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.
15
,
No
.
2
,
A
p
r
il
20
25
,
p
p
.
2
0
2
2
~
2
0
3
0
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
2
.
pp
2
0
2
2
-
2
0
3
0
2022
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Enha
ncing
spa
tiotempo
ra
l weath
er
f
o
recas
ting a
ccu
r
a
cy
with
3
D
co
nv
o
lutiona
l kernel
throug
h t
h
e
sequenc
e t
o
se
q
uence
m
o
del
Rena
ldy
F
re
dy
a
n
1
,
K
a
rli
E
ka
Set
ia
wa
n
1
,
K
elv
in Asclepiu
s
M
ino
r
2
1
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
D
e
p
a
r
t
me
n
t
,
S
c
h
o
o
l
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
B
i
n
a
N
u
s
a
n
t
a
r
a
U
n
i
v
e
r
si
t
y
,
J
a
k
a
r
t
a
,
I
n
d
o
n
e
si
a
2
M
a
t
h
e
ma
t
i
c
s De
p
a
r
t
me
n
t
,
S
c
h
o
o
l
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
B
i
n
a
N
u
sa
n
t
a
r
a
U
n
i
v
e
r
s
i
t
y
,
J
a
k
a
r
t
a
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
2
9
,
2
0
2
4
R
ev
is
ed
No
v
2
5
,
2
0
2
4
Acc
ep
ted
Dec
2
,
2
0
2
4
Ac
c
u
ra
te
we
a
th
e
r
fo
re
c
a
stin
g
i
s
imp
o
rta
n
t
w
h
e
n
d
e
a
li
n
g
wit
h
v
a
rio
u
s
se
c
to
rs,
su
c
h
a
s
re
tail,
a
g
ricu
l
tu
r
e
,
a
n
d
a
v
iatio
n
,
e
sp
e
c
iall
y
d
u
ri
n
g
e
x
trem
e
we
a
th
e
r
e
v
e
n
ts
li
k
e
h
e
a
t
wa
v
e
s,
d
ro
u
g
h
ts,
a
n
d
sto
rm
s
to
p
re
v
e
n
t
d
isa
ste
r
imp
a
c
t.
Trad
it
i
o
n
a
l
m
e
th
o
d
s
re
l
y
o
n
c
o
m
p
lex
,
p
h
y
sic
s
-
b
a
se
d
m
o
d
e
ls
to
p
re
d
ict
t
h
e
Eart
h
'
s
sto
c
h
a
stic
sy
ste
m
s.
Ho
we
v
e
r,
s
o
m
e
tec
h
n
o
l
o
g
ica
l
a
d
v
a
n
c
e
m
e
n
ts
a
n
d
t
h
e
a
v
a
il
a
b
i
li
t
y
o
f
e
x
te
n
siv
e
sa
telli
te
d
a
ta
fr
o
m
b
e
y
o
n
d
Earth
h
a
v
e
e
n
h
a
n
c
e
d
m
e
teo
ro
l
o
g
ica
l
p
re
d
icti
o
n
s
a
n
d
se
n
t
t
h
e
m
to
Eart
h
'
s
a
n
ten
n
a
e
.
De
e
p
lea
rn
i
n
g
m
o
d
e
ls
u
sin
g
th
is
h
ist
o
rica
l
d
a
ta
s
h
o
w
p
ro
m
ise
in
imp
ro
v
i
n
g
fo
re
c
a
st
a
c
c
u
ra
c
y
to
e
n
h
a
n
c
e
h
o
w
m
o
d
e
ls
lea
rn
t
h
e
d
a
ta
p
a
tt
e
rn
.
Th
is
stu
d
y
in
tro
d
u
c
e
s
a
n
o
v
e
l
a
rc
h
it
e
c
tu
re
,
c
o
n
v
o
lu
t
io
n
a
l
se
q
u
e
n
c
e
t
o
se
q
u
e
n
c
e
(Co
n
v
S
e
q
2
S
e
q
)
n
e
tw
o
r
k
,
wh
ich
e
m
p
l
o
y
s
3
D
c
o
n
v
o
lu
t
io
n
a
l
n
e
u
ra
l
n
e
two
rk
s
(CNN
)
to
a
d
d
re
ss
th
e
c
h
a
ll
e
n
g
e
s
o
f
sp
a
ti
o
tem
p
o
ra
l
f
o
re
c
a
stin
g
.
Un
li
k
e
re
c
u
rre
n
t
n
e
u
ra
l
n
e
two
r
k
(RNN
)
--
b
a
se
d
m
o
d
e
ls,
wh
ich
a
re
ti
m
e
-
c
o
n
su
m
in
g
d
u
e
to
se
q
u
e
n
t
ial
p
r
o
c
e
ss
in
g
,
3
D
CNN
s
c
a
p
tu
re
sp
a
ti
a
l
c
o
n
tex
t
m
o
re
e
fficie
n
tl
y
.
Co
n
v
S
e
q
2
S
e
q
o
v
e
rc
o
m
e
s
th
e
li
m
it
a
ti
o
n
s
o
f
trad
it
io
n
a
l
CNN
m
o
d
e
ls
b
y
e
n
su
ri
n
g
c
a
u
sa
l
c
o
n
stra
in
ts
a
n
d
g
e
n
e
ra
ti
n
g
flex
i
b
le
len
g
t
h
o
u
t
p
u
t
se
q
u
e
n
c
e
s.
O
u
r
e
x
p
e
rime
n
tal
re
su
lt
s
d
e
m
o
n
stra
te
t
h
a
t
C
o
n
v
S
e
q
2
S
e
q
o
u
t
p
e
rfo
rm
s
trad
i
ti
o
n
a
l
a
n
d
m
o
d
e
rn
RNN
-
b
a
se
d
a
rc
h
it
e
c
tu
re
s
in
b
o
th
p
re
d
ictio
n
a
c
c
u
ra
c
y
a
n
d
t
ime
e
fficie
n
c
y
,
lev
e
ra
g
i
n
g
h
isto
rica
l
m
e
teo
ro
l
o
g
ica
l
d
a
ta
to
p
r
o
v
i
d
e
a
r
o
b
u
st
so
lu
ti
o
n
fo
r
we
a
th
e
r
fo
re
c
a
stin
g
a
p
p
li
c
a
ti
o
n
s.
Th
e
p
ro
p
o
se
d
a
rc
h
i
tec
tu
re
o
u
tp
e
rf
o
rm
s
th
e
p
re
v
io
u
s
m
e
th
o
d
,
g
iv
i
n
g
n
e
w
in
sig
h
t
wh
e
n
d
e
a
li
n
g
wit
h
sp
a
t
i
o
tem
p
o
ra
l
with
h
ig
h
d
e
n
sity
.
K
ey
w
o
r
d
s
:
C
o
n
v
o
lu
tio
n
al
lay
e
r
Data
C
HI
R
PS
R
ain
f
all
an
aly
s
is
Seq
2
Seq
m
o
d
el
Sp
atio
tem
p
o
r
al
f
o
r
ec
asti
n
g
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
R
en
ald
y
Fre
d
y
an
C
o
m
p
u
ter
Scien
ce
Dep
ar
tm
e
n
t,
Sch
o
o
l o
f
C
o
m
p
u
ter
Scien
ce
,
B
in
a
Nu
s
an
tar
a
Un
iv
er
s
ity
Keb
o
n
J
er
u
k
R
ay
a
Stre
et
2
7
,
J
ak
ar
ta,
I
n
d
o
n
esia
E
m
ail:
r
en
ald
y
.
f
r
ed
y
a
n
@
b
in
u
s
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
I
n
s
itu
atio
n
s
o
f
s
ev
er
e
n
atu
r
al
o
cc
u
r
r
e
n
ce
s
lik
e
h
ea
t
wa
v
es
(
h
ig
h
tem
p
er
atu
r
es),
d
r
o
u
g
h
ts
,
an
d
s
to
r
m
s
,
r
eso
u
r
ce
p
lan
n
i
n
g
d
ep
en
d
s
cr
itically
o
n
wea
th
er
f
o
r
ec
asti
n
g
.
I
t
also
af
f
ec
ts
d
ec
is
i
o
n
-
m
ak
i
n
g
in
r
etail
m
ar
k
ets,
ag
r
icu
ltu
r
e,
a
v
iatio
n
,
an
d
o
th
e
r
in
d
u
s
tr
ies
s
in
ce
b
ad
wea
th
e
r
lo
wer
s
co
r
p
o
r
ate
r
ev
e
n
u
es
[
1
]
.
Me
teo
r
o
lo
g
ical
v
ar
iab
le
f
o
r
ec
asts
ar
e
b
ec
o
m
i
n
g
in
cr
ea
s
in
g
l
y
ac
cu
r
ate
d
u
e
t
o
tech
n
ical
a
d
v
an
ce
m
e
n
ts
o
v
er
tim
e.
Nev
er
th
eless
,
co
n
v
e
n
tio
n
al
f
o
r
ec
asti
n
g
n
ee
d
s
s
o
p
h
i
s
ticated
,
p
h
y
s
ics
-
b
ased
m
o
d
e
ls
to
f
o
r
ec
ast
th
e
wea
th
er
b
ec
au
s
e
o
f
th
e
s
to
ch
a
s
tic
d
em
ea
n
o
r
o
f
t
h
e
E
ar
th
s
y
s
tem
s
,
wh
ich
ar
e
co
n
tr
o
lled
b
y
p
h
y
s
ical
laws
[
2
]
.
His
to
r
ical
d
ata
i
s
av
ailab
le.
H
en
ce
,
r
esear
ch
er
s
ca
n
cr
ea
te
d
ee
p
-
lear
n
in
g
m
o
d
els
th
at
ca
n
f
o
r
ec
ast
th
e
wea
th
er
m
o
r
e
ac
cu
r
atel
y
[
3
]
.
Sh
i
et
a
l
.
[
4
]
s
u
g
g
ested
th
at
th
e
co
n
v
o
lu
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
C
o
n
v
L
STM
)
ar
ch
itectu
r
e
u
s
es
a
r
a
d
ar
ec
h
o
d
ataset
f
o
r
p
r
ec
i
p
itatio
n
f
o
r
ec
asti
n
g
in
o
r
d
e
r
to
o
v
er
co
m
e
th
e
s
eq
u
e
n
ce
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:
2088
-
8
7
0
8
E
n
h
a
n
ci
n
g
s
p
a
tio
temp
o
r
a
l wea
th
er fo
r
ec
a
s
tin
g
a
cc
u
r
a
cy
w
ith
3
D
co
n
v
o
lu
tio
n
a
l
…
(
R
en
a
ld
y
F
r
ed
ya
n
)
2023
p
r
ed
ictio
n
is
s
u
e.
T
o
co
n
cu
r
r
e
n
tly
lear
n
in
co
m
in
g
d
ata'
s
te
m
p
o
r
al
an
d
s
p
atial
co
n
tex
t
an
d
p
r
o
ject
th
e
f
u
tu
r
e
s
eq
u
en
ce
,
th
ey
m
er
g
e
d
th
e
co
n
v
o
lu
tio
n
o
p
er
ato
r
—
u
s
ed
b
y
th
e
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etw
o
r
k
(
C
NN)
—
with
a
r
ec
u
r
r
en
t
n
e
u
r
al
n
etwo
r
k
(
R
N
N)
.
E
v
en
wh
ile
C
o
n
v
L
STM
ar
ch
itectu
r
e
h
as
b
ee
n
s
ee
n
as
a
p
o
s
s
ib
le
m
eth
o
d
to
d
ev
elo
p
p
r
ed
ictio
n
m
o
d
els
f
o
r
g
eo
s
cien
ce
d
ata
[
5
]
,
f
r
es
h
p
r
o
s
p
ec
ts
h
a
v
e
s
u
r
f
ac
e
d
t
h
an
k
s
to
cu
r
r
en
t
d
ev
elo
p
m
e
n
ts
in
d
ee
p
lear
n
in
g
.
T
h
e
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
L
STM
)
u
n
it
was
en
h
an
ce
d
in
s
ev
er
al
s
tu
d
ies
to
m
em
o
r
ize
s
p
atial
-
tem
p
o
r
al
in
f
o
r
m
atio
n
.
R
NN
-
b
ased
ar
ch
itectu
r
es
ar
e
well
-
s
u
ited
f
o
r
m
u
lti
-
s
tep
f
o
r
ec
asti
n
g
task
s
in
v
o
lv
in
g
s
p
ati
o
tem
p
o
r
al
d
ata
d
u
e
to
th
eir
ab
ilit
y
to
p
r
ed
ict
lo
n
g
s
eq
u
en
ce
s
wh
ile
m
ain
tain
in
g
tem
p
o
r
al
o
r
d
er
(
ca
u
s
al
co
n
s
tr
ain
t)
[
6
]
.
Ho
wev
er
,
t
h
ese
m
o
d
els
r
ely
o
n
in
f
o
r
m
atio
n
f
r
o
m
p
r
e
v
io
u
s
t
im
e
s
tep
s
to
g
en
er
ate
o
u
tp
u
ts
,
lead
in
g
to
len
g
t
h
y
tr
ain
in
g
tim
es.
T
o
ad
d
r
ess
th
is
lim
itatio
n
,
we
p
r
o
p
o
s
e
a
n
o
v
el
ar
ch
itectu
r
e
th
at
ex
clu
s
iv
ely
u
tili
ze
s
3
D
C
NN
f
o
r
s
p
atio
tem
p
o
r
al
f
o
r
ec
asti
n
g
.
C
NNs
ar
e
h
ig
h
ly
e
f
f
ec
tiv
e
in
ca
p
tu
r
in
g
s
p
atial
co
n
te
x
t
an
d
h
av
e
d
e
m
o
n
s
tr
ated
s
tate
-
of
-
th
e
-
ar
t
p
er
f
o
r
m
a
n
ce
in
im
ag
e
class
if
icatio
n
with
2
D
k
er
n
els
[
7
]
.
R
ec
en
t
ad
v
an
ce
m
en
ts
h
av
e
ex
ten
d
ed
th
e
ap
p
licatio
n
o
f
C
NNs,
s
u
ch
as
u
s
in
g
1
D
k
er
n
els
f
o
r
task
s
lik
e
m
ac
h
in
e
tr
a
n
s
latio
n
[
8
]
,
wh
ich
en
a
b
les
th
e
ex
tr
ac
tio
n
o
f
tem
p
o
r
al
p
atter
n
s
in
s
eq
u
en
ce
s
.
Ap
p
licatio
n
s
lik
e
v
id
eo
an
aly
s
is
,
ac
tio
n
r
ec
o
g
n
itio
n
[
9
]
,
a
n
d
clim
ate
ev
en
t
d
etec
tio
n
[
1
0
]
h
ig
h
lig
h
t
th
e
p
o
ten
tial
o
f
3
D
C
NN
m
o
d
els.
Ho
wev
er
,
C
N
N
-
b
ased
ap
p
r
o
ac
h
es
f
ac
e
two
k
ey
ch
allen
g
es
f
o
r
m
u
lti
-
s
tep
f
o
r
ec
asti
n
g
:
th
ey
ar
e
u
n
ab
le
to
p
r
o
d
u
ce
o
u
t
p
u
t
s
eq
u
en
ce
s
lo
n
g
e
r
th
a
n
th
e
in
p
u
t
an
d
d
is
r
u
p
t
tem
p
o
r
al
o
r
d
er
b
y
in
co
r
p
o
r
atin
g
f
u
tu
r
e
in
f
o
r
m
atio
n
d
u
r
in
g
tem
p
o
r
al
r
ea
s
o
n
in
g
[
1
1
]
.
W
e
in
tr
o
d
u
ce
co
n
v
o
l
u
tio
n
al
s
eq
u
en
ce
-
to
-
s
eq
u
e
n
ce
(
C
o
n
v
Seq
2
Seq
)
n
etwo
r
k
,
a
s
p
atio
tem
p
o
r
al
p
r
ed
ictio
n
m
o
d
el
ex
p
licitly
d
esig
n
ed
f
o
r
m
u
lt
i
-
s
tep
f
o
r
ec
asti
n
g
task
s
to
ad
d
r
ess
th
ese
lim
itatio
n
s
.
As
f
ar
as
we
k
n
o
w,
C
o
n
v
Seq
2
Seq
is
th
e
f
ir
s
t
3
D
C
NN
-
b
ased
ar
ch
itectu
r
e
d
ev
elo
p
ed
as
an
en
d
-
to
-
e
n
d
t
r
ain
ab
le
m
o
d
el
th
at
ad
h
er
es
to
th
e
ca
u
s
al
co
n
s
tr
a
in
t
wh
ile
allo
win
g
th
e
p
r
ed
ic
tio
n
o
f
o
u
t
p
u
t
s
eq
u
en
ce
s
o
f
f
lex
ib
le
len
g
th
s
—
u
n
r
estricte
d
b
y
th
e
len
g
t
h
o
f
th
e
in
p
u
t seq
u
en
ce
.
T
h
r
o
u
g
h
e
x
p
er
im
e
n
tal
ev
alu
a
tio
n
s
,
we
ass
ess
ed
th
e
p
r
ed
i
ctiv
e
ac
cu
r
ac
y
an
d
tim
e
ef
f
icien
cy
o
f
C
o
n
v
Seq
2
Seq
in
co
m
p
ar
is
o
n
to
R
NN
-
b
ased
a
r
ch
itectu
r
es.
Usi
n
g
m
eteo
r
o
lo
g
ical
d
atasets
s
u
ch
as
clim
ate
h
az
ar
d
g
r
o
u
p
in
f
r
a
r
ed
p
r
ec
ip
i
tatio
n
s
atellite
(
C
HI
R
P
S)
th
at
co
m
b
in
e
s
atellite
d
ata
an
d
i
n
s
itu
s
tatio
n
[
1
2
]
m
ea
s
u
r
em
en
ts
,
th
e
p
r
o
p
o
s
ed
a
r
ch
itectu
r
e
m
atch
es
o
r
o
u
tp
e
r
f
o
r
m
s
ex
is
tin
g
tech
n
iq
u
es.
T
h
i
s
s
tu
d
y
co
n
tr
ib
u
tes
in
two
s
ig
n
if
ican
t
way
s
.
First,
it
in
tr
o
d
u
ce
s
v
ar
iatio
n
s
o
f
th
e
C
o
n
v
Seq
2
Seq
ar
ch
itectu
r
e
th
a
t
s
atis
f
y
th
e
ca
u
s
al
co
n
s
tr
ain
t.
On
e
ap
p
r
o
ac
h
in
v
o
lv
es
ad
ap
tin
g
ca
u
s
al
co
n
v
o
lu
ti
o
n
with
in
3
D
co
n
v
o
lu
tio
n
al
lay
er
s
,
wh
ile
an
o
th
er
ap
p
lies
a
n
o
v
el
tech
n
iq
u
e
th
at
r
ev
er
s
es
s
eq
u
en
ce
s
d
elib
er
atel
y
.
Seco
n
d
,
to
e
n
ab
le
l
o
n
g
e
r
o
u
tp
u
t
s
eq
u
e
n
ce
s
,
we
d
ev
elo
p
e
d
a
tem
p
o
r
al
g
en
er
at
o
r
b
lo
c
k
f
ea
tu
r
i
n
g
an
in
n
o
v
ativ
e
u
s
e
o
f
tr
a
n
s
p
o
s
ed
co
n
v
o
l
u
tio
n
al
lay
er
s
.
2.
RE
L
AT
E
D
WO
RK
S
His
to
r
ical
d
ata
r
eg
ar
d
in
g
tem
p
er
atu
r
e,
p
r
ec
ip
itatio
n
,
an
d
o
t
h
er
m
eteo
r
o
lo
g
ical
v
ar
iab
les
h
av
e
b
ee
n
u
s
ed
to
f
o
r
ec
ast
t
h
e
wea
th
er
u
s
in
g
s
ev
er
al
s
tatis
tical
an
d
m
ac
h
in
e
-
lear
n
i
n
g
a
p
p
r
o
ac
h
es
[
1
3
]
.
T
im
es
s
er
ies
an
aly
s
is
is
tr
ad
itio
n
ally
h
a
n
d
led
s
tatis
tically
u
s
in
g
au
to
-
r
e
g
r
ess
iv
e
in
teg
r
ated
m
o
v
in
g
a
v
er
ag
es
(
AR
I
MA
)
[
1
4
]
.
Oth
er
r
esear
ch
h
as
al
s
o
u
s
ed
ar
tific
ial
n
eu
r
al
n
e
two
r
k
s
(
ANN)
f
o
r
tim
e
s
er
ies
p
r
ed
ictio
n
in
m
eteo
r
o
lo
g
ical
d
ata,
in
clu
d
i
n
g
tem
p
er
atu
r
e
r
ea
d
in
g
s
[
1
5
]
.
Usi
n
g
L
STM
n
etwo
r
k
s
in
p
ar
t
icu
lar
,
m
an
y
wr
iter
s
h
av
e
b
ee
n
d
e
v
elo
p
i
n
g
n
o
v
el
d
ee
p
lea
r
n
in
g
-
b
ased
m
et
h
o
d
s
r
ec
en
tly
to
en
h
an
ce
tim
e
s
er
ies
f
o
r
ec
asti
n
g
p
er
f
o
r
m
an
ce
[
1
6
]
.
Ap
p
ly
i
n
g
L
STM
d
esig
n
s
ef
f
ec
tiv
ely
in
clu
d
es
tr
af
f
ic
f
lo
w
an
aly
s
is
[
1
7
]
,
lan
d
s
lid
e
d
is
p
lace
m
en
t
p
r
ed
ictio
n
[
1
8
]
,
p
etr
o
leu
m
p
r
o
d
u
ctio
n
[
1
9
]
,
an
d
s
ea
s
u
r
f
ac
e
tem
p
er
atu
r
e
f
o
r
ec
asti
n
g
[
2
0
]
.
Ho
wev
er
,
s
p
atial
r
elatio
n
s
h
ip
s
in
th
e
d
ata
ar
e
n
o
t
ca
p
tu
r
e
d
b
y
th
ese
m
eth
o
d
s
(
wh
ich
a
r
e
d
ir
ec
ted
at
tim
e
s
er
ies).
Sp
atio
tem
p
o
r
al
d
ee
p
lear
n
in
g
alg
o
r
ith
m
s
ef
f
ec
tiv
ely
ad
d
r
ess
b
o
th
g
eo
g
r
a
p
h
ic
an
d
tem
p
o
r
al
d
im
en
s
io
n
s
.
Sh
i
et
a
l
.
[
4
]
tr
ea
t
wea
th
er
f
o
r
ec
asti
n
g
as
a
s
eq
u
en
ce
-
to
-
s
eq
u
e
n
ce
p
r
o
b
lem
,
u
ti
lizin
g
s
eq
u
en
ce
s
o
f
2
D
r
ad
ar
m
a
p
s
as
b
o
th
in
p
u
t
a
n
d
o
u
t
p
u
t.
T
h
e
y
in
tr
o
d
u
ce
th
e
C
o
n
v
L
STM
ar
ch
itectu
r
e
to
c
r
ea
te
an
en
d
-
to
-
en
d
m
o
d
el
f
o
r
p
r
ec
ip
itatio
n
n
o
wc
asti
n
g
,
in
teg
r
atin
g
co
n
v
o
lu
tio
n
al
o
p
er
ati
o
n
s
in
to
th
e
L
STM
n
etwo
r
k
to
ca
p
tu
r
e
s
p
atial
p
atter
n
s
.
Similar
ly
,
Ki
m
et
a
l
.
[
2
1
]
e
m
p
lo
y
C
o
n
v
L
S
T
M
f
o
r
p
r
e
d
ictin
g
s
ev
e
r
e
clim
atic
ev
en
ts
,
f
r
a
m
in
g
th
eir
task
as
a
s
eq
u
en
ce
-
b
ased
p
r
o
b
lem
u
s
in
g
s
to
r
m
d
en
s
ity
m
ap
s
eq
u
en
ce
s
as
in
p
u
t.
So
u
t
o
et
a
l
.
[
2
2
]
p
r
o
p
o
s
e
a
s
p
atio
tem
p
o
r
al
-
awa
r
e
e
n
s
em
b
le
ap
p
r
o
ac
h
le
v
er
ag
in
g
C
o
n
v
L
STM
,
wh
ile
S
etiawa
n
e
t
a
l
.
[
2
3
]
wo
r
k
b
y
in
co
r
p
o
r
atin
g
a
n
o
v
el
L
STM
u
n
it
th
at
u
n
if
o
r
m
ly
h
an
d
les
te
m
p
o
r
al
an
d
s
p
atial
v
ar
iatio
n
s
in
its
m
em
o
r
y
p
o
o
l.
W
an
g
et
a
l
.
[
2
4
]
also
en
h
an
ce
m
em
o
r
y
f
u
n
ctio
n
ality
b
y
i
n
tr
o
d
u
ci
n
g
n
o
n
-
s
tatio
n
ar
ity
m
o
d
elin
g
with
in
th
e
L
STM
u
n
it.
Alth
o
u
g
h
th
ese
a
p
p
r
o
ac
h
es
co
m
b
i
n
e
L
STM
a
n
d
C
NN
f
o
r
clim
ate
an
d
wea
th
er
-
r
elate
d
task
s
,
o
u
r
m
o
d
el
ad
o
p
ts
a
p
u
r
ely
C
NN
-
b
ased
m
eth
o
d
o
l
o
g
y
,
av
o
id
i
n
g
t
h
e
h
y
b
r
id
s
tr
ateg
y
o
f
m
e
r
g
in
g
L
STM
with
C
NN.
A
f
ew
s
tu
d
ies
h
av
e
b
ee
n
d
o
n
e
o
n
ap
p
ly
in
g
s
p
atio
tem
p
o
r
al
co
n
v
o
l
u
tio
n
s
f
o
r
ac
tio
n
r
ec
o
g
n
itio
n
an
d
v
id
eo
a
n
aly
s
is
.
T
r
an
et
a
l.
[
2
5
]
d
e
m
o
n
s
tr
ate
t
h
at
f
ac
to
r
izin
g
th
e
3
D
c
o
n
v
o
lu
tio
n
al
k
er
n
e
l
in
to
d
is
tin
ct
an
d
co
n
s
ec
u
tiv
e
s
p
atial
an
d
tem
p
o
r
al
co
n
v
o
lu
tio
n
s
in
cr
ea
s
es
a
cc
u
r
ac
y
b
y
co
m
p
ar
in
g
m
u
ltip
le
s
p
atio
tem
p
o
r
al
d
esig
n
s
em
p
lo
y
in
g
ju
s
t
3
D
C
NN.
L
im
itatio
n
o
f
f
ac
t
o
r
ized
3
D
C
NN
as
well
as
3
D
C
NN
T
r
an
et
a
l.
[
2
5
]
v
io
lates
th
e
tem
p
o
r
al
o
r
d
er
b
y
lack
in
g
a
ca
u
s
al
r
eq
u
ir
em
e
n
t.
T
h
e
3
D
c
o
n
v
o
lu
tio
n
,
as
T
r
an
et
a
l.
[
2
5
]
,
is
f
ac
to
r
ized
b
y
Sin
g
h
an
d
C
u
zz
o
lin
[
2
6
]
an
d
C
h
en
g
et
a
l
.
[
2
7
]
.
T
h
e
ca
u
s
al
c
o
n
s
tr
ain
t
in
te
m
p
o
r
al
lea
r
n
in
g
f
o
r
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.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
0
2
2
-
2
0
3
0
2024
ac
tio
n
r
ec
o
g
n
itio
n
task
s
is
ad
d
r
ess
ed
b
y
Sin
g
h
an
d
C
u
zz
o
lin
[
2
6
]
u
s
in
g
a
r
ec
u
r
r
en
t
co
n
v
o
lu
tio
n
u
n
it
tech
n
iq
u
e;
th
e
ca
u
s
al
co
n
s
tr
ai
n
t
is
s
ati
s
f
ied
b
y
C
h
en
g
et
a
l.
[
2
7
]
b
y
u
s
in
g
ca
u
s
al
co
n
v
o
lu
tio
n
in
d
is
cr
ete
an
d
p
ar
allel
s
p
atial
an
d
tem
p
o
r
al
co
n
v
o
l
u
tio
n
s
.
Ho
wev
er
,
with
a
d
if
f
er
en
t
im
p
lem
en
tatio
n
,
we
lik
ewise
u
s
ed
a
f
ac
to
r
ized
3
D
C
NN.
W
e
u
s
e
a
f
u
ll
C
NN
tech
n
iq
u
e;
s
im
ilar
ly
,
we
p
r
o
v
id
e
a
n
o
v
el
wa
y
n
o
t
to
b
r
ea
k
th
e
tem
p
o
r
al
o
r
d
e
r
a
n
d
d
o
n
o
t
em
p
lo
y
p
ar
allel
c
o
n
v
o
lu
tio
n
s
w
h
en
ad
o
p
tin
g
a
ca
u
s
al
co
n
v
o
l
u
tio
n
.
Af
ter
Xu
et
a
l
.
[
2
8
]
s
u
cc
ess
f
u
lly
ca
p
tu
r
e
d
s
p
atial
co
r
r
elatio
n
in
p
ictu
r
es,
th
ey
p
r
o
p
o
s
ed
a
tech
n
iq
u
e
to
esti
m
ate
v
eh
icle
p
o
llu
tio
n
em
is
s
io
n
s
b
y
in
d
ep
e
n
d
en
tly
co
llectin
g
tem
p
o
r
al
a
n
d
s
p
atial
co
r
r
elatio
n
u
s
in
g
2
D
C
NN.
Mu
d
ig
o
n
d
a
et
a
l
.
[
2
9
]
id
en
tif
y
s
ev
er
e
clim
atic
ev
en
ts
b
y
u
s
in
g
a
3
D
C
NN
in
an
en
c
o
d
er
-
d
ec
o
d
e
r
ar
ch
i
tectu
r
e.
3.
M
E
T
H
O
D
3
.
1
.
Da
t
a
Gr
id
d
ed
r
ain
f
all
tim
e
s
er
ies
w
ith
d
aily
f
r
eq
u
en
cy
an
d
a
0
.
0
5
°
g
eo
g
r
ap
h
ic
r
eso
lu
tio
n
ar
e
p
r
o
d
u
ce
d
b
y
co
m
b
in
in
g
s
atellite
im
ag
es
an
d
in
s
itu
s
tatio
n
d
ata
in
th
e
C
HI
R
P
S
d
ataset.
I
n
th
is
w
o
r
k
,
we
p
e
r
f
o
r
m
ed
in
ter
p
o
latio
n
t
o
s
h
r
in
k
th
e
g
r
i
d
s
ize
to
5
0
×
5
0
u
s
in
g
a
r
ec
o
r
d
s
s
am
p
le
f
r
o
m
J
an
u
ar
y
1
9
8
1
to
Dec
em
b
er
2
0
2
0
.
Fig
u
r
e
1
d
is
p
lay
s
th
e
c
o
v
er
a
g
e
ar
ea
,
wh
ic
h
is
1
2
7
,
3
4
6
.
9
2
k
m
2
o
n
lan
d
a
n
d
2
5
,
6
5
6
k
m
2
o
n
wate
r
,
em
p
l
o
y
ed
in
o
u
r
s
tu
d
ies,
f
r
o
m
2
°3
3
′
N
o
r
th
L
atitu
d
e
-
2
°2
5
'
So
u
th
L
atitu
d
e,
1
1
3
°4
4
'
–
1
1
9
°0
0
'
E
ast
L
o
n
g
itu
d
e.
I
n
k
ee
p
in
g
with
Sh
i
et
a
l.
'
s
m
eth
o
d
o
lo
g
y
[
4
]
,
we
s
et
th
e
in
p
u
t
s
eq
u
en
ce
len
g
th
to
f
i
v
e,
i.e
.
,
th
e
n
ex
t
s
et
o
f
g
r
id
s
is
p
r
ed
icted
u
s
in
g
th
e
p
r
ev
io
u
s
f
iv
e
g
r
id
s
.
Fo
r
th
e
C
HI
R
PS
d
ata
s
et
(
h
ttp
s
:
//w
w
w
.
ch
c.
u
csb
.
ed
u
/d
a
ta
/ch
ir
p
s
)
,
th
u
s
,
th
e
in
p
u
t
d
ata
s
h
ap
es
f
o
r
th
e
d
ee
p
lear
n
in
g
ar
c
h
itectu
r
es
ar
e
5
×
50
×
50
×
1
.
Her
e,
1
d
en
o
te
s
th
e
s
in
g
le
ch
an
n
el
(
lik
e
a
g
r
ay
s
ca
le
p
ictu
r
e)
,
5
is
th
e
f
o
r
ec
asti
n
g
s
eq
u
en
ce
len
g
th
,
an
d
3
2
an
d
5
0
is
th
e
n
u
m
b
er
o
f
latitu
d
es
a
n
d
lo
n
g
itu
d
es u
tili
ze
d
to
b
u
ild
t
h
e
s
p
atial
g
r
id
f
o
r
ev
er
y
d
ataset.
Fro
m
th
e
r
ain
f
all
d
ataset,
we
p
r
o
d
u
ce
d
1
3
,
9
6
0
g
r
id
s
eq
u
en
c
es.
Af
ter
th
at,
n
o
n
-
o
v
er
lap
p
in
g
tr
ain
in
g
,
v
alid
atio
n
,
an
d
test
s
ets
wer
e
cr
ea
ted
f
r
o
m
b
o
th
d
atasets
in
p
r
o
p
o
r
tio
n
s
o
f
6
0
%,
2
0
%,
an
d
2
0
%,
r
esp
ec
tiv
ely
.
W
e
h
av
e
u
s
ed
r
ai
n
f
all
d
ata
s
ets
in
o
u
r
ex
p
er
im
en
tal
as
s
ess
m
en
t
b
ec
au
s
e
o
f
th
eir
i
m
p
o
r
tan
ce
as
m
ain
m
eteo
r
o
lo
g
ical
v
ar
iab
les.
Stu
d
y
in
g
t
h
eir
s
p
atio
tem
p
o
r
al
r
ep
r
esen
tatio
n
im
p
r
o
v
es
o
u
r
k
n
o
wled
g
e
o
f
lo
n
g
-
ter
m
clim
ate
v
ar
iab
ilit
y
an
d
is
ess
e
n
tial f
o
r
s
h
o
r
t
-
ter
m
f
o
r
ec
asti
n
g
.
Ho
wev
er
,
th
e
p
r
o
p
o
s
ed
ar
c
h
itectu
r
e
is
ad
ap
tab
le
an
d
m
ay
b
e
u
s
ed
f
o
r
o
t
h
er
m
eteo
r
o
lo
g
ical
v
a
r
iab
les
o
r
d
o
m
ain
s
,
p
r
o
v
id
ed
th
at
th
e
t
r
ain
in
g
d
ata
ca
n
b
e
o
r
g
an
ized
as d
escr
ib
ed
in
th
e
s
u
b
s
eq
u
en
t sectio
n
.
Fig
u
r
e
1
.
T
h
e
g
e
o
g
r
ap
h
ical
ex
ten
t e
n
co
m
p
ass
ed
b
y
th
e
d
atasets
u
s
ed
in
all
s
tu
d
ies.
T
h
e
g
r
i
d
b
elo
w
r
e
p
r
esen
ts
th
e
ch
o
s
en
s
eq
u
e
n
ce
f
o
r
Dec
e
m
b
er
2
0
2
0
,
s
h
o
win
g
th
e
r
ec
o
r
d
ed
r
ain
f
all
lev
els
3.
2
.
P
r
o
po
s
ed
m
o
dels
T
h
e
3
D
c
o
n
v
o
lu
tio
n
al
lay
er
th
r
o
u
g
h
Seq
2
Seq
m
o
d
el
is
a
co
m
p
r
eh
e
n
s
iv
e
d
ee
p
n
e
u
r
al
n
etwo
r
k
d
esig
n
ed
to
lear
n
an
d
p
r
ed
ict
p
atter
n
s
th
at
o
cc
u
r
in
b
o
th
s
p
ac
e
an
d
tim
e.
T
h
is
n
etwo
r
k
is
e
s
p
ec
ially
b
en
ef
icial
in
in
d
u
s
tr
ies
lik
e
wea
th
er
f
o
r
ec
asti
n
g
,
wh
er
e
th
ese
p
atter
n
s
ar
e
f
r
eq
u
e
n
tly
o
b
s
er
v
e
d
.
Ou
r
m
eth
o
d
o
l
o
g
y
allo
ws
f
o
r
th
e
p
r
ed
ictio
n
o
f
m
u
lti
-
s
tep
s
eq
u
en
ce
s
with
o
u
t
in
co
r
p
o
r
atin
g
th
e
an
ticip
ated
o
u
tco
m
e
b
ac
k
in
to
th
e
in
p
u
t
s
eq
u
en
ce
.
Ou
r
p
r
o
p
o
s
ed
d
ee
p
lear
n
in
g
f
r
am
ewo
r
k
is
co
m
p
r
e
h
en
s
iv
ely
illu
s
tr
ated
in
Fig
u
r
e
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:
2088
-
8
7
0
8
E
n
h
a
n
ci
n
g
s
p
a
tio
temp
o
r
a
l wea
th
er fo
r
ec
a
s
tin
g
a
cc
u
r
a
cy
w
ith
3
D
co
n
v
o
lu
tio
n
a
l
…
(
R
en
a
ld
y
F
r
ed
ya
n
)
2025
T
h
e
m
ajo
r
ity
o
f
wea
th
er
f
o
r
e
ca
s
tin
g
tech
n
iq
u
es
em
p
l
o
y
a
co
m
b
in
atio
n
o
f
2
D
C
NN
an
d
L
STM
to
lear
n
s
p
atial
an
d
tem
p
o
r
al
r
e
p
r
esen
tatio
n
s
.
Ho
wev
er
,
o
u
r
ap
p
r
o
ac
h
ex
clu
s
iv
ely
em
p
lo
y
s
3
D
co
n
v
o
lu
tio
n
al
lay
er
s
to
ac
q
u
ir
e
b
o
th
s
p
atial
an
d
tem
p
o
r
al
co
n
tex
ts
.
C
o
n
v
Seq
2
Seq
,
in
co
n
tr
ast
to
th
e
co
n
v
en
tio
n
al
co
n
v
o
l
u
tio
n
e
m
p
lo
y
e
d
in
ce
r
tain
3
D
C
NN
ar
ch
itectu
r
es,
g
u
ar
a
n
tees
th
at
it
d
o
es
n
o
t
d
ep
en
d
o
n
f
u
tu
r
e
k
n
o
wled
g
e
d
u
r
in
g
tem
p
o
r
al
lear
n
in
g
,
a
c
r
itical
r
eq
u
ir
em
en
t
f
o
r
task
p
r
ed
ictio
n
.
Ou
r
n
et
wo
r
k
ar
ch
itectu
r
e'
s
ab
ilit
y
to
ac
co
m
m
o
d
ate
v
ar
i
ab
le
o
u
tp
u
t
s
eq
u
en
ce
len
g
t
h
s
is
cr
itica
l.
T
h
is
s
u
g
g
ests
th
at
it
i
s
ca
p
ab
le
o
f
p
r
ed
ictin
g
a
s
ig
n
if
ican
t
n
u
m
b
er
o
f
f
u
tu
r
e
tim
e
s
tep
s
,
r
eg
ar
d
less
o
f
th
e
in
p
u
t
s
eq
u
en
ce
'
s
f
i
x
ed
d
u
r
atio
n
.
I
n
th
e
f
o
llo
win
g
s
ec
tio
n
,
we
o
f
f
er
ad
d
itio
n
al
d
etails r
eg
ar
d
i
n
g
th
e
c
o
m
p
o
n
en
ts
th
at
co
n
s
titu
te
o
u
r
ar
ch
itectu
r
e.
W
e
em
p
lo
y
a
f
ac
to
r
ized
3
D
k
er
n
el
th
at
is
in
s
p
ir
ed
b
y
t
h
e
(
2
+
1
)
n
etwo
r
k
in
tr
o
d
u
ce
d
b
y
T
r
an
et
a
l.
[
2
5
]
in
s
tead
o
f
a
tr
ad
itio
n
al
k
er
n
el
f
o
r
3
D
c
o
n
v
o
lu
tio
n
al
lay
er
s
,
wh
er
e
th
e
k
er
n
el
s
ize
is
d
ef
in
ed
b
y
d
in
th
e
s
p
atial
d
im
en
s
io
n
s
(
H
an
d
W
)
an
d
t in
th
e
tem
p
o
r
al
d
im
en
s
io
n
(
T
)
.
T
h
e
a
u
th
o
r
s
in
tr
o
d
u
ce
a
f
ac
to
r
ize
d
k
er
n
el,
d
en
o
ted
as
1
-
d
-
d
an
d
t
-
1
-
1
,
in
th
eir
wo
r
k
.
T
h
is
k
er
n
e
l
p
ar
titi
o
n
s
th
e
co
n
v
o
lu
tio
n
p
r
o
ce
d
u
r
e
o
f
a
s
in
g
le
lay
er
in
to
two
d
is
tin
ct
o
p
er
atio
n
s
:
a
s
p
atial
co
n
v
o
lu
tio
n
an
d
a
tem
p
o
r
al
c
o
n
v
o
lu
tio
n
.
W
e
em
p
lo
y
a
n
alter
n
ativ
e
m
eth
o
d
o
lo
g
y
in
o
u
r
in
n
o
v
ativ
e
d
esig
n
,
wh
ich
in
v
o
lv
es
th
e
n
o
n
-
s
eq
u
e
n
tial
ex
ec
u
tio
n
o
f
o
p
e
r
atio
n
s
with
in
ea
ch
co
n
v
o
lu
tio
n
al
la
y
er
.
T
h
e
f
ac
to
r
ize
d
k
er
n
els
ar
e
p
ar
titi
o
n
ed
in
to
two
g
r
o
u
p
s
,
wh
ich
lead
s
to
d
is
cr
ete
lear
n
in
g
ab
ilit
ies
f
o
r
ea
ch
.
T
h
e
tem
p
o
r
al
b
l
o
ck
e
m
p
lo
y
s
th
e
t
-
1
-
1
k
e
r
n
el
in
its
lay
er
s
to
u
n
d
er
s
tan
d
tem
p
o
r
al
r
elatio
n
s
h
ip
s
in
a
s
p
ec
if
ic
m
an
n
er
.
C
o
n
v
er
s
ely
,
th
e
s
p
atial
b
lo
ck
en
ca
p
s
u
lates
s
p
atial
d
ep
en
d
en
cies
b
y
em
p
l
o
y
in
g
a
1
-
d
-
d
k
er
n
el.
T
h
e
k
e
r
n
el
d
ec
o
m
p
o
s
itio
n
em
p
lo
y
ed
in
C
o
n
v
Seq
2
Seq
h
as
t
h
e
ad
v
an
tag
e
o
f
in
cr
ea
s
in
g
th
e
n
u
m
b
er
o
f
n
o
n
lin
ea
r
ities
in
th
e
n
etwo
r
k
,
i
n
co
n
tr
ast to
th
e
c
o
m
p
lete
3
D
k
er
n
el
u
s
ed
in
co
n
v
en
tio
n
al
c
o
n
v
o
lu
tio
n
s
.
T
h
is
is
ac
co
m
p
lis
h
ed
b
y
in
c
o
r
p
o
r
atin
g
s
u
p
p
lem
e
n
tar
y
ac
tiv
atio
n
f
u
n
ctio
n
s
b
etwe
en
f
a
cto
r
ized
c
o
n
v
o
lu
tio
n
s
,
wh
ich
lead
s
to
g
r
ea
ter
co
m
p
le
x
ity
in
th
e
p
atter
n
s
th
at
ca
n
b
e
r
ep
r
esen
ted
.
Ou
r
r
ec
o
m
m
en
d
ed
s
o
lu
tio
n
is
f
lex
ib
le
o
v
er
th
e
(
2
+1
)
D
b
lo
ck
.
T
h
is
is
d
u
e
to
th
e
f
ac
t
th
at
th
e
tem
p
o
r
al
an
d
s
p
atial
u
n
its
m
ay
h
a
v
e
v
ar
y
in
g
n
u
m
b
e
r
s
o
f
lay
er
s
,
wh
ich
f
ac
ilit
ates th
eir
o
p
tim
izatio
n
.
Fig
u
r
e
2
.
C
o
n
v
Seq
2
Seq
ar
c
h
itectu
r
e
3.
3
.
E
v
a
lua
t
i
o
n
m
a
t
rix
Po
s
tp
r
o
ce
s
s
in
g
is
s
o
u
g
h
t
to
o
b
tain
b
etter
r
ain
f
all
f
o
r
ec
asts
th
an
"r
aw"
(
u
n
p
r
o
ce
s
s
ed
)
h
y
d
r
o
lo
g
ical
m
o
d
els.
Fo
r
th
is
p
u
r
p
o
s
e,
it
is
im
p
o
r
tan
t
to
ass
ess
th
e
p
e
r
f
o
r
m
an
ce
o
f
th
e
m
o
d
els
an
d
c
o
n
t
r
ast
th
em
to
ch
o
o
s
e
th
e
b
est
o
n
e.
A
n
u
m
b
e
r
o
f
m
ea
s
u
r
es
ar
e
u
s
ed
to
ass
ess
f
o
r
ec
asts
f
o
r
v
ar
io
u
s
wait
d
u
r
atio
n
s
.
T
h
e
eq
u
atio
n
'
s
r
o
o
t
-
m
ea
n
-
s
q
u
ar
e
er
r
o
r
(
R
MSE
)
is
th
e
m
ai
n
ac
cu
r
ac
y
m
etr
i
c
f
o
r
a
d
eter
m
in
is
tic
f
o
r
ec
ast
b
ec
au
s
e
p
r
ec
is
e
a
n
d
tr
u
s
two
r
th
y
f
o
r
ec
asts
ar
e
s
o
im
p
o
r
tan
t
d
u
r
in
g
r
ain
f
all
ev
en
t
s
.
=
√
∑
(
−
)
2
wh
er
e
is
th
e
o
b
s
er
v
ed
d
aily
r
ain
f
all,
is
th
e
ℎ
tim
e
-
k
f
o
r
ec
ast
o
f
d
aily
r
ain
f
all,
an
d
is
th
e
to
tal
n
u
m
b
e
r
o
f
tim
e
-
k
m
o
n
th
ly
r
ain
f
all
p
r
ed
ictio
n
s
.
R
MSE
p
en
alize
s
m
o
r
e
s
u
b
s
tan
tial
m
is
tak
es
f
o
r
h
ig
h
r
ain
f
all
p
r
o
jectio
n
s
th
an
m
ea
n
a
b
s
o
lu
t
e
er
r
o
r
(
MA
E
)
m
ea
s
u
r
es.
W
h
er
e
th
e
to
tal
ef
f
ec
t
o
f
m
is
tak
es
is
p
r
o
p
o
r
tio
n
al
to
th
e
in
cr
ea
s
e
in
er
r
o
r
,
MA
E
,
a
lin
ea
r
s
tatis
tical
m
ea
s
u
r
e,
is
m
o
r
e
u
s
ef
u
l
th
an
R
MSE
,
wh
ich
ass
ig
n
s
a
co
m
p
ar
ativ
ely
la
r
g
e
weig
h
t t
o
b
ig
er
r
o
r
s
.
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.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
0
2
2
-
2
0
3
0
2026
=
1
∑
|
−
|
=
1
T
h
e
m
ea
n
s
q
u
ar
e
d
er
r
o
r
(
MSE
)
q
u
an
tifie
s
th
e
av
er
a
g
e
o
f
th
e
s
q
u
ar
ed
d
is
cr
ep
a
n
cies
b
etwe
en
th
e
o
b
s
er
v
ed
an
d
esti
m
ated
v
alu
e
s
.
T
h
e
m
etr
ic
m
ea
s
u
r
es
th
e
p
r
o
x
im
ity
o
f
th
e
p
r
e
d
ictio
n
s
to
th
e
ac
tu
al
r
esu
lts
,
wh
er
e
s
m
aller
v
alu
es
im
p
ly
s
u
p
er
io
r
p
e
r
f
o
r
m
an
ce
o
f
th
e
m
o
d
el.
W
h
er
e:
is
th
e
n
u
m
b
er
o
f
o
b
s
er
v
atio
n
s
,
an
d
r
ep
r
esen
ts
th
e
ac
tu
al
v
al
u
e.
T
h
e
v
ar
iab
le
r
ep
r
esen
ts
th
e
v
al
u
e
th
at
is
b
ein
g
f
o
r
ec
asted
.
=
1
∑
(
−
)
2
=
1
R
-
s
q
u
ar
ed
is
a
s
tati
s
tical
m
et
r
ic
th
at
q
u
an
tifie
s
th
e
p
er
ce
n
tag
e
o
f
th
e
v
ar
iatio
n
in
th
e
d
ep
en
d
e
n
t
v
ar
iab
le
th
at
th
e
in
d
e
p
en
d
e
n
t
v
ar
iab
les
ca
n
ac
co
u
n
t
f
o
r
in
a
r
eg
r
ess
io
n
m
o
d
el.
T
h
e
r
a
n
g
e
o
f
v
alu
es
is
f
r
o
m
0
to
1
,
wh
er
e
g
r
ea
ter
n
u
m
b
e
r
s
in
d
icate
a
s
tr
o
n
g
er
m
atch
.
W
h
e
r
e:
r
ep
r
esen
ts
th
e
cu
r
r
en
t
v
alu
e
T
h
e
v
ar
iab
le
r
ep
r
esen
ts
th
e
ex
p
ec
ted
v
alu
e.
T
h
e
ter
m
̅
r
ep
r
esen
ts
th
e
av
er
ag
e
o
f
th
e
ac
tu
al
n
u
m
b
er
s
.
r
ep
r
esen
ts
th
e
to
tal
co
u
n
t
o
f
o
b
s
er
v
atio
n
s
.
T
h
e
n
u
m
er
ato
r
in
th
is
f
o
r
m
u
la
in
d
icat
es
th
e
ag
g
r
e
g
ate
o
f
s
q
u
ar
ed
er
r
o
r
s
in
th
e
f
o
r
ec
ast,
wh
ile
th
e
d
en
o
m
i
n
ato
r
is
th
e
o
v
er
all
v
ar
ian
ce
in
t
h
e
d
ata
.
A
v
alu
e
o
f
1
f
o
r
R
²
s
ig
n
if
ies
a
c
o
m
p
lete
m
atch
,
wh
ile
a
v
al
u
e
o
f
0
s
h
o
ws
th
at
th
e
m
o
d
el
f
ails
to
ac
co
u
n
t
f
o
r
an
y
o
f
t
h
e
v
ar
iatio
n
s
in
th
e
r
e
s
p
o
n
s
e
d
ata
ar
o
u
n
d
its
av
er
ag
e.
2
=
1
−
∑
|
−
|
=
1
∑
|
−
̅
|
=
1
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
I
t
is
co
n
ce
iv
ab
le
f
o
r
a
Seq
2
Seq
m
o
d
el
to
h
av
e
a
h
ig
h
R
2
v
a
lu
e
b
u
t
r
elativ
ely
lo
w
MA
E
,
MSE
,
an
d
R
MSE
v
alu
es
f
o
r
a
n
u
m
b
e
r
o
f
d
if
f
e
r
en
t
r
ea
s
o
n
s
.
T
h
er
e
is
a
p
o
s
s
ib
ilit
y
th
at
Seq
2
Seq
is
a
u
s
ef
u
l
m
eth
o
d
f
o
r
id
en
tify
in
g
s
ev
er
e
v
ar
ian
ce
s
o
r
o
u
tlier
s
in
th
e
d
ata
th
at
h
av
e
a
s
u
b
s
tan
tial
im
p
ac
t
o
n
th
e
R
2
s
co
r
e.
A
h
ig
h
R
2
n
u
m
b
er
s
u
g
g
ests
th
at
th
e
m
o
d
el
is
ef
f
ec
tiv
e
in
ex
p
lain
in
g
b
ig
f
lu
ctu
atio
n
s
in
th
e
d
ata
as
a
wh
o
le,
b
u
t
if
th
ese
o
u
tlier
s
ar
e
n
o
t
d
elete
d
o
r
c
o
n
tr
o
lled
ef
f
icien
tl
y
,
th
e
y
h
a
v
e
t
h
e
p
o
te
n
tial
to
i
n
ter
f
er
e
with
t
h
e
MA
E
,
MSE
,
an
d
R
MSE
v
alu
es,
wh
ich
ar
e
m
o
r
e
s
en
s
itiv
e
to
ab
s
o
lu
te
e
r
r
o
r
th
an
R
2
.
Seq
u
e
n
ce
-
to
-
Seq
u
en
ce
(
Seq
2
Seq
)
m
o
d
els
o
f
ten
ef
f
ec
tiv
ely
ca
p
t
u
r
e
tem
p
o
r
al
co
n
n
ec
tio
n
s
an
d
th
e
co
m
p
lex
ity
o
f
tim
e
s
er
ies
d
ata.
Ad
d
itio
n
ally
,
th
is
m
ay
en
ab
le
th
e
m
o
d
el
to
m
a
k
e
c
o
r
r
ec
t
f
o
r
ec
asts
at
th
e
p
er
-
d
a
ta
p
o
in
t
p
r
ed
ictio
n
lev
el,
r
esu
ltin
g
in
in
cr
ea
s
e
d
p
r
ed
ictio
n
er
r
o
r
s
(
wo
r
s
en
in
g
MA
E
,
MSE
,
an
d
R
MSE
)
.
T
h
i
s
m
ig
h
t
b
e
a
co
n
s
eq
u
en
ce
o
f
t
h
e
m
o
d
el'
s
ab
ilit
y
to
p
r
o
d
u
ce
p
r
ed
ictio
n
s
th
at
f
o
llo
w
th
e
d
ata
tr
en
d
well
o
v
er
all
(
h
ig
h
er
R
2
)
.
B
o
th
th
e
m
a
g
n
it
u
d
e
o
f
th
e
d
ata
an
d
th
e
ab
s
o
lu
te
er
r
o
r
s
ig
n
if
ican
tly
im
p
ac
t th
e
MA
E
,
MSE
,
an
d
R
MSE
.
Sm
all
in
ac
cu
r
ac
ies in
f
o
r
ec
asts
m
ay
r
esu
lt
in
b
ig
in
cr
ea
s
es
in
th
ese
v
alu
es,
ev
en
if
th
e
m
o
d
el
n
o
r
m
ally
ad
ju
s
ts
to
th
e
p
atter
n
o
f
th
e
d
ata.
T
h
is
is
th
e
ca
s
e
wh
en
th
e
d
ata
is
h
u
g
e
in
s
ize
o
r
h
as a
h
ig
h
d
eg
r
ee
o
f
v
ar
ian
ce
.
A
co
m
p
lete
im
ag
e
o
f
r
ain
f
al
l
is
p
r
o
v
id
e
d
b
y
C
HI
R
PS
d
ata,
wh
ich
co
m
b
in
es
o
b
s
er
v
atio
n
s
f
r
o
m
s
atellite
s
an
d
g
r
o
u
n
d
s
tatio
n
s
.
T
h
is
p
ictu
r
e
m
ay
b
e
b
en
e
f
icial
f
o
r
m
o
d
els th
at
ar
e
u
s
ed
to
an
ticip
ate
o
r
ev
alu
ate
wea
th
er
-
r
elate
d
p
r
o
b
lem
s
s
u
ch
as
d
r
o
u
g
h
t.
On
th
e
b
asis
o
f
th
e
ab
u
n
d
a
n
t
an
d
in
tr
icate
c
h
ar
ac
ter
is
tics
o
f
th
e
C
HI
R
P
S
d
ata,
th
e
f
o
llo
win
g
a
r
e
a
f
ew
p
r
o
b
ab
le
ca
u
s
es
f
o
r
th
e
d
is
p
ar
ate
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
m
o
d
els
s
h
o
wn
in
th
e
tab
le.
Seq
2
Seq
m
o
d
els
ca
n
m
o
r
e
r
ea
d
ily
ex
p
lain
tem
p
o
r
al
(
tim
e)
a
n
d
s
p
atial
(
s
p
a
ce
)
f
lu
ctu
atio
n
s
in
C
HI
R
P
S
d
ata
th
an
C
NN
o
r
C
NN
-
L
STM
m
o
d
els.
T
h
is
is
b
e
ca
u
s
e
th
e
Seq
2
Seq
m
o
d
el
tak
e
s
in
to
ac
co
u
n
t
b
o
th
f
ac
to
r
s
s
im
u
ltan
eo
u
s
ly
.
T
h
is
m
ay
b
e
th
e
r
ea
s
o
n
wh
y
R
2
is
h
ig
h
(
th
e
m
o
d
el
ex
p
lain
s
a
s
ig
n
if
ican
t
am
o
u
n
t
o
f
th
e
v
ar
ian
ce
in
th
e
d
ata)
,
b
u
t
th
e
ab
s
o
lu
te
er
r
o
r
s
(
MA
E
,
MSE
,
an
d
R
MSE
)
ar
e
also
h
ig
h
b
ec
a
u
s
e
o
f
t
h
e
d
if
f
icu
lty
o
f
th
e
m
o
d
el
in
ca
p
t
u
r
in
g
s
p
ec
if
ic
f
ea
tu
r
es
o
f
th
e
l
o
ca
l
en
v
ir
o
n
m
en
t,
as
ca
n
b
e
s
ee
n
in
T
ab
le
1
.
I
t
is
p
o
s
s
ib
le
f
o
r
r
ain
f
all
s
tatis
tic
s
to
h
av
e
a
s
ig
n
if
ica
n
t
am
o
u
n
t
o
f
v
ar
iatio
n
,
d
ep
en
d
in
g
o
n
th
e
lo
ca
tio
n
an
d
th
e
tim
e
p
er
io
d
.
Mo
d
els
th
at
ca
n
ca
p
tu
r
e
o
v
er
all
tr
en
d
s
m
ay
o
n
ly
s
o
m
etim
es
b
e
s
u
cc
ess
f
u
l
wh
en
f
o
r
ec
asti
n
g
p
r
ec
is
e
v
alu
es
at
p
a
r
ticu
lar
ti
m
es
an
d
p
lace
s
,
w
h
ich
m
ig
h
t
r
esu
lt
in
lar
g
er
v
alu
e
er
r
o
r
s
.
B
ec
au
s
e
th
e
Seq
2
Seq
m
o
d
el
is
ab
le
to
ad
ju
s
t
to
th
e
o
v
er
all
tr
en
d
,
it
is
in
f
lu
e
n
ce
d
b
y
ex
tr
e
m
e
v
alu
es
in
th
e
MA
E
,
MSE
,
an
d
R
MSE
ca
lcu
latio
n
er
r
o
r
s
.
Fo
r
in
s
tan
ce
,
if
th
er
e
ar
e
o
u
tlier
s
in
th
e
r
ain
f
all
d
ata,
s
u
ch
as
v
er
y
u
n
co
m
m
o
n
h
ea
v
y
r
ain
f
all,
th
e
m
o
d
el
is
ab
le
to
ad
ap
t
to
th
e
g
e
n
er
al
tr
en
d
.
T
h
e
d
if
f
er
en
t
m
o
d
els'
ar
ch
itectu
r
al
co
m
p
lex
ity
an
d
p
ar
ticu
lar
o
p
e
r
atio
n
s
ac
co
u
n
t f
o
r
th
e
d
is
p
ar
ities
in
m
em
o
r
y
u
s
e
an
d
tr
ain
in
g
tim
e
o
b
s
er
v
ed
,
as seen
in
T
ab
le
2
.
C
o
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
(
C
NNs
)
,
o
f
ten
n
ee
d
a
lar
g
e
am
o
u
n
t
o
f
m
em
o
r
y
b
ec
au
s
e
o
f
its
m
an
y
co
n
v
o
l
u
tio
n
al
lay
er
s
,
wh
ich
r
e
co
r
d
s
p
atial
h
ier
ar
ch
ies
in
in
p
u
t,
s
u
ch
as
p
ictu
r
es.
Sti
l
l,
g
iv
en
th
eir
s
im
p
le
lay
er
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:
2088
-
8
7
0
8
E
n
h
a
n
ci
n
g
s
p
a
tio
temp
o
r
a
l wea
th
er fo
r
ec
a
s
tin
g
a
cc
u
r
a
cy
w
ith
3
D
co
n
v
o
lu
tio
n
a
l
…
(
R
en
a
ld
y
F
r
ed
ya
n
)
2027
s
tr
u
ctu
r
e
th
at
p
er
m
its
p
ar
allel
d
ata
p
r
o
ce
s
s
in
g
,
th
eir
tr
ain
in
g
tim
e
p
er
ep
o
c
h
is
q
u
ite
ef
f
icie
n
t.
L
o
n
g
er
tr
ain
in
g
p
er
io
d
s
an
d
h
ig
h
er
m
em
o
r
y
c
o
n
s
u
m
p
tio
n
ar
e
f
ea
tu
r
es
o
f
C
NN
-
L
STM
m
o
d
els,
wh
ich
m
ix
co
n
v
o
l
u
tio
n
al
an
d
L
STM
lay
er
s
.
T
h
e
m
o
d
el
b
ec
o
m
es
m
o
r
e
co
m
p
licated
a
n
d
r
eso
u
r
ce
-
in
te
n
s
iv
e
as
L
S
T
M
lay
er
s
ca
p
tu
r
e
tem
p
o
r
al
r
elatio
n
s
h
ip
s
an
d
C
NN
lev
els
ex
tr
ac
t
s
p
atia
l
d
ata.
C
o
m
p
ar
in
g
C
o
n
v
L
STM
m
o
d
els
to
d
is
tin
ct
C
NN
an
d
L
STM
lay
er
s
,
C
o
n
v
L
STM
m
o
d
els
h
an
d
le
s
p
atio
tem
p
o
r
al
in
p
u
t
d
i
r
ec
tly
b
y
co
m
b
in
in
g
co
n
v
o
lu
tio
n
p
r
o
ce
s
s
es with
L
STM
u
n
its
.
T
ab
le
1
.
Per
f
o
r
m
an
ce
d
ata
f
o
r
r
ain
f
all
f
o
r
ec
asti
n
g
p
r
ed
ictin
g
th
e
n
ex
t f
iv
e
o
b
s
er
v
atio
n
s
(
5
→
5
)
u
s
in
g
th
e
p
r
ev
io
u
s
f
i
v
e
o
b
s
er
v
atio
n
s
(
g
r
i
d
s
)
M
o
d
e
l
M
A
E
M
S
E
R
M
S
E
R
2
C
N
N
0
.
0
3
8
0
.
0
0
4
0
.
0
6
8
0
.
0
2
4
C
N
N
-
LSTM
0
.
0
3
6
0
.
0
0
4
0
.
0
6
6
0
.
0
4
7
C
o
n
v
LST
M
0
.
0
3
7
0
.
0
0
4
0
.
0
6
5
0
.
0
5
8
C
o
n
v
S
e
q
2
S
e
q
0
.
0
7
3
0
.
0
3
4
0
.
1
8
6
0
.
9
6
5
T
ab
le
2.
E
f
f
icien
cy
a
n
aly
s
is
o
f
m
o
d
el
m
em
o
r
y
u
s
ag
e
an
d
tr
ai
n
in
g
tim
e
M
o
d
e
l
M
e
m
o
r
y
u
s
a
g
e
(
M
B
)
Tr
a
i
n
i
n
g
t
i
me
(
s)
Tr
a
i
n
i
n
g
t
i
me
/
e
p
o
c
h
(
s)
C
N
N
2
0
0
5
.
0
7
1
6
.
5
6
0
.
8
2
C
N
N
-
LSTM
2
2
3
2
.
0
2
2
5
.
7
2
0
.
8
0
C
o
n
v
LST
M
1
4
5
7
.
7
6
1
0
.
9
4
0
.
5
4
C
o
n
v
S
e
q
2
S
e
q
5
6
5
.
0
7
1
0
3
7
.
7
0
2
0
.
7
5
T
o
d
em
o
n
s
tr
ate
h
o
w
clo
s
ely
th
e
m
o
d
el'
s
p
r
ed
ictio
n
s
m
atch
th
e
ac
tu
al
v
alu
es,
th
e
3
D
v
is
u
aliza
tio
n
in
Fig
u
r
e
3
co
n
tr
asts
th
e
ac
tu
al
an
d
p
r
o
jecte
d
d
ata.
Plo
ts
o
f
t
h
e
d
ata
s
h
o
w
v
al
u
es,
latitu
d
e,
an
d
lo
n
g
itu
d
e.
T
h
e
r
ea
l
d
ata
p
lo
t
o
n
th
e
lef
t
d
is
p
lay
s
a
co
m
p
licated
s
u
r
f
ac
e
wit
h
clea
r
v
alu
e
f
lu
ctu
atio
n
s
o
v
er
s
ev
er
al
g
e
o
g
r
a
p
h
ic
ar
ea
s
.
T
h
is
s
u
r
f
ac
e
is
tr
ied
to
b
e
d
u
p
licated
in
t
h
e
p
r
o
j
ec
ted
d
ata
p
lo
t
(
r
ig
h
t)
.
T
h
o
u
g
h
t
h
e
ac
tu
al
an
d
an
ticip
ated
d
ata
a
r
e
co
m
p
ar
a
b
le,
th
er
e
a
r
e
d
if
f
er
en
ce
s
in
ce
r
tain
p
lace
s
th
at
p
o
in
t
to
p
lace
s
wh
er
e
th
e
m
o
d
el'
s
p
r
ed
ictio
n
s
d
if
f
er
f
r
o
m
t
h
e
r
e
al
v
alu
es.
Fig
u
r
es
4
an
d
5
s
h
o
w
h
o
w
well
two
d
is
tin
ct
m
o
d
els
—
C
o
n
v
Seq
2
Seq
an
d
C
o
n
v
L
STM
—
p
r
ed
ict
r
ai
n
f
all
th
r
o
u
g
h
o
u
t
a
f
iv
e
-
m
o
n
th
test
s
et
o
f
th
e
C
HI
R
P
S
d
ata.
T
h
e
C
o
n
v
Seq
2
Seq
m
o
d
el'
s
p
r
o
jecte
d
an
d
ac
tu
al
r
ain
f
all
m
ap
s
ar
e
co
m
p
a
r
ed
in
Fig
u
r
e
3
f
o
r
ea
ch
o
f
th
e
f
iv
e
m
o
n
th
s
.
W
h
er
ea
s
th
e
an
ticip
ated
m
ap
s
in
d
icate
th
e
m
o
d
el'
s
f
o
r
ec
asts
,
th
e
g
r
o
u
n
d
tr
u
th
m
ap
s
r
ef
lect
th
e
ac
tu
al
r
ain
f
all
th
at
h
as
b
ee
n
m
ea
s
u
r
ed
.
Fig
u
r
e
3
.
3
D
Vis
u
aliza
tio
n
o
f
ac
tu
al
v
s
p
r
ed
icted
d
ata
V
a
lu
e
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.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
0
2
2
-
2
0
3
0
2028
Fig
u
r
e
4
.
An
ex
am
p
le
o
f
p
r
ed
i
ctin
g
r
ain
f
all
o
n
a
test
s
et
o
f
th
e
C
HI
R
P
S d
ataset
u
s
in
g
C
o
n
v
Seq
2
Seq
Fig
u
r
e
5
.
An
ex
am
p
le
o
f
p
r
ed
i
ctin
g
r
ain
f
all
o
n
a
test
s
et
o
f
th
e
C
HI
R
P
S d
ataset
u
s
in
g
C
o
n
v
L
STM
5.
CO
NCLU
SI
O
N
T
h
e
s
tu
d
y
h
as
s
h
o
wn
th
at
th
e
ef
f
ec
tiv
e
u
tili
za
tio
n
o
f
h
ar
d
war
e
s
ig
n
if
ican
tly
im
p
ac
t
s
m
ac
h
in
e
lear
n
in
g
m
o
d
els'
tr
ain
in
g
d
u
r
atio
n
an
d
r
eso
u
r
ce
co
n
s
u
m
p
tio
n
.
B
y
u
s
in
g
p
ar
allel
p
r
o
ce
s
s
in
g
an
d
h
ig
h
-
p
er
f
o
r
m
an
ce
g
r
ap
h
ics
p
r
o
ce
s
s
in
g
u
n
it
s
(
GPU
s
)
to
o
p
tim
ize
h
ar
d
war
e
u
s
e,
tr
ain
in
g
ti
m
e
an
d
o
p
er
atin
g
ex
p
en
s
es
m
ay
b
e
cu
t.
Mo
r
e
ex
ten
d
ed
tr
ain
i
n
g
p
er
i
o
d
s
an
d
g
r
ea
ter
m
e
m
o
r
y
u
s
e
ar
e
ty
p
ical
o
f
m
o
r
e
co
m
p
licated
ar
c
h
itectu
r
es,
as
th
e
tab
le
co
m
p
ar
in
g
s
ev
er
al
m
o
d
els
(
C
NN,
C
NN
-
L
STM
,
C
o
n
v
L
STM
,
an
d
C
o
n
v
Seq
2
Seq
)
d
em
o
n
s
tr
ates.
T
h
e
s
p
atio
tem
p
o
r
al
d
ata
p
r
o
c
ess
in
g
ef
f
icien
cy
o
f
m
o
d
els
s
u
ch
as
C
o
n
v
L
STM
lead
s
to
im
p
r
o
v
ed
m
em
o
r
y
u
s
e
an
d
tr
ain
in
g
tim
e
p
e
r
f
o
r
m
an
ce
.
T
h
e
3
D
v
is
u
aliza
tio
n
s
co
m
p
ar
in
g
ac
tu
al
an
d
p
r
ed
icted
d
ata
s
h
o
w
th
e
d
if
f
icu
lties
o
f
p
r
ec
is
ely
m
o
d
elin
g
r
ea
l
-
wo
r
ld
d
ata.
T
h
ese
d
if
f
e
r
en
ce
s
r
esu
lt
f
r
o
m
o
v
er
f
itti
n
g
o
r
u
n
d
er
f
itti
n
g
,
h
y
p
er
p
a
r
am
eter
tu
n
in
g
,
f
ea
tu
r
e
s
elec
tio
n
,
tr
ain
in
g
d
ata
co
n
s
tr
ain
ts
,
an
d
m
o
d
el
co
m
p
lex
ity
.
I
t
is
clea
r
f
r
o
m
co
m
p
ar
in
g
th
e
C
o
n
v
Seq
2
Seq
a
n
d
C
o
n
v
L
STM
m
o
d
els
f
o
r
r
ai
n
f
all
p
r
ed
ictio
n
o
n
th
e
C
HI
R
P
S
d
ataset
th
at,
wh
il
e
b
o
th
m
o
d
els
ca
p
tu
r
e
b
r
o
ad
p
atter
n
s
o
f
r
ain
f
all,
th
e
C
o
n
v
L
STM
m
o
d
el
d
o
es
a
b
etter
jo
b
o
f
p
r
ec
is
ely
ca
p
tu
r
in
g
g
eo
g
r
ap
h
ic
d
is
tr
ib
u
tio
n
a
n
d
i
n
ten
s
ity
.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
is
wo
r
k
is
s
u
p
p
o
r
ted
b
y
B
in
a
Nu
s
an
tar
a
Un
iv
er
s
ity
a
s
a
p
ar
t
o
f
B
in
a
Nu
s
an
tar
a
Un
iv
er
s
ity
'
s
B
I
NUS
R
esear
ch
f
o
r
E
ar
ly
C
ar
ee
r
R
esear
ch
er
s
en
titl
ed
"
A
n
a
lis
is
S
p
a
s
ia
l
-
Temp
o
r
a
l
d
en
g
a
n
LS
TM
u
n
tu
k
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:
2088
-
8
7
0
8
E
n
h
a
n
ci
n
g
s
p
a
tio
temp
o
r
a
l wea
th
er fo
r
ec
a
s
tin
g
a
cc
u
r
a
cy
w
ith
3
D
co
n
v
o
lu
tio
n
a
l
…
(
R
en
a
ld
y
F
r
ed
ya
n
)
2029
Men
in
g
ka
tka
n
P
r
ed
iksi
C
u
r
a
h
Hu
ja
n
B
u
la
n
a
n
"
with
co
n
t
r
ac
t
n
u
m
b
er
:
0
6
9
A/VR
R
T
T
/I
I
I
/2
0
2
4
an
d
c
o
n
tr
ac
t
d
ate:
Ma
r
ch
1
8
,
2
0
2
4
.
RE
F
E
R
E
NC
E
S
[
1
]
S
.
F
.
T
e
k
i
n
,
O
.
K
a
r
a
a
h
m
e
t
o
g
l
u
,
F
.
I
l
h
a
n
,
I
.
B
a
l
a
b
a
n
,
a
n
d
S
.
S
.
K
o
z
a
t
,
“
S
p
a
t
i
o
-
t
e
m
p
o
r
a
l
w
e
a
t
h
e
r
f
o
r
e
c
a
s
t
i
n
g
a
n
d
a
t
t
e
n
t
i
o
n
mec
h
a
n
i
sm
o
n
c
o
n
v
o
l
u
t
i
o
n
a
l
LS
TM
s
,
”
ArX
i
v
,
p
p
.
1
–
1
3
,
2
0
2
1
.
[
2
]
J.
A
.
B
r
o
t
z
g
e
e
t
a
l
.
,
“
C
h
a
l
l
e
n
g
e
s
a
n
d
o
p
p
o
r
t
u
n
i
t
i
e
s
i
n
n
u
m
e
r
i
c
a
l
w
e
a
t
h
e
r
p
r
e
d
i
c
t
i
o
n
,
”
Bu
l
l
e
t
i
n
o
f
t
h
e
Am
e
r
i
c
a
n
Me
t
e
o
ro
l
o
g
i
c
a
l
S
o
c
i
e
t
y
,
v
o
l
.
1
0
4
,
n
o
.
3
,
p
p
.
E
6
9
8
--
E7
0
5
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
7
5
/
B
A
M
S
-
D
-
22
-
0
1
7
2
.
1
.
[
3
]
R
.
F
r
e
d
y
a
n
a
n
d
K
.
E.
S
e
t
i
a
w
a
n
,
“
A
n
i
n
v
e
st
i
g
a
t
i
o
n
i
n
t
o
i
mp
r
o
v
i
n
g
e
l
n
i
ñ
o
-
s
o
u
t
h
e
r
n
o
sc
i
l
l
a
t
i
o
n
p
r
e
d
i
c
t
i
o
n
b
a
sed
o
n
t
e
m
p
o
r
a
l
t
r
a
n
sf
o
r
mer
a
r
c
h
i
t
e
c
t
u
r
e
,
”
C
o
m
m
u
n
i
c
a
t
i
o
n
s
i
n
M
a
t
h
e
m
a
t
i
c
a
l
B
i
o
l
o
g
y
a
n
d
N
e
u
r
o
sc
i
e
n
c
e
,
2
0
2
4
,
d
o
i
:
1
0
.
2
8
9
1
9
/
c
m
b
n
/
8
3
7
1
.
[
4
]
X
.
S
h
i
,
Z
.
C
h
e
n
,
H
.
W
a
n
g
,
D
.
-
Y
.
Y
e
u
n
g
,
W
.
-
K
.
W
o
n
g
,
a
n
d
W
.
W
o
o
,
“
C
o
n
v
o
l
u
t
i
o
n
a
l
LST
M
n
e
t
w
o
r
k
:
a
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
p
r
e
c
i
p
i
t
a
t
i
o
n
n
o
w
c
a
s
t
i
n
g
,
”
A
d
v
a
n
c
e
s i
n
n
e
u
r
a
l
i
n
f
o
rm
a
t
i
o
n
p
r
o
c
e
ssi
n
g
s
y
st
e
m
s
,
v
o
l
.
2
8
,
2
0
1
5
.
[
5
]
C.
-
A
.
D
i
a
c
o
n
u
,
S
.
S
a
h
a
,
S
.
G
u
n
n
e
ma
n
n
,
a
n
d
X
.
X
i
a
n
g
Z
h
u
,
“
U
n
d
e
r
st
a
n
d
i
n
g
t
h
e
r
o
l
e
o
f
w
e
a
t
h
e
r
d
a
t
a
f
o
r
e
a
r
t
h
s
u
r
f
a
c
e
f
o
r
e
c
a
s
t
i
n
g
u
si
n
g
a
C
o
n
v
LST
M
-
b
a
se
d
mo
d
e
l
,
”
i
n
2
0
2
2
I
EEE
/
C
V
F
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
rn
Re
c
o
g
n
i
t
i
o
n
Wo
rks
h
o
p
s
(
C
VPR
W)
,
J
u
n
.
2
0
2
2
,
p
p
.
1
3
6
1
–
1
3
7
0
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
W
5
6
3
4
7
.
2
0
2
2
.
0
0
1
4
2
.
[
6
]
Y
.
D
i
n
g
,
Y
.
Z
h
u
,
J.
F
e
n
g
,
P
.
Z
h
a
n
g
,
a
n
d
Z.
C
h
e
n
g
,
“
I
n
t
e
r
p
r
e
t
a
b
l
e
s
p
a
t
i
o
-
t
e
mp
o
r
a
l
a
t
t
e
n
t
i
o
n
LST
M
m
o
d
e
l
f
o
r
f
l
o
o
d
f
o
r
e
c
a
st
i
n
g
,
”
N
e
u
ro
c
o
m
p
u
t
i
n
g
,
v
o
l
.
4
0
3
,
p
p
.
3
4
8
–
3
5
9
,
A
u
g
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m
.
2
0
2
0
.
0
4
.
1
1
0
.
[
7
]
C.
-
F
.
R
.
C
h
e
n
e
t
a
l
.
,
“
D
e
e
p
a
n
a
l
y
s
i
s
o
f
C
N
N
-
b
a
s
e
d
s
p
a
t
i
o
-
t
e
m
p
o
r
a
l
r
e
p
r
e
s
e
n
t
a
t
i
o
n
s
f
o
r
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
2
0
2
1
I
E
E
E
/
C
V
F
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
V
i
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
(
C
V
P
R
)
,
J
u
n
.
2
0
2
1
,
p
p
.
6
1
6
1
–
6
1
7
1
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
4
6
4
3
7
.
2
0
2
1
.
0
0
6
1
0
.
[
8
]
C
.
W
a
n
g
,
K
.
C
h
o
,
a
n
d
J
.
G
u
,
“
N
e
u
r
a
l
m
a
c
h
i
n
e
t
r
a
n
sl
a
t
i
o
n
w
i
t
h
b
y
t
e
-
l
e
v
e
l
su
b
w
o
r
d
s,
”
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
AA
AI
C
o
n
f
e
r
e
n
c
e
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
3
4
,
n
o
.
0
5
,
p
p
.
9
1
5
4
–
9
1
6
0
,
A
p
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
6
0
9
/
a
a
a
i
.
v
3
4
i
0
5
.
6
4
5
1
.
[
9
]
M
.
C
.
S
c
h
i
a
p
p
a
e
t
a
l
.
,
“
A
l
a
r
g
e
-
sca
l
e
r
o
b
u
st
n
e
s
s
a
n
a
l
y
si
s
o
f
v
i
d
e
o
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
m
o
d
e
l
s,
”
i
n
2
0
2
3
I
E
EE/
C
V
F
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
Re
c
o
g
n
i
t
i
o
n
(
C
VP
R)
,
J
u
n
.
2
0
2
3
,
p
p
.
1
4
6
9
8
–
1
4
7
0
8
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
5
2
7
2
9
.
2
0
2
3
.
0
1
4
1
2
.
[
1
0
]
J
.
D
i
e
z
-
S
i
e
r
r
a
a
n
d
M
.
d
e
l
J
e
s
u
s
,
“
L
o
n
g
-
t
e
r
m
r
a
i
n
f
a
l
l
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
a
t
m
o
s
p
h
e
r
i
c
s
y
n
o
p
t
i
c
p
a
t
t
e
r
n
s
i
n
s
e
m
i
-
a
r
i
d
c
l
i
m
a
t
e
s
w
i
t
h
s
t
a
t
i
s
t
i
c
a
l
a
n
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
m
e
t
h
o
d
s
,
”
J
o
u
r
n
a
l
o
f
H
y
d
r
o
l
o
g
y
,
v
o
l
.
5
8
6
,
p
.
1
2
4
7
8
9
,
J
u
l
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
h
y
d
r
o
l
.
2
0
2
0
.
1
2
4
7
8
9
.
[
1
1
]
Y
.
J
i
n
,
D
.
Y
a
n
,
X
.
K
a
n
g
,
A
.
C
h
o
n
g
,
H
.
--
S
u
n
,
a
n
d
S
.
Z
h
a
n
,
“
F
o
r
e
c
a
s
t
i
n
g
b
u
i
l
d
i
n
g
o
c
c
u
p
a
n
c
y
:
a
t
e
m
p
o
r
a
l
-
s
e
q
u
e
n
t
i
a
l
a
n
a
l
y
s
i
s
a
n
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
i
n
t
e
g
r
a
t
e
d
a
p
p
r
o
a
c
h
,
”
E
n
e
r
g
y
a
n
d
B
u
i
l
d
i
n
g
s
,
v
o
l
.
2
5
2
,
p
.
1
1
1
3
6
2
,
D
e
c
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
b
u
i
l
d
.
2
0
2
1
.
1
1
1
3
6
2
.
[
1
2
]
Y
.
Zh
a
n
g
e
t
a
l
.
,
“
E
v
a
l
u
a
t
i
o
n
a
n
d
c
o
m
p
a
r
i
s
o
n
o
f
p
r
e
c
i
p
i
t
a
t
i
o
n
e
s
t
i
mat
e
s
a
n
d
h
y
d
r
o
l
o
g
i
c
u
t
i
l
i
t
y
o
f
C
H
I
R
P
S
,
TR
M
M
3
B
4
2
V
7
a
n
d
P
ER
S
I
A
N
N
-
C
D
R
p
r
o
d
u
c
t
s
i
n
v
a
r
i
o
u
s
c
l
i
m
a
t
e
r
e
g
i
mes
,
”
At
m
o
s
p
h
e
ri
c
Re
s
e
a
r
c
h
,
v
o
l
.
2
6
5
,
p
.
1
0
5
8
8
1
,
Jan
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
t
mo
sr
e
s.
2
0
2
1
.
1
0
5
8
8
1
.
[
1
3
]
“
S
p
a
t
i
o
t
e
m
p
o
r
a
l
c
o
n
v
o
l
u
t
i
o
n
a
l
LS
T
M
w
i
t
h
a
t
t
e
n
t
i
o
n
m
e
c
h
a
n
i
sm
f
o
r
mo
n
t
h
l
y
r
a
i
n
f
a
l
l
p
r
e
d
i
c
t
i
o
n
,
”
C
o
m
m
u
n
i
c
a
t
i
o
n
s
i
n
Ma
t
h
e
m
a
t
i
c
a
l
B
i
o
l
o
g
y
a
n
d
N
e
u
r
o
sc
i
e
n
c
e
,
v
o
l
.
1
1
8
,
n
o
.
1
,
p
p
.
1
–
2
5
,
2
0
2
2
,
d
o
i
:
1
0
.
2
8
9
1
9
/
c
m
b
n
/
7
7
6
1
.
[
1
4
]
M
.
M
.
H
.
K
h
a
n
,
N
.
S
.
M
u
h
a
mm
a
d
,
a
n
d
A
.
El
-
S
h
a
f
i
e
,
“
W
a
v
e
l
e
t
b
a
s
e
d
h
y
b
r
i
d
A
N
N
-
A
R
I
M
A
mo
d
e
l
s
f
o
r
met
e
o
r
o
l
o
g
i
c
a
l
d
r
o
u
g
h
t
f
o
r
e
c
a
st
i
n
g
,
”
J
o
u
rn
a
l
o
f
H
y
d
r
o
l
o
g
y
,
v
o
l
.
5
9
0
,
p
.
1
2
5
3
8
0
,
N
o
v
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
h
y
d
r
o
l
.
2
0
2
0
.
1
2
5
3
8
0
.
[
1
5
]
S
.
M
e
h
d
i
z
a
d
e
h
,
“
A
ssess
i
n
g
t
h
e
p
o
t
e
n
t
i
a
l
o
f
d
a
t
a
-
d
r
i
v
e
n
m
o
d
e
l
s
f
o
r
e
s
t
i
m
a
t
i
o
n
o
f
l
o
n
g
-
t
e
r
m
m
o
n
t
h
l
y
t
e
m
p
e
r
a
t
u
r
e
s,
”
C
o
m
p
u
t
e
rs
a
n
d
El
e
c
t
r
o
n
i
c
s i
n
Ag
r
i
c
u
l
t
u
re
,
v
o
l
.
1
4
4
,
p
p
.
1
1
4
–
1
2
5
,
J
a
n
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
m
p
a
g
.
2
0
1
7
.
1
1
.
0
3
8
.
[
1
6
]
R
.
F
r
e
d
y
a
n
,
M
.
R
.
N
.
M
a
j
i
i
d
,
a
n
d
G
.
P
.
K
u
su
ma,
“
S
p
a
t
i
o
t
e
m
p
o
r
a
l
a
n
a
l
y
s
i
s
f
o
r
r
a
i
n
f
a
l
l
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
e
x
t
r
e
me
l
e
a
r
n
i
n
g
m
a
c
h
i
n
e
c
l
u
st
e
r
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
n
A
d
v
a
n
c
e
d
S
c
i
e
n
c
e
,
En
g
i
n
e
e
ri
n
g
a
n
d
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
3
,
n
o
.
6
,
p
p
.
2
2
4
0
–
2
2
4
8
,
D
e
c
.
2
0
2
3
,
d
o
i
:
1
0
.
1
8
5
1
7
/
i
j
a
s
e
i
t
.
1
3
.
6
.
1
8
2
1
4
.
[
1
7
]
S
.
N
a
r
ma
d
h
a
a
n
d
V
.
V
i
j
a
y
a
k
u
m
a
r
,
“
S
p
a
t
i
o
-
t
e
m
p
o
r
a
l
v
e
h
i
c
l
e
t
r
a
f
f
i
c
f
l
o
w
p
r
e
d
i
c
t
i
o
n
u
si
n
g
m
u
l
t
i
v
a
r
i
a
t
e
C
N
N
a
n
d
LST
M
m
o
d
e
l
,
”
Ma
t
e
r
i
a
l
s T
o
d
a
y
:
Pr
o
c
e
e
d
i
n
g
s
,
v
o
l
.
8
1
,
p
p
.
8
2
6
–
8
3
3
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
mat
p
r
.
2
0
2
1
.
0
4
.
2
4
9
.
[
1
8
]
Q
.
G
e
,
H
.
S
u
n
,
Z.
L
i
u
,
a
n
d
X
.
W
a
n
g
,
“
A
d
a
t
a
‐
d
r
i
v
e
n
i
n
t
e
l
l
i
g
e
n
t
mo
d
e
l
f
o
r
l
a
n
d
sl
i
d
e
d
i
sp
l
a
c
e
m
e
n
t
p
r
e
d
i
c
t
i
o
n
,
”
G
e
o
l
o
g
i
c
a
l
J
o
u
r
n
a
l
,
v
o
l
.
5
8
,
n
o
.
6
,
p
p
.
2
2
1
1
–
2
2
3
0
,
J
u
n
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
2
/
g
j
.
4
6
7
5
.
[
1
9
]
I
.
K
u
m
a
r
,
B
.
K
.
Tr
i
p
a
t
h
i
,
a
n
d
A
.
S
i
n
g
h
,
“
A
t
t
e
n
t
i
o
n
-
b
a
se
d
LSTM
n
e
t
w
o
r
k
-
a
ss
i
st
e
d
t
i
me
s
e
r
i
e
s
f
o
r
e
c
a
st
i
n
g
mo
d
e
l
s
f
o
r
p
e
t
r
o
l
e
u
m
p
r
o
d
u
c
t
i
o
n
,
”
E
n
g
i
n
e
e
ri
n
g
Ap
p
l
i
c
a
t
i
o
n
s
o
f
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
2
3
,
p
.
1
0
6
4
4
0
,
A
u
g
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
a
p
p
a
i
.
2
0
2
3
.
1
0
6
4
4
0
.
[
2
0
]
A
.
B
a
r
t
h
,
A
.
A
l
v
e
r
a
-
A
z
c
á
r
a
t
e
,
C
.
Tr
o
u
p
i
n
,
a
n
d
J
.
-
M
.
B
e
c
k
e
r
s,
“
D
I
N
C
A
E
2
.
0
:
mu
l
t
i
v
a
r
i
a
t
e
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
w
i
t
h
e
r
r
o
r
e
st
i
mat
e
s
t
o
r
e
c
o
n
st
r
u
c
t
se
a
s
u
r
f
a
c
e
t
e
mp
e
r
a
t
u
r
e
sat
e
l
l
i
t
e
a
n
d
a
l
t
i
met
r
y
o
b
s
e
r
v
a
t
i
o
n
s,”
G
e
o
sc
i
e
n
t
i
f
i
c
Mo
d
e
l
D
e
v
e
l
o
p
m
e
n
t
,
v
o
l
.
1
5
,
n
o
.
5
,
p
p
.
2
1
8
3
–
2
1
9
6
,
M
a
r
.
2
0
2
2
,
d
o
i
:
1
0
.
5
1
9
4
/
g
m
d
-
15
-
2
1
8
3
-
2
0
2
2
.
[
2
1
]
S
.
K
i
m
e
t
a
l
.
,
“
D
e
e
p
-
h
u
r
r
i
c
a
n
e
-
t
r
a
c
k
e
r
:
t
r
a
c
k
i
n
g
a
n
d
f
o
r
e
c
a
s
t
i
n
g
e
x
t
r
e
m
e
c
l
i
mat
e
e
v
e
n
t
s
,
”
i
n
2
0
1
9
I
EEE
Wi
n
t
e
r
C
o
n
f
e
re
n
c
e
o
n
Ap
p
l
i
c
a
t
i
o
n
s
o
f
C
o
m
p
u
t
e
r
Vi
s
i
o
n
(
WA
C
V)
,
J
a
n
.
2
0
1
9
,
p
p
.
1
7
6
1
–
1
7
6
9
,
d
o
i
:
1
0
.
1
1
0
9
/
W
A
C
V
.
2
0
1
9
.
0
0
1
9
2
.
[
2
2
]
Y
.
M
.
S
o
u
t
o
,
F
.
P
o
r
t
o
,
A
.
M
.
M
o
u
r
a
,
a
n
d
E.
B
e
z
e
r
r
a
,
“
A
s
p
a
t
i
o
t
e
m
p
o
r
a
l
e
n
s
e
mb
l
e
a
p
p
r
o
a
c
h
t
o
r
a
i
n
f
a
l
l
f
o
r
e
c
a
st
i
n
g
,
”
i
n
2
0
1
8
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
r
e
n
c
e
o
n
N
e
u
ra
l
N
e
t
w
o
rks
(
I
J
C
N
N
)
,
Ju
l
.
2
0
1
8
,
p
p
.
1
–
8
,
d
o
i
:
1
0
.
1
1
0
9
/
I
JC
N
N
.
2
0
1
8
.
8
4
8
9
6
9
3
.
[
2
3
]
K
.
E
.
S
e
t
i
a
w
a
n
,
R
.
F
r
e
d
y
a
n
,
a
n
d
I
.
N
.
A
l
a
m,
“
E
n
h
a
n
c
i
n
g
e
l
n
i
ñ
o
-
so
u
t
h
e
r
n
o
s
c
i
l
l
a
t
i
o
n
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
a
n
a
t
t
e
n
t
i
o
n
-
b
a
se
d
seq
u
e
n
c
e
-
to
-
se
q
u
e
n
c
e
a
r
c
h
i
t
e
c
t
u
r
e
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
t
ri
c
a
l
a
n
d
C
o
m
p
u
t
e
r E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
4
,
n
o
.
6
,
pp
.
7
0
5
7
-
7
0
6
6
,
D
e
c
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
4
i
6
.
p
p
7
0
5
7
-
7
0
6
6
.
[
2
4
]
Y
.
W
a
n
g
e
t
a
l
.
,
“
P
r
e
d
R
N
N
:
a
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
sp
a
t
i
o
t
e
m
p
o
r
a
l
p
r
e
d
i
c
t
i
v
e
l
e
a
r
n
i
n
g
,
”
I
EE
E
T
ra
n
s
a
c
t
i
o
n
s
o
n
P
a
t
t
e
r
n
An
a
l
y
si
s
a
n
d
M
a
c
h
i
n
e
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
4
5
,
n
o
.
2
,
p
p
.
2
2
0
8
–
2
2
2
5
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
TPA
M
I
.
2
0
2
2
.
3
1
6
5
1
5
3
.
[
2
5
]
D
.
Tr
a
n
,
H
.
W
a
n
g
,
L
.
To
r
r
e
sa
n
i
,
J
.
R
a
y
,
Y
.
Le
C
u
n
,
a
n
d
M
.
P
a
l
u
r
i
,
“
A
c
l
o
ser
l
o
o
k
a
t
sp
a
t
i
o
t
e
m
p
o
r
a
l
c
o
n
v
o
l
u
t
i
o
n
s
f
o
r
a
c
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
2
0
1
8
I
E
EE/
C
V
F
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
a
n
d
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
,
J
u
n
.
2
0
1
8
,
p
p
.
6
4
5
0
–
6
4
5
9
,
d
o
i
:
1
0
.
1
1
0
9
/
C
V
P
R
.
2
0
1
8
.
0
0
6
7
5
.
[
2
6
]
G
.
S
i
n
g
h
a
n
d
F
.
C
u
z
z
o
l
i
n
,
“
R
e
c
u
r
r
e
n
t
c
o
n
v
o
l
u
t
i
o
n
s
f
o
r
c
a
u
s
a
l
3
D
C
N
N
s,”
i
n
2
0
1
9
I
E
EE/
C
V
F
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r
Vi
s
i
o
n
W
o
rks
h
o
p
(
I
C
C
V
W)
,
O
c
t
.
2
0
1
9
,
p
p
.
1
4
5
6
–
1
4
6
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
V
W
.
2
0
1
9
.
0
0
1
8
3
.
[
2
7
]
C
.
C
h
e
n
g
,
C
.
Z
h
a
n
g
,
Y
.
W
e
i
,
a
n
d
Y
.
-
G
.
Ji
a
n
g
,
“
S
p
a
r
se
t
e
mp
o
r
a
l
c
a
u
sa
l
c
o
n
v
o
l
u
t
i
o
n
f
o
r
e
f
f
i
c
i
e
n
t
a
c
t
i
o
n
m
o
d
e
l
i
n
g
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
7
t
h
A
C
M
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
M
u
l
t
i
m
e
d
i
a
,
O
c
t
.
2
0
1
9
,
p
p
.
5
9
2
–
6
0
0
,
d
o
i
:
1
0
.
1
1
4
5
/
3
3
4
3
0
3
1
.
3
3
5
1
0
5
4
.
[
2
8
]
Z.
X
u
,
Y
.
C
a
o
,
a
n
d
Y
.
K
a
n
g
,
“
D
e
e
p
sp
a
t
i
o
t
e
mp
o
r
a
l
r
e
si
d
u
a
l
e
a
r
l
y
-
l
a
t
e
f
u
s
i
o
n
n
e
t
w
o
r
k
f
o
r
c
i
t
y
r
e
g
i
o
n
v
e
h
i
c
l
e
e
mi
ssi
o
n
p
o
l
l
u
t
i
o
n
p
r
e
d
i
c
t
i
o
n
,
”
N
e
u
r
o
c
o
m
p
u
t
i
n
g
,
v
o
l
.
3
5
5
,
p
p
.
1
8
3
–
1
9
9
,
A
u
g
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m.
2
0
1
9
.
0
4
.
0
4
0
.
[
2
9
]
M
.
M
u
d
i
g
o
n
d
a
e
t
a
l
.
,
“
D
e
e
p
l
e
a
r
n
i
n
g
f
o
r
d
e
t
e
c
t
i
n
g
e
x
t
r
e
m
e
w
e
a
t
h
e
r
p
a
t
t
e
r
n
s
,
”
D
e
e
p
L
e
a
rn
i
n
g
f
o
r
t
h
e
Ea
rt
h
S
c
i
e
n
c
e
s
,
p
p
.
1
6
1
–
1
8
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
2
/
9
7
8
1
1
1
9
6
4
6
1
8
1
.
c
h
1
2
.
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.
15
,
No
.
2
,
Ap
r
il
20
25
:
2
0
2
2
-
2
0
3
0
2030
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Re
n
a
ld
y
Fre
d
y
a
n
re
c
e
iv
e
d
a
m
a
ste
r'
s
d
e
g
re
e
in
c
o
m
p
u
ter
s
c
ien
c
e
fro
m
Bin
a
Nu
sa
n
tara
Un
iv
e
rsity
in
I
n
d
o
n
e
s
ia
in
2
0
2
3
.
He
is
c
u
rre
n
t
ly
e
m
p
lo
y
e
d
b
y
In
d
o
n
e
sia
'
s
Bin
a
Nu
sa
n
tara
Un
iv
e
rsit
y
a
s
a
lec
tu
re
r
in
th
e
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
.
His
a
re
a
s
o
f
in
tere
st
in
re
se
a
rc
h
in
c
lu
d
e
ti
m
e
se
ries
,
m
a
c
h
in
e
lea
rn
in
g
,
p
a
tt
e
r
n
re
c
o
g
n
it
i
o
n
,
a
n
d
c
o
m
p
u
ter
v
isio
n
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
re
n
a
ld
y
.
fre
d
y
a
n
@b
i
n
u
s.a
c
.
i
d
.
K
a
r
li
Ek
a
S
e
ti
a
wa
n
re
c
e
iv
e
d
a
n
S
.
S
i
.
d
e
g
re
e
in
m
a
th
e
m
a
ti
c
s
fro
m
P
a
d
jad
jara
n
Un
iv
e
rsity
in
2
0
1
5
a
n
d
a
n
M
.
Ko
m
.
d
e
g
re
e
i
n
c
o
m
p
u
ter
sc
ie
n
c
e
fro
m
Bi
n
a
Nu
sa
n
tara
Un
iv
e
rsity
i
n
2
0
2
3
.
He
a
c
h
iev
e
d
o
u
tstan
d
in
g
g
ra
d
e
s
o
f
e
x
c
e
ll
e
n
c
e
,
su
m
m
a
c
u
m
lau
d
e
,
a
n
d
b
e
st
g
ra
d
u
a
te
in
h
is
m
a
ste
r'
s
stu
d
y
.
His
re
se
a
rc
h
in
tere
sts
c
o
v
e
r
a
rti
ficia
l
in
tel
li
g
e
n
c
e
,
c
o
m
p
u
ter
v
isio
n
,
d
a
ta
sc
ien
c
e
,
d
a
ta
se
c
u
rit
y
,
b
l
o
c
k
c
h
a
i
n
,
a
n
d
m
a
t
h
e
m
a
ti
c
a
l
m
o
d
e
li
n
g
.
He
is
a
n
a
ss
o
c
iate
lec
tu
re
r
sp
e
c
ialist
S
2
in
t
h
e
S
c
h
o
o
l
o
f
C
o
m
p
u
ter
S
c
ien
c
e
a
t
Bi
n
a
Nu
sa
n
tara
Un
iv
e
rsity
.
He
a
ls
o
h
a
s
e
x
p
e
rien
c
e
a
s
a
c
re
d
it
a
n
a
ly
st
a
t
a
c
o
n
v
e
n
t
io
n
a
l
b
a
n
k
,
w
h
ich
is
p
a
rt
o
f
a
sta
te
-
o
wn
e
d
e
n
terp
rise
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
k
a
rli
.
se
ti
a
wa
n
@b
i
n
u
s.a
c
.
i
d
.
K
e
lv
in
As
c
lepi
u
s
M
in
o
r
is
a
m
a
th
e
m
a
ti
c
s
lec
tu
re
r
a
t
Bin
a
Nu
s
a
n
tara
Un
iv
e
rsit
y
(BINU
S
).
Driv
e
n
b
y
a
stro
n
g
c
u
ri
o
sity
to
u
n
d
e
rsta
n
d
h
o
w t
h
i
n
g
s
w
o
rk
,
h
e
a
ims
t
o
c
o
n
tri
b
u
te
to
th
e
o
r
g
a
n
iza
ti
o
n
'
s
g
r
o
wth
b
y
lev
e
ra
g
in
g
h
is
k
n
o
wle
d
g
e
a
n
d
fo
c
u
si
n
g
o
n
d
e
tailed
d
e
sc
rip
ti
o
n
s
,
p
re
c
ise
sp
e
c
ifi
c
a
ti
o
n
s,
a
n
d
fa
c
t
u
a
l
i
n
sig
h
ts.
His
p
a
ss
io
n
fo
r
e
n
g
a
g
i
n
g
d
irec
tl
y
wit
h
h
is
e
n
v
iro
n
m
e
n
t
a
li
g
n
s with
h
is i
n
ter
e
st i
n
lea
rn
in
g
sy
ste
m
s a
n
d
u
n
d
e
r
sta
n
d
in
g
th
e
a
c
ti
o
n
s req
u
ired
to
a
c
h
iev
e
ta
n
g
i
b
le o
u
tco
m
e
s.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
k
e
lv
i
n
.
m
in
o
r@b
i
n
u
s.a
c
.
i
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.