I
nte
rna
t
io
na
l J
o
urna
l o
f
A
dv
a
nces in Applie
d Science
s
(
I
J
AAS)
Vo
l.
6
,
No
.
4
,
Dec
em
b
er
2
0
1
7
,
p
p
.
3
5
9
~3
6
7
I
SS
N:
2252
-
8814
359
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
J
AAS
Feature
Selec
tion
U
sing
Ev
o
lutiona
ry
F
unc
tiona
l Lin
k
Neura
l
Netw
o
rk
f
o
r Clas
sifica
tion
A
m
a
re
s
h Sa
h
u
1
,
Sa
by
a
s
a
chi P
a
t
t
na
ik
2
1
Co
m
p
u
ter S
c
ien
c
e
En
g
in
e
e
rin
g
De
p
a
rtme
n
t,
S
OA
Un
iv
e
rsit
y
Bh
u
b
a
n
e
sw
a
r,
Od
ish
a
7
5
1
0
3
0
,
In
d
ia
2
I
n
f
o
rm
a
ti
o
n
&
Co
m
m
u
n
ica
ti
o
n
T
e
c
h
n
o
lo
g
y
De
p
a
rt
m
e
n
t,
F
a
k
ir
M
o
h
a
n
U
n
iv
e
rsity
,
Ba
las
o
re
,
Od
ish
a
7
5
6
0
1
9
,
I
n
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
1
7
,
2
0
1
7
R
ev
i
s
ed
No
v
1
8
,
2
0
1
7
A
cc
ep
ted
No
v
2
3
,
2
0
1
7
Co
m
p
u
tatio
n
a
l
ti
m
e
is
h
ig
h
f
o
r
M
u
lt
il
a
y
e
r
p
e
rc
e
p
tro
n
(M
L
P
)
tra
in
e
d
w
it
h
b
a
c
k
p
ro
p
a
g
a
ti
o
n
lea
rn
in
g
a
lg
o
rit
h
m
(BP
)
a
lso
th
e
c
o
m
p
le
x
it
y
o
f
t
h
e
n
e
tw
o
rk
in
c
re
a
se
s
w
it
h
th
e
n
u
m
b
e
r
o
f
la
y
e
rs
a
n
d
n
u
m
b
e
r
o
f
n
o
d
e
s
i
n
la
y
e
rs.
In
c
o
n
tras
t
to
M
L
P
,
f
u
n
c
ti
o
n
a
l
li
n
k
a
rti
f
icia
l
n
e
u
ra
l
n
e
t
w
o
rk
(F
LA
N
N)
h
a
s
les
s
a
rc
h
it
e
c
tu
ra
l
c
o
m
p
lex
it
y
,
e
a
sie
r
to
tra
i
n
,
a
n
d
g
iv
e
s
b
e
tt
e
r
re
su
lt
i
n
t
h
e
c
las
si
f
ica
ti
o
n
p
ro
b
lem
s.
T
h
e
p
a
p
e
r
p
ro
p
o
se
d
a
n
e
v
o
lu
ti
o
n
a
ry
f
u
n
c
ti
o
n
a
l
li
n
k
a
rti
f
icia
l
n
e
u
ra
l
n
e
t
w
o
rk
(EF
LA
N
N)
u
sin
g
g
e
n
e
ti
c
a
lg
o
rit
h
m
(GA
)
b
y
e
li
m
in
a
ti
n
g
f
e
a
tu
re
s h
a
v
in
g
li
tt
le
o
r
n
o
p
re
d
ictiv
e
in
f
o
rm
a
ti
o
n
.
P
a
rt
icle
s
w
a
r
m
o
p
ti
m
iza
ti
o
n
(
P
S
O)
is
u
se
d
a
s
lea
rn
in
g
to
o
l
f
o
r
so
lv
in
g
th
e
p
ro
b
lem
o
f
c
las
si
f
ica
ti
o
n
in
d
a
ta m
in
in
g
.
EF
L
A
N
N o
v
e
rc
o
m
e
s
th
e
n
o
n
-
l
in
e
a
rit
y
n
a
tu
re
o
f
p
ro
b
lem
s
b
y
u
sin
g
th
e
f
u
n
c
ti
o
n
a
ll
y
e
x
p
a
n
d
e
d
se
lec
ted
f
e
a
tu
re
s,
w
h
ich
is
c
o
m
m
o
n
l
y
e
n
c
o
u
n
tere
d
i
n
sin
g
le
la
y
e
r
n
e
u
ra
l
n
e
tw
o
rk
s.
T
h
e
m
o
d
e
l
is
e
m
p
iri
c
a
ll
y
c
o
m
p
a
re
d
to
M
L
P
,
F
LA
NN
g
ra
d
ien
t
d
e
sc
e
n
t
lea
rn
in
g
a
lg
o
rit
h
m
,
Ra
d
ial
Ba
sis
F
u
n
c
ti
o
n
(RBF
)
a
n
d
H
y
b
rid
F
u
n
c
ti
o
n
a
l
L
in
k
Ne
u
ra
l
Ne
t
w
o
rk
(HFL
A
NN
)
.
T
h
e
re
su
lt
s
p
ro
v
e
d
th
a
t
t
h
e
p
ro
p
o
se
d
m
o
d
e
l
o
u
t
p
e
rf
o
r
m
s
th
e
o
th
e
r
m
o
d
e
ls.
K
ey
w
o
r
d
:
Fu
n
ctio
n
al
l
in
k
n
e
u
r
al
n
et
w
o
r
k
Gen
etic
al
g
o
r
ith
m
P
ar
ticle
s
w
ar
m
o
p
ti
m
izatio
n
C
o
p
yrig
h
t
©
201
7
I
n
s
t
itu
te
o
f
A
d
va
n
ce
d
E
n
g
i
n
ee
r
in
g
a
n
d
S
c
ien
ce
.
A
ll ri
g
h
ts
r
ese
r
ve
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Am
ar
es
h
Sa
h
u
,
Co
m
p
u
ter S
c
ien
c
e
En
g
in
e
e
rin
g
D
e
p
a
rtme
n
t
,
S
OA
Un
iv
e
rsit
y
Bh
u
b
a
n
e
sw
a
r,
Od
is
h
a
7
5
1
0
3
0
,
I
n
d
ia.
E
m
ail:
a
m
ar
es
h
_
s
ah
u
@
y
ah
o
o
.
co
m
1.
I
NT
RO
D
UCT
I
O
N
C
las
s
i
f
icatio
n
tas
k
i
s
o
n
e
o
f
t
h
e
m
o
s
t
i
m
p
o
r
tan
t
r
esear
ch
ar
ea
o
f
d
ata
m
i
n
i
n
g
.
A
lo
t
o
f
r
e
s
ea
r
ch
h
as
f
o
cu
s
ed
o
n
th
e
f
ield
o
v
er
th
e
l
ast
s
ev
er
al
y
ea
r
s
[
1
-
3
]
.
T
h
e
D
ata
m
in
in
g
is
a
p
r
o
ce
s
s
o
f
d
is
co
v
er
in
g
k
n
o
w
l
ed
g
e
f
r
o
m
lar
g
e
g
a
m
u
t
o
f
d
ata.
Var
io
u
s
clas
s
i
f
icatio
n
m
o
d
els
i
n
th
is
r
eg
ar
d
.
M.
J
am
e
s
[
4
]
h
a
s
u
s
ed
a
lin
ea
r
o
r
q
u
ad
r
atic
d
is
cr
i
m
i
n
ate
s
tec
h
n
iq
u
e
s
f
o
r
s
o
lv
i
n
g
p
r
o
b
le
m
s
in
c
lass
if
ica
tio
n
.
An
o
t
h
er
p
r
o
ce
d
u
r
e
h
as
b
ee
n
e
m
p
lo
y
ed
b
y
u
s
in
g
d
ec
is
io
n
t
r
ee
s
[
5
,
6
]
.
I
n
th
e
s
a
m
e
p
er
s
p
ec
tiv
e
D
u
d
a
et
al.
[
7
]
h
as
p
r
o
p
o
s
ed
d
is
cr
i
m
in
a
n
t
an
al
y
s
is
b
ased
o
n
th
e
B
a
y
e
s
ia
n
d
ec
is
io
n
t
h
eo
r
y
,
h
o
w
ev
er
h
e
h
as
m
e
n
tio
n
ed
tr
ad
itio
n
al
s
t
atis
tical
m
o
d
el
s
ar
e
b
u
ilt
m
ai
n
l
y
o
n
v
ar
io
u
s
li
n
ea
r
a
s
s
u
m
p
tio
n
s
.
T
o
o
v
er
co
m
e
s
u
ch
d
is
ad
v
a
n
ta
g
e,
ar
ti
f
icial
i
n
tellig
e
n
t
to
o
ls
h
av
e
b
ee
n
e
m
er
g
ed
to
s
o
lv
e
clas
s
i
f
icatio
n
p
r
o
b
le
m
s
in
d
ata
m
i
n
in
g
.
Fo
r
th
is
r
ea
s
o
n
,
m
a
n
y
g
en
et
ic
al
g
o
r
ith
m
s
m
o
d
el
s
w
er
e
u
s
ed
[
8
]
.
I
n
a
r
ec
en
t
s
t
u
d
y
,
Z
h
an
g
h
a
s
i
n
tr
o
d
u
ce
d
th
e
n
e
u
r
al
n
et
w
o
r
k
s
tec
h
n
iq
u
e
as
a
p
o
w
er
f
u
l
class
i
f
icatio
n
to
o
l
[
9
-
1
0
]
.
I
n
r
ec
en
t
y
ea
r
A
r
tific
ial
Neu
r
al
Net
w
o
r
k
s
(
A
NN
s
)
h
av
e
g
ain
ed
e
x
p
en
s
iv
e
p
o
p
u
lar
it
y
in
r
esear
c
h
ac
ti
v
iti
es.
ANNs
ar
e
g
e
n
er
all
y
i
m
p
le
m
en
ted
as
a
d
ata
-
b
ased
tech
n
i
q
u
es
in
p
er
f
o
r
m
i
n
g
n
o
n
li
n
ea
r
s
y
s
te
m
id
e
n
ti
f
icatio
n
o
f
co
m
p
lex
p
r
o
ce
s
s
e
s
.
A
N
Ns
h
a
v
e
b
ee
n
u
s
ed
in
cr
ea
s
i
n
g
l
y
as
a
p
r
o
m
i
s
in
g
m
o
d
eli
n
g
to
o
l
i
n
a
l
m
o
s
t
all
h
u
m
an
r
elate
d
ac
ti
v
it
y
ar
ea
s
w
h
e
r
e
q
u
an
ti
tati
v
e
ap
p
r
o
ac
h
es
ca
n
b
e
u
s
ed
to
h
e
lp
in
d
ec
is
io
n
m
a
k
i
n
g
e.
g
f
u
n
ctio
n
al
ap
p
r
o
x
i
m
atio
n
,
r
u
le
ex
tr
ac
tio
n
,
p
atter
n
r
ec
o
g
n
is
at
io
n
an
d
class
if
ica
tio
n
,
f
o
r
ec
asti
n
g
a
n
d
p
r
ed
ictio
n
,
b
u
s
i
n
ess
elec
tr
ical
e
n
g
in
ee
r
i
n
g
,
m
ed
ical
a
n
d
ci
v
il
e
n
g
i
n
ee
r
in
g
.
M
u
ltil
a
y
er
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
6
,
No
.
4,
Dec
em
b
er
201
7
:
3
5
9
–
3
6
7
360
p
er
ce
p
tr
o
n
s
(
ML
P
s
)
ar
e
th
e
m
o
s
t
co
m
m
o
n
t
y
p
e
o
f
n
eu
r
al
n
e
t
w
o
r
k
s
ap
p
lied
s
u
cc
ess
f
u
ll
y
f
o
r
id
en
tif
icatio
n
an
d
co
n
tr
o
l
o
f
a
v
ar
iet
y
o
f
n
o
n
li
n
e
ar
p
r
o
ce
s
s
es.
Var
io
u
s
r
esear
ch
ac
tiv
i
ties
h
a
v
e
b
ee
n
d
o
n
e
f
r
o
m
Mc
c
u
llo
c
h
-
p
itt
s
m
o
d
el
o
f
n
eu
r
o
n
[
1
1
]
to
h
i
g
h
e
r
o
r
d
er
n
eu
r
o
n
[
1
2
]
ar
e
co
n
s
id
er
ab
le
an
d
liter
atu
r
e
i
s
g
r
o
w
i
n
g
m
ass
iv
el
y
d
a
y
b
y
d
ay
.
Feed
f
o
r
w
ar
d
n
et
w
o
r
k
s
w
h
ic
h
b
ased
o
n
s
i
n
g
le
la
y
er
o
f
li
n
ea
r
t
h
r
es
h
o
ld
lo
g
ic
u
n
its
e
x
h
ib
it
f
a
s
t
lear
n
i
n
g
.
e.
g
th
e
s
i
m
p
le
p
er
ce
p
tr
o
n
[
1
3
]
ca
n
s
o
lv
e
a
lin
ea
r
l
y
s
ep
ar
ab
le
p
r
o
b
lem
in
a
f
in
i
te
n
u
m
b
er
o
f
lear
n
in
g
s
tep
s
,
h
o
w
ev
er
,
it
d
o
es
n
o
t
co
n
v
er
g
e
o
r
ap
p
r
o
x
im
a
te
n
o
n
li
n
ea
r
d
ec
is
io
n
m
ak
i
n
g
p
r
o
b
lem
s
.
B
y
u
s
in
g
g
r
ad
ien
t
d
escen
t le
ar
n
in
g
al
g
o
r
ith
m
to
ad
j
u
s
t th
e
w
ei
g
h
ts
ad
li
n
e
[
1
4
-
1
5
]
tr
ies to
f
in
d
a
s
o
l
u
tio
n
w
h
i
ch
m
in
i
m
i
s
e
a
m
ea
n
s
q
u
ar
e
er
r
o
r
cr
iter
io
n
.
B
u
t
it
d
is
cr
i
m
i
n
atio
n
ca
p
ab
ilit
y
is
al
s
o
li
m
ited
to
lin
ea
r
l
y
s
ep
ar
ab
le
p
r
o
b
lem
s
.
Ma
n
y
r
ea
l
p
r
o
b
lem
s
in
v
o
lv
e
ap
p
r
o
x
i
m
ati
n
g
n
o
n
li
n
ea
r
f
u
n
ctio
n
s
o
r
f
o
r
m
in
g
m
u
ltip
le
n
o
n
li
n
e
ar
d
ec
is
io
n
r
eg
io
n
s
li
m
it
s
t
h
e
ap
p
licab
ilit
y
o
f
s
i
m
p
le
s
in
g
le
la
y
er
ed
n
et
w
o
r
k
s
o
f
li
n
ea
r
th
r
es
h
o
ld
u
n
it.
B
y
ad
d
in
g
o
f
a
la
y
er
o
f
h
id
d
en
u
n
its
d
r
a
m
at
icall
y
i
n
c
r
ea
s
ed
th
e
p
o
w
er
o
f
la
y
er
ed
f
ee
d
f
o
r
w
ar
d
n
et
w
o
r
k
s
.
I
n
ML
P
u
s
i
n
g
t
h
e
b
ac
k
p
r
o
p
ag
atio
n
(
B
P)
lear
n
in
g
al
g
o
r
ith
m
h
as
b
ee
n
s
u
cc
e
s
s
f
u
ll
y
ap
p
lied
to
m
an
y
ap
p
licatio
n
p
r
o
b
lem
s
.
Ho
w
e
v
er
,
th
e
tr
ain
i
n
g
s
p
ee
d
o
f
ML
P
s
ar
e
ty
p
ical
l
y
s
lo
w
er
t
h
an
t
h
o
s
e
f
o
r
f
ee
d
f
o
r
w
ar
d
n
et
w
o
r
k
s
co
n
s
is
t
o
f
a
s
in
g
le
la
y
er
o
f
lin
ea
r
t
h
r
es
h
o
ld
u
n
i
ts
d
u
e
t
o
b
ac
k
p
r
o
p
ag
atio
n
o
f
er
r
o
r
i
n
tr
o
d
u
ce
d
b
y
m
u
lti
la
y
er
i
n
g
.
Ho
w
e
v
er
p
r
o
b
lem
s
o
cc
u
r
in
M
L
P
tr
ain
i
n
g
ar
e
lo
c
al
m
i
n
i
m
a
tr
ap
p
in
g
,
s
at
u
r
atio
n
,
w
ei
g
h
t
i
n
ter
f
er
e
n
ce
,
i
n
itial
w
ei
g
h
t
d
ep
en
d
en
ce
an
d
o
v
er
f
i
tti
n
g
.
Als
o
it i
s
v
er
y
d
if
f
ic
u
lt to
f
i
x
t
h
e
p
ar
a
m
eter
s
lik
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
i
n
a
l
a
y
er
an
d
n
u
m
b
er
o
f
h
id
d
en
la
y
er
s
i
n
a
n
et
w
o
r
k
.
So
d
ec
id
in
g
a
p
r
o
p
er
ar
ch
itectu
r
e
is
d
i
f
f
icu
l
t.
A
n
ea
s
y
wa
y
to
av
o
id
th
e
s
e
m
en
tio
n
ed
p
r
o
b
lem
s
co
n
s
i
s
ti
n
g
o
f
r
e
m
o
v
i
n
g
th
e
h
id
d
en
la
y
er
s
.
T
h
e
r
em
o
v
i
n
g
p
r
o
ce
s
s
s
h
o
u
ld
b
e
ca
r
r
ie
d
o
u
t
w
it
h
o
u
t
g
i
v
i
n
g
u
p
n
o
n
li
n
ea
r
it
y
.
P
r
o
v
id
ed
th
at
th
e
in
p
u
t
la
y
e
r
is
en
-
d
o
w
n
ed
w
it
h
ad
d
itio
n
a
l
h
ig
h
er
o
r
d
er
u
n
its
[
1
6
-
1
8
]
.
I
n
o
th
er
w
o
r
d
s
h
i
g
h
e
r
o
r
d
e
r
co
r
r
elatio
n
s
a
m
o
n
g
i
n
p
u
t
co
m
p
o
n
en
ts
ca
n
b
e
u
s
ed
to
co
n
s
tr
u
ct
a
h
i
g
h
er
o
r
d
er
n
et
w
o
r
k
s
to
p
er
f
o
r
m
n
o
n
li
n
ea
r
m
ap
p
in
g
u
s
i
n
g
o
n
l
y
a
s
in
g
le
la
y
er
o
f
u
n
it
s
[
1
6
]
.
A
l
l
th
ese
i
n
d
icate
t
h
e
f
ield
o
f
HON
s
[
1
9
-
2
0
]
lik
e
f
u
n
ctio
n
al
l
in
k
n
e
u
r
al
n
et
w
o
r
k
s
[
2
1
-
24
]
,
p
i
-
s
ig
m
a
n
eu
r
al
n
et
wo
r
k
s
[
2
5
-
2
6
]
,
r
a
d
ial
b
asis
f
u
n
ctio
n
s
[
2
7
,
2
8
]
an
d
p
o
l
y
n
o
m
ial
n
eu
r
al
n
et
w
o
r
k
s
(
R
P
NNs)
[
2
9
,
3
0
]
an
d
s
o
o
n
.
A
d
ir
ec
tio
n
is
g
i
v
e
n
b
y
p
ao
an
d
p
ao
et
al
,
th
at
f
u
n
ctio
n
al
lin
k
s
n
e
u
r
o
n
s
m
a
y
b
e
co
n
v
e
n
ie
n
t
l
y
u
s
ed
f
o
r
f
u
n
ctio
n
ap
p
r
o
x
i
m
atio
n
w
it
h
f
aster
co
n
v
er
g
en
ce
r
ate
a
n
d
les
s
er
co
m
p
u
tatio
n
al
lo
ad
th
a
n
ML
P
[
3
1
]
.
FL
A
N
N
w
it
h
g
r
ad
ien
t
d
esce
n
t
m
et
h
o
d
h
as
ac
h
ie
v
ed
g
o
o
d
r
esu
lt
s
i
n
c
lass
i
f
icatio
n
ta
s
k
o
f
d
ata
m
in
i
n
g
.
A
ls
o
,
d
i
f
f
er
e
n
t
s
et
o
f
o
r
th
o
g
o
n
al
b
asi
s
f
u
n
cti
o
n
s
h
a
v
e
b
ee
n
s
u
g
g
ested
f
o
r
f
ea
tu
r
e
ex
p
an
s
io
n
[
3
2
]
.
FLAN
N
is
b
asicall
y
a
f
lat
n
et
w
o
r
k
w
it
h
o
u
t
r
eq
u
ir
in
g
h
i
d
d
en
la
y
er
s
a
n
d
h
a
s
s
i
m
p
le
lear
n
in
g
r
u
le.
Fo
r
en
h
a
n
cin
g
th
e
clas
s
i
f
icatio
n
ac
cu
r
ac
y
t
w
o
F
L
A
NN
b
ased
class
i
f
ier
s
w
it
h
g
e
n
etic
al
g
o
r
ith
m
s
h
av
e
b
ee
n
d
ev
elo
p
ed
[
3
3
-
3
4
]
.
I
n
FLA
N
N
th
e
d
i
m
en
s
io
n
al
it
y
o
f
in
p
u
t
v
ec
to
r
is
a
ls
o
e
f
f
ec
ti
v
el
y
i
n
cr
ea
s
ed
b
y
u
s
i
n
g
f
u
n
ctio
n
al
e
x
p
an
s
io
n
o
f
t
h
e
in
p
u
t
v
ec
to
r
an
d
h
e
n
ce
t
h
e
h
y
p
er
p
lan
e
s
g
en
er
ated
b
y
t
h
e
F
L
A
N
N
p
r
o
v
id
in
g
g
r
ea
ter
d
is
cr
i
m
in
at
in
g
ca
p
ab
ilit
y
o
f
i
n
p
u
t
p
atter
n
s
.
A
lt
h
o
u
g
h
F
L
A
N
N
u
s
in
g
t
h
e
g
r
ad
ien
t
d
esce
n
t
g
i
v
e
s
p
r
o
m
is
i
n
g
r
es
u
lts
,
s
o
m
et
i
m
e
s
it
m
a
y
b
e
tr
ap
p
ed
in
lo
ca
l
o
p
ti
m
al
s
o
lu
tio
n
s
.
Mo
r
eo
v
er
FLA
N
N
co
u
p
led
w
it
h
g
en
et
ic
alg
o
r
it
h
m
s
m
a
y
s
u
f
f
er
w
i
th
p
r
o
b
le
m
s
lik
e
h
ea
v
y
co
m
p
u
tatio
n
al
b
u
r
d
e
n
s
an
d
lar
g
e
n
u
m
b
er
o
f
p
ar
a
m
et
er
s
tu
n
i
n
g
.
T
h
en
P
SO
h
as
b
ee
n
i
m
p
le
m
e
n
ted
f
o
r
class
i
f
icatio
n
[
3
5
]
s
u
f
f
er
s
f
r
o
m
p
r
o
b
lem
s
li
k
e
p
ar
a
m
eter
tu
n
in
g
a
n
d
co
m
p
lex
i
t
y
i
n
n
e
t
w
o
r
k
ar
ch
it
ec
t
u
r
e.
T
h
e
d
is
co
v
er
ed
k
n
o
w
led
g
e
f
r
o
m
t
h
e
d
ata
s
h
o
u
ld
b
e
p
r
e
d
ictiv
e
an
d
co
m
p
r
eh
en
s
ib
le.
T
h
e
ar
ch
itectu
r
al
co
m
p
le
x
it
y
o
f
F
L
A
NN
[
2
2
]
is
d
ir
ec
tl
y
p
r
o
p
o
r
tio
n
al
to
n
u
m
b
er
o
f
f
ea
t
u
r
es
an
d
th
e
f
u
n
c
tio
n
s
u
s
ed
f
o
r
ex
p
an
s
io
n
o
f
t
h
e
g
iv
e
n
f
ea
t
u
r
e
v
alu
e.
Kn
o
w
le
d
g
e
co
m
p
r
eh
e
n
s
ib
ili
t
y
is
u
s
e
f
u
l
f
o
r
at
least
t
w
o
r
elate
d
r
ea
s
o
n
s
.
Fir
s
t,
t
h
e
k
n
o
w
led
g
e
d
is
co
v
er
y
p
r
o
ce
s
s
u
s
u
all
y
ass
u
m
es
t
h
at
t
h
e
d
is
co
v
er
ed
k
n
o
w
led
g
e
w
ill
b
e
u
s
ef
u
l
f
o
r
s
u
p
p
o
r
tin
g
a
d
ec
is
io
n
to
b
e
m
ad
e
b
y
a
u
s
er
.
Seco
n
d
if
th
e
d
is
co
v
er
ed
k
n
o
w
led
g
e
is
n
o
t
co
m
p
r
eh
e
n
s
ib
le
to
th
e
u
s
er
,
h
e/s
h
e
w
il
l
n
o
t
b
e
ab
le
to
v
alid
ate
it,
o
b
s
tr
u
ctin
g
t
h
e
in
ter
ac
ti
v
e
f
ea
t
u
r
e
o
f
th
e
k
n
o
w
led
g
e
d
is
co
v
er
y
p
r
o
ce
s
s
i
n
cl
u
d
es
k
n
o
w
led
g
e
v
alid
atio
n
an
d
r
e
f
i
n
e
m
e
n
t.
T
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
f
o
r
cla
s
s
i
f
icatio
n
i
s
g
iv
e
n
an
e
q
u
al
i
m
p
o
r
tan
ce
to
b
o
th
p
r
ed
ictiv
e
ac
cu
r
ac
y
an
d
co
m
p
r
e
h
en
s
ib
ilit
y
.
W
e
m
ea
s
u
r
e
co
m
p
r
eh
e
n
s
ib
ilit
y
o
f
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
b
y
r
ed
u
cin
g
th
e
co
m
p
lex
i
t
y
i
n
a
r
ch
itect
u
r
e.
T
h
e
ex
tr
ac
ted
k
n
o
w
led
g
e
i
s
u
s
ed
b
y
u
s
f
o
r
s
u
p
p
o
r
tin
g
a
d
ec
is
io
n
m
ak
in
g
p
r
o
ce
s
s
th
a
t
is
t
h
e
u
l
ti
m
ate
g
o
al
o
f
d
ata
m
i
n
i
n
g
.
W
e
ch
o
o
s
e
ML
P
,
FLA
N
N
with
g
r
ad
ien
t
d
escen
t
lear
n
in
g
[
2
2
]
,
R
ad
ial
B
asis
Fu
n
ctio
n
(
R
B
F)
an
d
H
y
b
r
id
F
L
ANN
(
HFLA
NN)
[
4
3
]
f
o
r
r
esu
lt
s
co
m
p
ar
is
o
n
w
it
h
P
r
o
p
o
s
ed
m
et
h
o
d
.
2.
CO
NCEPT
S
AND
D
E
F
I
N
I
T
I
O
NS
2
.
1
.
G
enet
ic
a
lg
o
rit
h
m
s
Gen
etic
al
g
o
r
ith
m
s
d
ef
i
n
ed
as
a
s
ea
r
ch
te
c
h
n
iq
u
e
w
a
s
i
n
s
p
ir
ed
f
r
o
m
Dar
w
i
n
ia
n
T
h
eo
r
y
.
T
h
e
s
c
h
e
m
e
is
b
ased
o
n
t
h
e
th
eo
r
y
o
f
n
a
tu
r
al
s
e
lectio
n
.
Her
e
w
e
p
r
es
u
m
e
t
h
at
t
h
er
e
is
a
p
o
p
u
lati
o
n
co
m
p
o
s
ed
w
it
h
d
if
f
er
e
n
t
ch
ar
ac
ter
is
t
ics.
I
n
s
i
d
e
th
e
p
o
p
u
latio
n
th
e
s
tr
o
n
g
er
w
ill
b
e
ab
le
to
s
u
r
v
i
v
e
an
d
th
e
y
p
ass
t
h
ei
r
ch
ar
ac
ter
is
tic
s
to
th
eir
o
f
f
s
p
r
i
n
g
s
.
T
h
e
p
r
o
ce
s
s
o
f
G
A
is
d
escr
ib
e
d
as f
o
llo
w
s
:
a.
I
n
itial p
o
p
u
latio
n
i
s
g
e
n
er
ated
r
an
d
o
m
l
y
;
b.
Fit
n
e
s
s
f
u
n
ctio
n
ar
e
u
s
ed
to
ev
alu
ate
t
h
e
p
o
p
u
latio
n
;
c.
Gen
etic
o
p
er
ato
r
s
s
u
c
h
as sele
ctio
n
,
cr
o
s
s
o
v
er
,
an
d
m
u
tatio
n
ar
e
ap
p
lied
;
d.
R
ep
ea
t th
e
p
r
o
ce
s
s
“
E
v
al
u
atio
n
C
r
o
s
s
o
v
er
m
u
tatio
n
”
u
n
til s
t
o
p
p
in
g
cr
iter
ia
is
m
e
t
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
F
ea
tu
r
e
S
elec
tio
n
Usi
n
g
E
v
o
lu
tio
n
a
r
y
F
u
n
ctio
n
a
l Lin
k
N
eu
r
a
l …
(
A
ma
r
esh
S
a
h
u
)
361
2
.
2
.
P
a
rt
icle
Sw
a
rm
O
pti
m
i
za
t
io
n
P
SO
is
r
elati
v
el
y
a
n
e
w
p
o
p
u
latio
n
b
ased
s
ea
r
ch
tech
n
i
q
u
e
f
o
r
s
o
l
v
i
n
g
n
u
m
er
ical
o
p
ti
m
izatio
n
p
r
o
b
lem
s
.
I
t
in
s
p
ir
ed
f
r
o
m
t
h
e
co
m
p
le
x
s
o
cial
b
eh
av
i
o
r
th
at
is
s
h
o
w
n
b
y
t
h
e
n
at
u
r
a
l
s
p
ec
ies
lik
e
f
lo
ck
o
f
b
ir
d
s
.
E
ac
h
m
e
m
b
er
o
f
s
w
ar
m
ca
lled
p
ar
ticle
ass
o
ciate
d
w
it
h
a
p
o
s
itio
n
an
d
a
v
elo
cit
y
.
I
n
t
h
e
m
u
ltid
i
m
en
s
io
n
al
s
ea
r
ch
s
p
ac
e
th
e
p
ar
ticles
f
l
y
a
n
d
lo
o
k
f
o
r
a
p
o
ten
tial
s
o
lu
tio
n
.
E
ac
h
p
ar
ticle
ad
j
u
s
ts
its
p
o
s
itio
n
in
th
e
s
ea
r
c
h
s
p
ac
e
f
r
eq
u
en
tl
y
ac
co
r
d
in
g
to
th
e
f
l
y
i
n
g
e
x
p
er
ien
ce
o
f
its
o
w
n
a
n
d
its
n
ei
g
h
b
o
r
s
[
3
6
]
.
T
h
e
m
o
v
e
m
en
t
o
f
ea
c
h
p
ar
tic
le
is
g
o
v
er
n
ed
b
y
t
h
e
f
o
llo
w
i
n
g
t
w
o
E
q
u
atio
n
s
(
1
)
an
d
(
2
)
,
in
n
-
d
i
m
e
n
s
io
n
al
s
ea
r
ch
s
p
ac
e.
)
1
.
2
(
))
(
)
(
(
2
.
2
))
(
)
(
(
1
.
1
)
(
.
)
1
(
t
j
i
x
t
j
g
b
e
s
t
r
a
n
d
c
t
j
i
x
t
j
i
p
b
e
s
t
r
a
n
d
c
t
j
i
v
w
t
j
i
v
)
2
.
2
(
)
1
(
)
(
)
1
(
t
j
i
v
t
j
i
x
t
j
i
x
W
h
er
e,
x
i
r
ep
r
esen
ts
th
e
p
o
s
iti
o
n
o
f
th
e
i
th
p
ar
ticle;
v
i
r
ep
r
esen
ts
t
h
e
v
elo
cit
y
o
f
t
h
e
i
th
p
ar
ticle;
p
b
est
i
b
e
th
e
p
r
ev
io
u
s
b
est
p
ar
ticle
o
f
t
h
e
i
th
p
ar
ticle
a
n
d
g
b
est
is
t
h
e
g
lo
b
al
b
est
p
ar
ticle
f
o
u
n
d
b
y
a
ll
p
ar
ticles
u
n
til
n
o
w
.
r
a
n
d
1
an
d
r
an
d
2
ar
e
tw
o
r
an
d
o
m
v
ec
to
r
v
al
u
es
w
it
h
i
n
[
0
,
1
]
,
w
r
ep
r
esen
ts
a
in
er
tia
w
ei
g
h
t,
c
1
an
d
c
2
ar
e
t
w
o
lear
n
i
n
g
f
ac
to
r
s
,
t
h
e
v
alu
e
o
f
ea
c
h
v
elo
cit
y
v
ec
to
r
is
co
m
p
r
e
s
s
ed
w
i
th
i
n
th
e
r
an
g
e
[
v
max
,
v
m
in
]
to
d
ec
r
ea
s
e
th
e
li
k
eli
h
o
o
d
o
f
th
e
p
ar
ticle
leav
in
g
t
h
e
s
ea
r
ch
s
p
ac
e,
an
d
t=
1
,
2
,
3
…,
in
d
icate
s
th
e
n
u
m
b
e
r
o
f
iter
atio
n
s
.
T
h
e
co
n
s
tr
ictio
n
f
ac
to
r
ap
p
r
o
ac
h
o
f
P
SO
(
C
P
SO)
is
a
n
i
m
p
o
r
tan
t
v
ar
ia
n
t
o
f
s
ta
n
d
ar
d
PS
O,
I
t
w
a
s
p
r
o
p
o
s
ed
b
y
C
ler
c
a
n
d
Ken
n
e
d
y
[
3
7
-
3
8
]
.
T
h
e
v
elo
cit
y
o
f
C
P
SO is u
p
d
ated
b
y
t
h
e
f
o
llo
w
i
n
g
E
q
u
a
tio
n
:
)
3
.
2
)
)
)
(
(
)
(
.(
2
.
2
))
(
)
(
.(
1
.
1
)
(
.(
)
1
(
t
j
i
x
t
j
g
b
e
s
t
r
a
n
d
c
t
j
i
x
t
j
i
p
b
e
s
t
r
a
n
d
c
t
j
i
v
t
j
i
v
W
h
er
e,
is
ca
lled
a
co
n
s
tr
icti
o
n
f
ac
to
r
,
g
i
v
en
b
y
E
q
u
atio
n
:
)
4
.
2
(
|
4
2
2
|
2
W
h
er
e,
)
5
.
2
(
4
,
2
1
c
c
T
h
e
C
P
SO
en
s
u
r
e
s
th
e
co
n
v
er
g
en
ce
o
f
th
e
s
ea
r
c
h
p
r
o
ce
d
u
r
es
an
d
b
e
ab
le
to
g
en
er
ate
h
i
g
h
er
q
u
alit
y
s
o
lu
tio
n
s
t
h
an
t
h
e
s
ta
n
d
ar
d
P
SO is u
s
ed
in
t
h
is
s
tu
d
y
.
2
.
3
.
F
un
ct
io
na
l Li
n
k
Neura
l N
et
w
o
rk
T
h
e
FLA
NN
ar
ch
i
tectu
r
e
w
as
f
ir
s
t
p
r
o
p
o
s
ed
b
y
P
ao
et
al.
[
3
1
]
.
T
h
e
id
ea
o
f
th
is
m
o
d
el
is
t
o
ap
p
ly
an
ex
p
an
s
io
n
f
u
n
ctio
n
to
in
cr
ea
s
e
s
th
e
d
i
m
en
s
io
n
a
lit
y
o
f
i
n
p
u
t
v
ec
to
r
.
Hen
ce
th
e
h
y
p
er
-
p
la
n
es
g
en
er
ated
p
r
o
v
id
e
s
u
p
er
io
r
d
is
cr
i
m
in
a
tio
n
ca
p
ab
ilit
y
in
t
h
e
in
p
u
t
p
atter
n
s
p
ac
e.
T
h
e
n
ee
d
o
f
h
id
d
en
la
y
er
is
r
e
m
o
v
ed
b
y
ap
p
ly
i
n
g
e
x
p
an
s
io
n
f
u
n
ctio
n
t
h
u
s
t
h
e
lear
n
in
g
a
lg
o
r
it
h
m
b
ec
o
m
e
s
i
m
p
ler
.
T
h
is
m
o
d
el
h
a
s
th
e
b
e
n
ef
it
to
h
av
e
f
aster
co
n
v
er
g
en
ce
r
ate
a
n
d
less
er
co
m
p
u
tat
io
n
al
co
s
t
t
h
an
M
L
P
s
tr
u
ctu
r
e.
T
h
e
co
n
v
en
tio
n
al
n
o
n
li
n
ea
r
f
u
n
ctio
n
al
ex
p
a
n
s
io
n
s
ca
n
b
e
e
m
p
lo
y
ed
in
F
L
A
NN
ar
e
tr
i
g
o
n
o
m
e
tr
ic,
p
o
w
er
s
er
ies
o
r
C
h
eb
y
s
h
ev
t
y
p
e.
R
.
Ma
j
h
i
et
al.
[
3
9
]
,
illu
s
tr
ates
th
at
u
s
e
o
f
tr
ig
o
n
o
m
etr
ic
f
u
n
ct
io
n
e
x
p
an
s
io
n
p
r
o
v
id
es
s
u
p
er
io
r
p
r
ed
ictio
n
ca
p
ab
ilit
y
o
f
th
e
m
o
d
el.
T
h
er
ef
o
r
e,
in
th
e
p
r
esen
t
s
it
u
atio
n
,
tr
ig
o
n
o
m
etr
ic
ex
p
a
n
s
io
n
is
e
m
p
lo
y
ed
.
L
et
u
s
ass
u
m
e
ea
ch
ele
m
en
t
o
f
t
h
e
i
n
p
u
t
p
atter
n
b
ef
o
r
e
e
x
p
an
s
io
n
i
s
r
ep
r
esen
ted
a
s
Z
(
i)
,
1
<
i
<
I
w
h
er
e
I
is
th
e
to
tal
n
u
m
b
er
o
f
f
ea
t
u
r
es
a
n
d
ea
ch
e
le
m
e
n
t
Z
(
i)
is
f
u
n
c
tio
n
all
y
e
x
p
an
d
ed
as
Xn
(
i)
,
1
<
n
<
N
,
w
h
er
e
N
i
s
t
h
e
to
tal
n
u
m
b
er
o
f
ex
p
a
n
d
ed
p
o
in
ts
f
o
r
ea
ch
i
n
p
u
t
ele
m
en
t.
L
e
t
u
s
t
ak
e,
N=
5
i
n
t
h
i
s
ex
a
m
p
le.
T
h
e
ex
p
an
s
io
n
o
f
ea
c
h
in
p
u
t p
atter
n
is
r
ep
r
esen
ted
i
n
F
ig
u
r
e
1.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
6
,
No
.
4,
Dec
em
b
er
201
7
:
3
5
9
–
3
6
7
362
Fig
u
r
e
1
.
R
ep
r
esen
ts
t
h
e
E
x
p
a
n
s
io
n
o
f
E
ac
h
I
n
p
u
t P
atter
n
T
h
e
ex
p
an
d
ed
in
p
u
ts
ar
e
f
ed
to
th
e
s
in
g
le
la
y
e
r
n
eu
r
al
n
et
w
o
r
k
a
n
d
t
h
en
th
e
n
et
w
o
r
k
is
tr
ain
ed
b
y
u
s
i
n
g
tr
ain
i
n
g
alg
o
r
it
h
m
to
o
b
tain
th
e
d
es
ir
ed
o
u
tp
u
t.
2
.
4
.
P
ro
po
s
ed
M
et
ho
d
T
h
e
p
r
o
p
o
s
ed
s
in
g
le
la
y
er
A
N
N
n
a
m
ed
as
E
F
L
A
NN
w
it
h
a
g
en
et
icall
y
o
p
ti
m
ized
s
et
o
f
f
ea
tu
r
es.
I
t
h
as
t
h
e
p
o
ten
tial
o
f
g
e
n
er
atin
g
co
m
p
lex
d
ec
i
s
io
n
r
eg
io
n
s
b
y
n
o
n
li
n
ea
r
i
m
p
r
o
v
e
m
en
t
o
f
h
id
d
en
u
n
i
ts
r
ef
er
r
ed
to
as
f
u
n
c
tio
n
al
li
n
k
s
.
F
ig
u
r
e
2
s
h
o
w
s
t
h
e
to
p
o
lo
g
ical
s
tr
u
ct
u
r
e
o
f
t
h
e
E
F
L
A
NN.
T
h
is
m
et
h
o
d
is
ch
ar
ac
ter
ized
b
y
a
s
et
o
f
F
L
ANN
w
it
h
a
d
i
f
f
er
en
t
s
u
b
s
et
o
f
f
ea
tu
r
es.
L
et
tak
e
t
h
e
n
u
m
b
er
o
f
o
r
i
g
in
a
l
f
ea
tu
r
es
o
f
t
h
e
d
ata
d
o
m
ai
n
is
n
an
d
th
e
n
u
m
b
er
o
f
f
ea
t
u
r
es
s
elec
ted
as
a
ch
r
o
m
o
s
o
m
e
o
f
th
e
g
e
n
etic
p
o
p
u
lati
o
n
is
d
f,
w
h
er
e
d
f
≤
n
.
T
h
e
d
f
v
ar
ie
s
i
n
t
h
e
g
en
etic
p
o
p
u
latio
n
f
r
o
m
c
h
r
o
m
o
s
o
m
e
s
to
ch
r
o
m
o
s
o
m
e
s
(
i.e
.
1
≤
d
f
≤
n
)
.
Fo
r
s
i
m
p
licit
y
,
let
u
s
d
is
c
u
s
s
h
o
w
a
s
i
n
g
le
ch
r
o
m
o
s
o
m
e
w
it
h
d
f
f
ea
t
u
r
es i
s
wo
r
k
in
g
co
o
p
er
ativ
el
y
f
o
r
E
FLA
N
N.
Fig
u
r
e
2
.
T
o
p
o
lo
g
ical
Stru
ctu
r
e
o
f
th
e
E
F
L
A
NN
I
n
th
is
p
r
o
p
o
s
ed
m
et
h
o
d
,
th
e
g
en
er
al
tr
ig
o
n
o
m
etr
ic
f
u
n
ctio
n
is
u
s
ed
f
o
r
m
ap
p
in
g
t
h
e
d
f
f
ea
t
u
r
es
f
r
o
m
lo
w
to
h
i
g
h
d
i
m
e
n
s
io
n
.
I
n
o
r
d
er
to
u
s
e
a
f
u
n
ct
io
n
t
h
at
is
v
er
y
clo
s
e
to
th
e
u
n
d
er
l
y
i
n
g
d
is
tr
ib
u
tio
n
o
f
th
e
d
ata
s
o
m
e
p
r
io
r
d
o
m
ain
k
n
o
w
led
g
e
is
r
eq
u
ir
ed
.
H
er
e
f
iv
e
f
u
n
c
t
io
n
s
ar
e
tak
e
n
o
u
t
o
f
w
h
ic
h
f
o
u
r
f
u
n
ctio
n
s
ar
e
tr
ig
o
n
o
m
etr
ic
an
d
o
n
e
f
u
n
ctio
n
is
lin
ea
r
(
i.e
.
,
k
ee
p
s
th
e
o
r
ig
in
a
l
f
o
r
m
o
f
th
e
f
ea
t
u
r
e
v
a
lu
e)
.
W
ith
in
t
h
e
f
o
u
r
tr
ig
o
n
o
m
etr
ic
f
u
n
ctio
n
s
t
w
o
f
u
n
ct
io
n
s
ar
e
s
i
n
e
an
d
o
t
h
er
t
wo
f
u
n
ctio
n
s
ar
e
co
s
i
n
e.
B
y
u
s
i
n
g
th
e
tr
ig
o
n
o
m
etr
ic
f
u
n
ctio
n
s
t
h
e
d
o
m
ai
n
is
f
ea
tu
r
e
v
alu
e
s
an
d
r
an
g
e
is
a
r
ea
l n
u
m
b
er
lie
s
b
et
w
ee
n
[
-
1
,
1
]
.
I
t c
a
n
b
e
r
ep
r
esen
ted
as
f
: D
o
m
a
in
→R
a
n
g
e
[
-
1,
1]
∪
{x}
(
4
.
1
)
W
h
er
e,
Do
m
ai
n
={
x
i1
,x
i2
,
·
·
·
x
idf
},
a
n
d
d
f
r
ep
r
esen
ts
th
e
n
u
m
b
er
o
f
f
ea
t
u
r
es.
L
e
t
u
s
tak
e
f
1
,
f
2
,
….
,
f
v
b
e
th
e
n
u
m
b
er
o
f
f
u
n
c
tio
n
s
t
h
at
ar
e
u
s
ed
to
e
x
p
an
d
ea
c
h
f
ea
t
u
r
e
v
al
u
e
o
f
t
h
e
p
atter
n
.
Hen
ce
,
n
o
w
ea
c
h
in
p
u
t p
atter
n
ca
n
b
e
r
ep
r
esen
te
d
as
̅
i
=
{
x
i1
,x
i2
,
·
·
·
x
idf
}→{
{f
1
(
x
i1
)
,
f
2
(
x
i1
)
,
·
·
·
f
v
(
x
i1
)
},
{
f
1
(
x
i2
)
,
f
2
(x
i2
)
,
·
·
·
f
v
(
x
i2
)
}
·
·
·
{
f
1
(
x
idf
)
,
f
2
(
x
idf
)
,
·
·
·
f
v
(
x
idf
)
}}
={
{
y
11
,y
21
,
·
·
·
y
v1
},
{
y
12
,y
22
,
·
·
·
y
v2
}
·
·
·
{
y
1df
,
y
2
df
,
·
·
·
y
v
df
}
}
(
4
.
2
)
T
h
e
w
ei
g
h
t v
ec
to
r
ex
i
s
t b
et
w
e
en
h
id
d
en
la
y
er
an
d
o
u
tp
u
t
la
y
er
is
m
u
l
tip
lied
w
it
h
t
h
e
r
es
u
lt
an
t
s
e
ts
o
f
n
o
n
li
n
ea
r
o
u
tp
u
ts
a
n
d
ar
e
s
u
p
p
lied
to
th
e
o
u
tp
u
t
n
e
u
r
o
n
as a
n
i
n
p
u
t.
T
h
er
ef
o
r
e,
th
e
w
ei
g
h
t
ed
s
u
m
is
ca
lc
u
lated
as f
o
llo
w
s
:
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
F
ea
tu
r
e
S
elec
tio
n
Usi
n
g
E
v
o
lu
tio
n
a
r
y
F
u
n
ctio
n
a
l Lin
k
N
eu
r
a
l …
(
A
ma
r
esh
S
a
h
u
)
363
Su
m
=
∑
=
1
.
(
4
.
3
)
W
h
er
e,
i=
1
,
2
·
·
·
N
an
d
m
r
ep
r
esen
t
s
th
e
to
tal
n
u
m
b
er
o
f
ex
p
a
n
d
ed
f
ea
tu
r
e
s
.
T
h
e
n
et
w
o
r
k
h
a
s
t
h
e
ab
ilit
y
to
lear
n
b
y
u
s
i
n
g
P
SO
tr
ain
i
n
g
p
r
o
ce
s
s
.
T
h
e
n
et
w
o
r
k
tr
ain
in
g
n
ee
d
s
a
s
et
o
f
tr
ain
in
g
d
ata,
i.e
.
,
a
s
er
ie
s
o
f
i
n
p
u
t
a
n
d
r
elate
d
o
u
tp
u
t
v
ec
to
r
s
.
D
u
r
in
g
t
h
e
tr
ai
n
i
n
g
p
r
o
ce
s
s
,
th
e
d
ata
i
s
r
ep
ea
ted
ly
p
r
esen
ted
to
th
e
n
e
t
w
o
r
k
a
n
d
w
ei
g
h
t
s
ar
e
ad
j
u
s
te
d
b
y
P
SO
f
r
o
m
ti
m
e
to
ti
m
e
ti
ll
th
e
d
esire
d
in
p
u
t
to
o
u
tp
u
t
m
ap
p
in
g
i
s
o
b
tain
ed
.
T
h
er
ef
o
r
e
th
e
esti
m
ated
o
u
tp
u
t is ca
l
cu
la
ted
as
f
o
llo
w
s
:
y
o
u
t
i
(
t)
=
f
(
s
u
m
i
)
,
w
h
er
e,
i =
1
,
2
·
·
·
N
(
4
.
4
)
T
h
e
er
r
o
r
is
ca
lcu
lated
f
o
r
i
th
p
atter
n
o
f
tr
ai
n
in
g
s
et
a
s
er
r
o
r
i
(
t
)
=
y
i
(
t)
–
y
o
u
t
i
(
t)
,
w
h
er
e,
i =
1
,
2
·
·
·
N
(
4
.
5
)
So
,
th
e
er
r
o
r
cr
iter
io
n
f
u
n
ct
io
n
is
ca
lcu
lated
as
E
r
(
t)
=
∑
=
1
(
4
.
6
)
Her
e,
o
u
r
o
b
j
ec
tiv
e
f
u
n
ctio
n
i
s
to
m
i
n
i
m
ize
t
h
e
er
r
o
r
b
y
ad
j
u
s
tin
g
w
ei
g
h
ts
th
r
o
u
g
h
C
P
SO
u
n
til ε
≤
E
r
,
w
h
er
e
ε
i
s
v
er
y
s
m
a
ll
v
al
u
e.
T
h
is
p
r
o
ce
s
s
is
ap
p
lie
d
r
ep
ea
ted
l
y
f
o
r
ea
ch
c
h
r
o
m
o
s
o
m
es
o
f
t
h
e
G
A
a
n
d
s
u
b
s
eq
u
en
t
l
y
,
b
ased
o
n
th
e
p
er
f
o
r
m
a
n
ce
o
f
ea
c
h
c
h
r
o
m
o
s
o
m
e
w
i
ll b
e
as
s
ig
n
ed
w
it
h
a
f
it
n
e
s
s
v
al
u
e.
Usi
n
g
t
h
is
f
it
n
es
s
v
al
u
e
t
h
e
u
s
u
al
p
r
o
ce
s
s
o
f
G
A
is
e
x
ec
u
ted
u
n
til
w
e
f
in
d
s
o
m
e
g
o
o
d
to
p
o
lo
g
y
w
h
ich
ca
n
ac
h
ie
v
e
an
ac
ce
p
tab
le
p
r
ed
ictiv
e
ac
cu
r
ac
y
.
2
.
5
.
H
ig
h L
ev
el
Alg
o
rit
h
m
s
f
o
r
E
F
L
ANN
T
h
e
r
eq
u
ir
em
e
n
t
s
f
o
r
th
e
n
ea
r
o
p
tim
a
l
E
F
L
A
NN
ar
c
h
itect
u
r
e
an
d
r
elate
d
p
ar
am
eter
s
ca
n
b
e
o
b
tain
ed
b
y
u
s
i
n
g
b
o
th
g
en
e
tic
alg
o
r
it
h
m
s
an
d
P
ar
ticle
S
w
ar
m
Op
ti
m
izat
io
n
alg
o
r
it
h
m
f
o
r
lear
n
in
g
.
E
v
o
l
u
t
io
n
ar
y
g
en
et
ic
alg
o
r
it
h
m
u
s
e
s
s
to
ch
a
s
tic
s
ea
r
ch
a
n
d
o
p
tim
izatio
n
m
et
h
o
d
s
.
So
GA
b
ased
o
n
f
u
n
d
a
m
e
n
tal
p
r
o
ce
s
s
,
s
u
c
h
as
r
ep
r
o
d
u
ctio
n
,
r
ec
o
m
b
i
n
atio
n
a
n
d
m
u
tatio
n
.
An
alg
o
r
ith
m
o
f
t
h
is
t
y
p
e
s
tar
ts
w
it
h
a
s
et
(
p
o
p
u
latio
n
)
o
f
esti
m
ates
(
g
e
n
e
s
)
,
ca
lled
in
d
i
v
id
u
al
s
(
ch
r
o
m
o
s
o
m
es)
e
n
co
d
ed
p
r
o
p
er
ly
.
Fo
r
s
o
lv
i
n
g
th
e
c
lass
i
f
icatio
n
tas
k
o
f
d
ata
m
i
n
i
n
g
t
h
e
f
it
n
es
s
o
f
ea
c
h
o
n
e
is
e
v
alu
a
ted
co
r
r
ec
tl
y
.
T
h
e
b
est
f
it
i
n
d
i
v
id
u
al
s
ar
e
al
lo
w
ed
to
m
a
k
e
a
n
d
b
ea
r
o
f
f
s
p
r
in
g
d
u
r
i
n
g
ea
ch
i
t
er
atio
n
o
f
al
g
o
r
ith
m
.
T
h
e
P
SO
is
a
p
o
p
u
latio
n
b
ased
al
g
o
r
ith
m
u
s
ed
h
er
e
to
u
p
d
ate
th
e
w
ei
g
h
ts
i
n
lear
n
in
g
p
r
o
ce
s
s
.
Du
r
i
n
g
ev
o
l
u
tio
n
a
r
y
p
r
o
ce
s
s
th
e
le
n
g
th
o
f
ea
c
h
p
ar
ticle
h
as
u
p
p
er
b
o
u
n
d
n
,
r
ep
r
esen
ts
th
e
f
ea
t
u
r
e
v
ec
to
r
.
E
ac
h
ce
ll
o
f
th
e
ch
r
o
m
o
s
o
m
e
h
o
ld
s
b
in
ar
y
v
al
u
e
eith
er
0
o
r
1
,
w
h
er
e
th
e
ce
ll
v
al
u
e
1
r
ep
r
esen
ts
t
h
e
ac
tiv
atio
n
an
d
0
r
ep
r
esen
ts
d
ea
ctiv
atio
n
in
f
u
n
ctio
n
al
ex
p
a
n
s
io
n
o
f
i
n
d
iv
id
u
als.
Du
r
in
g
ev
o
l
u
tio
n
ea
ch
in
d
i
v
i
d
u
al
m
ea
s
u
r
es
it
s
ef
f
ec
ti
v
en
e
s
s
b
y
th
e
er
r
o
r
cr
ite
r
io
n
f
u
n
ct
io
n
th
at
is
g
iv
e
n
i
n
E
q
u
atio
n
(
4
.
6
)
an
d
th
en
t
h
e
p
r
ed
ictiv
e
a
cc
u
r
ac
y
is
as
s
i
g
n
ed
a
s
it c
o
r
r
esp
o
n
d
in
g
f
it
n
es
s
.
2
.
6
.
P
s
eudo
co
de
f
o
r
E
F
L
ANN
T
h
e
s
tep
s
th
at
ar
e
f
o
llo
w
ed
b
y
E
FLA
N
N
ca
n
b
e
d
escr
ib
ed
as f
o
llo
w
s
:
1
.
Div
is
io
n
o
f
Data
s
et
Div
id
e
th
e
d
ataset
i
n
to
tr
ain
in
g
a
n
d
test
i
n
g
p
ar
ts
2
.
R
an
d
o
m
I
n
itia
lizatio
n
o
f
I
n
d
iv
id
u
al
E
ac
h
in
d
iv
id
u
al
I
n
itia
lize
ea
ch
in
d
i
v
id
u
al
r
a
n
d
o
m
l
y
f
r
o
m
t
h
e
d
o
m
ain
{0
,
1
}
3
.
W
h
ile
(
T
er
m
i
n
atio
n
cr
iter
io
n
is
n
o
t
m
et
)
4
.
Fo
r
T
h
e
Po
p
u
latio
n
4
.
1
Fo
r
e
ac
h
s
a
m
p
le
o
f
th
e
tr
ain
i
n
g
s
e
t
4
.
2
Ma
p
p
in
g
o
f
I
n
p
u
t P
atte
r
n
Ma
p
ea
ch
p
atter
n
f
r
o
m
lo
w
to
h
ig
h
d
i
m
en
s
io
n
b
y
e
x
p
an
d
i
n
g
ea
ch
f
ea
t
u
r
e
v
al
u
e
ac
co
r
d
in
g
t
o
th
e
p
r
ed
ef
in
ed
s
et
o
f
f
u
n
ctio
n
s
.
4.
3
C
alcu
latio
n
o
f
W
eig
h
te
d
Su
m
C
alcu
late
t
h
e
w
ei
g
h
ted
s
u
m
an
d
f
ee
d
as a
n
i
n
p
u
t to
t
h
e
n
o
d
e
o
f
th
e
o
u
tp
u
t la
y
er
.
4
.
4
C
alcu
late
T
h
e
E
r
r
o
r
C
alcu
late
t
h
e
er
r
o
r
an
d
ac
c
u
m
u
late
i
t.
4
.
5
U
s
e
P
ar
ticle
S
w
ar
m
Op
ti
m
izat
io
n
f
o
r
L
ea
r
n
i
n
g
Min
i
m
ize
t
h
e
er
r
o
r
b
y
ad
j
u
s
tin
g
w
ei
g
h
ts
u
s
i
n
g
P
SO.
4
.
6
A
s
s
i
g
n
T
h
e
Fit
n
e
s
s
Val
u
e
5
.
Fo
r
T
h
e
P
o
p
u
latio
n
5
.
1
R
o
u
lette
W
h
ee
l Sele
cti
o
n
is
p
er
f
o
r
m
ed
to
o
b
tain
t
h
e
b
etter
ch
r
o
m
o
s
o
m
es.
6
.
Fo
r
T
h
e
P
o
p
u
latio
n
Ob
tain
ed
6
.
1
P
er
f
o
r
m
r
ec
o
m
b
i
n
atio
n
6
.
2
Mu
tatio
n
7
.
W
h
ile
E
n
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
6
,
No
.
4,
Dec
em
b
er
201
7
:
3
5
9
–
3
6
7
364
3.
E
XP
E
R
I
M
E
NT
A
L
ST
UDI
E
S
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
E
FLAN
N
m
o
d
el
w
a
s
ev
al
u
at
ed
u
s
i
n
g
a
s
et
o
f
s
ev
e
n
b
en
ch
m
ar
k
class
i
f
icatio
n
d
ataset
s
lik
e
I
R
I
S,
W
I
NE
,
P
I
MA
,
B
UP
A
,
E
C
OL
I
,
G
L
A
S
S
an
d
VOW
E
L
f
r
o
m
th
e
Un
i
v
er
s
it
y
o
f
C
ali
f
o
r
n
ia
at
I
r
v
in
e
(
UC
I
)
m
a
ch
in
e
lear
n
in
g
r
ep
o
s
ito
r
y
[
4
0
]
.
Fo
r
class
if
y
in
g
s
i
x
o
v
er
lap
p
in
g
v
o
w
el
class
e
s
,
w
e
h
av
e
al
s
o
tak
e
n
VOW
E
L
d
ataset
to
s
h
o
w
th
e
p
er
f
o
r
m
an
ce
o
f
E
F
L
A
NN
[
4
1
]
.
W
e
h
av
e
u
s
ed
M
L
P
,
FLAN
N
w
it
h
g
r
ad
ien
t
d
escen
t
lear
n
in
g
,
R
B
F
an
d
HF
L
ANN
r
esu
lt
s
co
m
p
ar
is
o
n
w
it
h
p
r
o
p
o
s
ed
m
eth
o
d
.
T
ab
le
1
s
u
m
m
ar
ize
s
th
e
m
ai
n
ch
ar
ac
ter
is
tics
o
f
t
h
e
d
atab
ases
t
h
at
h
av
e
b
ee
n
u
s
ed
in
t
h
is
p
ap
er
.
T
ab
le
.
1
.
Su
m
m
ar
y
o
f
t
h
e
Dat
asets
u
s
ed
in
S
i
m
u
lat
io
n
S
t
u
d
i
es
S
e
r
i
a
l
N
o
.
N
a
me
o
f
D
a
t
a
se
t
N
o
.
o
f
p
a
t
t
e
r
n
s
N
o
.
o
f
a
t
t
r
i
b
u
t
e
N
o
.
o
f
c
l
a
sse
s
1
I
R
I
S
1
5
0
4
3
2
W
I
N
E
1
7
8
13
3
3
P
I
M
A
7
6
8
8
2
4
B
U
P
A
3
4
5
6
2
5
EC
O
L
I
3
3
6
7
8
6
G
L
A
S
S
2
1
4
9
6
7
V
O
W
E
L
8
7
1
3
6
8
Z
O
O
1
0
1
16
7
3
.1
.
P
a
ra
m
et
er
Set
up
B
ef
o
r
e
ev
alu
ati
n
g
th
e
p
r
o
p
o
s
ed
E
FLA
NN
al
g
o
r
ith
m
,
w
e
n
ee
d
to
s
et
th
e
u
s
er
d
ef
i
n
ed
p
ar
am
eter
s
a
n
d
p
r
o
to
co
ls
r
elate
d
to
th
e
d
atase
t.
Her
e
w
e
h
a
v
e
u
s
ed
t
w
o
f
o
l
d
cr
o
s
s
v
a
lid
atio
n
f
o
r
all
th
e
d
ataset
b
y
r
an
d
o
m
l
y
d
iv
id
in
g
t
h
e
to
tal
s
et
o
f
p
atter
n
s
o
f
d
ataset
i
n
to
t
w
o
eq
u
al
p
ar
ts
a
s
d
ata
s
et1
.
d
at
an
d
d
ataset
2
.
d
at.
E
ac
h
o
f
t
h
es
e
t
w
o
s
ets
o
f
d
ata
s
et
w
a
s
alter
n
ati
v
el
y
u
s
ed
eit
h
er
as
a
tr
ai
n
in
g
s
et
o
r
a
test
s
et
[
2
2
]
.
T
h
e
q
u
al
it
y
o
f
ea
c
h
in
d
iv
id
u
al
o
f
d
ataset
is
m
ea
s
u
r
ed
b
y
th
e
p
r
ed
ictiv
e
p
er
f
o
r
m
an
ce
o
b
tain
ed
d
u
r
in
g
tr
ain
in
g
.
W
e
n
ee
d
to
s
et
th
e
o
p
tim
a
l
v
al
u
es
o
f
t
h
e
f
o
llo
w
i
n
g
p
ar
a
m
eter
s
to
r
ed
u
ce
th
e
l
o
ca
l
o
p
tim
al
it
y
.
T
h
e
p
ar
a
m
ete
r
s
w
e
h
a
v
e
u
s
ed
ar
e
d
escr
ib
ed
as
f
o
llo
w
s
:
P
o
p
u
latio
n
s
ize:
T
h
e
s
ize
o
f
th
e
p
o
p
u
latio
n
d
esi
g
n
ated
as
|
P
|
=5
0
is
f
i
x
ed
f
o
r
all
th
e
d
atasets
.
T
h
e
n
u
m
b
er
o
f
i
n
p
u
t
f
ea
t
u
r
es
i
s
n
.
He
n
ce
le
n
g
th
o
f
th
e
in
d
i
v
id
u
al
s
i
s
f
ix
ed
to
n
.
T
h
e
p
r
o
b
a
b
ilit
y
f
o
r
cr
o
s
s
o
v
er
is
tak
e
n
a
s
0
.
7
an
d
th
e
p
r
o
b
ab
ilit
y
f
o
r
m
u
ta
tio
n
is
0
.
0
2
.
T
h
e
to
tal
n
u
m
b
er
o
f
iter
atio
n
s
w
e
h
av
e
ap
p
lied
is
1
0
0
0
f
o
r
all
th
e
d
a
tasets
.
P
ar
a
m
eter
s
o
f
P
SO
f
o
r
E
FLA
NN
tr
a
i
n
i
n
g
p
r
o
ce
s
s
a
r
e
s
et
s
i
m
i
lar
l
y
as
n
eu
r
al
n
et
w
o
r
k
tr
ai
n
i
n
g
w
it
h
P
SO [
4
2
]
.
T
h
e
ML
P
u
s
es
t
w
o
h
id
d
en
la
y
er
s
f
o
r
all
d
atase
ts
.
I
n
ea
c
h
h
id
d
en
la
y
er
P
×
Q
n
u
m
b
er
s
o
f
n
eu
r
o
n
s
ar
e
tak
en
.
W
h
er
e,
P
an
d
Q
ar
e
th
e
n
u
m
b
er
o
f
in
p
u
t n
e
u
r
o
n
s
an
d
o
u
tp
u
t n
e
u
r
o
n
s
.
3
.
2
Co
m
p
a
ra
t
iv
e
P
er
f
o
rm
a
n
ce
T
h
e
p
r
e
d
ictiv
e
p
er
f
o
r
m
a
n
ce
ac
h
iev
ed
f
r
o
m
p
r
o
p
o
s
ed
E
FL
A
N
N
f
o
r
th
e
ab
o
v
e
m
en
tio
n
ed
d
atasets
w
er
e
co
m
p
ar
ed
w
it
h
t
h
e
r
es
u
lts
o
b
tain
ed
f
r
o
m
M
L
P
,
FLA
N
N
w
ith
g
r
ad
ien
t
d
esce
n
t
lear
n
in
g
,
R
B
F
a
n
d
HFLA
NN
r
es
u
lt
s
.
T
ab
le
2
an
d
T
ab
le
3
s
u
m
m
ar
ize
t
h
e
co
m
p
ar
ativ
e
av
er
a
g
e
tr
ai
n
in
g
a
n
d
te
s
t
p
er
f
o
r
m
a
n
ce
s
o
f
E
FLAN
N
w
it
h
o
th
er
ab
o
v
e
m
en
tio
n
ed
m
eth
o
d
s
.
T
ab
le
2
.
A
v
er
ag
e
C
o
m
p
ar
ativ
e
P
er
f
o
r
m
a
n
ce
o
f
E
F
L
A
NN,
ML
P
an
d
R
B
F
D
a
t
a
se
t
s
A
l
g
o
r
i
t
h
ms
EFL
A
N
N
M
L
P
RBF
T
r
a
i
n
i
n
g
T
e
st
i
n
g
T
r
a
i
n
i
n
g
T
e
st
i
n
g
T
r
a
i
n
i
n
g
T
e
st
i
n
g
IR
IS
9
8
.
4
7
1
0
9
7
.
4
5
0
5
9
7
.
1
5
1
0
9
4
.
0
0
0
0
3
8
.
5
0
0
0
3
8
.
5
0
0
0
W
I
N
E
9
9
.
7
5
6
0
9
0
.
5
6
0
0
9
6
.
1
4
1
0
9
2
.
2
9
2
0
8
5
.
3
9
3
5
7
9
.
2
1
3
0
P
I
M
A
8
0
.
8
2
2
0
7
2
.
6
3
4
3
7
6
.
6
1
0
0
7
7
.
1
9
2
0
7
7
.
4
7
4
0
7
6
.
0
4
1
5
B
U
P
A
7
7
.
9
8
3
0
6
9
.
3
7
9
8
6
7
.
5
2
2
0
6
7
.
3
9
0
0
7
1
.
0
1
2
5
6
6
.
9
5
3
0
EC
O
L
I
5
5
.
3
6
3
0
5
0
.
9
7
3
0
4
6
.
9
6
0
0
4
5
.
2
6
5
0
3
1
.
1
7
8
0
2
6
.
1
1
0
0
G
L
A
S
S
6
3
.
7
5
9
0
5
1
.
7
5
2
5
4
3
.
2
8
4
0
4
0
.
9
8
0
0
4
8
.
9
8
6
5
3
4
.
6
4
4
0
V
O
W
E
L
4
0
.
8
2
1
0
3
8
.
6
6
3
5
3
8
.
2
5
0
0
3
7
.
4
5
0
0
2
5
.
2
5
5
5
2
4
.
3
2
5
0
Z
O
O
9
9
.
4
5
3
2
8
6
.
4
6
2
4
9
6
.
3
5
0
0
9
2
.
3
2
0
0
9
6
.
1
5
4
0
8
1
.
0
8
3
0
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
F
ea
tu
r
e
S
elec
tio
n
Usi
n
g
E
v
o
lu
tio
n
a
r
y
F
u
n
ctio
n
a
l Lin
k
N
eu
r
a
l …
(
A
ma
r
esh
S
a
h
u
)
365
T
ab
le
3
.
A
v
er
ag
e
C
o
m
p
ar
ativ
e
P
er
f
o
r
m
a
n
ce
o
f
E
F
L
A
NN,
HFLA
NN
an
d
R
B
F
D
a
t
a
s
e
t
s
A
l
g
o
r
i
t
h
ms
EFL
A
N
N
H
F
L
A
N
N
F
L
A
N
N
T
r
a
i
n
i
n
g
T
e
st
i
n
g
T
r
a
i
n
i
n
g
T
e
st
i
n
g
T
r
a
i
n
i
n
g
T
e
st
i
n
g
I
R
I
S
9
8
.
4
7
1
0
9
7
.
4
5
0
5
9
8
.
0
0
0
1
9
7
.
3
3
3
5
9
6
.
6
6
6
4
9
6
.
6
6
6
4
W
I
N
E
9
9
.
7
5
6
0
9
0
.
5
6
0
0
9
9
.
4
3
8
0
9
0
.
4
4
9
5
9
7
.
1
9
1
0
8
8
.
7
6
4
0
P
I
M
A
8
0
.
8
2
2
0
7
2
.
6
3
4
3
8
0
.
7
2
9
0
7
2
.
1
3
5
5
7
9
.
5
5
7
0
7
2
.
1
3
5
5
B
U
P
A
7
7
.
9
8
3
0
6
9
.
3
7
9
8
7
7
.
6
8
2
0
6
9
.
2
7
8
5
7
7
.
9
7
2
5
6
9
.
2
8
0
0
EC
O
L
I
5
5
.
3
6
3
0
5
0
.
9
7
3
0
5
5
.
1
6
7
0
5
0
.
8
0
2
0
4
9
.
9
6
2
5
4
7
.
3
0
7
5
G
L
A
S
S
6
3
.
7
5
9
0
5
1
.
7
5
2
5
6
3
.
5
5
6
5
5
1
.
5
0
7
5
6
0
.
7
5
1
0
5
0
.
3
8
0
0
V
O
W
E
L
4
0
.
8
2
1
0
3
8
.
6
6
3
5
4
0
.
4
3
9
5
3
8
.
1
9
6
5
2
7
.
9
2
5
0
2
4
.
7
2
2
0
Z
O
O
9
9
.
4
5
3
2
8
6
.
4
6
2
4
9
9
.
0
3
8
5
8
6
.
1
8
5
0
9
7
.
1
1
5
5
8
5
.
1
6
4
5
Fro
m
T
ab
le
2
an
d
T
a
b
le
3
,
w
e
ca
n
ea
s
il
y
v
er
if
y
t
h
at
ex
c
ep
t
B
UP
A
ca
s
e
in
all
o
t
h
er
ca
s
es
o
n
a
n
av
er
ag
e
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
E
FLAN
N
is
g
i
v
in
g
as
s
u
r
in
g
r
esu
lt
s
i
n
b
o
th
tr
ain
i
n
g
an
d
te
s
t
ca
s
es.
F
L
A
NN
is
p
er
f
o
r
m
in
g
b
etter
i
n
ca
s
e
o
f
B
UP
A
d
ataset.
Fi
g
u
r
e
3
s
h
o
w
s
th
e
p
er
ce
n
ta
g
e
o
f
f
ea
t
u
r
e
s
elec
ted
b
y
t
h
e
E
FLAN
N
co
m
p
ar
e
to
HF
L
ANN
i
n
t
h
e
co
n
te
x
t
o
f
co
m
p
r
e
h
en
s
ib
ilit
y
.
T
h
e
d
ataset
s
ar
e
r
ep
r
esen
ted
b
y
u
s
in
g
X
-
a
x
is
an
d
p
er
ce
n
tag
e
o
f
ac
ti
v
e
b
it
s
i
n
t
h
e
o
p
ti
m
al
ch
r
o
m
o
s
o
m
e
o
b
tai
n
ed
d
u
r
i
n
g
t
h
e
tr
ai
n
in
g
ar
e
r
ep
r
esen
ted
b
y
u
s
i
n
g
Y
-
ax
i
s
.
Fig
u
r
e
3
.
R
ep
r
esen
ts
t
h
e
P
er
ce
n
tag
e
o
f
Op
ti
m
al
s
et
o
f
Select
ed
Featu
r
es in
E
FLAN
N
a
n
d
HFLA
NN
3
.
3
.
P
re
dict
iv
e
Acc
ura
c
y
by
K
no
w
ledg
e
I
nco
rpo
ra
t
io
n
L
et
u
s
ta
k
e
n
b
e
th
e
to
tal
n
u
m
b
er
o
f
f
ea
t
u
r
es
ex
i
s
t
in
t
h
e
d
ataset,
F
1
an
d
F
2
d
en
o
te
t
h
e
n
u
m
b
er
o
f
f
ea
t
u
r
e
s
elec
ted
b
y
u
s
i
n
g
t
h
e
tr
ain
i
n
g
s
et
1
a
n
d
test
i
n
g
s
et
2
alter
n
at
iv
el
y
.
No
tat
io
n
s
w
it
h
Me
an
i
n
g
:
I
NI
d
en
o
tes
th
e
to
tal
n
u
m
b
er
o
f
f
e
atu
r
es
i
n
t
h
e
d
ataset,
I
F
1
I
r
ep
r
esen
t
s
t
h
e
to
tal
n
u
m
b
er
o
f
s
e
le
cted
f
ea
tu
r
e
s
i
n
test
s
et
2
.
I
F
2
I
r
e
p
r
esen
ts
t
h
e
t
o
tal
n
u
m
b
er
o
f
s
elec
ted
f
e
atu
r
es
i
n
te
s
t
s
et
1
a
n
d
|
P
A|
r
ep
r
esen
t
s
t
h
e
p
r
ed
ictiv
e
ac
cu
r
ac
y
.
T
h
e
f
itn
e
s
s
o
f
t
h
e
ch
r
o
m
o
s
o
m
e
r
elate
d
to
F
1
is
f
(
F
1
)
=
|
|
×
|
|
−
×
|
F
1
|
|
|
(
5
.
1
)
Si
m
i
lar
l
y
,
t
h
e
f
i
tn
e
s
s
o
f
t
h
e
c
h
r
o
m
o
s
o
m
e
r
elate
d
to
F
2
is
f
(
F
2
)
=
|
|
×
|
|
−
×
|
F
2
|
|
|
(
5
.
2
)
r
ep
r
esen
ts
th
e
p
r
ed
ictiv
e
ac
c
u
r
ac
y
a
n
d
tr
ad
eo
f
f
b
et
w
ee
n
t
w
o
cr
iter
ia
an
d
its
v
al
u
e
is
ta
k
e
n
as
=
0
.
0
1
.
T
ab
le
4
.
P
r
ed
ictiv
e
A
cc
u
r
ac
y
o
f
E
FLAN
N
b
y
K
n
o
w
led
g
e
I
n
co
r
p
o
r
atio
n
w
it
h
=
0
.
0
1
D
a
t
a
se
t
s
N
P
r
e
d
i
c
t
i
v
e
A
c
c
u
r
a
c
y
o
f
Te
st
S
e
t
1
C
h
r
o
mo
so
me
P
r
e
d
i
c
t
i
v
e
A
c
c
u
r
a
c
y
o
f
Te
st
S
e
t
1
C
h
r
o
mo
so
me
I
R
I
S
4
9
6
.
1
0
0
1
9
7
.
3
2
6
0
W
I
N
E
13
9
0
.
0
1
1
0
9
1
.
1
2
6
4
P
I
M
A
8
7
1
.
6
5
2
5
7
2
.
9
6
4
8
B
U
P
A
6
7
0
.
5
3
4
6
6
8
.
9
6
0
2
EC
O
L
I
7
5
4
.
8
9
3
9
4
6
.
9
0
2
0
G
L
A
S
S
9
5
7
.
4
3
7
4
4
5
.
9
8
2
4
V
O
W
E
L
3
3
5
.
1
2
6
3
4
0
.
1
3
9
5
Z
O
O
16
8
7
.
9
4
9
2
8
4
.
9
2
1
2
0
10
0
IRIS
PI…
E
C…
VO…
Per
c
e
n
tage
o
f
Feat
u
r
e
S
e
l
e
c
te
d
D
atase
ts
Per
c
e
n
tage
o
f Feat
u
r
e
S
e
l
e
c
te
d
H
FLAN
N
E
FLAN
N
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
6
,
No
.
4,
Dec
em
b
er
201
7
:
3
5
9
–
3
6
7
366
4.
CO
NCLU
SI
O
N
I
n
th
is
s
tu
d
y
,
th
e
p
r
o
p
o
s
ed
E
FLAN
N
m
eth
o
d
is
u
s
ed
f
o
r
th
e
tas
k
o
f
d
ata
m
in
i
n
g
.
I
t
h
as
g
i
v
en
i
m
p
o
r
tan
ce
to
t
h
e
s
e
lectio
n
o
f
o
p
tim
a
l set
o
f
f
ea
t
u
r
es a
n
d
it
h
as a
ls
o
e
n
h
a
n
ce
d
t
h
e
clas
s
i
f
ica
tio
n
ac
cu
r
ac
y
.
T
h
e
class
i
f
icatio
n
p
er
f
o
r
m
an
ce
o
f
E
FLAN
N
m
o
d
el
is
p
r
o
m
is
i
n
g
as
d
em
o
n
s
tr
ated
in
e
x
p
er
i
m
e
n
tal
s
t
u
d
ies.
Her
e
th
e
ar
ch
itect
u
r
al
co
m
p
le
x
it
y
i
s
lo
w
b
u
t,
tr
ai
n
in
g
ti
m
e
is
m
o
r
e
th
an
F
L
A
NN.
A
l
s
o
it
is
s
h
o
w
n
th
at
E
F
L
A
N
N
h
as
b
etter
f
ea
t
u
r
e
s
el
ec
tio
n
ca
p
ab
ilit
y
th
a
n
HF
L
A
N
N.
T
h
er
ef
o
r
e,
th
is
m
o
d
el
ca
n
b
e
u
s
ed
in
class
i
f
icatio
n
task
s
o
f
d
ata
m
i
n
i
n
g
.
RE
F
E
R
E
NC
E
S
[1
]
F
a
y
y
a
d
,
U
.
A
n
d
Uth
u
ru
sa
m
y
,
R
.
:
Da
ta
M
in
in
g
a
n
d
Kn
o
w
le
d
g
e
Dis
c
o
v
e
r
y
in
Da
tab
a
se
s.
Co
m
m
u
n
.
ACM
,
3
9
:
2
4
-
2
7
.
1
9
9
6
.
[2
]
F
a
y
y
a
d
,
U
.
M
.,
P
iate
tsk
y
-
S
h
a
p
iro
,
G
.
,
A
n
d
S
m
y
th
,
P
.
:
F
r
o
m
Da
ta
M
in
i
n
g
to
Kn
o
w
led
g
e
Disc
o
v
e
r
y
:
A
n
O
v
e
rv
i
e
w
.
In
U
.
M
.
F
a
y
y
a
d
,
G
.
P
iate
tsk
y
-
S
h
a
p
iro
,
&
P
.
S
m
y
th
(
Ed
s
.
),
A
d
v
a
n
c
e
s
in
Kn
o
w
led
g
e
Disc
o
v
e
r
y
a
n
d
Da
ta
M
in
in
g
,
P
p
.
1
–
3
4
,
.
M
e
n
lo
P
a
rk
,
CA
:
AA
A
I
P
re
ss
,
1
9
9
6
.
[3
]
Ag
ra
w
a
l
,
R
.,
I
m
ielin
sk
i
,
T
.
,
A
n
d
S
w
a
m
i
A
.
:
Da
tab
a
s
e
M
in
in
g
:
A
P
e
rf
o
r
m
a
n
c
e
P
e
rsp
e
c
ti
v
e
.
IEE
E
T
r
a
n
s
.
Kn
o
w
led
g
e
Da
ta E
n
g
,
5
,
P
p
.
9
1
4
-
9
2
5
,
1
9
9
3
.
[4
]
Ja
m
e
s
,
M
.
:
Clas
sif
ic
a
ti
o
n
A
lg
o
rit
h
m
s
.
W
il
e
y
,
1
9
8
5
.
[5
]
Bre
i
man
,
L
.,
F
ried
m
a
n
,
J
.
H
.,
Olsh
e
n
,
R
.
A
.
,
A
n
d
S
t
o
n
e
C
.
J
.
:
C
las
sif
ic
a
ti
o
n
a
n
d
Re
g
re
ss
io
n
L
e
a
rn
in
g
.
M
o
rg
a
n
Ka
u
fm
a
n
,
1
9
8
4
.
[6
]
Krie
g
e
l
,
H
.
P
.
,
Et
A
l.
:
F
u
tu
re
T
re
n
d
s
in
Da
ta
M
in
in
g
.
Da
ta
M
in
in
g
a
n
d
Kn
o
w
led
g
e
Disc
o
v
e
r
y
,
1
5
(
1
),
8
7
–
9
7
.
Ne
th
e
rlan
d
s
:
S
p
rin
g
e
r
,
2
0
0
7
.
[7
]
Du
d
a
,
R
.
O
.
,
Ha
rt
,
P
.
E
.
a
n
d
S
to
rk
D
.
G
.
:
P
a
tt
e
rn
Clas
sif
ica
ti
o
n
.
W
il
e
y
,
Ne
w
Yo
rk
,
2
0
0
1
.
[8
]
G
o
ld
b
e
rg
,
D
.
E
.
:
G
e
n
e
ti
c
A
lg
o
rit
h
m
s in
S
e
a
rc
h
,
Op
ti
m
iza
ti
o
n
a
n
d
M
a
c
h
in
e
L
e
a
rn
in
g
.
M
o
rg
a
n
Ka
u
f
m
a
n
n
,
1
9
8
9
.
[9
]
Zh
a
n
g
,
G
.
P
.
:
Ne
u
ra
l
Ne
tw
o
rk
s
f
o
r
Clas
sif
ica
ti
o
n
,
A
S
u
rv
e
y
.
IEE
E
T
ra
n
sa
c
t
io
n
s
o
n
S
y
ste
m
s
,
M
a
n
,
C
y
b
e
rn
e
ti
c
s
-
P
a
rt
C
:
A
p
p
li
c
a
ti
o
n
a
n
d
Re
v
iew
s
,
3
0
(4
),
P
p
.
4
5
1
–
4
6
1
,
2
0
0
0
.
[1
0
]
Zh
a
n
g
,
G
.
P
.
:
A
v
o
id
in
g
P
it
f
a
ll
s
in
Ne
u
ra
l
Ne
t
w
o
rk
Re
se
a
r
c
h
.
IEE
E
T
ra
n
sa
c
ti
o
n
s
on
S
y
ste
m
s
,
M
a
n
a
n
d
Cy
b
e
rn
e
ti
c
s
-
P
a
rt
C
:
A
p
p
li
c
a
ti
o
n
s
a
n
d
Re
v
iew
s
,
3
7
(1
),
P
p
.
3
–
1
6
,
2
0
0
7.
[1
1
]
M
c
c
u
ll
o
c
h
,
W
.,
P
it
ts
,
W
.
:
A
L
o
g
ica
l
C
a
lcu
lu
s
o
f
th
e
Id
e
a
s
I
m
m
a
n
e
n
t
in
Ne
rv
o
u
s
A
c
ti
v
it
y
.
Bu
ll
e
ti
n
o
f
M
a
th
e
m
a
ti
c
a
l
Bio
p
h
y
sic
s
7
,
1
1
5
-
1
3
3
,
1
9
4
3
.
[1
2
]
G
il
e
s
C
.
L
.,
M
a
x
w
e
ll
,
T
.,
Lea
rn
in
g
In
v
a
rian
c
e
,
a
n
d
G
e
n
e
ra
li
z
a
ti
o
n
in
A
Hig
h
e
r
Ord
e
r
Ne
u
ra
l
N
e
tw
o
rk
.
A
p
p
li
e
d
Op
ti
c
s
2
6
(
2
3
)
,
4
9
7
2
-
4
9
7
8
,
1
9
8
7
.
[1
3
]
M
in
sk
y
,
M
.,
P
a
p
e
rt
,
S
.
:
P
e
rc
e
p
tr
o
n
s
.
T
h
e
M
IT
P
re
ss
,
1
9
6
9
.
[1
4
]
W
id
ro
w
,
B
.,
Ho
ff
,
M
.
E
.
:
A
d
a
p
ti
v
e
S
w
it
c
h
in
g
Circu
it
s.
IRE
W
ES
C
ON
Co
n
v
e
n
ti
o
n
Re
c
o
r
d
,
9
6
-
1
0
4
,
1
9
6
0
.
[1
5
]
W
id
ro
w
,
B
.,
L
e
h
r
,
M
.
:
3
0
Ye
a
rs
Of
A
d
a
p
ti
v
e
N
e
u
ra
l
Ne
t
w
o
rk
s
:
P
e
rc
e
p
tr
o
n
,
M
a
d
a
li
n
e
,
a
n
d
Ba
c
k
-
P
ro
p
a
g
a
ti
o
n
.
P
r
o
c
e
e
d
in
g
s
Of
IEE
E
7
8
(9
),
1
4
1
5
-
1
4
4
2
,
1
9
9
0
.
[1
6
]
G
il
e
s
,
C
.
L
.,
M
a
x
w
e
ll
,
T
.
:
L
e
a
rn
in
g
,
In
v
a
rian
c
e
a
n
d
G
e
n
e
ra
li
z
a
ti
o
n
in
Hig
h
e
r
-
Ord
e
r
Ne
u
ra
l
Ne
tw
o
rk
s.
A
p
p
l
.
Op
t
.
2
6
(
2
3
),
4
9
7
2
-
4
9
7
8
,
1
9
8
7
.
[1
7
]
[1
7
]
P
a
o
,
Y
.
H
.
:
A
d
a
p
ti
v
e
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
a
n
d
Ne
u
ra
l
Ne
tw
o
rk
.
Re
a
d
in
g
,
MA
:
A
d
d
iso
n
-
W
e
sle
y
,
1
9
8
9
.
[1
8
]
Yo
u
n
g
,
M
.
:
T
h
e
T
e
c
h
n
ica
l
W
rit
e
r'
s Ha
n
d
b
o
o
k
.
M
il
l
V
a
ll
e
y
,
CA
Un
iv
e
rsit
y
S
c
ien
c
e
,
1
9
8
9
.
[1
9
]
V
e
n
k
a
tes
h
,
S
.
S
.
Ba
l
d
i
,
P
.
:
P
r
o
g
ra
m
m
e
d
In
tera
c
ti
o
n
s
i
n
Hig
h
e
r
Ord
e
r
Ne
u
ra
l
Ne
tw
o
rk
s:
M
a
x
i
m
a
l
Ca
p
a
c
it
y
.
J
.
Co
m
p
lex
it
y
7
,
3
1
6
-
3
3
7
,
1
9
9
1
.
[2
0
]
A
n
t
y
o
m
o
v
,
E
.
A
n
d
P
e
c
h
t
,
O
.
Y
.
:
M
o
d
ifi
e
d
Hig
h
e
r
Ord
e
r
Ne
u
ra
l
Ne
tw
o
rk
f
o
r
In
v
a
ri
a
n
t
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
.
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
L
e
tt
e
rs
2
6
,
8
4
3
-
8
5
1
,
2
0
0
5
.
[2
1
]
P
a
o
,
Y
.
H
.
,
P
h
il
ip
s
,
S
.
M
.
:
T
h
e
F
u
n
c
ti
o
n
a
l
L
in
k
Ne
t
L
e
a
rn
in
g
Op
ti
m
a
l
Co
n
tro
l.
Ne
u
r
o
Co
m
p
u
ti
n
g
9
,
1
4
9
-
1
6
4
,
1
9
9
5
.
[2
2
]
M
isra
,
B
.
B
.,
De
h
u
ri
,
S
.
:
F
u
n
c
ti
o
n
a
l
L
in
k
Ne
u
ra
l
Ne
t
w
o
rk
F
o
r
Clas
sifi
c
a
ti
o
n
T
a
sk
in
Da
ta
M
in
in
g
.
Jo
u
r
n
a
l
o
f
Co
m
p
u
ter S
c
ien
c
e
3
(1
2
),
9
4
8
-
9
5
5
,
2
0
0
7
.
[2
3
]
M
irea
,
L
.,
M
a
rc
u
,
T
.
:
S
y
ste
m
Id
e
n
ti
fi
c
a
ti
o
n
Us
in
g
F
u
n
c
ti
o
n
a
l
L
in
k
Ne
u
ra
l
Ne
t
w
o
rk
s
W
it
h
D
y
n
a
m
ic
S
tru
c
tu
re
.
1
5
th
T
rien
n
ial
W
o
rld
Co
n
g
re
ss
,
B
a
rc
e
lo
n
a
,
S
p
a
in
,
2
0
0
2
.
[2
4
]
Ca
ss
,
R
.,
Ra
d
l
,
B
.
:
A
d
a
p
ti
v
e
P
ro
c
e
ss
Op
ti
m
iza
ti
o
n
Us
in
g
F
u
n
c
ti
o
n
a
l
L
in
k
Ne
t
w
o
rk
s
a
n
d
Ev
o
lu
ti
o
n
a
ry
A
lg
o
rit
h
m
s.
Co
n
tr
o
l
E
n
g
.
P
ra
c
ti
c
e
4
(1
1
),
1
5
7
9
-
1
5
8
4
,
1
9
9
6
.
[2
5
]
S
h
in
,
Y
.,
G
h
o
sh
,
J
.
:
T
h
e
Pi
-
S
ig
m
a
Ne
t
w
o
rk
s:
A
n
E
ffi
c
ien
t
Hig
h
e
r
Ord
e
r
Ne
u
ra
l
Ne
t
w
o
rk
f
o
r
P
a
tt
e
rn
Clas
sifi
c
a
ti
o
n
a
n
d
F
u
n
c
ti
o
n
A
p
p
ro
x
im
a
ti
o
n
.
P
ro
c
e
e
d
in
g
s
o
f
In
tern
a
ti
o
n
a
l
Jo
i
n
t
C
o
n
f
e
re
n
c
e
on
Ne
u
ra
l
Ne
tw
o
rk
s
,
1
3
-
1
8
,
1
9
9
1
.
[2
6
]
S
h
in
,
Y
.,
G
h
o
sh
,
J
.
:
C
o
m
p
u
tatio
n
a
ll
y
E
ffi
c
ien
t
In
v
a
rian
t
P
a
tt
e
rn
Re
c
o
g
n
it
i
o
n
W
it
h
Hig
h
e
r
O
rd
e
r
P
i
-
S
ig
m
a
Ne
tw
o
rk
s.
T
h
e
U
n
iv
e
rsit
y
o
f
T
e
x
a
s
A
t
A
u
stin
,
T
e
c
h
.
Re
p
o
rt
,
1
9
9
2
.
[2
7
]
Zh
u
,
Q
.,
Ca
i
,
Y
.,
L
iu
,
L
.
:
A
G
lo
b
a
l
L
e
a
rn
in
g
A
l
g
o
rit
h
m
F
o
r
A
RBF
Ne
tw
o
rk
.
Ne
u
ra
l
Ne
t
w
o
rk
s
1
2
,
5
2
7
-
5
4
0
,
1
9
9
9
.
[2
8
]
Li
,
M
.,
T
ian
,
J
.,
C
h
e
n
,
F
.
:
Im
p
ro
v
in
g
M
u
lt
icla
ss
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
W
it
h
A
Co
-
Ev
o
lu
ti
o
n
a
ry
RBF
NN
.
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
L
e
tt
e
rs
2
9
,
3
9
2
-
4
0
6
,
2
0
0
8
.
[2
9
]
S
h
in
,
Y
.
,
G
h
o
sh
,
J
.
:
Rid
g
e
P
o
ly
n
o
m
ial
N
e
t
w
o
rk
s.
IEE
E
T
ra
n
sa
c
ti
o
n
s Ne
u
ra
l
Ne
tw
o
rk
s
6
(2
),
6
1
0
-
6
2
2
.
1
9
9
5
.
[3
0
]
S
h
in
,
Y
.,
G
h
o
sh
,
J
.
:
A
p
p
ro
x
im
a
ti
o
n
Of
M
u
lt
iv
a
riate
F
u
n
c
ti
o
n
s
Us
in
g
Rid
g
e
P
o
ly
n
o
m
ial
Ne
t
w
o
rk
s.
P
r
o
c
e
e
d
in
g
s
o
f
In
tern
a
ti
o
n
a
l
Jo
in
t
C
o
n
f
e
re
n
c
e
o
n
Ne
u
ra
l
Ne
tw
o
rk
s II
,
3
8
0
-
3
8
5
,
1
9
9
2
.
[3
1
]
P
a
o
,
Y
.
H
.
,
P
h
il
li
p
s
,
S
.
M
.
,
S
o
b
a
j
ic
,
D
.
J
.
:
Ne
u
ra
l
-
Ne
t
C
o
m
p
u
ti
n
g
a
n
d
In
telli
g
e
n
t
Co
n
tro
l
S
y
ste
m
s,
In
tern
a
ti
o
n
a
l
Jo
u
rn
a
l
o
f
Co
n
tr
o
l
S
y
ste
m
s
5
6
(2
),
2
6
3
-
2
8
9
,
1
9
9
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
F
ea
tu
r
e
S
elec
tio
n
Usi
n
g
E
v
o
lu
tio
n
a
r
y
F
u
n
ctio
n
a
l Lin
k
N
eu
r
a
l …
(
A
ma
r
esh
S
a
h
u
)
367
[3
2
]
M
ish
ra
,
B
.
B
.,
De
h
u
r
i
,
S
.
:
F
u
n
c
t
io
n
a
l
L
in
k
A
rti
f
icia
l
Ne
u
ra
l
Ne
t
w
o
rk
F
o
r
Clas
si
f
ic
a
ti
o
n
T
a
sk
in
Da
ta
M
in
in
g
.
Jo
u
rn
a
l
o
f
Co
m
p
u
ter S
c
ien
c
e
3
,
9
4
8
-
9
5
5
,
2
0
0
7
.
[3
3
]
De
h
u
ri
,
S
.,
Ch
o
S
.
B
.
:
Ev
o
l
u
ti
o
n
a
r
y
Op
ti
m
i
z
e
d
F
e
a
tu
re
s
in
F
u
n
c
ti
o
n
a
l
L
in
k
Ne
u
ra
l
Ne
tw
o
rk
F
o
r
Clas
sif
ic
a
ti
o
n
.
Ex
p
e
rt
S
y
ste
m
A
n
d
A
p
p
li
c
a
ti
o
n
s
3
7
,
4
3
7
9
-
4
3
9
1
,
2
0
1
0
.
[3
4
]
De
h
u
ri
,
S
.
,
Ch
o
S
.
B
.
:
A
H
y
b
rid
Ge
n
e
ti
c
Ba
se
d
F
u
n
c
ti
o
n
a
l
L
in
k
A
rti
f
icia
l
Ne
u
ra
l
Ne
t
w
o
rk
W
it
h
A
S
tatisti
c
a
l
Co
m
p
a
riso
n
o
f
Clas
sif
iers
O
v
e
r
M
u
lt
i
p
le Da
tas
e
ts.
Ne
u
ra
l
Co
m
p
u
ti
n
g
A
p
p
li
c
a
ti
o
n
s
1
9
,
3
1
7
-
3
2
8
,
2
0
1
0
.
[3
5
]
De
h
u
ri
,
S
.,
Ro
y
R
.,
Ch
o
S
.
B
.,
G
h
o
sh
A
.
:
A
n
I
m
p
ro
v
e
d
S
w
a
r
m
Op
ti
m
i
z
e
d
F
u
n
c
ti
o
n
a
l
L
in
k
A
rti
f
icia
l
Ne
u
ra
l
Ne
tw
o
rk
(
IS
O
-
F
LA
NN
)
F
o
r
Clas
sif
ica
ti
o
n
.
T
h
e
Jo
u
rn
a
l
o
f
S
y
ste
m
s
A
n
d
S
o
f
tw
a
re
8
5
,
1
3
3
3
-
1
3
4
5
,
2
0
1
2
.
[3
6
]
Eb
e
rh
a
rt
,
R
.
C
.
,
Ke
n
n
e
d
y
J
.
:
A
Ne
w
Op
ti
m
ize
r
U
sin
g
P
a
rti
c
l
e
S
wa
r
m
T
h
e
o
r
y
.
P
ro
c
e
e
d
in
g
s
Of
T
h
e
S
ix
th
In
tern
a
ti
o
n
a
l
S
y
m
p
o
siu
m
On
M
ic
ro
M
a
c
h
i
n
e
A
n
d
Hu
m
a
n
S
c
ien
c
e
,
Na
g
o
y
a
,
Ja
p
a
n
,
P
p
.
3
9
–
4
3
,
1
9
9
5
.
[3
7
]
Clerc
M
.,
Ke
n
n
e
d
y
J
.
:
T
h
e
P
a
rti
c
le
S
wa
r
m
Ex
p
lo
sio
n
,
S
tab
il
it
y
a
n
d
Co
n
v
e
rg
e
n
c
e
In
A
M
u
lt
id
im
e
n
sio
n
a
l
Co
m
p
lex
S
p
a
c
e
.
IEE
E
T
ra
n
s E
v
o
l
C
o
m
p
u
t
2
0
0
2
,
6
(1
)
,
5
8
–
7
3
,
2
0
0
2
.
[3
8
]
Eb
e
rh
a
rt
,
R
.,
S
h
i
,
Y
.
:
Co
m
p
a
rin
g
In
e
rti
a
W
e
ig
h
ts
a
n
d
Co
n
stri
c
ti
o
n
F
a
c
to
rs
i
n
P
a
rti
c
le
S
w
a
rm
Op
ti
m
iza
ti
o
n
.
P
r
o
c
e
e
d
in
g
s
o
f
T
h
e
Co
n
g
re
ss
On
Ev
o
lu
ti
o
n
a
ry
Co
m
p
u
tatio
n
(
CEC
2
0
0
0
),
P
p
.
8
4
–
8
8
,
2
0
0
0
.
[3
9
]
M
a
jh
i
R
.,
P
a
n
d
a
G
.
A
n
d
S
a
h
o
o
G
.
:
De
v
e
lo
p
m
e
n
t
a
n
d
P
e
rf
o
rm
a
n
c
e
Ev
a
lu
a
ti
o
n
Of
F
LA
NN
Ba
s
e
d
M
o
d
e
l
F
o
r
F
o
re
c
a
stin
g
o
f
S
to
c
k
M
a
rk
e
ts,
Ex
p
e
rt
S
y
ste
m
s
W
it
h
A
p
p
li
c
a
ti
o
n
s
3
6
,
P
p
.
6
8
0
0
–
6
8
0
8
,
2
0
0
9
.
[4
0
]
Blak
e
,
C
.
L
.,
M
e
rz
,
C
.
J
.
:
UCI
Re
p
o
sit
o
ry
o
f
M
a
c
h
in
e
L
e
a
rn
in
g
Da
tab
a
se
s
,
Http
:/
/W
ww
.
Ic
s.Uc
i.
Ed
u
/~
M
lea
r
n
/M
lrep
o
sito
ry
.
Ht
m
l
[4
1
]
P
a
l
,
S
.
K
.,
M
a
j
u
m
d
a
r
,
D
.
D
.
:
F
u
z
z
y
S
e
ts
a
n
d
De
c
isio
n
M
a
k
in
g
A
p
p
ro
a
c
h
e
s
in
Vo
w
e
l
a
n
d
S
p
e
a
k
e
r
Re
c
o
g
n
it
io
n
.
IEE
E
T
ra
n
sa
c
ti
o
n
s
on
S
y
ste
m
s
,
M
a
n
,
A
n
d
Cy
b
e
rn
e
ti
c
s
7
,
6
2
5
–
6
2
9
,
1
9
7
7
.
[4
2
]
Ca
r
v
a
lh
o
,
M
.
A
n
d
L
u
d
e
r
m
ir
,
T
.
B
.
:
H
y
b
rid
T
ra
in
in
g
o
f
F
e
e
d
-
F
o
rw
a
rd
N
e
u
ra
l
Ne
t
w
o
rk
s
W
it
h
P
a
rti
c
le
S
w
a
r
m
Op
ti
m
iza
ti
o
n
,
I
.
Ki
n
g
Et
A
l.
(
Ed
s
.
):
ICON
IP
2
0
0
6
,
P
a
rt
II
,
L
NCS
4
2
3
3
,
P
p
.
1
0
6
1
–
1
0
7
0
,
2
0
0
6
.
[4
3
]
De
h
u
ri
,
S
.
,
M
ish
ra
,
B
.
B
.
,
Ch
o
S
.
B
.
:
G
e
n
e
ti
c
F
e
a
tu
re
S
e
lec
ti
o
n
F
o
r
Op
ti
m
a
l
F
u
n
c
t
io
n
a
l
L
in
k
A
rti
f
icia
l
Ne
u
ra
l
Ne
tw
o
rk
in
Clas
si
f
ica
ti
o
n
.
IDEA
L
2
0
0
8
,
L
NCS
5
3
2
6
,
P
p
.
1
5
6
-
1
6
3
,
2
0
0
8
.
[4
4
]
Hsu
,
C
.
W
.,
L
in
,
C
.
J
.
:
A
Co
m
p
a
riso
n
Of
M
e
th
o
d
s
F
o
r
M
u
lt
i
-
C
las
s
S
u
p
p
o
rt
V
e
c
to
r
M
a
c
h
i
n
e
s
.
IEE
E
T
ra
n
sa
c
ti
o
n
o
n
Ne
u
ra
l
Ne
tw
o
rk
s
.
3
(2
),
P
p
.
4
1
5
-
4
2
5
.
2
0
0
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.