I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
10
,
No
.
4
,
A
u
g
u
s
t
2
0
2
0
,
p
p
.
3
7
2
5
~
3
7
3
3
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/
ijece
.
v
1
0
i
4
.
p
p
3
7
25
-
3
7
3
3
3725
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m/in
d
ex
.
p
h
p
/I
JE
C
E
A hybrid
co
nst
ru
ctive al
g
o
rithm in
co
rpo
ra
ting
teaching
-
lea
rnin
g
bas
ed optimiza
ti
o
n f
o
r neural
network
training
M
a
hd
ie
h
K
ho
ra
s
ha
diza
de
1
,
M
o
rt
ez
a
J
o
uy
ba
n
2
,
M
o
ha
mm
a
drez
a
Asg
ha
ri
O
s
ko
ei
3
1
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
,
S
istan
a
n
d
Ba
lu
c
h
e
sta
n
Un
i
v
e
rsity
,
Ira
n
2,
3
De
p
a
rtme
n
t
o
f
C
o
m
p
u
ter S
c
ien
c
e
,
Allam
e
h
Tab
a
tab
a
’i
Un
iv
e
rsit
y
,
Ira
n
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
y
2
9
,
2
0
1
9
R
ev
is
ed
J
an
1
6
,
2
0
2
0
Acc
ep
ted
Feb
1
,
2
0
2
0
In
n
e
u
ra
l
n
e
two
rk
s,
sim
u
lt
a
n
e
o
u
s
d
e
term
in
a
ti
o
n
o
f
t
h
e
o
p
t
imu
m
stru
c
tu
re
a
n
d
we
ig
h
ts
is
a
c
h
a
ll
e
n
g
e
.
T
h
is
p
a
p
e
r
p
r
o
p
o
se
s
a
c
o
m
b
in
a
t
io
n
o
f
tea
c
h
in
g
-
lea
rn
in
g
b
a
se
d
o
p
ti
m
iza
ti
o
n
(TL
BO)
a
lg
o
rit
h
m
a
n
d
a
c
o
n
stru
c
ti
v
e
a
lg
o
rit
h
m
(CA) t
o
c
o
p
e
with
t
h
e
c
h
a
ll
e
n
g
e
.
In
li
tera
tu
re
,
TL
BO i
s u
se
d
t
o
c
h
o
o
se
p
ro
p
e
r
we
ig
h
ts,
wh
i
le
CA
is
a
d
o
p
ted
t
o
c
o
n
str
u
c
t
d
iffere
n
t
str
u
c
tu
re
s
i
n
o
rd
e
r
t
o
se
lec
t
th
e
p
r
o
p
e
r
o
n
e
.
I
n
th
is
stu
d
y
,
th
e
b
a
sic
TL
BO
a
lg
o
rit
h
m
a
lo
n
g
with
a
n
imp
ro
v
e
d
v
e
rsio
n
o
f
th
is
a
lg
o
rit
h
m
fo
r
n
e
two
r
k
we
ig
h
ts
se
lec
ti
o
n
a
re
u
ti
l
ize
d
.
M
e
a
n
wh
il
e
,
a
s
a
c
o
n
stru
c
ti
v
e
a
l
g
o
rit
h
m
,
a
n
o
v
e
l
m
o
d
ifi
c
a
ti
o
n
t
o
m
u
lt
ip
le
o
p
e
ra
ti
o
n
s,
u
sin
g
sta
ti
stica
l
tes
ts
(M
OST
)
,
is
a
p
p
li
e
d
a
n
d
tes
ted
to
c
h
o
o
se
th
e
p
ro
p
e
r
str
u
c
tu
re
.
Th
e
p
r
o
p
o
se
d
c
o
m
b
i
n
a
to
rial
a
lg
o
rit
h
m
s
a
re
a
p
p
li
e
d
t
o
ten
c
las
sifica
t
io
n
p
ro
b
lem
s
a
n
d
tw
o
-
t
ime
-
se
ries
p
re
d
ictio
n
p
ro
b
lem
s,
a
s
th
e
b
e
n
c
h
m
a
rk
.
T
h
e
re
su
lt
s
a
re
e
v
a
lu
a
ted
b
a
se
d
o
n
train
i
n
g
a
n
d
tes
ti
n
g
e
rro
r,
n
e
two
r
k
c
o
m
p
lex
it
y
a
n
d
m
e
a
n
-
sq
u
a
re
e
rro
r.
Th
e
e
x
p
e
rim
e
n
tal
re
su
lt
s
il
lu
stra
te
t
h
a
t
t
h
e
p
r
o
p
o
se
d
h
y
b
ri
d
m
e
th
o
d
o
f
th
e
m
o
d
i
fie
d
M
OS
T
c
o
n
stru
c
ti
v
e
a
lg
o
rit
h
m
a
n
d
t
h
e
imp
ro
v
e
d
T
LBO
(M
CO
-
ITL
BO)
a
lg
o
rit
h
m
o
u
t
p
e
rfo
rm
t
h
e
o
t
h
e
rs;
m
o
re
o
v
e
r,
th
e
y
h
a
v
e
b
e
e
n
p
ro
v
e
n
b
y
Wi
lco
x
o
n
sta
ti
stica
l
tes
ts
a
s
we
ll
.
Th
e
p
ro
p
o
se
d
m
e
th
o
d
d
e
m
o
n
stra
tes
les
s
a
v
e
ra
g
e
e
rro
r
with
les
s c
o
m
p
lex
it
y
in
th
e
n
e
tw
o
rk
stru
c
t
u
re
.
K
ey
w
o
r
d
s
:
Alg
o
r
ith
m
in
co
r
p
o
r
atin
g
T
ea
ch
in
g
-
lear
n
in
g
Op
tim
izatio
n
Neu
r
al
n
etwo
r
k
t
r
ain
in
g
Co
p
y
rig
h
t
©
2
0
2
0
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ma
h
d
ie
h
Kh
o
r
ash
ad
izad
e
,
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
Sis
tan
an
d
B
alu
ch
estan
Un
iv
er
s
ity
,
Z
ah
ed
an
,
I
r
an
.
E
m
ail:
k
h
o
r
ash
ad
iza
d
e.
m
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Ar
tific
ial
n
eu
r
al
n
etwo
r
k
s
(
ANN)
,
h
av
in
g
a
s
tr
o
n
g
s
im
ilit
u
d
e
to
b
io
l
o
g
ical
n
etwo
r
k
s
,
h
av
e
th
e
ab
ilit
y
to
lear
n
n
o
is
e
d
ata
as
well
as
th
e
ab
ilit
y
to
class
if
y
an
d
r
ec
o
g
n
ize
d
if
f
er
en
t
t
y
p
es
o
f
in
p
u
t
p
atter
n
s
.
T
h
ese
tak
e
p
lace
o
n
ly
if
th
e
n
eu
r
al
n
etwo
r
k
is
well
tr
ain
ed
with
o
u
t
an
o
v
er
-
f
itti
n
g
o
r
u
n
d
er
-
f
itti
n
g
m
o
d
el.
T
h
e
m
o
s
t
well
-
k
n
o
wn
t
r
ain
in
g
alg
o
r
ith
m
is
th
e
b
ac
k
p
r
o
p
ag
at
io
n
[
1
]
,
b
u
t
it
h
as
n
u
m
er
o
u
s
d
r
aw
b
ac
k
s
s
u
ch
as
tr
ap
p
in
g
i
n
lo
ca
l
m
in
im
a
[
2
]
.
He
n
ce
,
r
ese
ar
ch
er
s
h
av
e
d
ec
id
e
d
to
u
tili
ze
ev
o
lu
tio
n
ar
y
al
g
o
r
ith
m
s
i
n
s
t
ea
d
.
I
n
ad
d
itio
n
to
th
e
tr
ain
in
g
an
d
d
eter
m
in
atio
n
o
f
o
p
tim
al
weig
h
ts
,
an
o
th
er
c
r
itical
is
s
u
e
is
th
e
d
esig
n
o
f
a
n
ap
p
r
o
p
r
iate
ANN
ar
ch
itectu
r
e.
Ma
n
y
s
tu
d
ies
h
a
v
e
b
ee
n
c
o
n
d
u
cted
f
o
r
ar
ch
ite
ctu
r
e
as
well
as
weig
h
t
o
p
tim
i
za
tio
n
.
Fo
r
in
s
tan
ce
,
in
ap
p
ly
in
g
a
n
o
v
el
m
eth
o
d
b
ased
o
n
Gau
s
s
ian
-
PS
O
an
d
f
u
zz
y
r
ea
s
o
n
in
g
[
3
]
ANN
weig
h
t
an
d
s
tr
u
ctu
r
e
o
p
tim
izatio
n
is
p
r
esen
ted
.
I
n
liter
atu
r
e,
th
er
e
ar
e
o
th
er
m
eth
o
d
s
to
o
p
tim
ize
ANN
ar
ch
itectu
r
e,
n
am
ely
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
s
an
d
p
r
u
n
in
g
al
g
o
r
ith
m
s
.
C
o
n
s
tr
u
cti
v
e
alg
o
r
ith
m
s
h
a
v
e
m
a
n
y
a
d
v
an
tag
es
o
v
er
p
r
u
n
i
n
g
alg
o
r
ith
m
s
,
s
u
ch
as
ea
s
y
i
n
itiatio
n
,
less
co
m
p
lex
ity
o
f
th
e
f
in
al
s
o
lu
tio
n
,
a
n
d
lig
h
ter
lo
a
d
o
f
c
o
m
p
u
tatio
n
.
Fu
r
th
er
m
o
r
e
,
C
A’
s
ar
e
ab
le
to
f
r
ee
ze
th
e
ex
is
tin
g
weig
h
ts
in
th
e
n
eu
r
al
n
etwo
r
k
if
t
h
ey
ar
e
u
s
ef
u
l
in
o
u
tp
u
t;
as
a
r
esu
lt,
r
esu
ltin
g
in
th
e
r
ed
u
ctio
n
o
f
th
e
r
e
q
u
ir
ed
tim
e
an
d
m
em
o
r
y
.
I
n
p
r
u
n
i
n
g
alg
o
r
ith
m
s
,
s
ev
e
r
al
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
4
,
Au
g
u
s
t 2
0
2
0
:
3
7
2
5
-
3
7
3
3
3726
p
r
o
b
lem
-
d
ep
e
n
d
en
t
p
ar
a
m
eter
s
ar
e
to
b
e
p
r
o
p
er
ly
id
e
n
tifie
d
in
o
r
d
er
to
o
b
tain
an
ac
ce
p
tab
le
n
etw
o
r
k
with
a
s
atis
f
ac
to
r
y
p
er
f
o
r
m
an
ce
.
T
h
is
m
ak
es it d
if
f
icu
lt to
b
e
u
s
ed
in
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
[
4
].
T
h
is
p
ap
er
p
o
r
tr
a
y
s
a
co
m
b
in
atio
n
o
f
r
an
d
o
m
s
ea
r
ch
p
r
o
ce
d
u
r
es
an
d
s
y
s
tem
atic
m
eth
o
d
s
,
p
r
o
p
o
s
in
g
h
y
b
r
id
izin
g
im
p
r
o
v
e
d
teac
h
in
g
-
lear
n
in
g
alg
o
r
ith
m
s
with
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
s
f
o
r
th
e
p
u
r
p
o
s
e
o
f
ANN
d
esig
n
.
T
h
e
h
y
b
r
id
is
a
d
v
an
ta
g
eo
u
s
,
f
o
r
teac
h
i
n
g
-
lear
n
in
g
a
lg
o
r
ith
m
is
a
p
ar
am
eter
-
in
d
ep
en
d
en
t
o
p
tim
izatio
n
alg
o
r
ith
m
t
h
at
b
alan
ce
b
etwe
e
n
ex
p
l
o
r
atio
n
a
n
d
ex
p
l
o
itatio
n
.
Me
an
wh
ile,
c
o
n
s
tr
u
ctiv
e
alg
o
r
ith
m
s
ar
e
ad
o
p
ted
to
s
elec
t
an
ap
p
r
o
p
r
iate
ANN
ar
ch
itectu
r
e.
Sin
ce
u
s
in
g
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
s
is
co
s
t
-
ef
f
ec
tiv
e
in
ter
m
s
o
f
th
e
tr
ain
in
g
-
tim
e
an
d
co
m
p
lex
ity
o
f
ANN,
it
h
in
d
e
r
s
th
e
p
r
o
d
u
ctio
n
o
f
n
etwo
r
k
s
wit
h
an
i
n
ef
f
icien
t
v
e
r
y
la
r
g
e
ar
ch
itectu
r
e.
T
h
is
p
ap
e
r
,
with
th
e
aim
o
f
s
im
u
ltan
eo
u
s
ly
o
p
tim
izin
g
th
e
ANN
weig
h
ts
an
d
ar
ch
itectu
r
e,
co
m
b
in
es
tr
ain
i
n
g
a
n
d
c
o
n
s
tr
u
ctiv
e
alg
o
r
ith
m
s
ap
p
lie
d
to
ten
class
if
icatio
n
p
r
o
b
lem
s
a
n
d
two
-
tim
e
s
er
ies
p
r
ed
ictio
n
p
r
o
b
lem
s
,
as
th
e
b
en
ch
m
ar
k
.
Af
ter
ev
alu
atin
g
th
e
p
er
f
o
r
m
a
n
ce
o
f
p
r
o
p
o
s
ed
h
y
b
r
id
alg
o
r
ith
m
s
an
d
co
m
p
ar
in
g
th
eir
r
esu
lts
,
it
was
f
o
u
n
d
th
at
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
u
tp
e
r
f
o
r
m
ed
o
th
er
alg
o
r
it
h
m
s
.
T
h
e
p
r
o
p
o
s
ed
co
m
b
in
atio
n
m
eth
o
d
p
r
o
v
es
t
o
h
a
v
e
a
lo
wer
m
ea
n
e
r
r
o
r
in
m
o
s
t
ca
s
es.
T
h
e
r
est
o
f
t
h
is
s
tu
d
y
is
o
r
g
a
n
ized
as
f
o
llo
ws:
Sectio
n
2
p
r
o
v
id
es
a
b
r
ief
d
escr
ip
tio
n
o
f
th
e
al
g
o
r
ith
m
s
th
at
we
p
r
o
v
id
ed
.
T
h
en
,
in
Sect
io
n
3
,
a
h
y
b
r
id
p
r
o
p
o
s
ed
m
eth
o
d
to
ANN
o
p
tim
izatio
n
is
p
r
esen
t
ed
.
I
n
Sectio
n
4
,
th
e
e
x
p
er
im
en
tal
r
esu
lts
o
f
th
e
ap
p
licatio
n
o
f
th
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
es
to
th
e
ANN
p
r
o
b
l
em
s
ar
e
r
ep
o
r
ted
,
a
n
d
f
in
ally
,
th
e
co
n
clu
s
io
n
is
d
r
awn
in
th
e
last
s
ec
tio
n
.
2.
AL
G
O
RI
T
H
M
D
E
SCR
I
P
T
I
O
N
2
.
1
.
I
m
pro
v
ed
t
ea
ching
-
lea
r
nin
g
ba
s
ed
o
ptim
iza
t
io
n (
I
T
L
B
O
)
Alth
o
u
g
h
T
L
B
O
p
r
o
v
id
es
h
ig
h
-
q
u
ality
s
o
lu
t
io
n
s
in
th
e
least
am
o
u
n
t
o
f
tim
e
an
d
h
as
a
g
r
e
at
s
tab
ilit
y
in
co
n
v
e
r
g
en
ce
[
5
]
,
in
th
e
le
ar
n
er
p
h
ase
o
f
th
is
alg
o
r
ith
m
,
lear
n
er
s
r
an
d
o
m
ly
c
h
o
o
s
e
an
o
th
er
lear
n
e
r
f
r
o
m
th
e
p
o
p
u
latio
n
.
T
h
is
d
if
f
icu
l
ty
l
ea
d
s
to
a
lack
o
f
b
ala
n
ce
b
etwe
en
t
h
e
two
co
n
ce
p
t
s
o
f
d
iv
er
s
ity
an
d
co
n
v
er
g
en
ce
.
I
T
L
B
O
with
an
im
p
r
o
v
em
en
t
in
t
o
b
asic
T
L
B
O
o
v
er
co
m
es
th
is
d
if
f
ic
u
lty
.
I
n
t
h
is
alg
o
r
ith
m
,
th
e
teac
h
er
p
h
ase
is
th
e
s
am
e
as
th
e
teac
h
er
p
h
ase
in
t
h
e
b
asic
T
L
B
O
alg
o
r
ith
m
a
n
d
t
h
e
lear
n
er
p
h
ase
is
ex
p
r
ess
ed
as
f
o
llo
ws.
T
h
e
I
T
L
B
O
h
as
b
ee
n
d
ev
elo
p
e
d
t
o
im
p
r
o
v
e
t
h
e
wea
k
n
ess
es
o
f
T
L
B
O
alg
o
r
ith
m;
f
o
r
e
x
am
p
le,
i
n
T
L
B
O
r
an
d
o
m
ch
o
ices
d
u
e
to
lo
w
lo
ca
l
s
ea
r
ch
ca
p
ab
ilit
y
,
b
u
t
in
I
T
L
B
O
with
ad
d
itio
n
co
n
ce
p
t o
f
n
eig
h
b
o
r
h
o
o
d
we
t
r
y
in
g
to
r
ed
u
ce
r
an
d
o
m
ch
o
ice
s
an
d
u
tili
ze
o
f
n
eig
h
b
o
r
h
o
o
d
ab
ilit
ies.
T
h
is
is
s
u
e
in
cr
ea
s
es lo
ca
l sear
ch
an
d
g
l
o
b
al
s
ea
r
ch
ca
p
ab
ilit
y
.
T
h
e
m
ai
n
s
ec
tio
n
s
o
f
I
T
L
B
O
ar
e
as f
o
l
lo
ws:
2
.
1
.
1
.
I
T
L
B
O
lea
rner
ph
a
s
e
I
n
th
is
p
h
ase,
ea
ch
lear
n
er
is
e
n
co
d
ed
with
a
n
in
teg
er
a
n
d
p
lace
d
in
a
r
ec
tan
g
u
lar
ar
r
ay
.
le
ar
n
er
s
m
ay
lear
n
f
r
o
m
th
eir
n
eig
h
b
o
r
s
o
r
f
r
o
m
th
e
b
est
in
d
iv
i
d
u
al
in
w
h
o
le
class
.
T
h
is
p
r
o
ce
s
s
is
b
a
s
ed
o
n
lo
ca
l
s
ea
r
c
h
ab
ilit
y
;
f
u
r
th
er
m
o
r
e,
b
alan
ce
b
etwe
en
g
lo
b
al
s
ea
r
ch
an
d
lo
ca
l
s
ea
r
ch
ab
ilit
y
is
ap
p
lied
.
I
n
lo
ca
l
s
ea
r
c
h
,
ea
ch
lear
n
er
u
p
d
ates
h
is
p
o
s
itio
n
with
Pc
p
r
o
b
ab
ilit
y
b
y
th
e
b
est
lear
n
er
in
h
is
n
eig
h
b
o
r
h
o
o
d
(
o
r
,
ℎ
)
an
d
also
g
lo
b
al
b
est lea
r
n
er
th
at
in
p
o
p
u
latio
n
.
,
=
,
+
2
.
(
,
ℎ
−
,
)
+
3
.
(
ℎ
−
,
)
(
1
)
W
h
er
e
,
ℎ
is
th
e
teac
h
er
in
n
ei
g
h
b
o
r
h
o
o
d
,
ℎ
is
teac
h
er
o
f
wh
o
le
class
,
2
,
3
ar
e
r
an
d
o
m
n
u
m
b
er
s
in
th
e
r
a
n
g
e
of
(
0
,
1
)
.
T
h
e
n
ew
p
o
s
itio
n
o
f
ea
ch
lear
n
er
will
b
e
ac
ce
p
ted
if
its
f
itn
ess
v
alu
e
h
as
im
p
r
o
v
e
d
.
I
n
th
e
c
o
n
ce
p
t
o
f
g
lo
b
al
s
ea
r
ch
,
if
Pc
p
r
o
b
ab
ilit
y
d
o
n
’
t
m
ee
t,
ea
c
h
lear
n
er
ch
o
o
s
es
a
r
an
d
o
m
lear
n
er
(
)
f
r
o
m
th
e
wh
o
le
class
to
p
r
o
v
id
e
th
e
lear
n
in
g
g
o
al
,
if
is
b
etter
th
an
,
o
r
o
t
h
er
wis
e,
lear
n
in
g
o
cc
u
r
s
ac
co
r
d
in
g
to
lear
n
er
p
h
ase
i
n
b
asic
T
L
B
O.
T
h
er
ef
o
r
e,
u
s
in
g
th
ese
o
p
er
atio
n
s
b
o
t
h
lo
ca
l
an
d
g
lo
b
al
s
ea
r
ch
ca
p
ab
ilit
y
will
b
e
o
b
tain
ed
.
All
th
e
ac
ce
p
ted
lear
n
er
s
at
th
e
en
d
o
f
lear
n
er
p
h
ase
ar
e
p
r
eser
v
ed
.
Du
e
to
th
e
en
h
an
ce
d
ex
p
lo
itatio
n
ab
i
lity
alo
n
g
with
th
e
ex
p
lo
r
atio
n
ab
ilit
y
,
wh
ic
h
alr
ea
d
y
ex
is
ted
in
th
e
lear
n
in
g
p
h
ase
o
f
th
e
o
r
ig
i
n
al
alg
o
r
ith
m
,
we
u
s
e
th
e
c
o
n
ce
p
t
o
f
n
eig
h
b
o
r
h
o
o
d
in
th
e
class
r
o
o
m
.
F
o
r
ea
ch
in
d
iv
id
u
al
in
th
e
p
o
p
u
latio
n
ex
is
t
a
n
u
m
b
er
o
f
n
eig
h
b
o
r
h
o
o
d
m
em
b
e
r
th
at
lear
n
f
r
o
m
th
e
b
est
o
n
e.
Fo
r
m
ain
tain
o
f
d
iv
er
s
ity
af
te
r
a
n
u
m
b
er
o
f
ite
r
atio
n
s
th
e
n
eig
h
b
o
r
h
o
o
d
m
e
m
b
er
s
o
f
ea
c
h
in
d
iv
id
u
al
a
r
e
ch
an
g
ed
.
T
h
is
is
s
u
e
b
alan
ce
b
etwe
en
th
e
ex
p
l
o
r
at
io
n
an
d
ex
p
lo
itin
g
ab
ilit
ies.
Oth
er
ad
v
an
ta
g
e
th
er
e
is
in
t
h
is
alg
o
r
ith
m
,
wh
e
n
a
n
ew
p
o
s
itio
n
is
o
b
tain
ed
f
o
r
ea
ch
m
em
b
er
,
it
m
ay
lead
to
th
e
p
r
o
d
u
ctio
n
o
f
d
ec
is
io
n
v
ar
iab
les
v
alu
es
th
at
ar
e
o
u
t
o
f
t
h
e
r
an
g
e
o
f
th
e
d
ef
in
itio
n
in
ter
v
al.
I
n
th
is
ca
s
e,
m
o
s
t
r
esear
ch
er
s
u
s
e
th
e
co
n
v
er
g
en
ce
ap
p
r
o
ac
h
to
th
e
u
p
p
e
r
an
d
lo
wer
b
o
u
n
d
ac
co
r
d
in
g
to
alg
o
r
ith
m
,
b
u
t
th
is
m
eth
o
d
is
Old
an
d
d
is
ab
led
m
eth
o
d
witch
ca
u
s
e
alg
o
r
ith
m
to
lo
ca
l
o
p
tim
a.
I
n
th
e
im
p
r
o
v
ed
teac
h
in
g
-
lear
n
i
n
g
b
ased
o
p
tim
izatio
n
m
eth
o
d
,
we
u
s
e
m
o
d
if
ied
tech
n
iq
u
e
to
ch
ec
k
b
o
u
n
d
ar
ie
s
o
f
th
e
v
ar
iab
les
[
6
]
.
I
ts
ad
v
an
tag
e
is
av
o
id
in
g
eq
u
aliza
tio
n
o
f
th
e
d
ec
is
io
n
v
ar
iab
les.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8
7
0
8
A
h
yb
r
id
co
n
s
tr
u
ctive
a
lg
o
r
ith
m
in
co
r
p
o
r
a
tin
g
tea
c
h
in
g
-
le
a
r
n
in
g
b
a
s
ed
…
(
Ma
h
d
ie
h
K
h
o
r
a
s
h
a
d
iz
a
d
e
)
3727
2
.
2
.
M
o
dified
M
O
ST
a
lg
o
rit
hm
(
M
M
O
ST
)
Dete
r
m
in
in
g
th
e
ar
ch
itectu
r
e
o
f
ar
tific
ial
n
etwo
r
k
s
h
as
lu
r
e
d
m
an
y
r
esear
ch
e
r
s
in
th
e
f
iel
d
in
r
ec
en
t
y
ea
r
s
.
W
e
u
s
ed
Mu
ltip
le
Op
e
r
ato
r
s
u
s
in
g
Statis
tica
l
T
est
s
alg
o
r
ith
m
MO
ST
[
7
]
.
I
n
MO
ST
alg
o
r
ith
m
,
th
er
e
is
n
’
t
an
y
co
n
tr
o
lled
m
eth
o
d
f
o
r
ch
an
g
e
s
tr
u
tu
r
e.
T
h
is
alg
o
r
ith
m
m
ay
h
a
v
e
lar
g
e
c
h
an
g
es
i
n
n
etwo
r
k
s
tr
u
ctu
r
e
d
u
r
in
g
th
e
alg
o
r
ith
m
.
An
o
th
e
r
wea
k
n
ess
o
f
t
h
is
alg
o
r
ith
m
is
th
e
ad
d
itio
n
o
f
la
y
er
s
f
r
e
q
u
en
tly
with
o
u
t
an
y
co
n
d
itio
n
to
co
n
t
r
o
l.
I
n
m
o
d
if
ied
MO
ST
alg
o
r
ith
m
,
th
e
o
p
er
ato
r
p
o
o
l
was
r
em
o
v
ed
.
Fo
r
ch
an
g
in
g
s
tr
u
ctu
r
e
n
eu
r
o
n
s
ar
e
ad
d
e
d
o
n
e
af
ter
a
n
o
th
er
.
Selectin
g
t
h
e
n
ew
s
tr
u
ctu
r
es
is
d
o
n
e
m
o
r
e
ca
r
ef
u
lly
b
y
a
d
d
in
g
m
u
ltip
le
co
n
d
itio
n
s
.
At
th
e
b
eg
in
n
i
n
g
,
alg
o
r
ith
m
s
tar
ts
with
a
s
in
g
le
h
id
d
en
lay
er
n
etwo
r
k
b
y
th
e
m
in
im
u
m
n
u
m
b
er
o
f
n
eu
r
o
n
s
.
W
e
ch
o
s
e
o
n
e
o
f
p
o
p
u
lar
ap
p
r
o
ac
h
f
o
r
allo
wed
m
in
im
u
m
n
u
m
b
er
th
at
is
th
e
av
er
ag
e
o
f
n
u
m
b
er
o
f
o
u
tp
u
t
lay
er
an
d
in
p
u
t
lay
e
r
.
Netwo
r
k
in
th
e
f
ir
s
t
s
tep
h
as
a
s
in
g
le
h
id
d
en
lay
er
an
d
n
eu
r
o
n
s
ar
e
ad
d
ed
co
n
tin
u
ally
to
th
e
h
id
d
en
lay
er
to
o
b
tain
a
p
r
o
p
e
r
s
tr
u
ctu
r
e
o
f
th
e
n
etwo
r
k
.
T
o
av
o
id
c
r
ea
tin
g
v
er
y
lar
g
e
s
tr
u
ctu
r
es
f
o
r
n
etwo
r
k
s
,
th
e
n
eu
r
o
n
s
ar
e
ad
d
ed
to
s
in
g
le
h
i
d
d
en
lay
er
o
f
th
e
n
etwo
r
k
u
n
til
th
ey
d
o
n
’
t
ex
ce
ed
Max
-
h
id
d
e
n
n
u
m
b
er
.
I
n
f
ac
t,
n
etwo
r
k
s
with
v
er
y
lar
g
e
s
tr
u
ctu
r
e
n
o
t
o
n
ly
d
o
n
’
t
h
av
e
g
o
o
d
g
en
er
aliza
b
ilit
y
,
b
u
t
th
e
y
also
in
cr
ea
s
e
th
e
c
o
m
p
u
tatio
n
al
tim
e
o
f
t
h
e
al
g
o
r
ith
m
.
T
o
elim
in
ate
t
h
is
wea
k
n
ess
,
we
ad
d
th
e
s
ec
o
n
d
lay
er
to
n
etwo
r
k
s
tr
u
ctu
r
e
to
cr
ea
te
p
r
o
p
e
r
ar
ch
it
ec
tu
r
e
with
a
p
r
o
b
a
b
ilit
y
less
t
h
an
P.
af
ter
ad
d
in
g
th
e
s
ec
o
n
d
la
y
er
,
th
e
n
u
m
b
er
o
f
n
eu
r
o
n
s
in
ea
ch
h
id
d
en
la
y
er
is
s
et
b
y
m
i
n
-
h
id
d
en
.
M
MO
ST
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
ch
o
o
s
es
th
e
b
est
ar
ch
itectu
r
e
b
etwe
en
co
n
s
tr
u
cted
s
tr
u
ctu
r
es.
So
,
as
n
o
ted
ab
o
v
e,
th
e
d
if
f
e
r
en
ce
s
b
etwe
en
th
e
MM
OST
an
d
MO
ST
alg
o
r
ith
m
ar
e
as
f
o
l
lo
ws:
o
p
er
ato
r
’
s
p
o
o
l
is
d
elete
d
;
n
eu
r
o
n
s
ar
e
co
n
tin
u
ally
a
d
d
ed
;
a
n
d
th
er
e
i
s
a
m
o
r
e
p
r
ec
is
e
ch
o
ice
b
etw
ee
n
th
e
th
r
e
e
p
r
ev
i
o
u
s
,
cu
r
r
en
t
an
d
th
e
ca
n
d
id
ate
ar
ch
itectu
r
es.
3.
T
H
E
P
RO
P
O
SE
D
M
E
T
H
O
D
I
n
th
is
p
ap
er
,
we
p
r
o
p
o
s
ed
a
co
m
b
in
atio
n
alg
o
r
ith
m
f
o
r
p
r
o
d
u
cin
g
a
n
e
u
r
al
n
etwo
r
k
w
ith
p
r
o
p
er
s
tr
u
ctu
r
e
an
d
weig
h
ts
,
to
s
im
u
ltan
eo
u
s
o
p
tim
izatio
n
o
f
weig
h
ts
an
d
s
tr
u
ctu
r
e.
Fo
r
th
is
p
u
r
p
o
s
e,
a
co
m
b
in
atio
n
o
f
th
e
m
o
d
if
ied
MO
ST
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
with
an
i
m
p
r
o
v
e
d
v
er
s
io
n
o
f
th
e
tr
ain
in
g
alg
o
r
ith
m
was
p
r
o
p
o
s
ed
.
T
h
e
r
o
le
o
f
th
e
co
n
s
tr
u
ctiv
e
alg
o
r
it
h
m
in
t
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
is
to
co
n
s
tr
u
ct
d
if
f
er
en
t
s
tr
u
ct
u
r
es
in
o
r
d
er
to
s
elec
t
th
e
p
r
o
p
er
o
n
e,
wh
ich
is
ca
r
r
ied
o
u
t
b
y
u
s
in
g
a
s
witch
in
g
s
y
s
tem
atic
ap
p
r
o
ac
h
b
etwe
en
th
e
v
ar
io
u
s
s
tr
u
ctu
r
es
allo
wed
f
o
r
th
e
n
eu
r
al
n
etwo
r
k
.
On
t
h
e
o
t
h
er
h
an
d
,
th
e
r
o
le
o
f
tr
ain
i
n
g
alg
o
r
ith
m
is
to
f
in
d
o
p
tim
al
weig
h
ts
f
o
r
th
e
s
tr
u
ct
u
r
e
th
at
is
cr
ea
ted
b
y
th
e
c
o
n
s
tr
u
ctiv
e
alg
o
r
ith
m
.
Usi
n
g
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
s
in
cr
ea
tin
g
a
n
etwo
r
k
ar
ch
itec
tu
r
e
r
ed
u
ce
s
co
m
p
u
tatio
n
al
co
s
t
an
d
co
m
p
lex
ity
.
B
u
t
u
s
in
g
t
h
ese
alg
o
r
ith
m
s
in
s
o
lv
in
g
n
o
is
y
p
r
o
b
lem
s
[
8
]
h
as
f
ailed
,
wh
ich
in
co
m
b
in
atio
n
with
o
th
er
tech
n
iq
u
es,
s
u
ch
as
th
e
u
s
e
o
f
ev
o
lu
tio
n
ar
y
al
g
o
r
ith
m
s
,
ca
n
b
e
ef
f
ec
tiv
e
in
im
p
r
o
v
in
g
th
e
c
o
n
s
tr
u
ctiv
e
al
g
o
r
ith
m
.
I
n
a
d
d
it
io
n
,
we
h
av
e
m
a
d
e
s
o
m
e
m
o
d
if
icatio
n
s
o
n
t
h
e
MO
ST
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
.
F
o
r
a
m
o
r
e
d
etailed
d
e
s
cr
ip
tio
n
,
th
e
p
s
eu
d
o
co
d
e
o
f
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
alg
o
r
ith
m
s
is
s
h
o
wn
in
Fig
u
r
e
1
.
I
n
o
th
er
wo
r
d
s
,
in
o
r
d
er
to
clar
if
y
th
e
co
m
b
in
atio
n
o
f
ev
o
lu
tio
n
ar
y
tr
ain
in
g
alg
o
r
it
h
m
s
an
d
c
o
n
s
tr
u
ctiv
e
alg
o
r
it
h
m
s
,
we
s
h
o
wed
th
e
p
r
o
ce
s
s
in
f
lo
wch
ar
t
by
F
ig
u
r
e
2
.
Fi
g
u
re
1
.
P
se
u
d
o
c
o
d
e
c
o
m
b
i
n
e
d
a
lg
o
rit
h
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
4
,
Au
g
u
s
t 2
0
2
0
:
3
7
2
5
-
3
7
3
3
3728
F
ig
u
re
2
.
F
lo
wc
h
a
rt
o
f
h
y
b
rid
a
lg
o
rit
h
m
4.
CO
M
P
ARI
SO
N
RE
SUL
T
S
I
n
th
is
s
ec
tio
n
,
we
e
v
alu
ate
t
h
e
ef
f
ec
tiv
e
n
ess
o
f
p
r
o
p
o
s
ed
h
y
b
r
id
m
eth
o
d
s
.
T
h
ese
alg
o
r
ith
m
s
ar
e
ap
p
lied
t
o
ten
class
if
icatio
n
p
r
o
b
lem
a
n
d
two
tim
e
s
er
ies
p
r
ed
ictio
n
p
r
o
b
lem
s
.
W
e
co
m
p
a
r
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
alg
o
r
it
h
m
s
f
ir
s
t w
ith
ea
ch
o
th
er
an
d
t
h
en
with
o
th
e
r
av
ailab
le
m
et
h
o
d
s
.
4
.
1
.
Def
ini
t
io
n o
f
cla
s
s
if
ica
t
io
n a
nd
t
im
e
s
er
ies predict
io
n pro
blem
s
T
h
e
task
o
f
ass
ig
n
in
g
a
s
am
p
le
to
a
p
r
o
p
er
g
r
o
u
p
,
b
ased
o
n
th
e
ch
ar
ac
ter
is
tics
o
f
d
escr
ib
in
g
th
at
o
b
ject
in
a
p
r
o
b
lem
,
is
d
ef
i
n
e
d
as
class
if
icatio
n
.
T
h
e
class
if
icatio
n
p
r
o
b
lem
s
u
s
ed
in
t
h
is
ar
ticle
in
clu
d
e
i
r
is
,
d
iab
etes
d
iag
n
o
s
is
,
th
y
r
o
id
,
b
r
ea
s
t
ca
n
ce
r
,
cr
ed
it
ca
r
d
,
g
l
ass
,
h
ea
r
t,
win
e,
p
ag
e
b
lo
ck
s
,
an
d
liv
e
r
.
T
h
ese
class
if
icatio
n
p
r
o
b
lem
s
ar
e
tak
en
f
r
o
m
th
e
UC
I
m
ac
h
in
e
lear
n
in
g
r
e
p
o
s
ito
r
y
[
9
]
.
B
u
t th
e
ti
m
e
s
er
ies p
r
ed
ictio
n
p
r
o
b
lem
s
u
s
e
a
s
p
ec
if
ic
m
o
d
e
l
to
p
r
ed
ict
f
u
tu
r
e
v
alu
es
b
ase
d
o
n
th
eir
p
r
e
v
io
u
s
v
alu
es.
T
h
e
f
ir
s
t
is
th
e
Gas
Fu
r
n
ac
e
Data
s
et
[
10
]
,
wh
ich
is
co
m
p
iled
f
r
o
m
J
en
k
in
s
'
s
B
o
o
k
o
f
T
im
e
Ser
ies
An
aly
s
is
.
I
t
co
n
tain
s
g
as
co
n
ten
t
an
d
C
O2
p
e
r
ce
n
tag
e
in
g
as,
a
n
d
an
o
th
e
r
is
a
Ma
ck
ey
g
lass
d
ataset
o
b
tain
ed
f
r
o
m
th
e
b
elo
w
d
if
f
er
en
tial e
q
u
atio
n
:
(
)
=
−
(
)
+
(
−
)
1
+
10
(
−
)
(
2
)
All
p
r
o
p
o
s
ed
h
y
b
r
id
alg
o
r
ith
m
s
in
th
is
ar
ticle
h
a
v
e
b
ee
n
im
p
lem
en
ted
u
s
in
g
MA
T
L
AB
s
o
f
twar
e
an
d
h
av
e
u
s
ed
3
0
tim
e
r
u
n
to
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
th
ese
m
eth
o
d
s
.
T
h
e
4
-
f
o
ld
-
cr
o
s
s
-
v
al
id
atio
n
m
eth
o
d
h
as
b
ee
n
u
s
ed
to
d
iv
id
e
th
e
o
r
ig
i
n
al
d
ataset
in
to
two
tr
ain
in
g
a
n
d
test
in
g
se
ts
.
T
h
is
m
eth
o
d
ca
n
ef
f
ec
tiv
el
y
p
r
ev
en
t
tr
ap
p
in
g
t
o
lo
ca
l
m
in
im
a
.
B
ec
au
s
e
b
o
th
th
e
tr
ai
n
in
g
an
d
te
s
tin
g
s
am
p
les
co
n
tr
ib
u
te
to
le
ar
n
in
g
as
m
u
c
h
as
p
o
s
s
ib
le,
it
ca
n
p
r
o
v
i
d
e
a
s
ati
s
f
ac
to
r
y
lear
n
i
n
g
e
f
f
ec
t
.
T
h
e
av
er
ag
e
e
r
r
o
r
is
o
b
tain
e
d
f
r
o
m
th
e
4
-
f
o
ld
-
cr
o
s
s
-
v
alid
atio
n
wh
ich
is
p
r
esen
ted
as
th
e
f
in
al
er
r
o
r
o
f
th
e
n
et
wo
r
k
.
I
n
ad
d
itio
n
,
th
e
in
p
u
t
d
ata
s
et
to
th
e
n
eu
r
al
n
etwo
r
k
is
n
o
r
m
alize
d
u
s
in
g
th
e
m
in
-
m
ax
n
o
r
m
aliza
tio
n
m
eth
o
d
to
th
e
in
te
r
v
al
[
-
1
.
1
]
.
T
h
e
r
esu
lts
o
f
th
e
co
m
p
ar
is
o
n
ar
e
p
r
esen
ted
in
two
p
ar
ts
.
First,
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
s
ar
e
co
m
p
ar
ed
with
ea
ch
o
th
er
,
an
d
th
en
th
e
b
est p
r
o
p
o
s
ed
m
eth
o
d
is
co
m
p
ar
ed
with
t
h
e
ex
is
tin
g
m
eth
o
d
s
.
4.
2
.
Co
m
pa
ring
pro
po
s
ed
met
ho
ds
wit
h e
a
ch
o
t
her
E
ac
h
o
f
th
ese
alg
o
r
ith
m
s
h
as
b
ee
n
ex
ec
u
te
d
3
0
tim
es,
an
d
t
h
e
r
esu
lts
o
f
th
e
ex
p
er
im
e
n
ts
h
av
e
b
ee
n
co
m
p
ar
ed
with
ea
ch
o
t
h
er
ac
co
r
d
in
g
to
th
r
ee
cr
iter
ia:
class
if
icatio
n
er
r
o
r
p
er
ce
n
tag
e
o
f
tr
ain
in
g
an
d
test
in
g
d
ata
an
d
co
m
p
lex
ity
p
er
ce
n
ta
g
e.
T
h
e
f
u
n
ctio
n
o
f
er
r
o
r
ca
lc
u
latio
n
Fo
r
th
e
Ma
ck
ey
g
lass
i
s
R
M
SE
an
d
f
o
r
g
as
f
u
r
n
ac
e
is
MSE
.
First,
we
co
m
p
ar
e
th
e
p
e
r
f
o
r
m
an
ce
o
f
tw
o
k
in
d
o
f
tr
ain
in
g
alg
o
r
ith
m
t
h
at
co
n
s
is
t
o
f
class
ic
tr
ain
in
g
alg
o
r
ith
m
(
b
ac
k
-
p
r
o
p
ag
atio
n
)
an
d
ev
o
lu
tio
n
ar
y
tr
ain
in
g
alg
o
r
ith
m
(
im
p
r
o
v
ed
t
ea
c
h
in
g
lear
n
in
g
-
b
ased
o
p
tim
izatio
n
)
.
T
h
e
r
esu
lts
f
r
o
m
T
ab
le
1
s
h
o
w
th
at
th
e
I
T
L
B
O
alg
o
r
ith
m
h
as
a
h
ig
h
er
ef
f
icien
c
y
f
o
r
m
o
s
t
d
ata
s
ets.
Acc
o
r
d
in
g
to
T
ab
le
1
,
th
e
I
T
L
B
O
alg
o
r
ith
m
f
o
r
all
o
f
class
if
icatio
n
p
r
o
b
lem
s
h
as
b
etter
p
er
f
o
r
m
an
ce
th
an
th
e
B
p
alg
o
r
ith
m
,
th
en
in
p
ar
t2
f
r
o
m
T
ab
l
e
1
we
s
h
o
wed
th
e
r
esu
lts
o
f
c
o
m
p
ar
in
g
p
r
o
p
o
s
ed
h
y
b
r
id
alg
o
r
ith
m
s
with
ea
ch
o
th
er
.
All
th
e
r
esu
lts
ar
e
b
ased
o
n
t
h
r
ee
ch
a
r
ac
ter
is
tics
(
p
ar
am
eter
s
)
o
f
tr
ain
i
n
g
an
d
test
in
g
er
r
o
r
f
o
r
class
if
icatio
n
,
MSE
e
r
r
o
r
a
n
d
c
o
m
p
le
x
ity
.
T
o
b
etter
d
em
o
n
s
tr
ate
t
h
e
s
u
p
er
io
r
alg
o
r
ith
m
,
we
d
id
r
an
k
av
er
ag
e
test
,
an
d
th
e
r
an
k
av
er
ag
e
f
o
r
d
if
f
er
e
n
t
d
ata
s
et
wa
s
p
r
esen
ted
in
T
ab
le
2
.
As
ca
n
b
e
s
ee
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8
7
0
8
A
h
yb
r
id
co
n
s
tr
u
ctive
a
lg
o
r
ith
m
in
co
r
p
o
r
a
tin
g
tea
c
h
in
g
-
le
a
r
n
in
g
b
a
s
ed
…
(
Ma
h
d
ie
h
K
h
o
r
a
s
h
a
d
iz
a
d
e
)
3729
f
r
o
m
T
ab
le
2
,
MCO
-
I
T
L
B
O
h
as
g
ain
ed
th
e
f
ir
s
t
r
an
k
f
o
r
all
o
f
ch
ar
ac
ter
is
tic.
T
o
ev
alu
ate
wh
eth
er
th
e
MCO
-
ITL
B
O
r
esu
lts
ar
e
s
ig
n
if
ican
tly
b
etter
th
a
n
o
th
e
r
ap
p
r
o
ac
h
es
,
we
ca
lcu
lated
th
e
p
-
v
al
u
e
test
with
a
s
ig
n
if
ican
t
lev
el
o
f
0
.
0
5
f
o
r
d
ata
s
ets.
T
h
e
ca
lcu
lated
P
-
v
alu
es
f
o
r
M
C
O
-
I
T
L
B
O
ar
e
s
h
o
wn
in
T
ab
le
3
in
co
m
p
ar
is
o
n
with
o
th
er
alg
o
r
ith
m
s
.
T
h
e
b
est
r
esu
lts
ar
e
b
o
ld
ed
in
th
e
tab
les.
I
n
Fig
u
r
es
3
th
e
b
o
x
p
lo
t
g
r
ap
h
s
s
h
o
wed
th
e
r
esu
lts
o
f
th
e
d
is
tr
ib
u
tio
n
o
f
tr
ain
in
g
an
d
test
in
g
e
r
r
o
r
s
f
o
r
th
e
wh
o
le
d
ata
s
et
f
o
r
3
0
tim
es
r
u
n
n
in
g
.
T
h
e
ch
ar
ts
s
h
o
w
th
at
MCO
-
I
T
L
B
O
is
s
u
p
er
io
r
in
m
o
s
t c
ases
.
Tab
le
1
.
Av
e
ra
g
e
re
su
lt
s
o
f
3
0
r
u
n
s o
f
tw
o
k
in
d
trai
n
in
g
a
lg
o
rit
h
m
(p
a
rt1
)
a
n
d
e
a
c
h
h
y
b
rid
a
l
g
o
ri
th
m
(p
a
rt2
)
D
a
t
a
s
e
t
C
r
i
t
e
r
i
a
BP
I
TLB
O
(
p
a
r
t
1
)
M
C
O
-
TLB
O
M
C
O
-
I
TLB
O
M
C
O
-
BP
(
p
a
r
t
2)
1
.
I
r
i
s
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
5
.
3
1
2
5
7
.
5
4
6
8
0
.
0
0
2
4
0
.
4
8
9
2
f
i
x
0
.
9
5
1
6
2
.
1
4
4
0
3
.
5
6
4
2
e
-
09
1
.
7
7
2
9
e
-
06
f
i
x
0
.
7
8
1
9
3
.
2
7
5
2
6
.
7
3
2
4
e
-
05
0
.
0
7
2
1
2
6
.
4
6
6
7
0
.
0
1
3
8
2
.
2
9
7
5
6
.
2
4
8
0
e
-
07
3
.
3
9
1
7
e
-
03
2
1
.
5
8
0
6
7
.
6
0
6
1
9
.
8
9
9
0
1
.
0
2
9
0
e
-
02
1
.
2
8
6
4
2
3
.
4
6
6
7
2.
D
i
a
b
e
t
e
s
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
3
1
.
2
3
2
2
3
5
.
2
8
8
7
0
.
6
0
1
3
0
.
6
9
3
3
f
i
x
1
8
.
6
8
9
7
2
7
.
8
4
7
3
1
.
1
6
3
7
e
-
06
7
.
1
2
4
7
e
-
05
f
i
x
1
9
.
5
2
8
2
2
9
.
3
7
9
5
1
.
5
2
9
e
-
08
0
.
0
0
1
7
9
1
0
9
.
1
3
3
3
1
8
.
8
7
7
5
2
3
.
3
9
0
6
2
.
3
4
6
8
e
-
10
4
.
1
8
2
2
e
-
8
5
8
.
4
1
9
4
2
8
.
0
1
7
6
3
6
.
8
3
5
9
6
.
3
3
7
5
e
-
04
1
.
3
3
4
9
1
2
9
.
2
3
3
3
3.
Th
y
r
o
i
d
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
1
7
.
8
3
4
0
1
6
.
5
6
7
2
0
.
1
3
9
9
1
.
5
9
6
8
f
i
x
6
.
7
7
9
8
1
2
.
5
9
0
3
1
.
3
9
8
7
e
-
05
2
.
5
7
8
6
e
-
04
f
i
x
8
.
2
2
1
5
1
1
.
1
3
1
8
4
.
8
9
7
6
e
-
06
0
.
0
0
4
0
2
5
2
1
.
1
5
.
6
0
5
9
6
.
5
8
0
0
5
.
6
5
8
9
e
-
09
2
.
6
1
5
8
e
-
07
1
3
3
.
6
4
5
2
1
8
.
3
9
6
1
2
2
.
3
8
3
7
3
.
8
6
0
5
e
-
03
1
.
4
0
9
5
1
6
8
.
2
2
3
5
4.
C
a
n
c
e
r
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
2
0
.
9
4
1
9
1
6
.
3
5
1
0
0
.
1
4
3
7
0
.
8
2
2
4
f
i
x
1
.
9
6
7
2
4
.
5
0
1
8
3
.
0
6
0
0
e
-
04
0
.
0
2
2
0
f
i
x
1
.
7
2
1
3
8
.
5
9
4
2
.
3
6
0
4
e
-
07
0
.
0
0
3
3
6
2
5
7
.
9
3
3
3
1
.
0
8
4
3
2
.
0
7
2
9
2
.
5
6
9
5
e
-
11
7
.
3
6
7
9
e
-
6
6
7
.
4
1
9
4
2
1
.
5
3
5
0
2
9
.
5
1
3
1
1
.
2
6
2
6
e
-
03
7
.
4
2
8
5
7
7
.
4
5
4
5
5.
C
a
r
d
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
2
2
.
5
2
4
7
2
3
.
8
1
2
6
0
.
0
2
5
8
1
.
1
4
3
7
f
i
x
1
5
.
7
4
9
2
1
7
.
3
1
9
6
3
.
3
2
7
2
e
-
09
3
.
0
9
3
9
e
-
05
f
i
x
1
2
.
6
1
1
1
2
4
.
3
4
7
8
2
.
2
4
0
2
e
-
04
0
.
0
1
8
4
4
5
2
.
6
6
6
7
1
2
.
2
3
7
9
1
3
.
5
8
8
9
7
.
2
0
6
6
e
-
08
1
.
9
3
8
1
e
-
03
1
2
7
.
4
1
9
4
1
9
.
9
4
3
2
3
0
.
4
4
8
0
3
.
4
9
8
5
e
-
05
3
.
0
6
9
9
1
8
3
.
5
4
5
5
6.
G
l
a
ss
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
3
6
.
7
7
7
0
5
4
.
8
3
8
7
0
.
4
2
3
5
0
.
8
6
3
1
f
i
x
1
8
.
3
1
6
9
3
6
.
9
1
4
6
1
.
0
2
4
4
e
-
06
0
.
1
4
3
1
f
i
x
1
8
.
2
2
5
2
3
1
.
7
1
7
2
9
.
2
0
8
1
e
-
04
0
.
0
9
6
4
3
9
2
.
6
6
6
7
1
7
.
0
8
8
1
2
2
.
0
7
3
3
2
.
3
8
4
5
e
-
10
3
.
2
6
9
e
-
06
1
2
3
.
3
5
4
8
6
6
.
8
3
0
5
6
7
.
9
0
6
3
1
.
4
6
0
4
e
-
04
3
.
7
1
4
1
1
0
4
.
7
2
7
3
7.
H
e
a
r
t
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
2
1
.
2
2
0
8
2
3
.
6
7
4
6
0
.
0
0
2
2
0
.
8
0
8
6
f
i
x
1
0
.
4
4
7
3
2
0
.
9
8
7
7
7
.
4
4
4
1
e
-
11
0
.
0
1
5
6
f
i
x
1
2
.
1
5
8
7
20
9
.
4
2
5
9
e
-
05
8
.
7
6
6
1
e
-
03
1
9
8
.
7
3
3
3
1
1
.
2
6
4
7
1
3
.
3
3
2
5
8
.
2
2
7
1
e
-
12
1
.
0
2
6
0
e
-
07
9
3
.
3
5
4
8
2
0
.
2
9
8
2
2
3
.
4
2
5
4
2
.
4
2
8
3
e
-
06
0
.
2
2
1
3
1
0
0
.
9
0
9
1
8.
W
i
n
e
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
1
7
.
0
6
4
1
2
3
.
1
0
6
3
0
.
0
2
1
5
1
.
0
3
0
7
f
i
x
4
.
9
9
7
8
1
3
.
1
8
5
6
1
.
6
9
2
4
e
-
09
0
.
0
0
8
9
f
i
x
0
.
5
1
4
9
1
3
.
8
1
8
2
3
.
9
9
8
3
e
-
08
1
.
2
7
8
4
e
-
03
2
8
6
.
9
3
3
3
0
.
4
6
8
7
3
.
8
0
9
4
1
.
9
5
1
6
e
-
11
5
.
9
6
9
0
e
-
06
1
1
7
.
6
1
2
9
2
1
.
0
4
2
1
2
3
.
5
5
3
7
1
.
5
4
0
3
e
-
06
5
.
1
8
7
0
9
8
.
9
0
9
1
9.
P
a
g
e
-
b
l
o
c
k
s
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
1
0
.
0
0
0
0
1
2
.
2
1
5
1
0
.
0
9
9
3
0
.
8
6
6
6
f
i
x
6
.
1
2
6
0
7
.
2
9
6
0
3
.
3
0
0
1
e
-
07
5
.
1
3
1
2
e
-
04
f
i
x
6
.
9
1
8
1
3
.
3
0
9
2
.
6
7
0
7
e
-
09
0
.
0
2
8
4
8
3
2
8
.
9
3
3
3
6
.
3
4
7
1
6
.
4
5
7
5
2
.
1
1
4
3
e
-
19
1
.
0
8
6
1
e
-
08
1
0
7
.
7
0
9
7
1
2
.
8
0
4
3
1
6
.
6
0
3
0
1
.
3
9
5
4
e
-
10
0
.
0
5
8
4
6
4
.
0
9
0
9
1
0
.
L
i
v
e
r
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
S
E)
Te
st
i
n
g
e
r
r
o
r
(
M
S
E)
C
o
n
n
e
c
t
i
o
n
3
7
.
6
3
4
4
3
8
.
5
5
4
3
0
.
0
9
1
9
0
.
5
5
9
6
f
i
x
2
4
.
3
1
5
4
2
5
.
8
6
2
2
7
.
0
0
2
4
e
-
09
0
.
0
0
1
6
f
i
x
2
3
.
9
4
1
9
3
8
.
7
5
2
.
3
4
7
7
e
-
05
0
.
0
2
6
7
8
6
.
4
6
6
7
2
2
.
9
5
7
7
2
7
.
9
2
1
0
5
.
4
3
1
7
e
-
07
2
.
4
1
9
e
-
03
4
7
.
2
9
0
3
4
5
.
2
2
8
2
4
3
.
8
8
1
1
9
.
2
1
6
3
e
-
03
0
.
3
2
4
4
3
1
.
2
7
2
7
1
1
.
M
a
c
k
e
y
-
g
l
a
ss
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
C
o
n
n
e
c
t
i
o
n
8
.
3
0
5
4
e
-
06
0
.
0
0
7
0
f
i
x
3
.
7
7
7
5
e
-
04
0
.
0
2
8
0
f
i
x
2
.
2
4
8
9
e
-
04
2
.
3
5
0
5
e
-
04
4
7
.
7
5
1
.
5
4
7
5
e
-
05
2
.
0
7
7
0
e
-
04
4
0
.
8
7
1
0
3
.
8
4
6
8
e
-
06
0
.
0
5
6
0
1
6
.
4
8
3
9
1
2
.
G
a
s
f
u
r
n
a
c
e
Tr
a
i
n
i
n
g
e
r
r
o
r
(
%C
l
a
ss)
Te
st
i
n
g
e
r
r
o
r
(
%
C
l
a
ss)
C
o
n
n
e
c
t
i
o
n
3
.
5
3
2
9
e
-
04
4
.
6
3
5
2
f
i
x
0
.
0
0
7
1
0
.
1
9
8
7
f
i
x
1
.
5
9
3
2
e
-
3
1
.
6
6
7
1
e
-
3
6
2
.
6
1
.
2
1
0
1
e
-
04
3
.
2
0
0
1
e
-
03
4
7
.
1
9
3
5
6
.
8
4
6
e
-
04
0
.
0
2
2
8
1
5
.
4
5
1
6
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
4
,
Au
g
u
s
t 2
0
2
0
:
3
7
2
5
-
3
7
3
3
3730
T
ab
le
2
.
Av
e
r
ag
e
o
f
th
e
r
a
n
k
s
f
o
r
p
r
o
p
o
s
ed
alg
o
r
ith
m
s
Tr
a
i
n
i
n
g
e
r
r
o
r
(
C
l
a
ssi
f
i
c
a
t
i
o
n
)
A
l
g
o
r
i
t
h
m
M
C
O
-
TL
B
O
M
C
O
-
I
TLB
O
M
C
O
-
Bp
R
a
n
k
2
.
0
0
0
0
1
.
0
0
0
0
3
.
0
0
0
0
Te
st
i
n
g
e
r
r
o
r
(
C
l
a
ss
i
f
i
c
a
t
i
o
n
)
A
l
g
o
r
i
t
h
m
M
C
O
-
TL
B
O
M
C
O
-
I
TLB
O
M
C
O
-
Bp
R
a
n
k
2
.
0
0
0
0
1
.
0
0
0
0
3
.
0
0
0
0
Tr
a
i
n
i
n
g
e
r
r
o
r
(
M
s
e
)
A
l
g
o
r
i
t
h
m
M
C
O
-
TL
B
O
M
C
O
-
I
TLB
O
M
C
O
-
Bp
R
a
n
k
2
.
4
0
0
0
1
.
0
0
0
0
2
.
6
0
0
0
Te
st
i
n
g
e
r
r
o
r
(
M
se)
A
l
g
o
r
i
t
h
m
M
C
O
-
TL
B
O
M
C
O
-
I
TLB
O
M
C
O
-
Bp
R
a
n
k
2
.
0
0
0
0
1
.
0
0
0
0
3
.
0
0
0
0
N
u
mb
e
r
o
f
c
o
n
n
e
c
t
i
o
n
s
A
l
g
o
r
i
t
h
m
M
C
O
-
TL
B
O
M
C
O
-
I
TLB
O
M
C
O
-
Bp
R
a
n
k
2
.
8
0
0
0
1
.
4
0
0
0
1
.
8
0
0
0
T
ab
le
3
.
P
-
v
al
u
e
r
esu
lts
f
o
r
p
a
ir
wis
e
co
m
p
ar
is
o
n
o
f
MCO
-
I
T
L
B
O
v
er
s
u
s
o
th
er
alg
o
r
ith
m
s
b
y
wilco
x
o
n
test
D
a
t
a
s
e
t
C
r
i
t
e
r
i
a
M
C
O
-
TL
B
O
M
C
O
-
Bp
1
.
I
r
i
s
t
e
st
i
n
g
e
r
r
o
r
0
.
6
5
6
4
1
9
.
0
7
3
3
e
-
12
c
o
n
n
e
c
t
i
o
n
0
.
1
2
5
8
1
.
2
1
9
6
e
-
11
2
.
D
i
a
b
e
t
e
s
t
e
st
i
n
g
e
r
r
o
r
0
.
0
9
3
7
7
9
2
.
9
1
7
4
e
-
11
c
o
n
n
e
c
t
i
o
n
0
.
3
0
6
1
7
9
.
8
3
7
5
e
-
09
3
.
Th
y
r
o
i
d
t
e
st
i
n
g
e
r
r
o
r
1
.
2
3
6
9
e
-
05
5
.
9
9
4
1
e
-
07
c
o
n
n
e
c
t
i
o
n
0
.
0
0
0
1
0
0
1
2
7
.
0
6
4
3
e
-
09
4
.
C
a
n
c
e
r
t
e
st
i
n
g
e
r
r
o
r
0
.
4
0
8
9
8
2
.
6
7
5
7
e
-
11
c
o
n
n
e
c
t
i
o
n
0
.
3
1
6
0
4
5
.
0
4
2
2
e
-
11
5
.
C
a
r
d
t
e
st
i
n
g
e
r
r
o
r
0
.
7
7
2
3
9
5
.
5
8
9
3
e
-
11
c
o
n
n
e
c
t
i
o
n
0
.
1
4
3
3
2
.
9
3
2
1
e
-
11
6
.
G
l
a
ss
t
e
st
i
n
g
e
r
r
o
r
0
.
0
3
9
3
5
2
.
8
5
0
2
e
-
11
c
o
n
n
e
c
t
i
o
n
0
.
0
0
6
6
3
5
3
.
9
7
8
7
e
-
10
7
.
H
e
a
r
t
t
e
st
i
n
g
e
r
r
o
r
0
.
3
4
1
6
2
.
7
7
2
2
e
-
11
c
o
n
n
e
c
t
i
o
n
0
.
5
1
0
1
8
1
.
4
9
7
e
-
10
8
.
W
i
n
e
t
e
st
i
n
g
e
r
r
o
r
0
.
0
4
3
9
2
4
2
.
3
4
8
1
e
-
11
c
o
n
n
e
c
t
i
o
n
0
.
0
2
6
3
4
3
.
1
3
7
2
e
-
11
9
.
P
a
g
e
-
b
l
o
c
k
e
s
t
e
st
i
n
g
e
r
r
o
r
0
.
0
0
3
3
8
3
.
0
1
2
3
e
-
11
c
o
n
n
e
c
t
i
o
n
0
.
0
4
8
1
9
2
.
9
9
9
1
e
-
11
1
0
.
l
i
v
e
r
t
e
st
i
n
g
e
r
r
o
r
0
.
7
4
9
6
5
2
.
8
9
9
1
e
-
11
c
o
n
n
e
c
t
i
o
n
0
.
6
9
9
2
5
3
.
6
3
0
5
e
-
09
1
1
.
M
a
c
k
e
y
-
G
l
a
ss
t
e
st
i
n
g
e
r
r
o
r
0
.
0
0
1
1
1
4
3
0
.
0
0
7
9
5
9
c
o
n
n
e
c
t
i
o
n
0
.
1
9
0
7
3
0
.
0
0
0
5
8
7
3
7
1
2
.
G
a
s
F
u
r
n
a
c
e
t
e
st
i
n
g
e
r
r
o
r
0
.
1
4
5
3
2
5
.
4
6
2
e
-
06
c
o
n
n
e
c
t
i
o
n
3
.
3
5
2
e
-
08
0
.
0
0
5
0
8
4
2
Fig
u
r
e
3
.
B
o
x
p
lo
ts
o
f
tr
ai
n
in
g
an
d
test
in
g
er
r
o
r
s
f
o
r
all
d
atasets
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8
7
0
8
A
h
yb
r
id
co
n
s
tr
u
ctive
a
lg
o
r
ith
m
in
co
r
p
o
r
a
tin
g
tea
c
h
in
g
-
le
a
r
n
in
g
b
a
s
ed
…
(
Ma
h
d
ie
h
K
h
o
r
a
s
h
a
d
iz
a
d
e
)
3731
Fig
u
r
e
3
.
B
o
x
p
lo
ts
o
f
tr
ai
n
in
g
an
d
test
in
g
er
r
o
r
s
f
o
r
all
d
atasets
(
co
n
tin
u
e
)
4
.
3
.
Resul
t
s
o
f
co
m
pa
ring
t
h
e
bes
t
pro
po
s
ed
hy
bri
d m
et
ho
d wit
h o
t
her
m
et
ho
ds
I
n
th
is
s
ec
tio
n
,
we
co
m
p
ar
e
o
u
r
h
y
b
r
id
alg
o
r
ith
m
s
with
o
th
er
liter
atu
r
e
m
eth
o
d
s
in
T
ab
le
4
.
T
h
e
p
e
r
ce
n
tag
e
o
f
tr
ain
in
g
er
r
o
r
a
n
d
test
in
g
er
r
o
r
c
o
llectio
n
in
t
h
is
tab
le.
E
ac
h
ar
ticle
wo
r
k
s
o
n
a
b
atch
o
f
d
atasets
.
T
h
e
ce
lls
o
f
th
is
tab
l
e
th
at
d
o
n
’
t h
av
e
an
y
v
alu
e
(
th
at
in
d
icate
with
an
-
ico
n
)
s
h
o
ws th
at
th
ese
v
alu
es
ar
e
m
is
s
in
g
d
ata
o
r
b
elo
n
g
t
o
a
d
ataset
th
at
ar
ticles
d
o
n
’
t
wo
r
k
o
n
t
h
is
.
W
e
g
iv
e
a
b
r
ief
d
escr
ip
tio
n
o
f
th
e
co
m
p
ar
ativ
e
a
p
p
r
o
ac
h
es
as
f
o
llo
ws.
W
e
r
ef
er
en
ce
all
th
e
ap
p
r
o
ac
h
es
th
at
we
co
m
p
ar
e
d
o
u
r
b
est
p
r
o
p
o
s
ed
m
eth
o
d
with
th
e
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
4
,
Au
g
u
s
t 2
0
2
0
:
3
7
2
5
-
3
7
3
3
3732
T
ab
le
4
.
C
o
m
p
a
r
in
g
th
e
r
esu
lts
o
f
b
est
alg
o
r
ith
m
with
o
th
er
m
eth
o
d
s
in
liter
atu
r
e
5.
CO
NCLU
SI
O
N
I
n
th
is
p
ap
er
,
we
p
r
o
p
o
s
ed
a
h
y
b
r
id
izatio
n
o
f
tr
ain
in
g
alg
o
r
ith
m
s
an
d
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
s
t
o
s
im
u
ltan
eo
u
s
ly
d
eter
m
in
e
th
e
weig
h
t
a
n
d
s
tr
u
ct
u
r
e
o
f
th
e
n
eu
r
al
n
et
wo
r
k
.
T
h
e
g
o
al
is
to
e
x
am
in
e
h
y
b
r
id
izatio
n
o
f
a
d
eter
m
in
i
s
tic
an
d
s
y
s
tem
atic
p
r
o
ce
d
u
r
e
(
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
)
w
ith
r
an
d
o
m
s
ea
r
ch
(
ev
o
lu
tio
n
a
r
y
alg
o
r
ith
m
)
f
o
r
n
eu
r
al
n
etwo
r
k
o
p
tim
izatio
n
.
C
o
m
b
in
ed
m
eth
o
d
s
in
clu
d
e
th
e
b
ase
an
d
im
p
r
o
v
ed
v
er
s
io
n
o
f
th
e
T
L
B
O
alg
o
r
i
th
m
with
th
e
MM
OST
alg
o
r
ith
m
s
.
T
h
en
we
co
m
p
ar
e
d
h
y
b
r
id
alg
o
r
ith
m
s
,
an
d
s
elec
ted
th
e
s
u
p
er
io
r
alg
o
r
ith
m
in
class
if
icatio
n
an
d
tim
e
s
er
ies
p
r
ed
icti
o
n
p
r
o
b
le
m
s
.
T
h
e
r
esu
lts
o
f
th
e
co
m
p
a
r
is
o
n
illu
s
tr
ate
th
e
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
b
el
o
n
g
s
to
th
e
MCO
-
IT
L
B
O
a
lg
o
r
ith
m
.
T
h
is
v
e
r
s
io
n
h
as
a
p
o
wer
f
u
l
tr
ai
n
in
g
alg
o
r
ith
m
ag
ain
s
t
ea
r
ly
co
n
v
er
g
e
n
ce
,
a
n
d
b
alan
ce
s
b
etwe
en
ex
p
lo
itati
o
n
a
n
d
ex
p
lo
r
atio
n
.
T
h
is
alg
o
r
ith
m
i
n
co
m
b
in
atio
n
with
th
e
MM
OST
co
n
s
tr
u
ctiv
e
alg
o
r
ith
m
,
m
o
r
e
ef
f
ec
tiv
ely
s
elec
ts
th
e
o
p
tim
al
n
etwo
r
k
s
tr
u
ctu
r
e.
W
e
h
av
e
a
ls
o
v
er
if
ied
th
ese
r
esu
lts
with
s
tatis
tica
l
test
s
,
an
d
f
in
ally
t
h
is
alg
o
r
ith
m
was
co
m
p
ar
ed
with
o
th
er
m
eth
o
d
s
in
liter
atu
r
e
a
n
d
it
h
as
b
ee
n
p
r
o
v
en
th
at
it
is
m
o
r
e
co
n
v
en
ien
t
th
a
n
o
t
h
e
r
alg
o
r
ith
m
s
f
o
r
class
if
icatio
n
an
d
tim
e
s
er
ies
p
r
ed
ictio
n
e
r
r
o
r
.
T
h
ese
p
r
o
m
is
in
g
r
esu
lts
m
o
tiv
ate
u
s
to
f
in
d
way
s
to
ch
a
n
g
e
o
u
r
p
ath
to
f
u
tu
r
e
wo
r
k
.
T
h
is
d
ev
elo
p
m
e
n
t
ca
n
b
e
u
s
in
g
c
h
ao
tic
(
d
is
o
r
d
er
)
m
ap
p
in
g
s
i
n
th
is
m
eth
o
d
.
RE
F
E
R
E
NC
E
S
[1
]
N.
Zh
a
n
g
,
“
An
o
n
li
n
e
g
ra
d
ie
n
t
m
e
th
o
d
wit
h
m
o
m
e
n
t
u
m
fo
r
tw
o
-
lay
e
r
fe
e
d
fo
rwa
rd
n
e
u
ra
l
n
e
tw
o
rk
s,”
A
p
p
li
e
d
M
a
t
h
e
ma
ti
c
s a
n
d
Co
mp
u
ta
ti
o
n
,
v
o
l.
2
1
2
,
n
o
.
2
,
p
p
.
4
8
8
–
4
9
8
,
2
0
0
9
.
[2
]
M
.
G
o
ri,
A.
Tes
i,
“
On
th
e
p
r
o
b
le
m
o
f
lo
c
a
l
m
i
n
ima
in
b
a
c
k
-
p
ro
p
a
g
a
ti
o
n
,
”
IEE
E
T
ra
n
s
a
c
ti
o
n
s
o
n
P
a
tt
e
rn
An
a
lys
is
&
M
a
c
h
i
n
e
In
tell
ig
e
n
c
e
,
v
o
l
.
1
4
,
n
o
.
1
,
p
p
.
7
6
–
8
6
,
1
9
9
2
.
[3
]
H.
M
e
lo
,
J.
Wata
d
a
,
“
G
a
u
ss
ian
-
P
S
O
with
fu
z
z
y
re
a
so
n
i
n
g
b
a
se
d
o
n
stru
c
tu
ra
l
lea
rn
i
n
g
,
”
Ne
u
ro
c
o
m
p
u
ti
n
g
,
v
o
l.
1
7
2
,
p
p
.
4
0
5
-
4
1
2
,
2
0
1
6
.
[4
]
S
.
Ya
n
g
,
Y.
Ch
e
n
,
“
An
e
v
o
l
u
ti
o
n
a
ry
c
o
n
stru
c
ti
v
e
a
n
d
p
ru
n
in
g
a
l
g
o
rit
h
m
fo
r
a
rti
ficia
l
n
e
u
ra
l,
”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l.
8
6
,
p
p
.
1
4
0
–
1
4
9
,
2
0
1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8
7
0
8
A
h
yb
r
id
co
n
s
tr
u
ctive
a
lg
o
r
ith
m
in
co
r
p
o
r
a
tin
g
tea
c
h
in
g
-
le
a
r
n
in
g
b
a
s
ed
…
(
Ma
h
d
ie
h
K
h
o
r
a
s
h
a
d
iz
a
d
e
)
3733
[5
]
R.
V.
Ra
o
,
V.
J.
S
a
v
sa
n
i,
D.
P
.
V
a
k
h
a
ria,
“
Tea
c
h
in
g
–
lea
rn
in
g
-
b
a
se
d
o
p
t
imiz
a
ti
o
n
:
A
n
o
v
e
l
m
e
th
o
d
f
o
r
c
o
n
stra
in
e
d
m
e
c
h
a
n
ica
l
d
e
sig
n
o
p
t
imiz
a
ti
o
n
p
ro
b
lem
s,”
Co
mp
u
ter
-
Ai
d
e
d
De
sig
n
,
v
o
l.
4
3
,
n
o
.
3
,
p
p
.
3
0
3
-
3
1
5
,
2
0
1
1
.
[6
]
V.
Ho
-
Hu
u
,
T.
Ng
u
y
e
n
-
T
h
o
i,
T.
Vo
-
Du
y
,
T.
Ng
u
y
e
n
-
Tran
g
,
“
An
a
d
a
p
ti
v
e
e
li
ti
st
d
iffere
n
ti
a
l
e
v
o
lu
ti
o
n
f
o
r
o
p
ti
m
iza
ti
o
n
o
f
tr
u
ss
stru
c
t
u
re
s
with
d
isc
re
te
d
e
si
g
n
v
a
riab
les
,
”
Co
mp
u
ter
s
a
n
d
S
tr
u
c
tu
re
s
,
v
o
l.
1
6
5
,
p
p
.
5
9
–
7
5
,
2
0
1
6
.
[7
]
O.
Ara
n
,
E.
Alp
a
y
d
ın
,
“
An
in
c
re
m
e
n
tal
n
e
u
ra
l
n
e
two
rk
c
o
n
stru
c
t
io
n
a
lg
o
rit
h
m
f
o
r
trai
n
in
g
m
u
lt
i
lay
e
r
p
e
rc
e
p
tro
n
s,”
Arti
fi
c
ia
l
Ne
u
ra
l
Ne
two
rk
s a
n
d
Ne
u
ra
l
In
fo
rm
a
ti
o
n
P
ro
c
e
ss
in
g
,
2
0
0
3
.
[8
]
M
.
Az
e
v
e
d
o
,
Bra
g
a
,
A.
P
á
d
u
a
,
d
e
M
e
n
e
z
e
s,
B.
Ro
d
rig
u
e
s,
“
Im
p
r
o
v
i
n
g
n
e
u
ra
l
n
e
two
r
k
s
g
e
n
e
ra
li
z
a
ti
o
n
wit
h
n
e
w
c
o
n
stru
c
ti
v
e
a
n
d
p
r
u
n
i
n
g
m
e
th
o
d
s,”
J
o
u
rn
a
l
o
f
I
n
telli
g
e
n
t
&
F
u
zz
y
S
y
ste
ms
,
v
o
l.
1
3
,
n
o
.
2
,
p
p
.
7
5
-
8
3
,
2
0
0
2
.
[9
]
C.
L.
Blak
e
,
C.
J.
M
e
rz
,
“
UCI
Re
p
o
sito
r
y
o
f
M
a
c
h
i
n
e
Lea
rn
i
n
g
Da
tab
a
se
s,
Un
i
v
e
rsity
o
f
Ca
li
fo
rn
ia
a
t
Irv
i
n
e
,
”
1
9
9
8
.
[1
0
]
Av
a
il
a
b
le at
h
t
tp
:/
/d
a
ta
se
ts.co
n
n
e
c
tmv
.
c
o
m
/d
a
tas
e
ts/
[1
1
]
M
.
Kh
ish
e
,
M
.
R.
M
o
sa
v
i,
M
.
Ka
v
e
h
,
“
Im
p
ro
v
e
d
m
ig
ra
ti
o
n
m
o
d
e
ls
o
f
b
i
o
g
e
o
g
ra
p
h
y
-
b
a
se
d
o
p
ti
m
iz
a
ti
o
n
fo
r
so
n
a
r
d
a
tas
e
t
c
las
sifica
ti
o
n
b
y
u
si
n
g
n
e
u
ra
l
n
e
two
r
k
,
”
A
p
p
l
ied
Aco
u
stics
,
v
o
l
.
1
1
8
,
p
p
.
1
5
–
2
9
,
2
0
1
7
.
[1
2
]
Q.
F
a
n
,
Z
.
Wa
n
g
,
H.
Zh
a
,
D.
G
a
o
,
“
M
REKLM
:
A
fa
st
m
u
l
ti
p
le
e
m
p
iri
c
a
l
k
e
r
n
e
l
lea
rn
i
n
g
m
a
c
h
in
e
,
”
P
a
tt
e
r
n
Rec
o
g
n
it
io
n
,
v
o
l.
6
1
,
p
p
.
1
9
7
-
2
0
9
2
0
1
7
.
[1
3
]
N.
S
.
Ja
d
d
i
,
S
,
A
b
d
u
ll
a
h
,
A.
R
.
H
a
m
d
a
n
,
“
A so
lu
ti
o
n
re
p
re
se
n
tati
o
n
o
f
g
e
n
e
ti
c
a
lg
o
r
it
h
m
fo
r
n
e
u
ra
l
n
e
two
rk
we
ig
h
ts
a
n
d
stru
c
t
u
re
,
”.
I
n
fo
rm
a
t
io
n
Pro
c
e
ss
in
g
L
e
tt
e
rs
,
v
o
l.
1
1
6
,
n
o
.
1
,
p
p
.
2
2
-
2
5
,
2
0
1
6
.
[1
4
]
R.
M
.
Cru
z
,
R.
S
a
b
o
u
rin
,
G
.
D.
Ca
v
a
lca
n
ti
,
“
M
ET
A
-
DES
.
Ora
c
le:
M
e
ta
-
lea
rn
in
g
a
n
d
fe
a
tu
re
se
lec
ti
o
n
fo
r
d
y
n
a
m
ic
e
n
se
m
b
le se
lec
ti
o
n
,
”
In
f
o
rm
a
ti
o
n
Fu
sio
n
,
v
o
l
.
3
8
,
p
p
.
84
-
1
0
3
,
2
0
1
7
.
[1
5
]
H.
Ba
d
e
m
,
A.
Ba
stu
r
k
,
A
.
Ca
li
sk
a
n
,
M
.
E
.
Y
u
k
se
l,
“
A
n
e
w
e
fficie
n
t
trai
n
i
n
g
stra
teg
y
fo
r
d
e
e
p
n
e
u
r
a
l
n
e
two
r
k
s
b
y
h
y
b
rid
iza
ti
o
n
o
f
a
rti
ficia
l
b
e
e
c
o
l
o
n
y
a
n
d
l
imited
--
m
e
m
o
ry
B
F
G
S
o
p
ti
m
iza
ti
o
n
a
l
g
o
ri
th
m
s,”
Ne
u
r
o
c
o
m
p
u
ti
n
g
,
v
o
l
.
2
6
6
,
p
p
.
5
0
6
–
5
2
6
,
2
0
1
7
.
[1
6
]
M
.
De
G
re
g
o
rio
M
.
G
io
rd
a
n
o
,
“
An
e
x
p
e
rime
n
tal
e
v
a
lu
a
ti
o
n
o
f
we
ig
h
tl
e
ss
n
e
u
ra
l
n
e
two
r
k
s
fo
r
m
u
lt
i
-
c
las
s
c
las
sifica
ti
o
n
,
”
Ap
p
l.
S
o
f
t
Co
m
p
u
t
.
,
v
o
l.
7
2
,
p
p
.
3
3
8
-
3
5
4
,
2
0
1
8
.
[1
7
]
C.
Z
h
a
n
g
,
C
.
L
iu
,
X.
Z
h
a
n
g
,
G
.
Alm
p
a
n
i
d
is
,
“
A
n
u
p
-
to
-
d
a
te
c
o
m
p
a
riso
n
o
f
sta
te
-
of
-
th
e
-
a
rt
c
las
sifica
ti
o
n
a
lg
o
rit
h
m
s
,
”
Exp
e
rt S
y
ste
ms
wit
h
Ap
p
li
c
a
ti
o
n
s
,
v
o
l
.
82
,
p
p
.
1
2
8
-
1
5
0
,
2
0
1
7
.
[1
8
]
N.
S
.
Ja
d
d
i,
S
.
Ab
d
u
ll
a
h
,
“
A
c
o
o
p
e
ra
ti
v
e
-
c
o
m
p
e
ti
ti
v
e
m
a
ste
r
-
sla
v
e
g
lo
b
a
l
-
b
e
st
h
a
rm
o
n
y
se
a
rc
h
fo
r
AN
N
o
p
ti
m
iza
ti
o
n
a
n
d
wa
ter
-
q
u
a
li
ty
p
r
e
d
ictio
n
,
”
Ap
p
li
e
d
S
o
ft
C
o
mp
u
ti
n
g
,
v
o
l.
51
,
p
p
.
2
0
9
-
2
2
4
,
2
0
1
7
.
[1
9
]
J.
Va
sh
ish
th
a
,
P
.
G
o
y
a
l,
J.
Ah
u
ja,
a
n
d
o
t
h
e
rs,
“
A
No
v
e
l
F
it
n
e
s
s
Co
m
p
u
tatio
n
F
ra
m
e
wo
rk
fo
r
Na
tu
re
In
sp
ire
d
Clas
sifica
ti
o
n
Alg
o
rit
h
m
s,”
Pro
c
e
d
ia
C
o
mp
u
t.
S
c
i.
,
v
o
l.
1
3
2
,
p
p
.
2
0
8
–
2
1
7
,
2
0
1
8
.
[2
0
]
D.
P
.
F
.
Cr
u
z
,
R
.
D.
M
a
i
a
,
L.
A.
d
a
S
il
v
a
,
L.
N
.
d
e
Ca
stro
,
”
Be
e
RBF
:
a
b
e
e
-
i
n
sp
ired
d
a
ta
c
l
u
ste
rin
g
a
p
p
r
o
a
c
h
t
o
d
e
sig
n
RB
F
n
e
u
ra
l
n
e
two
r
k
c
las
si
fiers
,
”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l
.
1
7
2
,
p
p
.
4
2
7
-
4
3
7
,
2
0
1
6
.
[2
1
]
Z.
Xie
,
Y.
Xu
,
Q.
H
u
,
“
Un
c
e
rtai
n
d
a
ta
c
las
sifica
ti
o
n
wit
h
a
d
d
it
iv
e
k
e
rn
e
l
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
,
”
Da
ta
Kn
o
wl
.
En
g
.
v
o
l.
1
1
7
,
p
p
.
8
7
-
9
7
,
2
0
1
8
.
[2
2
]
N.
S
.
Ja
d
d
i
,
S
.
A
b
d
u
ll
a
h
,
M
.
A
b
d
u
l
M
a
lek
,
“
M
a
ste
r
-
Lea
d
e
r
-
S
lav
e
Cu
c
k
o
o
S
e
a
rc
h
wi
th
P
a
ra
m
e
ter
Co
n
tr
o
l
f
o
r
a
n
n
Op
ti
m
iza
ti
o
n
a
n
d
I
ts Rea
l
-
Wo
rl
d
Ap
p
li
c
a
ti
o
n
t
o
Wate
r
Qu
a
li
ty
P
re
d
ictio
n
,
"
PL
o
S
ONE
.
v
o
l.
1
2
(
1
)
,
2
0
1
7
.
[2
3
]
A.
Ca
li
sk
a
n
,
M
.
E
.
Yu
k
se
l,
H
.
B
a
d
e
m
,
A.
Ba
stu
r
k
,
“
P
e
rfo
rm
a
n
c
e
imp
ro
v
e
m
e
n
t
o
f
d
e
e
p
n
e
u
ra
l
n
e
t
wo
rk
c
las
sifiers
b
y
a
sim
p
le
train
i
n
g
stra
teg
y
,
”
E
n
g
in
e
e
rin
g
Ap
p
li
c
a
ti
o
n
s o
f
Arti
fi
c
i
a
l
In
telli
g
e
n
c
e
.
v
o
l.
6
7
,
p
p
.
1
4
–
2
3
,
2
0
1
8
.
[2
4
]
M
.
F
.
M
o
h
a
m
m
e
d
,
C.
P
.
Li
m
,
“
Im
p
ro
v
i
n
g
th
e
F
u
z
z
y
M
i
n
-
M
a
x
n
e
u
ra
l
n
e
two
rk
with
a
K
-
n
e
a
re
st
h
y
p
e
rb
o
x
e
x
p
a
n
sio
n
r
u
le fo
r
p
a
tt
e
rn
c
las
sifica
ti
o
n
,
”
A
p
p
li
e
d
S
o
ft
C
o
mp
u
ti
n
g
,
v
o
l.
5
2
,
p
p
.
1
3
5
-
1
4
5
,
2
0
1
7
.
[2
5
]
B.
Y.
Hie
w,
S
.
C.
Tan
,
W
.
S
.
Li
m
,
“
A
d
o
u
b
le
-
e
li
m
i
n
a
ti
o
n
-
to
u
rn
a
m
e
n
t
-
b
a
se
d
c
o
m
p
e
ti
ti
v
e
c
o
-
e
v
o
l
u
ti
o
n
a
r
y
a
rti
ficia
l
n
e
u
ra
l
n
e
tw
o
rk
c
las
sifier
,
”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l.
2
4
9
,
p
p
.
3
4
5
-
3
5
6
,
2
0
1
7
.
[2
6
]
K.
S
u
n
,
J.
Zh
a
n
g
,
C.
Z
h
a
n
g
,
J.
Hu
,
“
G
e
n
e
ra
li
z
e
d
e
x
trem
e
lea
rn
i
n
g
m
a
c
h
in
e
a
u
to
e
n
c
o
d
e
r
a
n
d
a
n
e
w
d
e
e
p
n
e
u
ra
l
n
e
two
rk
,
”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l.
230
,
p
p
.
3
7
4
-
3
8
1
,
2
0
1
7
.
[2
7
]
Z.
G
a
o
,
C.
M
a
,
D.
S
o
n
g
,
Y.
Li
u
,
“
De
e
p
q
u
a
n
tu
m
in
s
p
ired
n
e
u
ra
l
n
e
two
rk
wit
h
a
p
p
li
c
a
ti
o
n
to
a
ircr
a
ft
fu
e
l
sy
ste
m
fa
u
lt
d
ia
g
n
o
sis
,
”
Ne
u
ro
c
o
m
p
u
ti
n
g
,
v
o
l.
2
3
8
,
p
p
.
13
-
2
3
,
2
0
1
7
.
[2
8
]
J.
Wan
g
,
Q.
Ca
i,
Q.
Ch
a
n
g
,
J.
M
.
Zu
ra
d
a
,
“
Co
n
v
e
rg
e
n
c
e
a
n
a
l
y
se
s
o
n
s
p
a
rse
fe
e
d
fo
rwa
rd
n
e
u
ra
l
n
e
t
wo
rk
s
v
ia
g
r
o
u
p
las
so
re
g
u
lariz
a
ti
o
n
,
”
In
fo
rm
a
t
io
n
S
c
ien
c
e
s
,
v
o
l.
3
8
1
,
p
p
.
2
5
0
-
2
6
9
,
2
0
1
7
.
[2
9
]
S
.
S
h
in
d
e
,
U.
Ku
l
k
a
rn
i
,
“
Ex
ten
d
e
d
fu
z
z
y
h
y
p
e
rli
n
e
-
se
g
m
e
n
t
n
e
u
r
a
l
n
e
two
r
k
wit
h
c
las
sifica
ti
o
n
ru
le
e
x
trac
ti
o
n
,
”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l
.
2
6
0
,
p
p
.
7
9
–
9
1
,
2
0
1
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.