I
AE
S
I
nte
rna
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
7
,
No
.
4
,
Dec
em
b
er
201
8
,
p
p
.
170
~
178
I
SS
N:
2252
-
8938
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ai.
v
7
.
i4
.
p
p
1
70
-
1
78
170
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e.
co
m/jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JA
I
I
m
pro
v
ed
Ti
m
e
T
ra
ining
w
ith
Accuracy
o
f
Ba
tch
Ba
c
k
Propa
g
a
tion Alg
o
r
ith
m
Via
Dyna
m
ic Lea
rning
Rate
a
nd
Dy
na
m
ic Mo
m
en
tu
m
F
a
c
tor
M
o
ha
m
m
e
d Sa
rha
n Al_
Dua
is
,
F
a
t
m
a
Su
s
ila
w
a
t
i.
M
o
ha
m
a
d
F
a
c
u
lt
y
o
f
In
f
o
r
m
a
ti
c
s a
n
d
Co
m
p
u
ti
n
g
,
U
n
iv
e
rsiti
S
u
lt
a
n
Zai
n
a
l
A
b
id
in
,
T
e
re
n
g
g
a
n
u
,
M
a
lay
sia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l
11
,
2
0
1
8
R
ev
i
s
ed
No
v
2
,
2
0
1
8
A
cc
ep
ted
No
v
26
,
2
0
1
8
T
h
e
m
a
n
p
ro
b
lem
o
f
b
a
tch
b
a
c
k
p
ro
p
a
g
a
ti
o
n
(BB
P
)
a
lg
o
ri
th
m
is
slo
w
train
in
g
a
n
d
t
h
e
re
a
re
se
v
e
ra
l
p
a
r
a
m
e
ters
n
e
e
d
s
to
b
e
a
d
j
u
ste
d
m
a
n
u
a
ll
y
,
a
lso
s
u
ff
e
rs
f
ro
m
sa
tu
ra
ti
o
n
train
i
n
g
.
T
h
e
lea
rn
in
g
ra
te
a
n
d
m
o
m
e
n
tu
m
fa
c
to
r
a
re
sig
n
if
ica
n
t
p
a
ra
m
e
ters
f
o
r
in
c
re
a
sin
g
th
e
e
ff
icie
n
c
y
o
f
th
e
(BB
P
)
.
In
th
is
stu
d
y
,
w
e
c
r
e
a
ted
a
n
e
w
d
y
n
a
m
ic
f
u
n
c
ti
o
n
o
f
e
a
c
h
lea
rn
in
g
ra
te
a
n
d
m
o
m
e
n
tu
m
f
a
c
o
r.
W
e
p
re
s
e
n
t
th
e
DBBP
L
M
a
lg
o
rit
h
m
,
w
h
ich
train
s
w
it
h
a
d
y
n
a
m
ic
f
u
n
c
ti
o
n
f
o
r
e
a
c
h
th
e
lea
rn
in
g
ra
te
a
n
d
m
o
m
e
n
tu
m
fa
c
to
r.
A
S
ig
m
o
id
f
u
n
c
ti
o
n
u
se
d
a
s
a
c
ti
v
a
ti
o
n
f
u
n
c
ti
o
n
.
T
h
e
X
OR
p
ro
b
le
m
,
b
a
lan
c
e
,
b
re
a
st
c
a
n
c
e
r
a
n
d
iri
s
d
a
tas
e
t
we
r
e
u
se
d
a
s
b
e
n
c
h
m
a
r
k
s
f
o
r
tes
ti
n
g
th
e
e
ffe
c
ts
o
f
th
e
d
y
n
a
m
ic
DBBP
L
M
a
lg
o
rit
h
m
.
A
ll
th
e
e
x
p
e
ri
m
e
n
ts
w
e
r
e
p
e
r
f
o
r
m
e
d
o
n
M
a
tl
a
b
2
0
1
2
a
.
T
h
e
sto
p
trai
n
in
g
w
a
s
d
e
ter
m
in
e
d
ten
p
o
w
e
r
-
5
.
F
ro
m
th
e
e
x
p
e
ri
m
e
n
tal
re
su
lt
s,
th
e
DBB
P
L
M
a
lg
o
rit
h
m
p
r
o
v
id
e
s
su
p
e
ri
o
r
p
e
rf
o
r
m
a
n
c
e
in
term
s
o
f
train
in
g
,
a
n
d
f
a
ste
r
train
in
g
w
it
h
h
ig
h
e
r
a
c
c
u
ra
c
y
c
o
m
p
a
re
d
to
th
e
BBP
a
lg
o
r
it
h
m
a
n
d
w
it
h
e
x
isti
n
g
w
o
rk
s.
K
ey
w
o
r
d
:
A
cc
u
r
ac
y
tr
ai
n
i
n
g
B
atch
B
ac
k
-
p
r
o
p
ag
atio
n
alg
o
r
ith
m
D
y
n
a
m
ic
lear
n
i
n
g
r
ate
D
y
n
a
m
ic
m
o
m
en
t
u
m
f
ac
to
r
Sp
ee
d
u
p
T
r
ain
in
g
Co
p
y
rig
h
t
©
2
0
1
8
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
:
Mo
h
a
m
m
ed
Sar
h
a
n
A
l_
D
u
ais
Facu
lt
y
o
f
I
n
f
o
r
m
a
tics
a
n
d
C
o
m
p
u
ti
n
g
,
Un
i
v
er
s
iti
Su
lta
n
Z
ai
n
al
A
b
id
i
n
,
T
er
en
g
g
an
u
,
Ma
la
y
s
ia.
E
m
ail: sar
h
a
n
2
w
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
b
atch
B
P
alg
o
r
ith
m
i
s
c
o
m
m
o
n
l
y
u
s
ed
in
m
a
n
y
ap
p
l
icatio
n
s
i
n
cl
u
d
in
g
r
o
b
o
tics
,
au
to
m
at
io
n
,
an
d
w
ei
g
h
t
ch
a
n
g
es
in
A
NNs
[
1
]
.
T
h
e
B
P
alg
o
r
ith
m
h
as
l
ed
to
tr
e
m
en
d
o
u
s
b
r
ea
k
t
h
r
o
u
g
h
s
i
n
ap
p
licatio
n
s
in
v
o
l
v
i
n
g
m
u
lti
la
y
er
p
er
ce
p
tio
n
s
[
2]
.
Gr
ad
ien
t
d
esce
n
t
is
co
m
m
o
n
l
y
u
s
ed
to
ad
j
u
s
t
th
e
w
e
ig
h
t
u
s
i
n
g
a
c
h
a
n
g
e
th
e
er
r
o
r
tr
ain
in
g
;
h
o
w
ev
er
,
t
h
is
ap
p
r
o
ac
h
is
n
o
t
g
u
ar
an
tee
d
to
f
i
n
d
th
e
g
lo
b
al
m
i
n
i
m
u
m
er
r
o
r
[
3
]
.
T
h
e
B
B
P
alg
o
r
ith
m
is
a
cc
u
r
ate
in
ter
m
s
o
f
tr
ai
n
in
g
[
4
]
.
T
h
e
b
atch
BP
alg
o
r
ith
m
is
a
n
e
w
s
t
y
le
f
o
r
u
p
d
at
in
g
w
ei
g
h
t,
it
is
w
id
el
y
u
s
ed
in
tr
ain
i
n
g
al
g
o
r
it
h
m
s
a
s
it
is
ac
cu
r
ate
f
o
r
tr
ai
n
i
n
g
[
5
]
.
I
t u
ti
lizes
th
e
g
r
ad
ie
n
t
d
escen
t,
w
h
ic
h
d
o
es
n
o
t
en
s
u
r
e
to
r
ea
ch
th
e
g
lo
b
al
m
i
n
i
m
u
m
er
r
o
r
b
ec
au
s
e
i
t
m
a
y
r
esu
lt
i
n
lead
in
g
t
h
e
lo
ca
l
m
i
n
i
m
u
m
[
6
,
7
]
.
Desp
ite
t
h
e
tr
ain
in
g
r
ate
a
n
d
m
o
m
e
n
t
u
m
f
ac
to
r
b
ein
g
s
i
g
n
i
f
ican
t
p
ar
a
m
eter
s
f
o
r
co
n
tr
o
llin
g
th
e
u
p
d
ated
w
ei
g
h
t,
it
is
d
if
f
ic
u
lt
to
s
ele
ct
th
e
b
est
v
al
u
ed
u
r
i
n
g
tr
ain
i
n
g
[
8
]
.
Gen
er
all
y
,
t
h
er
e
ar
e
t
w
o
tech
n
iq
u
e
s
f
o
r
s
elec
ti
n
g
t
h
e
v
al
u
es
f
o
r
ea
ch
t
r
ain
in
g
r
at
e
an
d
m
o
m
e
n
tu
m
f
ac
to
r
.
T
h
e
f
ir
s
t
is
s
et
to
b
e
a
s
m
al
l
co
n
s
tan
t
v
al
u
e
f
r
o
m
in
ter
v
al
[
1
]
,
th
e
s
ec
o
n
d
t
h
e
s
ele
cted
s
er
ie
s
v
al
u
e
f
r
o
m
[
9
]
.
T
h
e
lear
n
i
n
g
r
ate
s
h
o
u
ld
b
e
s
u
f
f
icie
n
tl
y
lar
g
e
to
allo
w
f
o
r
e
s
ca
p
in
g
th
e
lo
c
al
m
in
i
m
u
m
[
1
0
]
.
B
u
t
th
e
b
i
g
g
e
s
t
v
al
u
e
lead
s
to
f
ast
tr
ai
n
in
g
w
i
th
o
s
cil
latio
n
er
r
o
r
tr
ain
in
g
.
T
o
en
s
u
r
e
a
S
u
itab
le
lear
n
i
n
g
B
P
alg
o
r
ith
m
,
th
e
l
ea
r
n
i
n
g
r
ate
m
u
s
t
b
e
s
m
al
l
[
1
1
]
.
A
n
o
th
er
r
eq
u
ir
e
m
en
ts
f
o
r
s
p
ee
d
i
n
g
u
p
o
f
t
h
e
b
atc
h
B
B
P
alg
o
r
ith
m
is
ad
ap
ti
v
e
tr
ai
n
i
n
g
r
ate
a
n
d
m
o
m
en
tu
m
f
ac
to
r
to
g
eth
er
[
1
2
].
T
h
e
m
ai
n
p
r
o
b
le
m
o
f
B
B
P
alg
o
r
ith
m
,
i
s
s
lo
w
tr
ai
n
in
g
,
o
r
s
tu
c
k
tr
ai
n
i
n
g
ar
o
u
n
d
t
h
e
lo
ca
l
m
i
n
i
m
u
m
a
n
d
s
u
f
f
er
s
f
r
o
m
s
atu
r
atio
n
tr
ai
n
i
n
g
[
1
3
]
.
I
n
ad
d
itio
n
p
r
o
b
lem
o
f
t
h
e
B
P
alg
o
r
it
h
m
,
s
e
v
er
al
p
ar
am
eter
s
n
ee
d
to
b
e
ad
j
u
s
ted
m
a
n
u
all
y
,
s
u
c
h
as lea
r
n
i
n
g
r
ate
an
d
m
o
m
e
n
t
u
m
f
ac
to
r
[
1
4
].
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
I
mp
r
o
ve
d
Time
Tr
a
in
in
g
W
ith
A
cc
u
r
a
cy
o
f B
a
tch
B
a
ck
P
r
o
p
a
g
a
tio
n
…
(
Mo
h
a
mme
d
S
a
r
h
a
n
A
l_
Du
a
is
)
171
C
u
r
r
en
t
w
o
r
k
f
o
r
s
o
lv
i
n
g
t
h
e
s
lo
w
tr
ain
in
g
o
f
th
e
B
B
P
al
g
o
r
ith
m
is
t
h
r
o
u
g
h
ad
ap
tin
g
o
f
a
s
o
m
e
s
ig
n
i
f
ica
n
t
p
ar
a
m
e
ter
s
,
s
u
c
h
a
s
le
ar
n
in
g
r
ate
an
d
m
o
m
en
t
u
m
f
ac
to
r
.
Fo
r
t
h
ese
ca
s
es
m
a
n
y
s
t
u
d
ies
h
a
s
b
ee
n
d
o
n
e
s
u
c
h
a
s
[
1
5
]
i
m
p
r
o
v
ed
th
e
B
P
a
lg
o
r
it
h
m
t
h
r
o
u
g
h
t
w
o
tec
h
n
iq
u
e
s
,
th
e
tr
ain
in
g
r
ate
an
d
m
o
m
e
n
t
u
m
f
ac
to
r
,
th
e
v
al
u
es
o
f
tr
ai
n
i
n
g
r
ate
w
er
e
f
i
x
ed
at
d
if
f
er
en
t
v
al
u
es.
T
h
e
id
ea
o
f
th
is
s
t
u
d
y
i
s
to
s
et
th
e
v
al
u
e
o
f
tr
ain
i
n
g
,
r
ate
to
b
e
la
r
g
e
in
itia
ll
y
,
an
d
th
e
n
to
lo
o
k
at
th
e
v
alu
e
o
f
er
r
o
r
t
r
ain
in
g
af
ter
iter
at
io
n
.
I
f
th
e
er
r
o
r
(
e)
tr
ain
i
n
g
is
i
n
cr
ea
s
ed
,
th
e
f
it
p
r
o
d
u
ce
d
ch
an
g
e
s
t
h
e
v
al
u
e
o
f
t
r
ain
in
g
,
r
ate
m
u
lt
ip
lied
b
y
les
s
t
h
an
o
n
e
an
d
th
e
n
r
ec
alcu
lated
in
t
h
e
o
r
ig
i
n
al
d
ir
ec
tio
n
.
I
f
th
e
iter
atio
n
er
r
o
r
ca
n
b
e
r
ed
u
ce
d
,
th
e
f
it
p
r
o
d
u
ce
d
ch
an
g
es
t
h
e
v
al
u
e
o
f
tr
ain
in
g
r
ate
b
y
m
u
ltip
lied
b
y
a
co
n
s
ta
n
t
g
r
ea
ter
t
h
an
o
n
e,
t
h
e
n
ex
t
iter
at
io
n
is
ca
lcu
lated
co
n
ti
n
u
o
u
s
l
y
.
I
n
[
1
6
]
co
m
p
ar
e
s
ev
er
al
tech
n
i
q
u
es
f
o
r
i
m
p
r
o
v
ed
B
P
alg
o
r
ith
m
.
T
h
e
B
P
alg
o
r
ith
m
w
it
h
ad
ap
tiv
e
lear
n
i
n
g
r
ate
an
d
m
o
m
e
n
tu
m
f
ac
to
r
g
a
v
e
s
u
p
er
io
r
ac
cu
r
ac
y
tr
ain
in
g
at
1
0
0
0
ep
o
ch
s
.
I
n
[
1
7
]
m
o
d
if
y
i
n
g
t
h
e
tr
ai
n
i
n
g
r
ate
an
d
m
o
m
e
n
t
u
m
.
T
h
e
v
al
u
e
o
f
th
e
tr
ain
i
n
g
r
ate
s
elec
ted
d
ep
en
d
s
o
n
th
e
r
atio
b
et
w
ee
n
t
h
e
n
e
w
er
r
o
r
an
d
th
e
p
r
ev
io
u
s
er
r
o
r
tr
ain
i
n
g
.
T
h
e
s
i
m
u
latio
n
r
e
s
u
l
ts
s
h
o
w
a
n
o
p
ti
m
izatio
n
o
f
t
h
e
tr
ai
n
i
n
g
s
p
ee
d
an
d
an
o
s
c
illatio
n
r
ed
u
ctio
n
d
u
r
atio
n
tr
ain
i
n
g
.
I
n
[
1
8
]
,
cr
ea
ted
d
y
n
a
m
ic
tr
ai
n
i
n
g
t
h
at
co
n
s
i
s
ts
o
f
m
u
l
ti
-
s
tep
s
.
T
h
e
v
alu
e
o
f
th
e
lear
n
in
g
r
ate
an
d
m
o
m
en
tu
m
f
ac
to
r
ar
e
s
et
as
m
u
n
au
le
v
al
u
e
.
Fro
m
t
h
e
e
x
p
er
i
m
e
n
tal
r
e
s
u
lt
s
,
t
h
e
i
m
p
r
o
v
ed
alg
o
r
ith
m
w
as o
v
er
all
ef
f
icie
n
t,
b
o
th
in
v
i
s
u
a
l e
f
f
ec
t a
n
d
q
u
alit
y
.
T
h
e
r
em
a
in
i
n
g
p
o
r
tio
n
o
f
th
is
p
ap
er
is
o
r
g
an
ized
a
s
f
o
llo
w
s
:
Sectio
n
2
is
th
e
m
ater
ia
ls
a
n
d
m
et
h
o
d
;
Sectio
n
3
is
cr
ea
ted
th
e
d
y
n
a
m
ic
p
ar
a
m
eter
s
;
Sect
io
n
4
,
is
ex
p
er
i
m
e
n
tal
r
es
u
lts
;
Secti
o
n
5
d
is
cu
s
s
io
n
to
v
alid
ate
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
i
m
p
r
o
v
ed
alg
o
r
it
h
m
;
Secti
o
n
6
,
ev
alu
ate
t
h
e
p
er
f
o
r
m
an
c
e
o
f
i
m
p
r
o
v
e
DB
B
P
alg
o
r
ith
m
.
Fi
n
all
y
,
Sectio
n
7
th
e
co
n
clu
s
io
n
s
.
2.
M
AT
E
RIAL
S AN
D
M
E
T
H
O
D
T
h
is
k
i
n
d
o
f
th
is
r
esear
c
h
b
elo
n
g
s
to
th
e
h
eu
r
i
s
tic
m
et
h
o
d
.
T
h
is
m
et
h
o
d
is
in
c
lu
d
es
t
h
e
le
ar
n
in
g
r
ate
an
d
m
o
m
e
n
t
u
m
f
ac
to
r
.
T
o
I
n
v
esti
g
a
te
th
e
ai
m
s
o
f
th
i
s
s
t
u
d
y
th
er
e
ar
e
m
an
y
s
tep
s
as f
o
llo
ws
.
2
.
1
.
Da
t
a
s
et
T
h
e
d
ata
s
et
i
s
v
er
y
i
m
p
o
r
ta
n
t
f
o
r
v
er
if
ica
ti
o
n
to
i
m
p
r
o
v
e
t
h
e
B
B
P
alg
o
r
ith
m
.
I
n
th
i
s
s
tu
d
y
,
all
d
ata
ar
e
tak
en
f
r
o
m
UC
I
Ma
ch
in
e
L
ea
r
n
i
n
g
R
ep
o
s
ito
r
y
t
h
r
o
u
g
h
th
e
li
n
k
h
ttp
s
:/
/ar
ch
i
v
e.
ics.
u
ci.
ed
u
/
m
l
/in
d
e
x
.
h
t
m
l
.
A
ll
r
ea
l
d
ata
s
et
ch
a
n
g
e
to
b
ec
o
m
e
n
o
r
m
izatio
n
d
atase
t
b
et
w
ee
n
[
0
,
1
]
.
A
ll
d
ata
s
et
d
i
v
id
ed
in
to
t
w
o
s
et
tr
ain
i
n
g
s
e
t a
n
d
test
in
g
s
et.
2.
2
.
Neura
l N
et
w
o
rk
M
o
del
W
e
p
r
o
p
o
s
e
an
A
NN
m
o
d
el,
w
h
ic
h
co
n
s
is
t
o
f
t
h
r
ee
-
la
y
er
n
eu
r
al
n
e
t
w
o
r
k
t
h
at
h
as
a
n
i
n
p
u
t,
h
id
d
en
,
an
d
o
u
tp
u
t
la
y
er
.
T
h
e
in
p
u
t
la
y
er
is
co
n
s
id
er
ed
to
b
e
{
1
x
,
2
x
,
i
x
},
w
h
ic
h
r
ep
r
esen
ts
th
e
n
o
d
es;
th
e
n
o
d
es
d
ep
en
d
o
n
th
e
ty
p
es
o
r
attr
ib
u
tes
o
f
th
e
d
ata.
T
h
e
h
id
d
e
n
la
y
er
is
m
ad
e
o
f
t
w
o
la
y
e
r
s
w
it
h
f
o
u
r
n
o
d
es.
W
h
er
ea
s
th
e
h
L
an
d
k
LL
ar
e
th
e
f
ir
s
t
an
d
s
ec
o
n
d
la
y
er
r
esp
ec
ti
v
el
y
.
T
h
e
o
u
tp
u
t
la
y
er
r
Y
is
m
ad
e
o
f
o
n
e
la
y
er
w
ith
o
n
e
n
eu
r
o
n
.
T
h
r
ee
b
asis
,
t
w
o
o
f
t
h
e
m
ar
e
u
s
ed
i
n
th
e
h
id
d
en
an
d
o
n
e
in
t
h
e
o
u
t
p
u
t
la
y
er
,
w
h
ic
h
is
d
en
o
ted
b
y
0
j
u
,
0
k
v
an
d
0
r
w
.
hj
v
is
th
e
w
ei
g
h
t
b
et
w
ee
n
n
eu
r
o
n
h
f
r
o
m
h
id
d
en
la
y
er
L
an
d
n
e
u
r
o
n
j
f
r
o
m
th
e
h
id
d
en
la
y
er
LL
.
ih
u
is
t
h
e
w
ei
g
h
t b
et
w
ee
n
n
e
u
r
o
n
i
in
t
h
e
i
n
p
u
t la
y
er
a
n
d
n
e
u
r
o
n
h
in
th
e
h
id
d
en
la
y
er
.
Fin
all
y
,
t
h
e
s
i
g
m
o
id
f
u
n
ctio
n
i
s
e
m
p
lo
y
ed
as a
n
ac
tiv
a
tio
n
f
u
n
ctio
n
.
3.
CREA
T
E
D
T
H
E
D
YNAM
I
C
L
E
A
RNIN
G
RA
T
E
AND
M
O
M
E
NT
UM
F
ACTOR
T
h
e
w
ei
g
h
t
u
p
d
ate
b
et
w
ee
n
n
e
u
r
o
n
k
f
r
o
m
t
h
e
o
u
tp
u
t
la
y
er
a
n
d
n
e
u
r
o
n
j
f
r
o
m
t
h
e
h
id
d
en
la
y
er
is
as
f
o
llo
w
s
:
(
1
)
(
)
(
)
(
1
)
j
k
j
k
j
k
j
k
w
t
w
t
w
t
w
t
(
1
)
W
h
er
e
()
jk
wt
is
a
w
ei
g
h
t
c
h
an
g
e
th
e
w
e
ig
h
t
is
u
p
d
ated
f
o
r
ea
c
h
ep
o
c
h
in
E
q
u
atio
n
1
.
Sp
ee
d
u
p
tr
ain
i
n
g
d
ep
en
d
s
o
n
a
p
ar
a
m
e
ter
th
at
af
f
ec
t
s
t
h
e
u
p
d
atin
g
o
f
t
h
e
w
e
ig
h
t.
B
e
f
o
r
e
g
o
in
g
to
cr
ea
te
th
e
d
y
n
a
m
ic
f
u
n
ctio
n
f
o
r
lear
n
i
n
g
r
ate
a
n
d
m
o
m
en
t
u
m
f
ac
to
r
.
T
h
e
ex
p
o
n
en
t
ial
i
s
m
o
n
o
to
n
e
f
u
n
ctio
n
,
w
e
ca
n
cr
ea
te
t
h
e
lear
n
in
g
r
ate
as b
o
u
n
d
ar
y
f
u
n
c
tio
n
as f
o
llo
w
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
7
,
No
.
4
,
Dec
em
b
er
201
8
:
1
7
0
–
1
7
8
172
(
sin
2
)
kE
dmic
e
(
2
)
f
r
o
m
ab
o
v
e
th
e
d
y
n
a
m
ic
lear
n
in
g
r
ate
dmic
.
I
n
th
is
ca
s
e
th
e
p
r
o
p
er
ty
o
f
f
u
n
ctio
n
ex
p
o
n
en
tia
l
d
ep
en
d
o
n
o
f
th
e
v
alu
e
o
f
s
i
n
2
ke
.
s
i
n
e
is
th
e
b
o
u
n
d
ar
y
f
u
n
ctio
n
o
n
d
ef
i
n
in
g
s
et
o
f
e
(
er
r
o
r
tr
ain
in
g
)
also
s
i
n
e
h
as
a
b
o
u
n
d
ar
y
a
s
1
s
i
n
1
e
e
[0
,
1
]
.
T
h
e
E
q
u
atio
n
2
,
is
b
o
u
n
d
ed
f
u
n
ct
io
n
.
T
h
e
w
eig
h
t
u
p
d
ated
u
n
d
er
ef
f
ec
ted
b
o
u
n
d
ar
y
o
f
lear
n
in
g
r
ate.
T
o
g
et
s
m
o
o
th
tr
ai
n
i
n
g
a
n
d
av
o
id
in
f
latio
n
in
t
h
e
g
r
o
s
s
w
ei
g
h
t
o
f
t
h
e
ad
d
ed
v
al
u
es
f
o
r
m
o
m
e
n
t
u
m
fa
cto
r
,
th
e
f
it
tin
g
p
r
o
d
u
ce
r
th
r
o
u
g
h
cr
ea
ti
n
g
d
y
n
a
m
ic
m
o
m
en
t
u
m
f
ac
to
r
an
d
i
m
p
licate
t
h
e
dmic
.
Dep
en
d
ab
o
v
e
w
e
ca
n
cr
ea
ted
th
e
d
y
n
a
m
ic
m
o
m
e
n
t
u
m
f
ac
to
r
as
f
o
llo
w
:
(
sin
2
)
1
sin(
[
(
1
)
]
)
sin(
)
dm
ic
r
r
kE
e
Y
Y
e
(
3
)
W
h
er
e
th
e
i
s
t
h
e
p
en
alt
y
,
t
h
e
d
m
i
c
is
b
o
u
n
d
ar
y
f
u
n
ca
tio
n
.
I
n
s
er
t
th
e
E
q
u
atio
n
2
an
d
3
in
t
o
E
q
u
atio
n
1
,
th
e
n
th
e
w
ei
g
h
t is
u
p
d
ated
b
et
w
ee
n
a
n
y
la
y
er
as
b
elo
w
(
s
in
2
)
(
s
in
2
)
1
sin
(
[
(
1
)
]
(
1
)
(
)
(
)
)
si
(
(
1
)
n)
kE
rr
jk
jk
jk
j
k
k
E
e
w
t
w
t
w
t
w
t
e
Y
Y
e
(
4
)
T
h
e
w
eig
h
ted
u
p
d
ated
u
n
d
er
ef
f
ec
ted
u
n
d
er
d
y
n
a
m
ic
lear
n
i
n
g
r
ate
an
d
m
o
m
en
tu
m
f
ac
to
r
3
.
1
.
Dy
na
m
ic
ba
t
ch
B
a
c
k
pr
o
pa
g
a
t
io
n (
D
B
B
P
L
M
)
a
lg
o
ri
t
h
m
Up
d
ate
W
eig
h
t P
h
ase
,
t
h
e
w
ei
g
h
t
s
a
r
e
ad
j
u
s
ted
s
i
m
u
lta
n
eo
u
s
l
y
,
as
f
o
llo
w
s
Fo
r
ea
ch
o
u
tp
u
t la
y
er
,
Y,
th
e
o
u
t p
u
t a
t n
e
u
r
o
n
r
,
j
LL
th
e
s
ec
o
n
d
la
y
er
(
s
in
2
)
(
s
in
2
)
1
(
1
)
(
)
(
)
[
sin
(
(
(
1
)
)
sin
(
)
(
1
)
k
E
j
rr
r
jr
r
j
jr
E
k
w
t
w
t
e
L
YY
L
e
w
t
e
(
5
)
Fo
r
b
ias
(
s
in
2
)
0
0
0
(
s
in
2
)
1
(
1
)
(
)
(
)
[
sin
(
(
(
1
)
)
sin
(
)
(
1
)
E
rr
k
r
rr
k
E
r
w
t
w
t
e
e
w
t
e
YY
(
6
)
Fo
r
ea
ch
h
id
d
en
la
y
er
h
L
i=0
,
…,
n
;
h
=1
,
…,
q
(
s
in
2
)
(
s
in
2
)
1
(
1
)
(
)
(
)
[
sin
(
(
(
1
)
)
sin
(
)
(
1
)
E
ih
ih
j
i
r
r
i
E
k
k
h
u
t
w
t
e
x
e
Y
Y
u
t
e
(
7
)
Fo
r
b
ias
(
s
in
2
)
0
0
0
(
s
in
2
)
1
(
1
)
(
)
(
)
[
sin
(
(
(
1
)
)
sin
(
)
(
1
)
E
h
k
h
h
r
r
k
h
E
u
t
u
t
e
e
Y
Y
u
t
e
(
8
)
4.
E
XP
E
R
I
M
E
NT
A
L
RE
SUL
T
S
W
e
ca
lcu
late
th
e
ac
c
u
r
ac
y
o
f
t
r
ain
in
g
as
f
o
l
lo
w
s
[
1
9
]
,
A
c
c
ura
c
y
(
%
)
=
1
−
ab
s
o
l
ut
(
T
i
−
O
i
)
UP
−
LW
∗
100
w
h
er
e
UP
=1
an
d
L
W
=
0
ar
e
th
e
u
p
p
er
b
o
u
n
d
an
d
lo
w
er
b
o
u
n
d
o
f
th
e
ac
tiv
a
tio
n
f
u
n
ctio
n
.
4
.
1
.
E
x
peri
m
ent
re
s
ult
o
f
t
he
DB
B
P
L
M
a
lg
o
rit
h
m
w
it
h XO
R
pro
ble
m
T
h
e
DB
B
P
L
M
alg
o
r
i
th
m
is
t
r
ain
in
g
u
n
d
er
ef
f
ec
ted
d
y
n
a
m
ic
lear
n
i
n
g
r
ate
a
n
d
m
o
m
e
n
tu
m
f
ac
to
r
w
h
ic
h
c
r
ea
ted
in
ea
ch
E
q
u
ati
o
n
2
an
d
3
.
T
en
ex
p
er
i
m
en
t
s
h
as
b
ee
n
d
o
n
e
an
d
th
e
n
tak
e
t
h
e
av
er
ag
e
o
f
ti
m
e,
ep
o
ch
an
d
ac
cu
r
a
c
y
.
T
h
e
r
esu
l
t r
ec
o
r
d
e
d
in
th
e
T
ab
le
1
.
j
0
,1
,
2
,
p
;
r
1
,
m
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
I
mp
r
o
ve
d
Time
Tr
a
in
in
g
W
ith
A
cc
u
r
a
cy
o
f B
a
tch
B
a
ck
P
r
o
p
a
g
a
tio
n
…
(
Mo
h
a
mme
d
S
a
r
h
a
n
A
l_
Du
a
is
)
173
T
ab
le
1
.
A
v
er
ag
e
t
h
e
P
er
f
o
r
m
an
ce
o
f
DB
B
P
L
M
alg
o
r
it
h
m
w
it
h
XO
R
F
i
r
st
st
r
u
c
t
u
r
e
se
c
o
n
d
s
t
r
u
c
t
u
r
e
Ex
T
i
me
-
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
T
i
me
-
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
Av
1
.
9
5
6
9
2
7
4
1
0
.
9
8
3
4
1
.
6
2
6
7
2
8
3
2
0
.
9
8
5
8
S
.
D
0
.
1
9
8
0
2
0
0
0
.
1
2
0
3
2
8
0
0
Fro
m
T
ab
le
1
,
f
o
r
f
ir
s
t
s
tr
u
c
tu
r
e
th
e
av
er
a
g
e
tr
ain
in
g
ti
m
e
i
s
t
=
1
.
9
5
6
9
s
ec
o
n
d
s
w
ith
2
7
4
1
ep
o
ch
.
Fo
r
s
ec
o
n
d
s
tr
u
ct
u
r
es
t
h
e
av
er
a
g
e
ti
m
e
tr
ai
n
in
g
is
t
=
1
.
6
2
6
7
s
ec
o
n
d
s
,
w
it
h
2
8
3
2
e
p
o
ch
.
No
m
o
r
e
d
if
f
er
en
t
b
et
w
ee
n
b
o
th
s
tr
u
ctu
r
e
s
f
o
r
ac
cu
r
ac
y
tr
ain
i
n
g
.
T
h
e
ac
cu
r
ac
y
tr
ai
n
i
n
g
is
v
er
y
h
i
g
h
f
o
r
b
o
th
s
t
u
r
ac
tu
r
e.
T
h
e
cu
r
v
e
o
f
t
h
e
tr
ain
i
n
g
i
s
s
h
o
w
n
i
n
th
e
f
o
llo
win
g
Fi
g
u
r
e
1.
(
a)
(
b
)
Fig
u
r
e
1
.
C
u
r
v
e
T
r
ain
i
n
g
o
f
D
y
n
a
m
ic
al
g
o
r
ith
m
w
i
th
XO
R
Fro
m
Fig
u
r
e
1
,
th
e
cu
r
v
e
(
a)
,
is
d
ais
y
q
u
ic
k
l
y
w
it
h
in
d
ex
ep
o
ch
to
m
ee
t
g
lo
b
al
m
i
n
i
m
u
m
.
W
h
ile
th
e
cu
r
v
e
(
b
)
,
th
e
w
e
ig
h
t
tr
ain
in
g
ch
a
n
g
e
n
ea
r
est
1
0
0
0
ep
o
ch
s
,
th
at
m
ea
n
i
n
g
th
e
DB
B
P
L
M
alg
o
r
ith
m
,
it
h
a
s
s
atu
r
atio
n
tr
ai
n
in
g
,
b
u
t
a
f
te
r
th
at,
t
h
e
c
u
r
v
e
tr
ain
i
n
g
i
s
co
n
v
er
g
e
s
q
u
ic
k
l
y
to
o
b
t
ain
t
h
e
m
i
n
i
m
u
m
er
r
o
r
tr
ian
in
g.
4
.
2
.
E
x
peri
m
ent
re
s
ult
o
f
t
he
B
B
P
a
lg
o
rit
h
m
w
it
h XO
R
pro
ble
m
T
h
e
B
B
P
alg
o
r
ith
m
is
tr
ai
n
i
n
g
w
i
th
m
u
n
u
al
v
alu
e
f
o
r
ea
ch
le
ar
n
i
n
g
r
ate
an
d
m
o
m
e
n
t
u
m
f
ac
to
r
f
r
o
m
[
0
,
1
]
.
E
ig
h
t
v
al
u
e
f
o
r
e
ac
h
lear
n
in
g
r
ate
a
n
d
m
o
m
e
n
tu
m
f
ac
to
r
w
er
e
u
s
ed
.
T
h
e
e
x
p
er
i
m
e
n
t
r
e
s
u
lt
s
is
r
ec
o
r
d
e
d
in
th
e
T
ab
le
2
.
T
a
b
l
e
2
.
A
v
e
r
a
g
e
P
e
r
f
o
r
m
a
n
c
e
o
f
B
B
P
a
l
g
o
r
i
t
h
m
w
i
t
h
X
O
R
V
a
l
u
e
s o
f
F
i
r
st
st
r
u
c
t
u
r
e
S
e
c
o
n
d
s
t
r
u
c
t
u
r
e
T
i
me
-
se
c
Ep
o
c
h
T
i
me
-
se
c
Ep
o
c
h
Av
2
3
5
1
.
9
8
5
2
5
1
7
8
2
2
9
2
1
7
2
.
4
9
5
7
5
5
0
2
6
2
2
S
.
D
2
4
7
2
.
5
4
1
3
5
3
1
4
2
0
5
5
.
6
4
6
5
2
5
9
6
.
8
6
8
0
2
9
4
9
9
4
2
1
.
8
4
6
4
Fro
m
T
ab
le
2
,
f
o
r
f
ir
s
t
s
tr
u
c
tu
r
e
th
e
a
v
er
ag
e
tr
ai
n
in
g
ti
m
e
i
s
2
3
5
1
.
9
8
5
2
5
s
ec
o
n
d
s
w
it
h
1
7
8
2
2
9
ep
o
ch
.
Fo
r
th
e
s
ec
o
n
d
s
tr
u
ct
u
r
e
,
th
e
av
er
ag
e
tr
ain
in
g
ti
m
e
2
1
7
2
.
4
9
5
7
5
s
ec
o
n
d
w
it
h
5
0
2
6
2
2
e
p
o
c
h
.
T
h
e
S.D
f
o
r
b
o
th
s
tr
u
ct
u
r
e
is
g
r
ea
ter
t
h
an
o
n
e
.
4
.
3
.
E
x
peri
m
ent
s
re
s
ult
o
f
t
he
DB
B
P
L
M
a
lg
o
rit
hm
w
it
h
B
a
la
nce
-
T
ra
ini
ng
s
et
W
e
i
m
p
le
m
en
t
th
e
DB
B
P
L
M
alg
o
r
it
h
m
u
s
i
n
g
b
alan
ce
-
t
r
ain
in
g
s
et.
T
h
e
ex
p
er
i
m
e
n
ts
r
esu
lt
s
i
s
tab
u
latio
n
in
t
h
e
T
ab
le
3
.
Fro
m
T
ab
le
3
,
f
o
r
f
ir
s
t
s
tr
u
ct
u
r
e
t
h
e
a
v
er
ag
e
tr
ain
i
n
g
ti
m
e
is
2
.
6
0
3
4
s
ec
o
n
d
s
w
it
h
4
5
ep
o
ch
s
.
Fo
r
s
ec
o
n
d
s
tr
u
ctu
r
e
th
e
av
er
ag
e
tr
ain
in
g
ti
m
e
is
3
.
0
1
4
8
s
ec
o
n
d
s
w
it
h
ep
o
ch
is
8
5
ep
o
ch
s
.
B
o
th
s
tr
u
ct
u
r
es
g
av
e
h
i
g
h
ac
c
u
r
ac
y
tr
ain
i
n
g
.
T
h
e
a
v
er
ag
e
S.D
o
f
ti
m
e
f
o
r
b
o
th
s
tr
u
ct
u
r
es
ar
e
l
ess
t
h
an
o
n
e.
T
h
e
cu
r
v
e
o
f
t
h
e
tr
ai
n
in
g
is
s
h
o
w
n
in
th
e
Fig
u
r
e
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
7
,
No
.
4
,
Dec
em
b
er
201
8
:
1
7
0
–
1
7
8
174
T
h
e
tr
ain
in
g
cu
r
v
e
o
f
(
b
)
s
tar
ted
w
it
h
f
l
at
-
s
p
o
t
tr
ain
i
n
g
,
wh
ile
t
h
e
cu
r
v
e
o
f
tr
ai
n
i
n
g
i
n
(
a)
,
s
tar
ted
w
it
h
o
u
t
f
lat
s
p
o
t.
T
h
e
cu
r
v
e
(
a
)
atten
d
to
t
h
e
g
lo
b
al
m
i
n
i
m
u
m
ar
o
u
n
d
3
5
ep
o
ch
s
,
w
h
ile
t
h
e
cu
r
v
e
(
b
)
atte
n
d
to
th
e
g
lo
b
al
m
in
i
m
u
m
a
f
ter
s
p
e
n
d
8
0
ep
o
ch
s
.
A
ls
o
ea
c
h
c
u
r
v
e
(
a)
an
d
(
b
)
h
av
e
d
if
f
er
en
t
ti
m
e
f
o
r
tr
a
in
i
n
g
to
r
ea
ch
th
e
g
lo
b
al
m
i
n
i
m
u
m
.
T
ab
le
3
.
A
v
er
ag
e
t
h
e
P
er
f
o
r
m
an
ce
o
f
DB
B
P
L
M
alg
o
r
it
h
m
w
it
h
b
alan
ce
-
tr
ai
n
i
n
g
s
et
F
i
r
st
st
r
u
c
t
u
r
e
se
c
o
n
d
st
r
u
c
t
u
r
e
Ex
T
i
me
-
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
T
i
me
-
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
Av
2
.
6
0
3
4
45
0
.
9
9
9
9
6
3
.
0
1
4
8
85
0
.
9
8
6
S
.
D
0
.
5
5
0
4
1
-
2
.
8
4
2
E
-
14
4
.
8
9
8
E
-
05
0
.
6
7
7
0
3
0
0
(
a)
(
b
)
Fig
u
r
e
2
.
T
r
ain
in
g
cu
r
v
e
o
f
t
h
e
B
P
alg
o
r
ith
m
w
it
h
b
alan
ce
-
t
r
ain
4
.
3
.
1
.
E
x
peri
m
e
nts o
f
t
he
ba
t
ch
B
P
a
lg
o
rit
h
m
w
it
h B
a
la
nce
-
T
ra
ing
s
et
Sev
er
al
v
al
u
e
f
o
r
ea
ch
an
d
w
er
e
u
s
ed
f
r
o
m
]
0
,
1
]
.
T
h
e
ex
p
er
i
m
en
t
s
r
es
u
lt
s
ar
e
tab
u
lated
in
T
ab
le
4
T
a
b
l
e
4
.
P
e
r
f
o
r
m
a
n
c
e
o
f
b
a
t
c
h
B
P
a
l
g
o
r
i
t
h
m
w
i
t
h
b
a
l
a
n
c
e
-
t
r
a
i
n
s
e
t
s
e
t
V
a
l
u
e
s o
f
F
i
r
st
st
r
u
c
t
u
r
e
S
e
c
o
n
d
s
t
r
u
c
t
u
r
e
T
i
me
-
se
c
Ep
o
c
h
T
i
me
-
se
c
Ep
o
c
h
Av
1
0
6
6
.
5
4
5
3
4
1
6
4
4
3
.
0
4
7
5
4
8
3
4
S
.
D
2
0
2
5
.
9
5
6
1
0
2
3
5
7
7
.
9
6
5
0
9
5
3
2
7
.
7
7
4
8
6
2
9
3
4
3
1
.
4
3
7
7
1
7
Fro
m
T
ab
le
4
,
f
o
r
f
ir
s
t
s
tr
u
ct
u
r
e,
th
e
av
er
g
e
o
f
ti
m
e
i
s
1
0
6
6
.
5
4
5
1
0
6
7
s
w
it
h
a
v
er
ag
e
ep
o
ch
is
3
4
1
6
,
w
h
i
le
th
e
s
ec
o
n
d
s
tr
u
ct
u
r
e
th
e
av
er
g
e
o
f
ti
m
e
tr
ain
i
n
g
i
s
4
4
3
.
0
4
7
5
4
4
3
s
w
it
h
4
8
3
8
ep
o
ch
.
4
.
3
.
2
.
E
x
peri
m
e
nts
re
s
u
lt
o
f
t
he
DB
B
P
L
M
a
lg
o
rit
h
m
w
it
h B
a
la
nce
-
T
esting
s
et
T
h
e
ex
p
er
im
e
n
t
s
r
esu
lt is
tab
u
lated
in
th
e
T
ab
le
5.
T
a
b
l
e
5.
A
v
e
r
a
g
e
t
h
e
P
e
r
f
o
r
m
a
n
c
e
o
f
D
B
B
P
L
M
w
i
t
h
B
a
l
a
n
c
e
-
T
e
s
t
i
n
g
s
e
t
F
i
r
st
st
r
u
c
t
u
r
e
se
c
o
n
d
st
r
u
c
t
u
r
e
Ex
T
i
me
–
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
T
i
me
-
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
Av
4
.
6
9
7
5
92
0
.
9
9
0
8
4
.
5
9
0
6
1
0
4
0
.
9
8
6
0
S
.
D
0
.
7
6
9
5
1
4
4
0
0
0
.
4
1
9
1
7
4
9
0
0
Fro
m
T
ab
le
5
,
f
o
r
f
ir
s
t
s
tr
u
ct
u
r
e
th
e
av
er
ag
e
tr
ain
i
n
g
ti
m
e
i
s
4
.
6
9
7
5
s
ec
o
n
d
s
at
an
av
er
ag
e
ep
o
ch
o
f
is
9
2
ep
o
ch
.
Fo
r
s
ec
o
n
d
s
tr
u
ctu
r
e
th
e
av
er
ag
e
tr
ai
n
i
n
g
ti
m
e
i
s
4
.
4
8
5
0
s
ec
o
n
d
s
at
an
av
er
ag
e
ep
o
ch
o
f
is
1
0
4
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
I
mp
r
o
ve
d
Time
Tr
a
in
in
g
W
ith
A
cc
u
r
a
cy
o
f B
a
tch
B
a
ck
P
r
o
p
a
g
a
tio
n
…
(
Mo
h
a
mme
d
S
a
r
h
a
n
A
l_
Du
a
is
)
175
ep
o
ch
.
B
o
th
s
tr
u
ct
u
r
es
g
a
v
e
h
ig
h
ac
c
u
r
ac
y
tr
ai
n
i
n
g
.
T
h
e
av
er
ag
e
S.D
o
f
ti
m
e
f
o
r
b
o
th
s
tr
u
ctu
r
es
ar
e
n
ea
r
s
t
to
ze
r
o
.
B
o
th
s
tr
u
ct
u
r
es g
a
v
e
h
ig
h
ac
cu
r
ac
y
tr
ain
i
n
g
.
T
h
e
cu
r
v
e
o
f
tr
ain
i
n
g
s
h
o
w
n
in
Fig
u
r
e
3
.
(
a)
(
b
)
Fig
u
r
e
3
.
C
u
r
v
e
T
r
ain
i
n
g
o
f
t
h
e
DB
B
L
M
alg
o
r
ith
m
f
o
r
B
alan
ce
-
T
esti
n
g
s
e
t
4
.
3
.
2
.
E
x
peri
m
e
nts r
esu
lt
o
f
t
he
b
a
t
ch
B
P
a
lg
o
rit
h
m
w
it
h B
a
la
nce
-
T
esting
s
et
W
e
r
u
n
th
e
b
atch
B
P
alg
o
r
ith
m
w
it
h
s
e
v
er
al
m
u
n
al
v
a
lu
e
,
an
d
u
s
ed
th
e
b
alan
ce
-
te
s
tin
g
s
et.
T
h
e
ex
p
er
im
e
n
t
s
r
esu
lt r
ec
o
d
ed
in
th
e
T
ab
le
6
.
T
a
b
l
e
6
.
T
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
t
h
e
t
r
a
i
n
i
n
g
o
f
b
a
t
c
h
B
P
a
l
g
o
r
i
t
h
m
w
i
t
h
B
a
l
a
n
c
e
-
T
e
s
t
i
n
g
s
e
t
V
a
l
u
e
s o
f
F
i
r
st
st
r
u
c
t
u
r
e
S
e
c
o
n
d
s
t
r
u
c
t
u
r
e
T
i
me
-
se
c
Ep
o
c
h
T
i
me
-
se
c
Ep
o
c
h
Av
2
3
5
1
.
5
9
6
8
6
7
2
1
8
1
1
.
0
0
5
5
1
9
0
9
6
S
.
D
2
3
7
7
.
3
2
7
9
1
7
5
.
2
5
3
3
8
1
2
0
4
3
.
0
2
9
1
1
1
7
8
3
5
.
4
9
1
2
Fro
m
T
ab
le
6
,
f
o
r
f
ir
s
t
s
tr
u
ct
u
r
e
th
e
av
er
ag
e
ti
m
e
tr
ain
in
g
is
2
3
5
1
.
5
9
6
s
ec
o
n
d
w
it
h
8
6
7
2
ep
o
ch
.
Fo
r
s
ec
o
n
d
s
tr
u
ct
u
r
e
th
e
a
v
er
ag
e
ti
m
e
is
1
8
1
1
.
0
0
5
5
s
ec
o
n
d
s
w
it
h
1
9
0
9
6
ep
o
ch
.
4
.
3
.
3
.
E
x
peri
m
e
nts D
B
B
L
M
a
lg
o
rit
h
m
w
it
h B
re
a
s
t
-
T
ra
i
nin
g
s
et
W
e
w
ill r
u
n
t
h
e
DB
B
L
M
al
g
o
r
i
th
m
,
t
h
e
e
x
p
er
ien
ce
r
esu
lts
a
r
e
g
iv
e
n
in
t
h
e
T
ab
le
7
.
T
ab
le
7
.
A
v
er
ag
e
t
h
e
p
er
f
o
r
m
an
ce
o
f
DB
B
P
L
M
alg
o
r
it
h
m
w
it
h
b
r
ea
s
t.
-
T
r
ain
in
g
s
et
F
i
r
st
st
r
u
c
t
u
r
e
se
c
o
n
d
st
r
u
c
t
u
r
e
Ex
T
i
me
–
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
T
i
me
-
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
Av
2
.
3
5
6
62
0
.
9
9
9
2
.
3
0
3
4
59
0
.
9
9
8
2
S
.
D
0
.
1
0
7
0
9
6
2
1
0
0
0
.
1
0
6
8
5
3
3
5
0
0
Fro
m
T
ab
le
7
,
ea
s
ily
ca
n
s
ee
p
er
f
o
r
m
an
ce
o
f
DB
B
P
L
M
al
g
o
r
ith
m
.
B
o
t
h
th
e
s
tr
u
ctu
r
e
s
th
e
av
er
ag
e
o
f
th
e
tr
ain
in
g
ti
m
e
is
v
e
r
y
s
h
o
r
t
.
T
h
e
av
er
ag
e
S.D
o
f
ti
m
e
f
o
r
b
o
th
s
tr
u
ct
u
r
es a
r
e
n
ea
r
s
t to
ze
r
o
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
7
,
No
.
4
,
Dec
em
b
er
201
8
:
1
7
0
–
1
7
8
176
(
a)
(
b
)
F
i
g
u
r
e
4
.
C
u
r
v
e
T
r
a
i
n
i
n
g
o
f
t
h
e
D
B
B
L
M
a
l
g
o
r
i
t
h
m
f
o
r
B
r
e
a
s
t
-
T
r
a
i
n
i
n
g
s
e
t
Fig
u
r
e
4
f
r
o
m
b
o
th
s
tr
u
ct
u
r
e,
o
f
th
e
DB
B
P
L
M
alg
o
r
ith
m
t
h
e
tr
ain
in
g
(
a)
an
d
(
b
)
h
av
e
s
m
o
o
th
cu
r
v
e
tr
ain
in
g
.
B
o
th
C
u
r
v
es a
r
e
atten
d
ed
f
a
s
t
w
it
h
i
n
d
ex
ti
m
e
to
th
e
g
lo
b
al
m
i
n
i
m
u
m
.
4
.
3
.
4
.
E
x
peri
m
e
nts r
esu
lt
o
f
t
he
B
B
P
a
l
g
o
rit
h
m
w
it
h B
re
a
s
t
-
T
ra
ini
ng
s
et
W
e
u
s
ed
3
7
4
p
atter
n
s
f
o
r
tr
ain
in
g
s
et
.
T
h
e
r
esu
lt
s
ar
e
s
h
o
wn
in
t
h
e
T
ab
le
8
.
Fr
o
m
T
ab
le
8
f
o
r
f
ir
s
t
s
tr
u
ct
u
r
e
th
e
a
v
er
ag
e
ti
m
e
tr
ai
n
in
g
is
1
5
4
7
.
8
0
7
5
s
ec
o
n
d
w
it
h
1
2
4
3
0
ep
o
ch
s
w
h
il
s
ec
o
n
d
s
tr
u
ctu
r
e
t
h
e
av
er
ag
e
ti
m
e
i
s
1
3
6
1
.
4
8
6
6
6
7
s
ec
o
n
d
s
w
it
h
1
5
9
5
3
.
T
a
b
l
e
8
.
P
e
r
f
o
r
m
a
n
c
e
o
f
B
B
P
a
l
g
o
r
i
t
h
m
w
i
t
h
B
r
e
a
s
t
-
T
r
a
i
n
i
n
g
s
e
t
V
a
l
u
e
s o
f
F
i
r
st
st
r
u
c
t
u
r
e
S
e
c
o
n
d
s
t
r
u
c
t
u
r
e
T
i
me
-
se
c
Ep
o
c
h
T
i
me
-
se
c
Ep
o
c
h
Av
1
5
4
7
.
8
0
7
5
1
2
4
3
0
1
3
6
1
.
4
8
6
6
6
7
1
5
9
5
3
S
.
D
2
0
9
4
.
2
4
7
3
2
9
1
8
6
1
7
.
6
9
2
2
7
1
8
2
9
.
0
9
6
8
0
7
1
5
3
8
5
.
1
8
2
8
4
4
.
3
.
5
.
E
x
peri
m
e
nts D
B
B
P
L
M
a
lg
o
rit
hm
w
it
h B
re
a
s
t
-
T
esting
s
et
Fro
m
T
ab
le
9
,
th
e
d
y
n
a
m
ic
t
r
ain
in
g
r
ate
an
d
m
o
m
en
t
u
m
f
ac
to
r
h
elp
s
t
h
e
DB
B
P
L
M
alg
o
r
ith
m
f
o
r
r
ed
u
cin
g
t
h
e
ti
m
e
tr
ain
i
n
g
.
B
o
th
th
e
s
tr
u
c
tu
r
es
,
th
e
av
er
ag
e
o
f
th
e
tr
ain
in
g
t
i
m
e
is
v
er
y
s
h
o
r
t
.
Fo
r
f
ir
s
t
s
tr
u
c
t
u
r
e
th
e
a
v
er
ag
e
ti
m
e
i
s
0
.
8
4
4
s
ec
o
n
d
s
w
ith
a
v
er
ag
e
3
3
ep
o
ch
s
,
w
h
ile
th
e
s
co
n
d
s
tr
u
ctu
r
e
t
h
e
av
er
a
g
e
ti
m
e
i
s
1
.
6
1
7
7
w
it
h
av
er
a
g
e
6
1
ep
o
ch
s
.
T
a
b
l
e
9
.
A
v
e
r
a
g
e
t
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
D
B
B
P
L
M
a
l
g
o
r
i
t
h
m
w
i
t
h
B
r
e
a
s
t
-
T
e
s
t
i
n
g
s
et
F
i
r
st
st
r
u
c
t
u
r
e
se
c
o
n
d
st
r
u
c
t
u
r
e
Ex
T
i
me
-
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
T
i
me
-
se
c
Ep
o
c
h
A
c
c
u
r
a
c
y
T
r
a
i
n
i
n
g
Av
0
.
8
4
4
33
0
.
9
4
4
2
0
6
1
.
6
1
7
7
61
0
.
9
8
7
S
.
D
1
.
1
1
0
2
E
-
16
0
0
0
.
0
9
2
1
7
4
8
8
0
0
4
.
3
.
6
.
E
x
peri
m
e
nts r
esu
lt
s
o
f
B
B
P
a
lg
o
rit
hm
w
it
h B
re
a
s
t
-
T
esting
s
et
W
e
u
s
ed
2
5
1
p
atter
n
s
f
o
r
te
s
tin
g
th
e
p
er
f
o
r
m
a
n
ce
o
f
B
B
P
alg
o
r
ith
m
.
T
h
e
ex
p
er
m
en
t
s
r
esu
lt
is
tab
u
lated
in
t
h
e
T
ab
le
1
0
.
T
a
b
l
e
1
0
.
P
e
r
f
o
r
m
a
n
c
e
o
f
B
B
P
a
l
g
o
r
i
t
h
m
w
i
t
h
B
r
e
a
s
t
-
T
e
s
t
i
n
g
s
e
t
V
a
l
u
e
s o
f
F
i
r
st
st
r
u
c
t
u
r
e
S
e
c
o
n
d
st
r
u
c
t
u
r
e
T
i
me
-
se
c
Ep
o
c
h
T
i
me
-
se
c
Ep
o
c
h
Av
1
7
4
1
.
0
1
7
7
1
4
1
7
7
8
5
.
4
2
8
5
7
1
9
2
0
.
9
8
4
1
4
3
1
0
7
0
9
S
.
D
2
3
3
9
.
4
7
0
1
1
9
1
5
5
1
5
.
2
9
4
0
8
2
0
1
3
.
9
5
2
5
4
7
9
7
8
1
.
1
9
2
9
8
9
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
I
mp
r
o
ve
d
Time
Tr
a
in
in
g
W
ith
A
cc
u
r
a
cy
o
f B
a
tch
B
a
ck
P
r
o
p
a
g
a
tio
n
…
(
Mo
h
a
mme
d
S
a
r
h
a
n
A
l_
Du
a
is
)
177
Fo
r
m
th
e
T
ab
le
1
0
,
th
e
r
an
g
e
o
f
th
e
tr
ain
i
n
g
ti
m
e
f
o
r
b
o
th
s
tr
u
ct
u
r
e
is
1
0
0
.
3
1
2
0
≤
t
≤
6
3
0
0
s
ec
o
n
d
s
an
d
6
0
.
1
6
7
0
s
e
c
on
ds
≤
t
≤
4560
s
ec
o
n
d
s
,
th
is
m
ea
n
s
t
h
e
r
an
g
e
o
f
ti
m
e
tr
ai
n
in
g
i
s
w
id
el
y
ti
m
e
tr
ai
n
i
n
g
5.
DIS
CU
SS
I
O
N
T
O
VA
L
I
DA
T
E
T
H
E
P
E
RF
O
RM
ANCE O
F
I
M
P
RO
VE
D
AL
G
O
RI
T
H
M
T
o
v
alid
ate
th
e
ef
f
icie
n
c
y
o
f
th
e
i
m
p
r
o
v
ed
alg
o
r
it
h
m
,
th
r
o
u
g
h
co
m
p
ar
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
DB
B
P
ML
alg
o
r
ith
m
w
it
h
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
b
atc
h
B
P
alg
o
r
ith
m
b
ased
o
n
ce
r
tai
n
cr
i
ter
ia
.
W
e
ca
lcu
late
th
e
s
p
ee
d
u
p
tr
ain
i
n
g
u
s
in
g
t
h
e
f
o
llo
w
in
g
f
o
r
m
u
la
[
2
0
]
:
Sp
ee
d
u
p
=
Execut
i
o
n
t
i
m
e
of
ℎ
Execut
i
o
n
t
i
m
e
of
ℎ
5.
1
.
P
r
o
ess
ing
T
i
m
e
o
f
DB
B
P
L
M
Alg
o
rit
hm
Ver
s
u
s
t
he
B
B
P
Alg
o
ri
t
h
m
f
o
r
w
it
h diff
er
ent
Str
uct
ure
T
o
v
alid
ate
th
e
i
m
p
r
o
v
ed
alg
o
r
ith
m
o
r
DB
B
P
L
M
alg
o
r
ith
m
,
w
e
co
m
p
ar
e
th
e
p
er
f
o
r
m
a
n
ce
b
et
w
ee
n
th
e
DB
B
P
L
M
alg
o
r
ith
m
a
n
d
th
e
B
B
P
alg
o
r
ith
m
.
T
h
e
s
p
ee
d
-
u
p
o
b
tain
ed
in
tr
a
in
in
g
is
s
h
o
w
n
i
n
T
ab
le
1
1
.
T
ab
le
1
1
.
Sp
ee
d
u
p
th
e
DB
B
PL
M
al
g
o
r
ith
m
v
er
s
u
s
B
B
P
w
it
h
d
if
f
er
en
t
s
tr
u
ct
u
r
e
F
i
r
st
st
r
u
c
t
u
r
e
S
e
c
o
n
d
s
t
r
u
c
t
u
r
e
D
B
B
P
L
M
a
l
g
o
r
i
t
h
m
B
B
P
a
l
g
o
r
i
t
h
m
S
p
e
e
d
u
p
R
a
t
e
(
B
B
P
/
D
B
B
P
L
M
)
D
B
B
P
L
M
a
l
g
o
r
i
t
h
m
B
B
P
a
l
g
o
r
i
t
h
m
S
p
e
e
d
u
p
R
a
t
e
(
B
B
P
/
D
B
B
P
L
M
)
A
V
t
i
me
-
sc
A
V
t
i
me
-
sc
A
V
t
i
me
-
sc
XOR
1
.
9
5
6
9
2
3
5
1
.
9
8
5
1
2
0
1
.
8
9
3
1
.
6
2
7
2
1
7
2
.
4
9
5
1
3
3
5
.
5
2
3
B
a
l
a
n
c
e
T
r
a
i
n
i
n
g
B
a
l
a
n
c
e
T
e
st
i
n
g
2
.
6
0
3
4
1
0
6
6
.
5
4
5
4
0
9
.
6
7
4
3
.
0
1
5
4
4
3
.
0
4
7
1
4
6
.
9
5
8
4
.
6
9
7
5
2
3
0
1
.
5
9
6
4
8
9
.
9
6
1
9
4
.
5
9
1
1
7
5
6
.
0
0
5
3
8
2
.
5
2
2
B
r
e
a
st
T
r
a
i
n
i
n
g
B
r
e
a
st
T
e
st
i
n
g
2
.
3
5
6
1
5
4
7
.
8
0
8
6
5
6
.
9
6
4
1
2
.
3
0
3
1
3
6
1
.
4
8
7
5
9
1
.
0
7
7
1
0
.
8
4
4
1
7
4
1
.
0
1
8
2
0
6
2
.
8
1
7
1
.
6
1
8
1
9
0
0
.
9
8
4
1
1
7
5
.
1
1
5
Fro
m
T
ab
le
1
1
,
it is
ev
id
en
t t
h
at
th
e
d
y
n
a
m
ic
al
g
o
r
ith
m
p
r
o
v
id
es su
p
er
io
r
p
er
f
o
r
m
an
ce
o
v
er
th
e
B
B
P
alg
o
r
ith
m
f
o
r
all
d
atasets
with
b
o
th
s
tr
u
ctu
r
e.
Ho
w
ev
er
f
o
r
f
ir
s
t
s
tr
u
c
tu
r
e,
th
e
DB
B
P
L
M
alg
o
r
ith
m
i
s
2
0
6
2
.
8
1
7
2
0
6
3
s
tim
e
s
f
a
s
ter
th
an
t
h
e
B
B
P
alg
o
r
ith
m
at
m
ax
i
m
u
m
tr
ain
i
n
g
,
an
d
also
th
e
DB
B
P
L
M
alg
o
r
ith
m
is
4
0
5
.
7
3
8
4
0
6
s
tim
es
f
aster
th
a
n
th
e
B
B
P
alg
o
r
ith
m
at
m
in
i
m
u
m
tr
ain
i
n
g
.
Fo
r
s
ec
o
n
d
s
tr
u
ct
u
r
e
T
h
e
DB
B
P
L
M
alg
o
r
it
h
m
is
1
3
3
5
.
5
2
3
1
3
3
6
tim
es
f
as
ter
th
a
n
t
h
e
B
B
P
alg
o
r
ith
m
at
m
a
x
i
m
u
m
tr
ain
i
n
g
,
an
d
also
th
e
DB
B
P
L
M
alg
o
r
it
h
m
i
s
1
4
6
.
9
5
8
1
4
7
s
tim
e
s
f
a
s
ter
th
an
th
e
B
B
P
alg
o
r
ith
m
a
t
m
i
n
i
m
u
m
tr
ai
n
i
n
g
.
6.
E
VA
L
UA
T
I
O
N
O
F
T
H
E
P
E
RF
O
RM
ANCE O
F
I
M
P
R
O
VE
D
B
AT
CH
B
P
AL
G
O
R
I
T
H
M
T
o
ev
alu
ated
th
e
p
er
f
o
r
m
an
ce
s
o
f
t
h
e
i
m
p
r
o
v
ed
al
g
o
r
ith
m
o
r
DB
B
P
ML
alg
o
r
it
h
m
f
o
r
s
p
ee
d
in
g
u
p
tr
ain
i
n
g
w
h
ic
h
p
r
ese
n
ted
i
n
th
is
s
t
u
d
y
.
T
h
e
p
er
f
o
r
m
a
n
ce
s
o
f
t
h
e
DB
B
P
ML
al
g
o
r
ith
m
ar
e
co
m
p
ar
ed
to
p
r
ev
io
u
s
r
e
s
ea
r
ch
w
o
r
k
s
[
1
3
]
[
1
6
]
.
T
h
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
i
m
p
r
o
v
e
al
g
o
r
ith
m
w
h
ic
h
p
r
o
p
o
s
ed
in
t
h
i
s
s
tu
d
y
g
iv
e
s
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
th
an
ex
i
s
ts
w
o
r
k
s
.
7.
CO
NCLU
SI
O
N
T
h
is
p
ap
er
in
tr
o
d
u
ce
d
th
e
D
B
B
P
L
M
alg
o
r
ith
m
,
w
h
ich
tr
a
in
s
b
y
a
d
y
n
a
m
ic
f
u
n
ct
io
n
f
o
r
ea
ch
th
e
lear
n
in
g
r
ate
an
d
m
o
m
e
n
t
u
m
f
ac
to
r
.
T
h
is
f
u
n
c
tio
n
in
f
l
u
e
n
c
es
o
n
th
e
w
ei
g
h
t
f
o
r
ea
ch
h
id
d
en
la
y
er
an
d
o
u
tp
u
t
la
y
er
.
Fro
m
ex
p
er
i
m
en
ts
r
es
u
l
tin
g
t
h
e
DB
B
P
L
M
al
g
o
r
ith
m
g
iv
e
s
s
u
p
er
io
r
tr
ain
in
g
t
h
a
n
B
B
P
alg
o
r
ith
m
f
o
r
all
d
ata
s
et,
w
it
h
b
o
th
s
tr
u
ct
u
r
e.
On
e
o
f
t
h
e
m
ai
n
ad
v
an
ta
g
e
s
o
f
th
e
d
y
n
a
m
ic
tr
ai
n
in
g
is
th
at
i
t
r
ed
u
ce
s
t
h
e
tr
ain
i
n
g
ti
m
e
an
d
r
ed
u
ce
s
t
h
e
er
r
o
r
tr
ain
in
g
,
n
u
m
b
er
o
f
ep
o
ch
s
a
n
d
e
n
h
a
n
ce
m
e
n
t
t
h
e
ac
cu
r
ac
y
o
f
th
e
tr
ain
i
n
g
.
T
h
e
p
er
f
o
r
m
an
ce
o
f
DB
B
P
L
M
alg
o
r
ith
m
w
h
ich
p
r
ese
n
ted
in
th
i
s
s
t
u
d
y
g
av
e
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
co
m
p
ar
e
w
it
h
e
x
is
t
s
w
o
r
k
.
RE
F
E
RE
NCES
[1
]
R.
Ka
laiv
a
n
i,
K.S
u
d
h
a
g
a
r
K,
L
a
k
sh
m
i
P
.
Ne
u
ra
l
Ne
tw
o
rk
b
a
se
d
V
ib
ra
ti
o
n
Co
n
tro
l
f
o
r
V
e
h
icle
A
c
ti
v
e
S
u
sp
e
n
sio
n
S
y
st
e
m
.
In
d
ia
n
J
o
u
rn
a
l
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
.
9
(1
)
,
2
0
1
6
.
[2
]
P
.
M
o
a
ll
e
m
.
I
m
p
ro
v
in
g
Ba
c
k
‐P
ro
p
a
g
a
ti
o
n
V
IA
a
n
e
ff
icie
n
t
Co
m
b
in
a
ti
o
n
o
f
A
S
a
tu
ra
ti
o
n
S
u
p
p
r
e
ss
io
n
M
e
th
o
d
.
Ne
u
ra
l
Ne
two
rk
W
o
rld
.
2
0
(2
)
,
2
0
1
0
.
[3
]
l.
v
Ka
m
b
le,
D.R
P
a
n
g
a
v
h
a
n
e
,
&
T
.
P
S
i
n
g
h
,
Im
p
ro
v
in
g
t
h
e
P
e
rf
o
rm
a
n
c
e
o
f
Ba
c
k
-
P
ro
p
a
g
a
ti
o
n
T
ra
in
in
g
A
lg
o
rit
h
m
b
y
Us
in
g
AN
N.
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Co
m
p
u
ter
,
El
e
c
trica
l,
A
u
to
ma
ti
o
n
,
Co
n
tro
l
a
n
d
In
f
o
rm
a
ti
o
n
E
n
g
i
n
e
e
rin
g
,
9
(1
),
1
8
7
-
1
9
2
,
2
0
1
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
7
,
No
.
4
,
Dec
em
b
er
201
8
:
1
7
0
–
1
7
8
178
[4
]
J.M
.
Riz
w
a
n
,
P
N.Krish
n
a
n
,
R.
Ka
rth
ik
e
y
a
n
,
S
R.
Ku
m
a
r.
M
u
lt
i
la
y
e
r
p
e
rc
e
p
ti
o
n
ty
p
e
a
rti
f
icia
l
n
e
u
ra
l
n
e
tw
o
rk
b
a
se
d
traff
ic co
n
tro
l.
I
n
d
i
a
n
J
o
u
r
n
a
l
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
9
(
5
),
2
0
1
6
.
[5
]
R.
Ka
laiv
a
n
i,
K.
S
u
d
h
a
g
a
r
,
P
.
L
a
k
sh
m
i,
Ne
u
ra
l
Ne
tw
o
rk
b
a
se
d
V
ib
ra
ti
o
n
Co
n
tr
o
l
f
o
r
V
e
h
icle
A
c
ti
v
e
S
u
sp
e
n
sio
n
S
y
st
e
m
.
In
d
ia
n
J
o
u
rn
a
l
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
9
(1
)
,
2
0
1
6
.
[6
]
M
.
S
.
A
l_
Du
a
is,
&
F
.
S
.
M
o
h
a
m
a
d
,
A
Re
v
ie
w
o
n
En
h
a
n
c
e
m
e
n
ts
to
S
p
e
e
d
u
p
T
ra
in
in
g
o
f
th
e
B
a
tch
Ba
c
k
P
r
o
p
a
g
a
ti
o
n
A
lg
o
rit
h
m
,
In
d
ia
n
J
o
u
rn
a
l
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
lo
g
y
,
9
(4
6
),
1
-
1
0
,
2
0
1
6
.
[7
]
H..
M
o
,
J.W
a
n
g
,
H.
Niu
,
Ex
p
o
n
e
n
t
b
a
c
k
p
ro
p
a
g
a
ti
o
n
n
e
u
ra
l
n
e
t
w
o
rk
f
o
re
c
a
stin
g
f
o
r
f
in
a
n
c
ial
c
ro
ss
-
c
o
rre
latio
n
re
latio
n
sh
i
p
.
Exp
e
rt S
y
ste
ms
wit
h
Ap
p
li
c
a
ti
o
n
s
,
5
3
,
1
0
6
-
1
0
1
6
,
2
0
1
6
.
[8
]
Q.A
b
b
a
s
,
F
.
A
h
m
a
d
,
M
.
Im
ra
n
,
V
a
riab
le
lea
rn
in
g
ra
te
b
a
se
d
m
o
d
if
ica
ti
o
n
i
n
b
a
c
k
p
ro
p
a
g
a
ti
o
n
a
lg
o
ri
t
h
m
(M
BP
A
)
o
f
a
rti
f
icia
l
n
e
u
ra
l
n
e
tw
o
rk
f
o
r
d
a
ta cla
ss
i
f
ica
ti
o
n
.
S
c
ien
c
e
I
n
ter
n
a
t
io
n
a
l
,
2
8
(
3
),
2
3
6
9
-
2
3
7
8
,
2
0
1
6
.
[9
]
W
u
S
X
,
L
u
o
DL
,
Zh
o
u
ZW
,
Ca
i
JH
,
S
h
i
YX
.
A
k
in
d
o
f
BP
n
e
u
ra
l
n
e
tw
o
rk
a
lg
o
rit
h
m
b
a
s
e
d
o
n
g
re
y
in
terv
a
l.
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
S
y
ste
ms
S
c
ien
c
e
,
4
2
(3
)
,
3
8
9
-
9
6
,
2
0
1
1
.
[1
0
]
H.
A
z
a
m
i,
S
.
S
a
n
e
i
,
M
o
h
a
m
m
a
d
i
K.
Im
p
ro
v
in
g
th
e
n
e
u
ra
l
n
e
tw
o
rk
train
in
g
f
o
r
f
a
c
e
re
c
o
g
n
it
io
n
u
sin
g
a
d
a
p
ti
v
e
lea
rn
in
g
ra
te,
re
sili
e
n
t
b
a
c
k
p
ro
p
a
g
a
ti
o
n
a
n
d
c
o
n
j
u
g
a
te
g
ra
d
ien
t
a
lg
o
rit
h
m
.
J
o
u
rn
a
l
o
f
Co
m
p
u
te
r
Ap
p
li
c
a
ti
o
n
s
,
3
4
(
2
):2
2
-
6
2
0
1
1
.
[1
1
]
J.
G
e
,
J.S
h
a
,
&
Y.F
a
n
g
,
A
n
n
e
w
b
a
c
k
p
ro
p
a
g
a
ti
o
n
a
lg
o
rit
h
m
w
it
h
c
h
a
o
ti
c
lea
rn
i
n
g
ra
te.
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
S
o
f
twa
re
En
g
i
n
e
e
rin
g
a
n
d
S
e
rv
ice
S
c
ien
c
e
s
,
1
6
,
4
0
4
-
4
0
7
,
2
0
1
5
.
[1
2
]
J.
G
u
,
G
.
Yin
,
P
.
Hu
a
n
g
,
J.
G
u
o
,
L
.
Ch
e
n
,
A
n
im
p
ro
v
e
d
b
a
c
k
p
ro
p
a
g
a
ti
o
n
n
e
u
ra
l
n
e
tw
o
rk
p
re
d
ictio
n
m
o
d
e
l
f
o
r
su
b
su
rf
a
c
e
d
rip
irri
g
a
ti
o
n
sy
ste
m
.
Co
mp
u
ter
s
a
n
d
El
e
c
trica
l
E
n
g
in
e
e
rin
g
,
1
-
8
,
2
0
1
7
.
[1
3
]
A
.
A
.
Ha
m
e
e
d
,
B.
Ka
rli
k
,
M
.
S
.
S
a
l
m
a
n
,
Ba
c
k
-
p
ro
p
a
g
a
ti
o
n
a
lg
o
rit
h
m
w
it
h
v
a
riab
le
a
d
a
p
ti
v
e
m
o
m
e
n
tu
m
.
Kn
o
wle
d
g
e
-
Ba
se
d
S
y
ste
ms
,
1
4
1
,
7
9
–
87
,
2
0
1
6
.
[1
4
]
E.
No
e
rsa
so
n
g
k
o
,
F
.
T
.
Ju
lf
ia,
A
.
S
y
u
k
u
r,
R.
A
.
P
ra
m
u
n
e
n
d
a
r,
&
C
.
S
u
p
riy
a
n
to
,
A
to
u
rism
a
rriv
a
l
f
o
re
c
a
stin
g
u
sin
g
g
e
n
e
ti
c
a
lg
o
rit
h
m
b
a
se
d
n
e
u
ra
l
n
e
tw
o
rk
.
In
d
ia
n
J
o
u
rn
a
l
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
9
(4
)
,
2
0
1
6
.
[1
5
]
W
.
Zh
a
n
g
,
Z.
L
i
,
W
.
X
u
,
H.Z
h
o
u
,
A
c
la
ss
if
ier
o
f
sa
telli
te
sig
n
a
ls
b
a
se
d
o
n
th
e
b
a
c
k
-
p
r
o
p
a
g
a
ti
o
n
n
e
u
ra
l
n
e
two
rk
.
I
n
8
th
I
n
tern
a
ti
o
n
a
l
C
o
n
g
re
ss
o
n
Im
a
g
e
a
n
d
S
ig
n
a
l
P
ro
c
e
ss
in
g
(CIS
P
)
,
1
3
5
3
-
1
3
5
7
,
2
0
1
5
.
[1
6
]
H..
A
z
a
m
i,
&
J.E
sc
u
d
e
ro
,
A
c
o
mp
a
ra
t
ive
stu
d
y
o
f
b
re
a
st
c
a
n
c
e
r
d
ia
g
n
o
sis
b
a
se
d
o
n
n
e
u
ra
l
n
e
two
rk
e
n
se
mb
le
v
ia
imp
ro
v
e
d
tra
i
n
in
g
a
l
g
o
rit
h
ms
.
P
r
o
c
e
e
d
in
g
s
in
3
7
th
A
n
n
u
a
l
In
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
f
th
e
IEE
E
En
g
in
e
e
rin
g
in
M
e
d
icin
e
a
n
d
Bi
o
lo
g
y
S
o
c
iet
y
(EM
BC)
,
2
0
1
5
,
2
8
3
6
-
2
8
3
9
,
2
0
1
5
.
[1
7
]
L
.
.
Ru
i,
Y.
X
i
o
n
g
,
X
iao
,
K.,
&
Qiu
,
X
.
BP
n
e
u
r
a
l
n
e
two
rk
-
b
a
se
d
we
b
se
rv
ice
se
lec
ti
o
n
a
lg
o
rith
m
in
th
e
sm
a
rt
d
istrib
u
ti
o
n
g
ri
d
.
p
ro
c
e
e
d
i
n
g
s
1
6
t
h
A
sia
-
P
a
c
if
ic
In
Ne
t
w
o
rk
Op
e
ra
t
io
n
s
a
n
d
M
a
n
a
g
e
m
e
n
t
S
y
m
p
o
siu
m
(
A
P
NO
M
S
),
1
-
4
,
2
0
1
4
.
[1
8
]
D,
Yo
n
g
h
a
o
Z.
P
e
n
g
,
Yu
m
in
g
S
,
S
a
n
y
u
a
n
Z
.
Imp
ro
v
e
me
n
ts
o
f
c
o
e
ff
icie
n
t
lea
rn
i
n
g
in
BP
NN
fo
r
im
a
g
e
re
sto
ra
ti
o
n
re
sto
ra
ti
o
n
ICS
AI
.
I
n
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
S
y
ste
m
s,
Ya
n
tai,
2
6
9
2
-
9
4
,
2
0
1
2
.
[1
9
]
N.
M
,
Na
w
i,
N.
A
Ha
m
id
,
R.
S
,
Ra
n
sin
g
,
R,
G
h
a
z
a
li
&
M
.
N.
S
a
ll
e
h
,
En
h
a
n
c
i
n
g
Ba
c
k
P
ro
p
a
g
a
ti
o
n
Ne
u
ra
l
Ne
tw
o
rk
A
l
g
o
rit
h
m
w
it
h
A
d
a
p
ti
v
e
Ga
in
o
n
Clas
sif
ic
a
ti
o
n
P
r
o
b
lem
s
”
.
Ne
two
rk
s
,
v
o
l.
4
,
n
o
.
2
,
2
0
1
1
.
[2
0
]
H.S
a
k
i
,
A
.
T
a
h
m
a
sb
i
,
H
.
S
a
lt
a
n
ian
-
Zad
a
h
,
a
S
.
B
S
h
o
k
o
u
h
i
.
F
a
st
o
p
p
o
site
w
e
ig
h
t
lea
rn
in
g
ru
les
w
it
h
a
p
p
li
c
a
ti
o
n
in
b
re
a
st ca
n
c
e
r.
Co
mp
u
ter
s i
n
b
io
l
o
g
y
a
n
d
me
d
ici
n
e
.
4
3
(1
)
:
32
-
4
1
,
2
0
1
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.