
im
pleme
nt
thes
e
methods
in
other
da
tas
e
ts
;
t
he
r
e
f
or
e
,
the
model
is
e
xa
mi
ne
d
in
dif
f
e
r
e
nt
w
a
ys
with
the
r
e
s
ult
that
i
t
would
be
c
ons
ider
e
d
a
s
the
a
ppr
op
r
iate
method
whic
h
gives
a
n
a
c
c
ur
a
te
diagnos
is
.
AC
KNOWL
E
DGE
M
E
NT
S
T
his
r
e
s
e
a
r
c
h
wa
s
f
ull
s
uppor
ted
f
inanc
ially
by
Unive
r
s
it
y
of
I
ndone
s
ia,
with
a
DR
P
M
P
UT
I
Q2
2020
r
e
s
e
a
r
c
h
g
r
a
nt
s
c
he
me
.
T
he
a
utho
r
s
wo
uld
li
ke
to
a
ppr
e
c
iate
a
ll
the
r
e
view
e
r
s
invo
lved
in
the
im
pr
ove
ment
o
f
thi
s
a
r
ti
c
le.
RE
F
E
RE
NC
E
S
[1
]
W
an
g
W
.
C
.
,
L
u
M
.
L
.
,
Ch
en
V
.
C
.
H
.
,
N
g
M
.
H
.
,
H
u
an
g
K
.
Y
.
,
H
s
i
eh
M
.
H
.
,
H
s
i
e
h
M
.
J
.
,
McIn
t
y
re
R
.
S
.
,
L
ee
Y
.
,
L
ee
C
.
T
.
C
,
“
A
s
t
h
ma,
co
rt
i
co
s
t
er
o
i
d
u
s
e,
an
d
s
ch
i
zo
p
h
ren
i
a:
A
n
a
t
i
o
n
w
i
d
e
p
o
p
u
l
at
i
o
n
-
b
a
s
e
d
s
t
u
d
y
i
n
T
a
i
w
a
n
,”
P
Lo
S
O
n
e
,
v
o
l
.
1
2
,
n
o
.
3
,
p
p
.
1
-
12
,
2
0
1
7
.
[2
]
T
h
e
N
a
t
i
o
n
a
l
In
s
t
i
t
u
t
e
o
f
Me
n
t
a
l
H
eal
t
h
.
[O
n
l
i
n
e
]
.
A
v
a
i
l
ab
l
e
:
h
t
t
p
s
:
/
/
w
w
w
.
n
i
mh
.
n
i
h
.
g
o
v
/
.
A
cce
s
s
e
d
:
2
7
J
u
n
e
2
0
1
9
.
[3
]
W
o
r
l
d
H
ea
l
t
h
O
r
g
an
i
zat
i
o
n
.
[O
n
l
i
n
e]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
w
w
w
.
w
h
o
.
i
n
t
/
.
A
cces
s
ed
:
2
7
J
u
n
e
2
0
1
9
.
[4
]
Ro
y
a
l
Co
l
l
eg
e
o
f
Ps
y
ch
i
at
r
i
s
t
s
,
“
Rep
o
rt
o
f
t
h
e
N
a
t
i
o
n
a
l
A
u
d
i
t
o
f
Sch
i
zo
p
h
re
n
i
a
(
N
A
S)
2
0
1
2
,”
L
o
n
d
o
n
:
H
ea
l
t
h
c
a
r
e
Q
u
a
l
i
t
y
Im
p
r
o
vem
e
n
t
P
a
r
t
n
er
s
h
i
p
,
p
p
.
2
3
,
2
0
1
2
.
[5
]
K
i
r
k
b
r
i
d
e
J
.
B
.
,
Fearo
n
P
.
,
Mo
rg
an
C
.
,
et
al
.
,
“
H
et
ero
g
en
ei
t
y
i
n
i
n
ci
d
en
ce
rat
e
s
o
f
s
ch
i
zo
p
h
re
n
i
a
a
n
d
o
t
h
er
p
s
y
c
h
o
t
i
c
s
y
n
d
r
o
mes
:
Fi
n
d
i
n
g
s
fro
m
t
h
e
3
-
ce
n
t
er
A
eS
O
P
s
t
u
d
y
,”
A
r
c
h
G
en
P
s
yc
h
i
a
t
r
y
,
v
o
l
.
6
3
,
n
o
.
3
,
p
p
.
2
5
0
-
2
5
8
,
2
0
0
6
.
[6
]
Mi
k
o
l
as
P
.
,
H
l
i
n
k
a
J
.
,
Sk
o
ch
A
.
,
Pi
t
ra
Z
.
,
Fro
d
l
T
.
,
Sp
an
i
el
F
.
,
H
aj
e
k
T
.
,
“
Mach
i
n
e
l
ear
n
i
n
g
c
l
as
s
i
f
i
cat
i
o
n
o
f
fi
rs
t
-
ep
i
s
o
d
e
s
ch
i
zo
p
h
re
n
i
a
s
p
ec
t
ru
m
d
i
s
o
r
d
ers
a
n
d
co
n
t
ro
l
s
u
s
i
n
g
w
h
o
l
e
b
rai
n
w
h
i
t
e
mat
t
er
frac
t
i
o
n
a
l
an
i
s
o
t
r
o
p
y
,”
B
M
C
P
s
ych
i
a
t
r
y
,
v
o
l
.
18
,
n
o
.
9
7
,
p
p
.
1
-
7
,
2
0
1
8
.
[7
]
Y
i
l
aci
o
g
l
u
K
.
,
K
o
n
u
k
M.
,
“
Cl
as
s
i
f
i
cat
i
o
n
o
f
s
ch
i
z
o
p
h
ren
i
a
p
at
i
e
n
t
s
b
y
u
s
i
n
g
g
en
o
mi
c
d
a
t
a:
A
d
at
a
mi
n
i
n
g
ap
p
r
o
ach
,”
Th
e
Jo
u
r
n
a
l
o
f
Neu
r
o
b
e
h
a
v
i
o
r
a
l
S
c
i
en
ce
s
,
v
o
l
.
2
,
n
o
.
3
,
p
p
.
1
-
3
,
2
0
1
5
.
[8
]
Say
l
a
n
C
.
C
.
,
Y
i
l
an
ci
o
g
l
u
K
.
,
“
Cl
as
s
i
fi
cat
i
o
n
o
f
s
c
h
i
z
o
p
h
ren
i
a
an
d
b
i
p
o
l
ar
d
i
s
o
r
d
er
b
y
u
s
i
n
g
mach
i
n
e
l
ear
n
i
n
g
al
g
o
ri
t
h
m
s
,”
Th
e
Jo
u
r
n
a
l
o
f
Neu
r
o
b
eh
a
vi
o
r
a
l
S
ci
e
n
ces
,
v
o
l
.
3
,
n
o
.
3
,
p
p
.
92
-
95
,
2
0
1
6
.
[9
]
Ch
i
n
R
.
,
Y
o
u
A
.
X
.
,
Men
g
F
.
,
Z
h
o
u
J
.
,
Si
m
K
.
,
“
Rec
o
g
n
i
t
i
o
n
o
f
s
c
h
i
zo
p
h
re
n
i
a
w
i
t
h
reg
u
l
ar
i
zed
s
u
p
p
o
rt
v
e
ct
o
r
mach
i
n
e
an
d
s
e
q
u
e
n
t
i
al
re
g
i
o
n
o
f
i
n
t
er
es
t
s
e
l
ect
i
o
n
u
s
i
n
g
s
t
ru
c
t
u
ra
l
mag
n
et
i
c
res
o
n
a
n
ce
i
ma
g
i
n
g
,
”
S
ci
e
n
t
i
f
i
c
R
ep
o
r
t
s
,
v
o
l
.
8
,
p
p
.
1
-
10
,
2
0
1
8
.
[1
0
]
Ramp
i
s
el
a
T
.
V
.
,
Ru
s
t
am
Z
.
,
“
Cl
as
s
i
fi
ca
t
i
o
n
o
f
s
c
h
i
z
o
p
h
re
n
i
a
d
a
t
a
u
s
i
n
g
s
u
p
p
o
rt
v
ec
t
o
r
mach
i
n
e
(SV
M),
”
I
O
P
Co
n
f
.
S
er
i
es
:
Jo
u
r
n
a
l
o
f
P
h
y
s
i
c
s
:
Co
n
f
.
S
e
r
i
e
s
,
v
o
l
.
1
1
0
8
,
p
p
.
1
-
7.
2
0
1
8
.
[1
1
]
Ru
s
t
am
Z
.
,
Ramp
i
s
e
l
a
T
.
V.
,
“
Su
p
p
o
rt
v
ect
o
r
mac
h
i
n
es
an
d
t
w
i
n
s
u
p
p
o
r
t
v
ec
t
o
r
mac
h
i
n
es
f
o
r
cl
a
s
s
i
fi
ca
t
i
o
n
o
f
s
ch
i
zo
p
h
re
n
i
a
d
a
t
a
,”
In
t
e
r
n
a
t
i
o
n
a
l
Jo
u
r
n
a
l
o
f
E
n
g
i
n
eer
i
n
g
a
n
d
Tec
h
n
o
l
o
g
y
,
v
o
l
.
7
,
n
o
.
4
,
p
p
.
6
8
7
3
-
6
8
7
7
,
2
0
1
8
.
[1
2
]
W
an
g
L
.
,
K
o
g
an
A
.
,
Co
b
i
a
D
.
,
A
l
p
er
t
K
.
,
K
o
l
a
s
n
y
A
.
,
Mi
l
l
er
M
.
I
.
,
Marcu
s
D
.
,
“
N
o
rt
h
w
es
t
ern
U
n
i
v
er
s
i
t
y
Sch
i
z
o
p
h
ren
i
a
D
at
a
a
n
d
So
f
t
w
are
T
o
o
l
(
N
U
SD
A
ST
)
,”
F
r
o
n
t
i
er
s
i
n
Neu
r
o
i
n
f
o
r
m
a
t
i
cs
,
v
o
l
.
7
,
n
o
.
25
,
2
0
1
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
T
he
c
ompar
is
on
s
tudy
of
k
e
r
ne
l
K
C
-
me
ans
and
s
u
ppor
t
v
e
c
tor
mac
hine
s
for
c
las
s
if
y
ing
(
Sr
i
Har
ti
ni
)
1649
[1
3
]
K
o
h
av
i
R.
,
“
A
s
t
u
d
y
o
f
cro
s
s
-
v
al
i
d
at
i
o
n
an
d
b
o
o
t
s
t
ra
p
f
o
r
accu
racy
es
t
i
ma
t
i
o
n
an
d
mo
d
e
l
s
e
l
ect
i
o
n
,”
P
r
o
ceed
i
n
g
s
o
f
t
h
e
1
4
th
In
t
er
n
a
t
i
o
n
a
l
J
o
i
n
t
Co
n
f
e
r
e
n
ce
o
n
A
r
t
i
f
i
c
i
a
l
In
t
e
l
l
i
g
en
ce
(IJCA
I).
M
o
n
t
r
e
a
l
,
v
o
l
.
2
,
p
p
.
1
1
3
7
-
1
1
4
3
,
1
9
9
5
.
[1
4
]
Raj
u
S
.
K
.
,
Mu
rt
y
M
.
R
.
,
Rao
M
.
V
.
,
Sat
ap
at
h
y
S
.
C.
,
“
Su
p
p
o
rt
v
ect
o
r
mach
i
n
e
w
i
t
h
k
-
f
o
l
d
cro
s
s
v
al
i
d
a
t
i
o
n
mo
d
e
l
fo
r
s
o
ft
w
are
fau
l
t
p
re
d
i
c
t
i
o
n
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
Jo
u
r
n
a
l
o
f
P
u
r
e
a
n
d
A
p
p
l
i
e
d
M
a
t
h
em
a
t
i
cs
,
v
o
l
.
1
1
8
,
n
o
.
2
0
,
p
p
.
3
2
1
-
3
3
4
,
2
0
1
8
.
[1
5
]
Sal
zb
er
g
S
.
L.
,
“
O
n
co
mp
ar
i
n
g
cl
a
s
s
i
fi
er
s
:
P
i
t
fa
l
l
s
t
o
a
v
o
i
d
an
d
a
reco
mme
n
d
e
d
ap
p
ro
ac
h
,
”
D
a
t
a
M
i
n
i
n
g
a
n
d
Kn
o
wl
e
d
g
e
D
i
s
co
ve
r
y
,
v
o
l
.
1
,
p
p
.
3
1
7
-
3
2
8
,
1
9
9
7
.
[1
6
]
Cri
s
t
i
a
n
i
n
i
N
.
,
Sh
a
w
e
T
ay
l
o
r
J
.
,
“
A
n
i
n
t
r
o
d
u
ct
i
o
n
t
o
s
u
p
p
o
rt
v
ec
t
o
r
mac
h
i
n
es
an
d
o
t
h
er
k
er
n
e
l
-
b
a
s
ed
l
ear
n
i
n
g
met
h
o
d
s
,”
Camb
ri
d
g
e
:
Ca
m
b
r
i
d
g
e
U
n
i
ve
r
s
i
t
y
P
r
es
s
,
2
6
-
5
1
,
2
0
1
4
.
[1
7
]
Bi
s
h
o
p
C
.
M.
,
“
Pat
t
er
n
reco
g
n
i
t
i
o
n
a
n
d
mach
i
n
e
l
earn
i
n
g
,”
N
ew
Y
o
rk
:
S
p
r
i
n
g
er
,
2
9
1
-
2
9
9
,
2
0
0
6
.
[1
8
]
L
i
u
L
.
,
Sh
en
B
.
,
W
an
g
X
.
,
“
Res
earch
o
n
k
ern
e
l
fu
n
c
t
i
o
n
o
f
s
u
p
p
o
rt
v
ec
t
o
r
mac
h
i
n
e.
In
:
H
u
a
n
g
Y
M,
Ch
ao
H
C,
D
en
g
D
J
,
Park
J
.
E
d
i
t
o
r
s
.
A
d
v
an
c
ed
T
ec
h
n
o
l
o
g
i
es
,
E
mb
ed
d
ed
,
an
d
Mu
l
t
i
m
e
d
i
a
fo
r
H
u
man
-
ce
n
t
r
i
c
Co
m
p
u
t
i
n
g
,
”
Lect
u
r
e
No
t
e
s
i
n
E
l
ec
t
r
i
ca
l
E
n
g
i
n
ee
r
i
n
g
,
S
p
r
i
n
g
er
, v
ol
.
2
6
0
,
2
0
1
4
.
[1
9
]
Ru
s
t
am
Z
.
,
H
art
i
n
i
S.
,
“
Cl
as
s
i
f
i
cat
i
o
n
o
f
b
rea
s
t
can
cer
u
s
i
n
g
fas
t
f
u
zz
y
cl
u
s
t
er
i
n
g
b
as
e
d
o
n
k
ern
e
l
,
”
IO
P
C
o
n
f
.
S
er
i
es
:
M
a
t
e
r
i
a
l
s
S
c
i
en
ce
a
n
d
E
n
g
i
n
eer
i
n
g
,
v
o
l
.
5
4
6
,
p
p
.
1
-
10,
2
0
1
9
.
[2
0
]
A
t
i
y
a
h
I
.
A
.
,
Mo
h
ammad
p
o
u
r
A
.
,
T
ah
eri
S
.
M.
,
“
KC
-
mean
s
:
A
fas
t
fu
zzy
cl
u
s
t
er
i
n
g
,”
H
i
n
d
a
w
i
:
A
d
v
a
n
ce
i
n
F
u
z
z
y
S
ys
t
em
s
,
p
p
.
1
-
8
,
2
0
1
8
.
[2
1
]
Bezd
ek
J
.
C.
,
“
Pat
t
ern
rec
o
g
n
i
t
i
o
n
w
i
t
h
f
u
zzy
o
b
j
e
ct
i
v
e
fu
n
ct
i
o
n
al
g
o
r
i
t
h
ms
,”
N
e
w
Y
o
rk
:
P
l
en
u
m
P
r
e
s
s
,
p
p
.
1
7
4
-
1
9
1
.
1
9
8
1
.
[2
2
]
St
i
t
s
o
n
M
.
O
.
,
W
es
t
o
n
J
.
A
.
E
.
,
G
ammerman
A
.
,
V
o
v
k
V
.
,
V
ap
n
i
k
V
.
,
“
Th
eo
r
y
o
f
s
u
p
p
o
rt
v
ec
t
o
r
mac
h
i
n
es
,
”
Tech
n
i
ca
l
R
ep
o
r
t
CS
D
-
TR
-
96
-
17
.
Ro
y
al
H
o
l
l
o
w
ay
,
U
n
i
v
ers
i
t
y
o
f
L
o
n
d
o
n
.
1
9
9
6
.
[2
3
]
L
eal
N
.
,
L
eal
E
.
,
San
ch
ez
G
.
,
“
Mari
n
e
v
es
s
e
l
reco
g
n
i
t
i
o
n
b
y
aco
u
s
t
i
c
s
i
g
n
a
t
u
re
,”
A
R
P
N
Jo
u
r
n
a
l
o
f
E
n
g
i
n
ee
r
i
n
g
a
n
d
A
p
p
l
i
ed
S
c
i
en
ce
s
,
v
o
l
.
1
0
,
n
o
.
2
0
,
p
p
.
9
6
3
6
,
2
0
1
5
.
[2
4
]
A
rfi
a
n
i
,
Ru
s
t
am
Z
.
,
Pan
d
el
ak
i
J
.
,
Si
ah
aa
n
A
.
,
“
K
ern
e
l
s
p
h
er
i
cal
k
-
mea
n
s
an
d
s
u
p
p
o
rt
v
ec
t
o
r
mach
i
n
e
fo
r
ac
u
t
e
s
i
n
u
s
i
t
i
s
c
l
as
s
i
f
i
cat
i
o
n
,”
IO
P
C
o
n
f
.
S
er
i
es
:
M
a
t
e
r
i
a
l
s
S
c
i
en
ce
a
n
d
E
n
g
i
n
ee
r
i
n
g
,
n
o
.
5
4
6
,
p
p
.
1
-
1
0
,
2
0
1
9
.
[2
5
]
Ru
s
t
am
Z
.
,
Z
ah
ras
D
.
,
“
Co
mp
ari
s
o
n
b
et
w
een
s
u
p
p
o
r
t
v
e
ct
o
r
mach
i
n
es
an
d
fu
zzy
c
-
mean
s
as
cl
as
s
i
f
i
er
fo
r
i
n
t
r
u
s
i
o
n
d
et
ec
t
i
o
n
s
y
s
t
em
,
”
IO
P
C
o
n
f
.
S
er
i
es
:
Jo
u
r
n
a
l
o
f
P
h
y
s
i
c
s
:
Co
n
f
.
S
e
r
i
e
s
,”
n
o
.
1
0
2
8
p
p
.
1
-
7
,
2
0
1
8
.
[2
6
]
Ru
s
t
am
Z
.
,
Sy
ari
fah
M
.
A
.
,
Si
s
w
an
t
i
n
i
n
g
T
.
,
“
Recu
rs
i
v
e
p
art
i
c
l
e
s
w
arm
o
p
t
i
mi
za
t
i
o
n
(RPSO
)
s
c
h
emed
s
u
p
p
o
r
t
v
ect
o
r
mach
i
n
e
(SV
M)
i
mp
l
emen
t
at
i
o
n
fo
r
m
i
cro
arra
y
d
at
a
an
a
l
y
s
i
s
o
n
ch
r
o
n
i
c
k
i
d
n
ey
d
i
s
eas
e
(C
K
D
)
,”
IO
P
C
o
n
f
.
S
er
i
es
:
M
a
t
e
r
i
a
l
s
S
c
i
en
ce
a
n
d
E
n
g
i
n
eer
i
n
g
,
n
o
.
5
4
6
,
p
p
.
1
-
7
,
2
0
1
9
.
[2
7
]
H
o
s
s
i
n
M
.
,
Su
l
ai
man
M
.
N.
,
“
A
rev
i
ew
o
n
ev
al
u
at
i
o
n
met
ri
cs
fo
r
d
at
a
cl
as
s
i
f
i
cat
i
o
n
ev
al
u
at
i
o
n
s
,
”
In
t
e
r
n
a
t
i
o
n
a
l
Jo
u
r
n
a
l
o
f
D
a
t
a
M
i
n
i
n
g
&
Kn
o
wl
e
d
g
e
M
a
n
a
g
em
e
n
t
P
r
o
ces
s
(IJD
K
P
)
,
v
o
l
.
5
,
n
o
.
2
,
p
p
.
1
-
1
1
,
2
0
1
5
.
[2
8
]
Park
er
C.
,
“
A
n
an
a
l
y
s
i
s
o
f
p
erf
o
rman
ce
mea
s
u
re
s
fo
r
b
i
n
ar
y
cl
as
s
i
f
i
ers
,
”
I
E
E
E
In
t
e
r
n
a
t
i
o
n
a
l
Co
n
f
e
r
en
ce
o
n
D
a
t
a
M
i
n
i
n
g
,
p
p
.
5
1
7
-
5
2
6
,
2
0
1
1
.
[2
9
]
So
k
o
l
o
v
a
M
.
,
J
ap
k
o
w
i
cz
N
.
,
an
d
Szp
ak
o
w
i
cz
S.
,
“
Bey
o
n
d
accu
ra
cy
,
F
-
s
co
re
an
d
RO
C:
A
fami
l
y
o
f
d
i
s
cr
i
mi
n
an
t
meas
u
re
s
fo
r
p
erf
o
rman
ce
e
v
al
u
at
i
o
n
.
In
:
Sa
t
t
ar
A
,
K
an
g
B.
E
d
i
t
o
r
s
.
A
I
2
0
0
6
:
A
d
v
a
n
ces
i
n
A
r
t
i
f
i
ci
al
In
t
el
l
i
g
en
ce.
A
I
2
0
0
6
.
Lect
u
r
e
N
o
t
e
s
i
n
C
o
m
p
u
t
er
S
c
i
en
ce
.
v
ol
.
4
3
0
4
.
Berl
i
n
,
H
e
i
d
e
l
b
er
g
:
S
p
ri
n
g
er
,
2
0
0
6
.
[3
0
]
Rah
i
d
eh
A
.
,
Sh
ah
eed
M
.
H.
,
“
Can
cer
cl
as
s
i
fi
ca
t
i
o
n
u
s
i
n
g
cl
u
s
t
er
i
n
g
b
as
e
d
g
en
e
s
el
ec
t
i
o
n
an
d
art
i
f
i
ci
a
l
n
eu
ra
l
n
et
w
o
r
k
s
,
”
P
r
o
ceed
i
n
g
o
f
t
h
e
2
nd
In
t
er
n
a
t
i
o
n
a
l
Co
n
f
e
r
e
n
ce
o
n
C
o
n
t
r
o
l
,
In
s
t
r
u
m
en
t
a
t
i
o
n
a
n
d
A
u
t
o
m
a
t
i
o
n
(ICCI
A
)
,
p
p
.
1
1
7
5
-
1
1
8
0
,
2
0
1
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.