I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
10
,
No
.
6
,
Decem
b
er
2
0
2
0
,
p
p
.
6
5
7
4
~
6
5
8
1
I
SS
N:
2088
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
10
i
6
.
pp
6
5
7
4
-
6
5
8
1
6574
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
Predic
tion o
f
a
t
mo
spheric po
llutio
n using
neural
ne
tw
o
rk
s
m
o
del o
f
fin
e part
icles in
the
tow
n o
f
K
ennedy
i
n Bo
g
o
tá
J
ua
n Ca
m
ilo
P
edra
za
,
O
s
w
a
ldo
Albert
o
Ro
m
er
o
,
H
elber
t
E
du
a
rdo
E
s
pitia
F
a
c
u
lt
a
d
d
e
I
n
g
e
n
iería
,
Un
iv
e
rsid
a
d
Distrit
a
l
F
ra
n
c
isc
o
Jo
sé
d
e
Ca
l
d
a
s,
Co
l
o
m
b
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Feb
17
,
2
0
20
R
ev
i
s
ed
J
u
n
1
,
2020
A
cc
ep
ted
J
u
n
1
4
,
2
0
2
0
T
h
is
w
o
rk
sh
o
w
s
a
n
a
p
p
li
c
a
ti
o
n
b
a
se
d
o
n
n
e
u
ra
l
n
e
tw
o
rk
s
to
d
e
term
in
e
th
e
p
re
d
ictio
n
o
f
a
ir
p
o
ll
u
t
i
o
n
,
e
sp
e
c
ially
p
a
rti
c
u
late
m
a
teria
l
o
f
2
.
5
m
icro
m
e
ters
len
g
th
.
T
h
is
a
p
p
li
c
a
ti
o
n
is
c
o
n
si
d
e
re
d
o
f
g
re
a
t
im
p
o
rtan
c
e
d
u
e
t
o
t
h
e
im
p
a
c
t
o
n
h
u
m
a
n
h
e
a
lt
h
a
n
d
h
ig
h
im
p
a
c
t
d
u
e
to
th
e
a
g
g
lo
m
e
ra
ti
o
n
o
f
p
e
o
p
le
i
n
c
it
ies
.
T
h
e
im
p
le
m
e
n
tatio
n
is
p
e
rf
o
rm
e
d
u
sin
g
d
a
ta
c
a
p
tu
re
d
f
ro
m
se
v
e
ra
l
d
e
v
ice
s
th
a
t
c
a
n
b
e
i
n
sta
ll
e
d
i
n
sp
e
c
if
ic
lo
c
a
ti
o
n
s
f
o
r
a
p
a
rti
c
u
lar
g
e
o
g
ra
p
h
ica
l
e
n
v
iro
n
m
e
n
t,
e
sp
e
c
ially
in
th
e
lo
c
a
li
ty
o
f
Ke
n
n
e
d
y
in
B
o
g
o
tá.
T
h
e
m
o
d
e
l
o
b
tain
e
d
c
a
n
b
e
u
se
d
f
o
r
th
e
d
e
sig
n
o
f
p
u
b
li
c
p
o
li
c
ies
t
h
a
t
c
o
n
tr
o
l
a
ir
q
u
a
li
ty
.
K
ey
w
o
r
d
s
:
A
ir
m
o
n
ito
r
i
n
g
A
t
m
o
s
p
h
er
ic
p
o
llu
tio
n
Neu
r
al
n
et
w
o
r
k
s
P
ar
ticu
late
m
ater
ial
S
m
ar
t c
itie
s
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
:
Os
w
ald
o
A
lb
er
to
R
o
m
er
o
,
Facu
ltad
d
e
I
n
g
e
n
ier
ía,
Un
i
v
er
s
id
ad
Dis
tr
ital Fr
a
n
cisc
o
J
o
s
é
d
e
C
ald
as,
C
r
a.
7
#
4
0
b
-
5
3
,
B
o
g
o
tá,
C
o
lo
m
b
ia,
E
m
ail: o
r
o
m
er
o
v
@
u
d
i
s
tr
ital.e
d
u
.
co
1.
I
NT
RO
D
UCT
I
O
N
A
ir
p
o
llu
tio
n
i
s
an
ac
u
te
t
h
r
e
at,
it
is
a
p
h
e
n
o
m
en
o
n
th
a
t
h
as
a
p
ar
ticu
lar
i
m
p
ac
t
o
n
h
u
m
an
h
ea
lt
h
.
T
h
e
ch
an
g
e
s
th
at
o
cc
u
r
in
th
e
ch
e
m
ical
co
m
p
o
s
itio
n
o
f
t
h
e
at
m
o
s
p
h
er
e
ca
n
c
h
an
g
e
th
e
cli
m
ate,
p
r
o
d
u
ce
ac
id
r
ain
o
r
d
estro
y
t
h
e
o
zo
n
e
la
y
er
,
all
p
h
en
o
m
en
a
o
f
g
r
ea
t
g
l
o
b
al
i
m
p
o
r
tan
ce
.
T
h
e
W
o
r
ld
Hea
lth
Or
g
a
n
izatio
n
(
W
HO)
co
n
s
id
er
s
air
p
o
llu
tio
n
as o
n
e
o
f
th
e
m
o
s
t i
m
p
o
r
tan
t
g
lo
b
al
p
r
io
r
ities
[
1
]
.
T
h
e
u
s
e
o
f
n
o
n
-
r
e
n
e
w
ab
le
r
eso
u
r
ce
s
in
t
h
e
p
r
o
d
u
ctio
n
o
f
en
er
g
y
,
s
u
ch
as
o
il
o
r
co
al,
g
en
er
ate
s
i
m
p
o
r
tan
t
e
m
i
s
s
io
n
s
o
f
s
u
lf
u
r
d
io
x
id
e
(
SO2
)
,
ca
r
b
o
n
m
o
n
o
x
id
e
(
C
O)
,
a
m
o
n
g
o
t
h
er
s
.
On
t
h
e
o
t
h
er
h
an
d
,
th
e
m
ea
n
s
o
f
tr
a
n
s
p
o
r
t u
s
ed
in
ev
er
y
d
a
y
li
f
e
co
n
s
t
itu
te
an
o
t
h
er
alar
m
in
g
s
o
u
r
ce
o
f
co
n
ta
m
i
n
atio
n
.
A
lar
g
e
p
ar
t
o
f
th
e
s
e
p
o
llu
tan
ts
e
m
itted
in
t
o
th
e
en
v
ir
o
n
m
en
t i
s
g
e
n
er
ate
d
b
y
au
t
o
m
o
b
iles
[
2
]
.
I
n
co
n
s
eq
u
e
n
ce
,
th
r
ee
m
ai
n
s
tep
s
ar
e
p
r
o
p
o
s
ed
to
b
e
f
o
llo
w
ed
to
ad
d
r
ess
air
p
o
llu
tio
n
p
r
o
b
lem
s
.
T
h
e
d
ef
in
itio
n
o
f
air
q
u
alit
y
p
ar
a
m
eter
s
t
h
at
n
ee
d
to
b
e
co
n
t
r
o
lled
;
th
e
m
o
n
ito
r
in
g
o
f
t
h
ese
th
r
o
u
g
h
t
h
e
u
s
e
o
f
s
p
ec
if
ic
h
ar
d
w
ar
e
an
d
s
o
f
t
w
ar
e;
an
d
t
h
e
ad
o
p
tio
n
o
f
s
o
l
u
tio
n
s
ai
m
ed
at
r
ed
u
ci
n
g
t
h
e
co
n
ce
n
tr
atio
n
o
f
h
ar
m
f
u
l
s
u
b
s
ta
n
ce
s
a
n
d
en
s
u
r
i
n
g
clea
n
lo
n
g
-
ter
m
air
[
3
]
.
A
s
f
o
r
air
q
u
alit
y
p
ar
a
m
e
ter
s
,
t
h
ese
ar
e
d
eter
m
i
n
ed
b
y
th
e
e
n
v
ir
o
n
m
en
tal
p
o
licies
o
u
tlin
ed
b
y
t
h
e
cit
y
g
o
v
er
n
m
e
n
t
.
T
h
e
m
o
n
i
to
r
in
g
,
f
o
r
t
h
is
ex
p
er
i
m
en
t,
is
ca
r
r
ied
o
u
t
b
y
t
h
e
s
u
r
v
e
illa
n
ce
s
tat
i
o
n
s
t
h
at
ar
e
i
n
s
talled
i
n
d
i
f
f
er
en
t
p
o
in
ts
o
f
t
h
e
cit
y
.
I
n
t
h
e
ca
s
e
o
f
B
o
g
o
tá,
d
esp
ite
th
e
ef
f
o
r
ts
m
ad
e
b
y
en
v
ir
o
n
m
e
n
tal
a
u
t
h
o
r
ities
,
ac
ad
em
ic
in
s
tit
u
tio
n
s
an
d
ci
tizen
s
i
n
g
en
er
al,
th
e
ca
p
ital
cit
y
o
f
C
o
lo
m
b
ia
i
s
to
d
a
y
o
n
e
o
f
t
h
e
m
o
s
t
p
o
llu
t
ed
cities
in
L
ati
n
Am
er
ica.
O
n
e
o
f
t
h
e
p
o
ll
u
tan
ts
o
f
g
r
ea
test
c
o
n
ce
r
n
i
n
th
e
cit
y
i
s
th
e
p
ar
ticu
late
m
ater
ial,
s
in
ce
its
le
v
els
f
r
eq
u
en
tl
y
e
x
ce
ed
th
e
air
q
u
ality
s
tan
d
ar
d
s
[
2
]
.
T
h
e
Secr
etar
ia
Dis
tr
ital
d
el
A
m
b
ien
te
o
f
B
o
g
o
tá
h
as
t
h
e
“R
ed
d
e
Mo
n
ito
r
eo
d
e
C
alid
ad
d
el
A
ir
e
d
e
B
o
g
o
tá
-
R
M
C
A
B
”,
w
h
ich
co
m
p
r
i
s
es
1
3
f
i
x
ed
m
o
n
ito
r
i
n
g
s
tatio
n
s
an
d
a
m
o
b
ile
s
tatio
n
o
p
er
atin
g
in
d
if
f
er
en
t
p
ar
ts
o
f
t
h
e
ci
t
y
,
eq
u
ip
p
ed
w
it
h
c
u
tti
n
g
-
ed
g
e
tec
h
n
o
lo
g
y
to
a
llo
w
co
n
ti
n
u
o
u
s
m
o
n
ito
r
in
g
o
f
th
e
co
n
ce
n
tr
atio
n
s
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
-
8708
P
r
ed
ictio
n
o
f a
tmo
s
p
h
eric p
o
llu
tio
n
u
s
in
g
n
eu
r
a
l n
etw
o
r
ks mo
d
el
o
f fin
e
p
a
r
ticles
.
..
(
Ju
a
n
C
a
milo
P
ed
r
a
z
a
)
6575
o
f
p
ar
ticu
late
m
atter
(
P
M1
0
,
P
S
T
,
P
M
2
.
5
)
,
p
o
llu
tan
t
g
a
s
es
(
SO2
,
NO2
,
C
O,
O3
)
an
d
w
ea
t
h
er
v
ar
iab
les
p
r
ec
ip
itatio
n
,
w
in
d
s
p
ee
d
an
d
d
ir
ec
tio
n
,
te
m
p
er
atu
r
e,
s
o
lar
r
ad
iatio
n
,
r
elativ
e
h
u
m
id
it
y
,
an
d
b
ar
o
m
etr
ic
p
r
ess
u
r
e
[
2
]
.
Fig
u
r
e
1
estab
lis
h
e
s
a
co
m
p
ar
is
o
n
o
f
a
h
u
m
a
n
h
air
v
s
P
M1
0
an
d
P
M2
.
5
to
o
b
s
er
v
e
th
e
i
m
p
o
r
ta
n
ce
o
f
th
e
p
ar
tic
u
la
te
m
atter
.
A
t
th
e
g
lo
b
al
le
v
el,
d
if
f
er
en
t
s
o
u
r
ce
s
o
f
i
n
f
o
r
m
atio
n
h
a
v
e
b
e
en
u
s
ed
to
m
iti
g
ate
an
d
p
r
ed
ict
lev
el
s
o
f
air
p
o
llu
tio
n
an
d
o
n
e
o
f
th
e
m
o
s
t
p
r
o
m
is
i
n
g
to
o
ls
ar
e
th
e
ar
tif
icia
l
n
e
u
r
al
n
et
w
o
r
k
s
,
s
in
ce
th
e
s
e
allo
w
an
al
y
z
in
g
a
lar
g
e
a
m
o
u
n
t
o
f
d
ata
an
d
m
a
k
i
n
g
p
r
ed
ictiv
e
m
o
d
els
t
h
at
allo
w
f
o
r
ec
ast
in
g
f
u
t
u
r
e
p
o
llu
tio
n
le
v
el
s
.
I
n
th
is
w
o
r
k
,
a
n
ar
ti
f
icial
n
e
u
r
al
n
et
w
o
r
k
w
as c
h
o
s
e
n
f
o
r
th
e
p
r
ed
ictio
n
o
f
s
tati
s
tical
d
ata
a
n
d
th
e
r
e
s
u
l
t
w
as a
n
an
al
y
s
is
o
f
a
p
r
o
t
o
t
y
p
e
i
n
a
m
ac
h
in
e
lear
n
i
n
g
m
o
d
el
f
o
r
t
h
e
p
r
ed
ictio
n
o
f
air
p
o
llu
ta
n
t
s
o
f
p
ar
ticu
la
te
m
atter
(
P
M)
in
an
ar
ea
o
f
B
o
g
o
tá
(
Ke
n
n
ed
y
)
.
Fig
u
r
e
1
.
C
o
m
p
ar
is
o
n
o
f
a
h
u
m
an
h
air
v
s
.
P
M1
0
v
s
.
P
M2
.
5
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
is
w
o
r
k
e
m
p
lo
y
s
n
eu
r
al
n
et
w
o
r
k
s
s
i
n
ce
th
e
y
ar
e
w
id
el
y
u
s
ed
in
t
h
e
f
o
r
ec
asti
n
g
p
r
o
ce
s
s
in
d
if
f
er
e
n
t
ar
ea
s
,
as
ca
n
b
e
s
e
en
in
[
4
-
6
]
.
T
o
ca
r
r
y
o
u
t
th
e
d
esig
n
o
f
t
h
i
s
ar
tif
icia
l
n
e
u
r
al
n
et
w
o
r
k
(
NN)
,
th
e
f
o
llo
w
i
n
g
m
e
th
o
d
o
lo
g
ical
s
tep
s
w
er
e
f
o
llo
w
ed
:
-
Selectio
n
o
f
i
n
p
u
t v
ar
iab
les.
-
No
r
m
a
lizatio
n
.
-
A
r
c
h
itect
u
r
e
s
elec
tio
n
: a
m
o
u
n
t o
f
NN
f
o
r
h
id
d
en
la
y
er
.
-
Selectio
n
o
f
t
h
e
ac
ti
v
atio
n
f
u
n
ctio
n
.
-
Selectio
n
o
f
t
h
e
lear
n
in
g
al
g
o
r
ith
m
.
-
T
r
ain
in
g
an
d
v
alid
atio
n
o
f
NN
.
2
.
1
.
Select
io
n o
f
inp
ut
v
a
ria
bles
Data
p
r
o
v
id
ed
b
y
th
e
Ken
n
e
d
y
air
q
u
alit
y
m
o
n
ito
r
in
g
s
tat
io
n
,
as
s
h
o
w
n
i
n
T
ab
le
1
w
a
s
u
s
ed
f
o
r
th
e
s
e
lectio
n
o
f
i
n
p
u
t
v
ar
iab
les;
t
h
is
s
tat
io
n
i
s
u
n
d
er
t
h
e
ad
m
i
n
is
tr
atio
n
o
f
th
e
Secr
e
tar
ía
Dis
tr
ital
d
el
Am
b
ien
te
o
f
B
o
g
o
tá
[
7
]
,
th
r
o
u
g
h
th
e
air
q
u
alit
y
m
o
n
ito
r
in
g
n
et
w
o
r
k
o
f
B
o
g
o
tá
R
MC
A
B
,
s
ee
Fig
u
r
e
2
.
E
s
p
ec
iall
y
,
i
n
th
i
s
p
ap
er
is
u
n
d
er
s
tu
d
y
th
e
d
ata
ta
k
en
f
r
o
m
th
e
m
o
n
ito
r
in
g
s
tatio
n
in
Ke
n
n
ed
y
(
d
eli
m
ited
i
n
r
ed
in
Fig
u
r
e
2
)
,
co
n
s
id
er
in
g
t
h
e
v
al
u
e
o
f
I
B
OC
A
(
Í
n
d
ice
B
o
g
o
tan
o
d
e
C
alid
ad
d
e
A
ir
e)
.
T
h
e
in
f
o
r
m
atio
n
f
r
o
m
t
h
i
s
s
tatio
n
w
as
co
m
p
iled
t
h
r
o
u
g
h
t
h
e
m
o
n
t
h
l
y
r
ep
o
r
ts
p
u
b
lis
h
ed
b
y
th
e
B
o
g
o
ta
Air
Q
u
alit
y
M
o
n
ito
r
in
g
Net
w
o
r
k
,
w
h
ic
h
s
p
ec
if
ies
th
e
m
ai
n
at
m
o
s
p
h
e
r
ic
p
o
llu
tan
ts
th
a
t
th
e
m
o
n
ito
r
in
g
s
tatio
n
i
s
ca
p
ab
le
o
f
m
ea
s
u
r
i
n
g
,
a
n
d
th
e
a
v
er
ag
e
d
ata
o
f
ea
ch
p
o
llu
tan
t,
s
ee
T
a
b
le
2
;
th
is
p
ap
er
is
p
ar
ticu
lar
l
y
f
o
cu
s
ed
o
n
P
M2
.
5
p
a
r
ticu
late
m
ater
ial.
T
h
e
p
u
b
lic
in
f
o
r
m
atio
n
a
v
ailab
l
e
f
r
o
m
th
e
Ke
n
n
ed
y
m
o
n
ito
r
i
n
g
s
tatio
n
w
as
co
n
s
o
lid
ated
u
n
til
2
0
1
8
;
o
n
e
o
f
th
e
f
ir
s
t
s
tep
s
ta
k
en
w
ith
th
ese
d
ata
w
a
s
its
n
o
r
m
aliza
t
io
n
,
a
p
r
o
ce
d
u
r
e
th
at
co
n
s
is
ted
o
f
clea
n
i
n
g
t
h
e
d
ata
o
f
e
m
p
t
y
f
ield
s
s
o
t
h
at
th
e
m
o
d
el
d
id
n
o
t
p
r
esen
t e
r
r
o
r
s
.
A
cc
o
r
d
in
g
to
t
h
e
a
v
ailab
le
d
a
ta
o
f
T
ab
le
2
th
e
in
p
u
t
v
ar
iab
les
ar
e:
m
o
n
t
h
,
y
ea
r
,
m
ax
i
m
u
m
v
alu
e
o
f
P
M2
.
5
[
/
3
]
,
n
u
m
b
er
o
f
e
x
ce
ed
an
ce
s
to
th
e
s
tan
d
ar
d
o
f
2
4
h
o
u
r
s
(
2
4
H)
,
th
e
p
er
ce
n
tag
e
o
f
v
alid
d
ata,
th
e
p
er
ce
n
t
o
f
I
B
OC
A
f
av
o
r
a
b
le
,
m
o
d
er
ate
an
d
r
eg
u
lar
(
th
e
s
u
m
o
f
t
h
ese
th
r
ee
m
u
s
t
b
e
e
q
u
al
to
1
0
0
%),
an
d
f
i
n
all
y
i
s
u
s
ed
th
e
d
ata
ass
o
ci
ated
to
th
e
n
u
m
b
er
o
f
a
n
o
m
al
ies.
T
h
e
o
u
t
v
ar
iab
le
co
r
r
esp
o
n
d
s
to
th
e
av
er
a
g
e
v
alu
e
o
f
P
M2
.
5
.
T
a
b
le
3
p
r
esen
ts
t
h
e
g
e
n
er
al
s
ta
tis
tic
s
o
f
t
h
e
tr
ain
in
g
d
ata.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
6
,
Decem
b
er
2020
:
6
5
7
4
-
6
5
8
1
6576
T
ab
le
1
.
C
h
ar
ac
ter
is
tics
o
f
t
h
e
m
o
n
ito
r
i
n
g
s
tatio
n
i
n
Ke
n
n
ed
y
R
MC
A
B
[
7
]
C
o
d
e
a
ss
i
g
n
e
d
t
o
t
h
e
st
a
t
i
o
n
i
n
t
h
e
R
M
C
A
B
9
(
F
i
x
e
d
st
a
t
i
o
n
)
Ty
p
e
o
f
S
V
C
A
A
u
t
o
mat
i
c
L
o
c
a
t
i
o
n
C
a
r
r
e
r
a
8
6
#
4
0
-
5
5
S
u
r
G
e
o
g
r
a
p
h
i
c
a
l
c
o
o
r
d
i
n
a
t
e
s
L
a
t
i
t
u
d
e
:
0
4
.
3
7
.
2
9
,
9
L
o
n
g
i
t
u
d
e
:
-
7
4
.
0
9
.
4
0
,
7
Ty
p
e
o
f
z
o
n
e
U
r
b
a
n
Ty
p
e
o
f
st
a
t
i
o
n
B
a
c
k
g
r
o
u
n
d
Eq
u
i
p
me
n
t
P
a
r
a
me
t
e
r
N
a
me
A
n
a
l
y
t
i
c
a
l
p
r
i
n
c
i
p
l
e
P
M
1
0
M
e
t
o
n
e
b
a
m
1
0
2
0
B
e
t
a
a
t
t
e
n
u
a
t
i
o
n
P
M
2
5
M
e
t
o
n
e
b
a
m
1
0
2
0
B
e
t
a
a
t
t
e
n
u
a
t
i
o
n
CO
T
h
e
r
mo
sci
e
n
t
i
f
i
c
4
8
i
I
n
f
r
a
r
e
d
a
b
so
r
p
t
i
o
n
w
i
t
h
g
a
s
f
i
l
t
e
r
S
O
2
T
a
p
i
1
0
0
E
P
u
l
se
d
f
l
u
o
r
e
sce
n
c
e
NO2
T
h
e
r
mo
sci
e
n
t
i
f
i
c
4
2
i
C
h
e
mi
l
u
mi
n
e
sce
n
c
e
L
o
c
a
t
i
o
n
o
f
t
h
e
s
a
mp
l
i
n
g
p
o
i
n
t
G
r
e
e
n
a
r
e
a
H
e
i
g
h
t
o
f
t
h
e
sam
p
l
i
n
g
p
o
i
n
t
7
.
0
m
H
e
i
g
h
t
o
f
me
t
e
o
r
o
l
o
g
i
c
a
l
st
a
t
i
o
n
1
0
.
0
m
Fig
u
r
e
2
.
Statio
n
s
o
f
m
o
n
ito
r
in
g
n
et
w
o
r
k
i
n
B
o
g
o
tá,
R
MC
A
B
[
7
]
T
ab
le
2
.
E
x
am
p
le
o
f
d
ata
u
s
ed
M
o
n
t
h
Y
e
a
r
A
v
e
r
a
g
e
P
M
2
.
5
M
a
x
P
M
2
.
5
S
t
a
n
d
a
r
d
42H
V
a
l
i
d
D
a
t
a
I
B
O
C
A
V
a
l
u
e
A
n
o
mal
i
e
s
F
a
v
o
r
a
b
l
e
M
o
d
e
r
a
t
e
R
e
g
u
l
a
r
1
2
0
1
8
26
39
0
1
0
0
1
89
10
0
2
2
0
1
8
33
48
0
1
0
0
0
63
37
0
3
2
0
1
8
3
2
.
2
52
1
1
0
0
0
57
43
0
4
2
0
1
8
2
6
.
6
39
0
1
0
0
0
94
6
0
5
2
0
1
8
2
1
.
3
34
0
94
0
1
0
0
0
0
6
2
0
1
8
1
6
.
9
28
0
1
0
0
22
78
0
0
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
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
-
8708
P
r
ed
ictio
n
o
f a
tmo
s
p
h
eric p
o
llu
tio
n
u
s
in
g
n
eu
r
a
l n
etw
o
r
ks mo
d
el
o
f fin
e
p
a
r
ticles
.
..
(
Ju
a
n
C
a
milo
P
ed
r
a
z
a
)
6577
T
ab
le
3
.
Gen
er
al
s
tatis
tic
s
o
f
t
h
e
in
p
u
t
tr
ain
in
g
d
ata
V
a
r
i
a
b
l
e
A
mo
u
n
t
M
e
a
n
S
T
D
M
i
n
M
a
x
M
o
n
t
h
29
6
,
3
1
0
,
3
4
5
3
,
5
0
6
,
1
5
2
1
12
Y
e
a
r
29
2
.
0
1
6
,
9
3
1
0
3
4
0
,
8
4
2
2
3
5
2
0
1
6
2
0
1
8
M
a
x
P
M
2
.
5
29
4
0
,
8
2
0
,
4
2
4
5
,
3
5
0
,
1
2
8
2
8
.
3
52
S
t
a
n
d
a
r
d
2
4
H
29
0
,
1
3
5
2
7
9
0
,
1
8
1
1
6
6
0
1
V
a
l
i
d
D
a
t
a
29
9
6
,
6
8
9
,
6
5
5
3
,
1
4
0
,
7
8
5
87
1
0
0
I
B
O
C
A
F
a
v
o
r
a
b
l
e
29
3
4
6
,
9
3
5
4
,
3
0
6
,
3
3
0
0
22
M
o
d
e
r
a
t
e
29
7
5
,
3
1
9
,
7
4
9
8
,
0
1
8
,
3
3
9
57
1
0
0
R
e
g
u
l
a
r
29
2
1
,
2
1
6
,
3
0
9
1
0
,
1
3
7
,
8
7
5
0
43
A
n
o
mal
i
e
s
29
0
,
0
3
4
4
8
3
0
,
1
8
5
6
9
5
0
1
2
.
2
.
No
rm
a
liza
t
io
n
T
h
er
e
ar
e
d
if
f
er
en
t
s
ca
le
s
a
n
d
r
an
g
e
s
w
i
th
i
n
t
h
e
co
llected
d
ata,
alth
o
u
g
h
,
t
h
e
m
o
d
el
co
u
l
d
co
n
v
er
g
e
w
it
h
th
e
in
f
o
r
m
a
tio
n
i
n
th
i
s
w
a
y
,
m
ak
in
g
t
h
e
tr
ai
n
i
n
g
m
o
r
e
d
if
f
icu
lt
an
d
m
a
k
i
n
g
th
e
m
o
d
el
d
ep
en
d
en
t
o
n
th
e
u
n
i
ts
u
s
ed
f
o
r
th
e
in
p
u
t
[
8
]
.
I
n
ad
d
itio
n
,
to
n
o
r
m
alize
t
h
e
v
alu
e
s
i
s
u
s
ed
t
h
e
Sta
n
d
ar
d
No
r
m
al
Dis
tr
ib
u
tio
n
m
et
h
o
d
(
1
)
,
w
h
ic
h
allo
w
s
to
r
ed
u
ce
an
y
n
o
r
m
al
d
i
s
tr
ib
u
tio
n
to
th
e
s
ta
n
d
ar
d
n
o
r
m
al
f
o
r
m
at,
s
i
n
ce
it
h
as
a
n
ar
ith
m
etic
av
er
a
g
e
an
d
u
n
iq
u
e
s
tan
d
ar
d
d
ev
iatio
n
s
,
w
h
o
s
e
v
a
lu
es a
r
e
ze
r
o
an
d
o
n
e,
r
esp
ec
tiv
el
y
[
9
]
.
=
−
(
1
)
w
h
er
e
co
r
r
esp
o
n
d
s
to
th
e
v
alu
e
o
f
th
e
v
ar
iab
le
to
b
e
s
tan
d
ar
d
ized
,
r
ep
r
esen
ts
th
e
ar
ith
m
etic
av
er
a
g
e
an
d
th
e
s
ta
n
d
ar
d
d
ev
iatio
n
o
f
t
h
e
v
ar
iab
le,
th
is
e
n
s
u
r
es t
h
at
t
h
er
e
ar
e
o
n
l
y
v
al
u
es
w
i
th
i
n
t
h
e
r
an
g
e
o
f
0
an
d
1
.
2
.
3
.
Arc
hite
ct
ure
s
elec
t
io
n
C
ar
d
in
alit
y
s
elec
tio
n
o
f
t
h
e
h
i
d
d
en
la
y
er
s
(
i
f
th
er
e
ar
e
m
o
r
e
th
a
n
o
n
e)
s
ee
m
s
to
b
e
a
n
u
n
c
lear
is
s
u
e
w
h
e
n
d
esi
g
n
in
g
a
n
eu
r
al
n
e
t
wo
r
k
(
NN)
,
to
o
m
an
y
u
n
its
ca
n
lead
to
lo
w
g
en
er
aliza
tio
n
ca
p
ac
it
y
.
O
n
th
e
o
th
er
h
an
d
,
f
e
w
u
n
it
s
ca
n
lead
in
g
to
th
e
NN
d
o
n
o
t
h
av
e
s
u
f
f
icie
n
t
ca
p
ac
it
y
to
s
o
lv
e
th
e
p
r
o
b
lem
[
1
0
]
.
R
eg
ar
d
in
g
th
e
n
et
w
o
r
k
to
p
o
lo
g
y
,
d
eter
m
i
n
i
n
g
t
h
e
n
u
m
b
er
o
f
la
y
er
s
t
h
at
in
te
g
r
ate
i
t
a
n
d
t
h
e
n
u
m
b
er
o
f
h
id
d
en
n
eu
r
o
n
s
to
b
e
i
n
cl
u
d
ed
in
ea
ch
la
y
er
is
a
co
m
p
le
x
tas
k
t
h
at
d
ir
ec
tl
y
a
f
f
ec
ts
th
e
g
e
n
er
al
izatio
n
ca
p
ac
it
y
o
f
th
e
m
o
d
el.
Si
n
ce
ev
er
y
n
eu
r
al
n
et
w
o
r
k
n
ec
e
s
s
ar
il
y
h
a
s
an
in
p
u
t
la
y
er
w
h
ich
r
ec
ei
v
es
ex
ter
n
al
s
t
i
m
u
li
,
th
e
p
r
o
b
lem
i
s
li
m
ited
to
estab
lis
h
in
g
t
h
e
n
u
m
b
er
o
f
e
x
te
n
s
io
n
o
f
t
h
e
h
id
d
en
la
y
er
s
[
1
1
]
.
A
lt
h
o
u
g
h
i
t
h
as
b
ee
n
d
e
m
o
n
s
tr
ated
th
at
t
h
e
u
n
iv
er
s
a
l
ap
p
r
o
ac
h
in
g
p
r
o
p
er
ty
o
f
ML
P
(
Mu
ltil
a
y
er
P
er
ce
p
tr
o
n
)
n
et
w
o
r
k
f
u
n
ctio
n
s
r
eq
u
ir
es
a
m
a
x
i
m
u
m
o
f
t
w
o
h
id
d
en
la
y
er
s
,
in
m
o
s
t
ca
s
es
a
s
in
g
le
h
id
d
en
la
y
er
is
s
u
f
f
icien
t
to
ac
h
ie
v
e
o
p
ti
m
a
l
r
esu
l
ts
.
L
ip
p
m
a
n
n
[
1
2
]
co
n
s
i
d
er
s
th
at
n
et
w
o
r
k
s
w
it
h
a
s
in
g
le
h
id
d
en
la
y
er
ar
e
s
u
f
f
icie
n
t
to
s
o
lv
e
ar
b
itra
r
il
y
co
m
p
le
x
p
r
o
b
lem
s
,
p
r
o
v
id
ed
th
at
th
e
h
id
d
en
la
y
er
i
n
cl
u
d
es
at
least
th
r
ee
ti
m
es
th
e
n
u
m
b
er
o
f
i
n
p
u
t
n
o
d
es.
Me
an
w
h
ile,
Hec
h
t
-
Niel
s
en
a
n
d
L
ip
p
m
a
n
ap
p
l
y
a
n
e
x
te
n
s
io
n
o
f
Ko
l
m
o
g
o
r
o
v
th
eo
r
e
m
to
s
h
o
w
t
h
at
a
n
et
w
o
r
k
w
it
h
a
s
i
n
g
le
h
id
d
en
la
y
er
i
n
teg
r
ated
2
+
1
n
eu
r
o
n
s
a
n
d
tr
an
s
f
er
f
u
n
ctio
n
o
f
co
n
tin
u
o
u
s
,
n
o
n
-
li
n
ea
r
an
d
m
o
n
o
to
n
icall
y
i
n
cr
ea
s
i
n
g
i
s
s
u
f
f
icie
n
t
to
co
m
p
u
te
an
y
co
n
t
in
u
o
u
s
f
u
n
ctio
n
o
f
in
p
u
t
v
ar
iab
les [
1
3
]
.
Usu
al
l
y
,
ad
h
o
c
r
u
les
ar
e
u
s
ed
to
d
eter
m
in
e
t
h
e
n
u
m
b
er
o
f
h
id
d
en
n
eu
r
o
n
s
i
n
ea
ch
la
y
er
.
A
lt
h
o
u
g
h
th
e
y
ar
e
n
o
t
m
at
h
e
m
atica
ll
y
j
u
s
ti
f
iab
le,
th
e
y
h
av
e
s
h
o
w
n
g
o
o
d
b
eh
av
io
r
in
v
ar
io
u
s
p
r
ac
tical
ap
p
licatio
n
s
.
Ma
s
ter
s
[
1
4
]
p
r
o
p
o
s
ed
a
m
et
h
o
d
th
at
h
e
ca
lled
th
e
g
eo
m
etr
ic
p
y
r
a
m
id
r
u
le,
w
h
ic
h
is
b
ased
o
n
th
e
as
s
u
m
p
tio
n
th
at
th
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
i
n
t
h
e
h
id
d
en
l
a
y
er
m
u
s
t
b
e
less
t
h
a
n
t
h
e
to
t
al
n
u
m
b
er
o
f
i
n
p
u
t
v
ar
iab
les,
b
u
t
g
r
ea
ter
t
h
an
th
e
n
u
m
b
er
o
f
o
u
tp
u
t
v
ar
iab
les.
I
t
is
co
n
s
id
er
ed
th
at
t
h
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
in
ea
ch
la
y
er
f
o
llo
w
s
a
g
eo
m
etr
ic
p
r
o
g
r
ess
io
n
,
s
u
c
h
th
at
f
o
r
a
n
et
w
o
r
k
w
it
h
a
s
i
n
g
le
h
id
d
en
la
y
er
,
th
e
n
u
m
b
er
o
f
in
ter
m
ed
iate
n
eu
r
o
n
s
m
u
s
t
b
e
clo
s
e
to
√
∙
,
w
h
er
e
is
th
e
n
u
m
b
er
o
f
v
a
r
iab
les
o
f
in
p
u
t
an
d
t
h
e
to
ta
l
o
u
tp
u
t
n
e
u
r
o
n
s
;
th
is
p
r
o
j
ec
t
ta
k
es
a
to
tal
o
f
9
in
p
u
t
v
ar
iab
les
an
d
o
n
e
o
u
tp
u
t;
t
h
u
s
,
9
n
eu
r
o
n
s
w
er
e
d
ef
i
n
ed
in
th
e
h
id
d
en
la
y
er
ac
co
r
d
in
g
to
[
1
0
]
.
2
.
4
.
Select
io
n o
f
t
he
a
ct
i
v
a
t
i
o
n f
un
ct
io
n
I
n
b
o
th
ar
tif
icial
a
n
d
b
io
lo
g
i
c
al
n
eu
r
al
n
et
w
o
r
k
s
,
a
n
eu
r
o
n
n
o
t
o
n
l
y
tr
a
n
s
m
it
s
th
e
i
n
p
u
t
i
t
r
ec
eiv
es.
T
h
er
e
is
an
ad
d
itio
n
a
l
s
tep
,
an
ac
ti
v
atio
n
f
u
n
ctio
n
,
w
h
ich
is
a
n
alo
g
o
u
s
to
t
h
e
ac
t
io
n
p
o
ten
tial
r
ate
[
1
5
]
.
T
h
er
e
ar
e
m
an
y
ac
ti
v
atio
n
f
u
n
ctio
n
s
,
f
o
r
th
is
p
r
o
j
ec
t
is
u
s
e
d
th
e
ac
tiv
at
io
n
f
u
n
ctio
n
ca
lle
d
R
ec
tif
ied
L
i
n
ea
r
Un
it
ab
b
r
ev
iated
as R
e
L
U,
wh
ich
i
s
d
ef
i
n
ed
in
(
2
)
an
d
r
ep
r
esen
ted
i
n
Fi
g
u
r
e
3
.
(
)
=
ma
x
(
0
,
)
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
6
,
Decem
b
er
2020
:
6
5
7
4
-
6
5
8
1
6578
Fig
u
r
e
3
.
R
eL
U
ac
t
iv
at
io
n
f
u
n
ctio
n
T
h
e
s
u
p
er
io
r
it
y
o
f
R
e
L
U
i
s
b
a
s
ed
o
n
e
m
p
ir
ical
r
esear
c
h
,
p
r
o
b
ab
ly
b
ec
au
s
e
it
h
as
a
m
o
r
e
u
s
ef
u
l r
an
g
e
o
f
r
esp
o
n
s
e
ca
p
ac
it
y
[
1
6
]
.
I
n
o
th
er
w
o
r
d
s
,
R
e
L
Us
al
lo
w
s
al
l
p
o
s
itiv
e
v
al
u
e
s
to
p
ass
w
ith
o
u
t
ch
a
n
g
i
n
g
th
e
m
,
b
u
t
ass
i
g
n
all
n
e
g
ativ
e
v
al
u
e
s
to
0
.
A
lth
o
u
g
h
th
er
e
ar
e
ev
en
m
o
r
e
r
ec
en
t
ac
tiv
atio
n
f
u
n
ct
io
n
s
,
m
o
s
t
o
f
th
e
cu
r
r
en
t
n
eu
r
al
n
et
w
o
r
k
s
u
s
e
R
eL
U
o
r
o
n
e
o
f
its
v
ar
ian
t
s
[
1
5
]
.
I
n
f
ac
t,
an
y
m
at
h
e
m
ati
ca
l
f
u
n
ctio
n
ca
n
b
e
u
s
ed
as
an
ac
ti
v
atio
n
f
u
n
c
tio
n
.
Su
p
p
o
s
e
th
at
Fi
g
u
r
e
3
r
ep
r
e
s
en
t
s
th
e
ac
ti
v
atio
n
f
u
n
c
tio
n
(
R
eL
U,
s
ig
m
o
id
o
r
an
y
o
th
er
)
,
to
d
e
f
i
n
e
t
h
e
ac
t
iv
atio
n
f
u
n
ctio
n
o
f
t
h
e
n
e
u
r
al
n
et
w
o
r
k
m
o
d
el,
s
i
m
p
l
y
u
s
i
n
g
t
h
e
o
n
e
p
r
o
v
id
ed
b
y
th
e
T
en
s
o
r
Flo
w
l
ib
r
ar
y
[
1
6
]
.
2
.
5
.
Select
io
n o
f
t
he
lea
rning
a
lg
o
rit
hm
Fo
r
m
o
d
elin
g
th
e
A
r
ti
f
icial
N
eu
r
al
Net
w
o
r
k
it
w
as
u
s
ed
t
h
e
Ker
as
l
ib
r
ar
y
[
1
7
]
,
w
h
ic
h
i
s
w
r
itten
i
n
P
y
t
h
o
n
,
w
h
ic
h
s
u
p
p
o
r
ts
w
o
r
k
i
n
g
to
g
et
h
er
w
it
h
T
en
s
o
r
Flo
w
[
1
8
]
an
d
allo
w
s
t
h
e
m
o
d
eli
n
g
o
f
ar
tif
ic
ial
n
e
u
r
al
n
et
w
o
r
k
s
.
Ker
as
h
as
t
w
o
t
y
p
es
o
f
m
o
d
els;
f
o
r
th
is
p
r
o
ject,
it
w
a
s
d
ec
id
ed
to
u
s
e
th
e
s
eq
u
e
n
tial
t
y
p
e,
d
ef
in
ed
as
a
p
ip
elin
e
w
it
h
its
r
a
w
d
ata
en
ter
ed
in
th
e
lo
w
er
p
ar
t
an
d
th
e
p
r
ed
ictio
n
s
th
at
co
m
e
o
u
t
in
t
h
e
u
p
p
er
p
ar
t.
T
h
is
is
a
u
s
ef
u
l
co
n
ce
p
ti
o
n
in
Ker
as,
s
in
ce
t
h
e
y
w
er
e
tr
ad
itio
n
all
y
as
s
o
ciate
d
w
i
th
a
lay
er
ca
n
al
s
o
b
e
d
iv
id
ed
an
d
ad
d
ed
as
s
ep
ar
ate
la
y
er
s
[
1
9
]
.
T
h
e
co
m
p
ilatio
n
r
eq
u
ir
es
t
h
at
t
h
e
p
ar
a
m
eter
s
o
f
th
e
n
et
w
o
r
k
b
e
s
p
ec
if
ied
,
as
w
e
ll
as
th
e
o
p
tim
izatio
n
alg
o
r
ith
m
to
b
e
u
s
e
d
to
tr
ain
th
e
n
et
w
o
r
k
an
d
th
e
f
u
n
ctio
n
u
s
ed
to
ev
alu
a
te
th
e
n
et
w
o
r
k
[
1
9
]
.
Fig
u
r
e
4
s
h
o
w
s
a
f
r
ag
m
e
n
t
o
f
th
e
n
eu
r
al
n
et
w
o
r
k
m
o
d
e
lin
g
co
d
e
u
s
i
n
g
a
s
m
en
tio
n
ed
ab
o
v
e,
th
e
Ker
as a
n
d
T
en
s
o
r
f
lo
w
l
ib
r
ar
ies to
g
eth
er
.
Fo
r
th
i
s
ca
s
e,
it
w
as
d
ec
id
ed
to
u
s
e
t
h
e
R
MSp
r
o
p
(
R
o
o
t
Me
an
Sq
u
ar
e
P
r
o
p
ag
atio
n
)
o
p
ti
m
izer
,
w
h
ic
h
is
an
alg
o
r
it
h
m
u
s
ed
f
o
r
co
m
p
lete
b
atch
o
p
ti
m
izati
o
n
[
2
0
]
.
R
MSp
r
o
p
t
r
ies
to
s
o
lv
e
th
e
p
r
o
b
le
m
o
f
th
e
p
o
s
s
ib
le
w
id
e
v
ar
iatio
n
i
n
m
a
g
n
itu
d
e
o
f
t
h
e
g
r
ad
ien
t
s
.
So
m
e
g
r
ad
ien
ts
ca
n
b
e
s
m
a
ll
an
d
o
th
er
s
h
u
g
e,
w
h
ic
h
is
a
v
er
y
d
if
f
ic
u
lt i
s
s
u
e
tr
y
in
g
to
f
i
n
d
a
u
n
iq
u
e
g
lo
b
al
l
ea
r
n
in
g
r
ate
f
o
r
th
e
al
g
o
r
ith
m
.
T
h
is
ad
j
u
s
t
m
e
n
t i
s
u
s
e
f
u
l
f
o
r
th
e
s
u
p
p
o
r
t
p
o
i
n
ts
a
n
d
f
lat
s
e
g
m
e
n
t
s
,
s
i
n
ce
lar
g
e
e
n
o
u
g
h
s
tep
s
ar
e
ta
k
e
n
,
ev
e
n
with
s
m
al
l
g
r
ad
ie
n
ts
;
th
e
s
tep
s
ize
ad
ap
ts
i
n
d
iv
id
u
all
y
w
it
h
ti
m
e,
s
o
th
at
lear
n
in
g
i
s
ac
ce
ler
ated
in
th
e
r
eq
u
ir
ed
d
ir
ec
tio
n
[
2
1
]
.
An
o
th
er
p
ar
a
m
eter
a
s
s
i
g
n
ed
f
o
r
th
e
co
m
p
ilatio
n
is
t
h
e
lo
s
s
f
u
n
ctio
n
,
w
h
ich
i
n
d
icate
s
th
e
v
al
u
e
o
f
th
e
p
r
ed
ictio
n
er
r
o
r
w
h
ich
co
n
s
i
s
ts
o
f
t
h
e
s
u
m
o
f
all
t
h
e
er
r
o
r
s
o
b
tain
ed
;
th
e
m
etr
ics
v
al
u
e
w
as
also
d
ef
i
n
ed
,
w
h
ic
h
co
n
s
is
t
s
o
f
t
h
e
li
s
t
o
f
m
etr
ics
to
b
e
ev
a
lu
ated
b
y
t
h
e
m
o
d
el
d
u
r
i
n
g
tr
ain
i
n
g
a
n
d
test
in
g
.
Fi
g
u
r
e
5
s
h
o
w
s
a
co
d
e
f
r
ag
m
e
n
t
w
i
th
t
h
e
co
n
f
ig
u
r
atio
n
o
f
th
e
af
o
r
e
m
en
tio
n
ed
p
ar
a
m
e
ter
s
.
Ma
k
in
g
t
h
e
r
esp
ec
ti
v
e
co
n
f
i
g
u
r
atio
n
s
o
f
th
e
n
e
u
r
al
n
et
w
o
r
k
a
s
r
esu
lt
it
is
o
b
tain
ed
th
e
r
ep
o
r
t
s
h
o
w
n
i
n
Fi
g
u
r
e
6
,
w
h
er
e
th
e
la
y
er
s
ar
e
d
is
p
la
y
ed
,
th
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
in
ea
c
h
la
y
er
an
d
th
e
to
tal
n
u
m
b
er
o
f
tr
ain
in
g
p
ar
a
m
eter
s
.
Fig
u
r
e
4
.
Fra
g
m
e
n
t o
f
t
h
e
n
eu
r
al
n
et
w
o
r
k
m
o
d
eli
n
g
co
d
e
Fig
u
r
e
5
.
P
ar
am
eter
s
co
n
f
i
g
u
r
atio
n
co
d
e
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
-
8708
P
r
ed
ictio
n
o
f a
tmo
s
p
h
eric p
o
llu
tio
n
u
s
in
g
n
eu
r
a
l n
etw
o
r
ks mo
d
el
o
f fin
e
p
a
r
ticles
.
..
(
Ju
a
n
C
a
milo
P
ed
r
a
z
a
)
6579
Fig
u
r
e
6
.
Mo
d
el
s
u
m
m
ar
y
2
.
6
.
T
ra
ini
ng
a
nd
v
a
lid
a
t
io
n
I
n
th
e
co
n
tex
t
o
f
n
e
u
r
al
n
et
w
o
r
k
s
,
lear
n
i
n
g
ca
n
b
e
s
ee
n
as
th
e
p
r
o
ce
s
s
o
f
ad
j
u
s
t
in
g
th
e
f
r
ee
p
ar
am
eter
s
o
f
t
h
e
n
et
w
o
r
k
[
2
2
,
2
3
]
.
Star
tin
g
f
r
o
m
a
s
et
o
f
r
an
d
o
m
s
y
n
ap
tic
w
ei
g
h
ts
,
t
h
e
lear
n
in
g
p
r
o
ce
s
s
lo
o
k
s
f
o
r
a
s
et
o
f
w
ei
g
h
t
s
th
at
allo
w
th
e
n
et
w
o
r
k
to
co
r
r
e
ctly
d
e
v
elo
p
a
ce
r
tain
task
.
T
h
e
lear
n
in
g
p
r
o
ce
s
s
i
s
iter
ativ
e,
in
w
h
ic
h
th
e
s
o
l
u
tio
n
is
r
ef
i
n
ed
u
n
til
r
ea
ch
i
n
g
a
s
u
f
f
icie
n
tl
y
g
o
o
d
lev
el
o
f
o
p
er
atio
n
[
2
4
]
.
T
h
e
g
o
al
o
f
a
n
y
tr
ain
in
g
a
lg
o
r
it
h
m
i
s
to
m
in
i
m
ize
t
h
e
m
ea
n
s
q
u
ar
e
er
r
o
r
MSE
(
o
r
an
y
o
th
er
lo
s
s
f
u
n
c
tio
n
)
,
b
u
t
e
x
p
er
ien
ce
h
as
s
h
o
w
n
t
h
at
n
et
w
o
r
k
s
ten
d
to
o
v
er
ad
j
u
s
t
d
ata
[
2
5
]
.
Fo
r
th
is
r
ea
s
o
n
,
t
h
e
co
llected
d
ata
w
er
e
d
iv
id
ed
in
to
8
0
-
2
0
:
8
0
%
o
f
t
h
e
d
ata
w
a
s
r
an
d
o
m
l
y
d
i
v
id
ed
t
o
b
e
u
s
ed
as
tr
ai
n
i
n
g
d
ata
f
o
r
th
e
n
eu
r
al
n
et
w
o
r
k
an
d
th
e
r
e
m
a
in
i
n
g
2
0
% f
o
r
th
e
test
in
g
o
r
v
alid
atio
n
s
ta
g
es.
3.
RE
SU
L
T
S
T
w
o
d
if
f
er
en
t
m
ac
h
i
n
es
w
er
e
u
s
ed
f
o
r
m
o
d
eli
n
g
t
h
e
n
e
u
r
al
n
et
w
o
r
k
:
t
h
e
f
ir
s
t
is
a
W
in
d
o
w
s
d
es
k
to
p
co
m
p
u
ter
w
it
h
8
GB
o
f
R
A
M
an
d
A
M
D
R
y
ze
n
2
6
0
0
p
r
o
ce
s
s
o
r
w
it
h
s
i
x
co
r
es
at
3
.
4
0
GHz
;
th
e
d
ev
elo
p
m
e
n
t
en
v
ir
o
n
m
e
n
t
o
f
P
y
t
h
o
n
3
an
d
J
u
n
y
p
er
w
as
in
s
talled
o
n
t
h
i
s
m
ac
h
in
e.
I
n
th
e
o
t
h
er
d
ev
elo
p
m
en
t
en
v
ir
o
n
m
e
n
t,
a
Go
o
g
le
C
o
llab
o
r
ativ
e
in
s
ta
n
ce
w
as
m
a
n
ag
ed
w
h
er
e
r
eso
u
r
ce
s
ar
e
h
an
d
l
ed
ac
co
r
d
in
g
to
av
ailab
ili
t
y
i
n
a
v
ir
t
u
al
m
ac
h
i
n
e
w
i
th
J
u
n
y
p
e
r
.
T
h
e
r
esu
lt
o
f
t
h
e
er
r
o
r
in
b
o
th
m
ac
h
in
e
s
w
as
al
m
o
s
t
t
h
e
s
a
m
e;
f
ir
s
t
m
ac
h
i
n
e
MSE
3
.
5
6
an
d
th
e
s
ec
o
n
d
MSE
o
f
3
.
6
7
.
Du
r
in
g
t
h
e
tr
ain
i
n
g
o
f
th
e
m
o
d
el,
it
is
a
s
u
itab
le
p
r
ac
tice
to
v
er
if
y
t
h
e
iter
atio
n
s
w
h
il
e
tr
ain
in
g
th
e
ar
ti
f
icial
n
eu
r
al
n
et
w
o
r
k
,
th
is
to
s
to
p
p
in
g
t
h
e
iter
atio
n
s
(
ea
r
ly
s
to
p
)
,
s
in
ce
t
h
is
p
r
ev
e
n
ts
t
h
e
m
o
d
el
f
r
o
m
b
ein
g
o
v
er
tr
ain
ed
m
o
r
e
th
a
n
n
ec
es
s
ar
y
.
I
n
itiall
y
,
it
w
a
s
d
ef
i
n
ed
5
0
0
0
ep
o
ch
s
f
o
r
tr
ain
in
g
p
r
o
ce
s
s
,
Fig
u
r
e
7
s
h
o
w
s
t
h
at
a
f
ter
ap
p
r
o
x
i
m
a
tel
y
3
0
0
0
ep
o
ch
s
t
h
e
er
r
o
r
d
o
es
n
o
t
d
ec
r
ea
s
e,
in
s
tead
,
it
r
e
m
ai
n
s
s
tab
le
[
1
6
]
.
I
n
ad
d
itio
n
,
f
o
r
th
e
f
ir
s
t
iter
atio
n
s
th
e
v
al
id
atio
n
er
r
o
r
is
less
t
h
an
th
e
tr
ain
in
g
er
r
o
r
,
w
h
i
le
f
o
r
th
e
las
t
iter
atio
n
s
t
h
e
o
p
p
o
s
ite
o
cc
u
r
s
.
I
n
t
h
e
r
es
u
lt
o
f
F
ig
u
r
e
7
,
t
h
e
tr
ain
i
n
g
p
r
o
ce
s
s
s
to
p
s
u
n
til
f
i
n
alizi
n
g
t
h
e
to
tal
n
u
m
b
er
o
f
ep
o
ch
s
,
to
ac
ce
ler
ate
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
as
w
ell
a
s
to
a
v
o
id
t
h
e
o
v
er
f
itti
n
g
t
h
e
v
ar
iab
le
“
ea
r
l
y
s
to
p
”
is
u
s
ed
to
v
er
i
f
y
ev
er
y
5
0
ep
o
ch
s
if
t
h
e
er
r
o
r
w
as
r
ed
u
ce
d
,
th
e
n
t
h
e
tr
ai
n
in
g
i
s
s
to
p
p
ed
,
o
th
er
w
is
e,
i
t
w
o
u
ld
co
n
ti
n
u
e.
T
h
e
r
esu
lt
u
s
in
g
t
h
i
s
o
p
tio
n
i
s
p
r
esen
ted
i
n
F
i
g
u
r
e
8
,
w
h
er
e
i
t
ca
n
b
e
o
b
s
er
v
ed
th
a
t
at
t
h
e
en
d
o
f
t
h
e
tr
ain
i
n
g
p
r
o
ce
s
s
th
e
v
alid
atio
n
an
d
t
r
ain
in
g
er
r
o
r
ten
d
to
b
e
th
e
s
a
m
e.
Fin
al
l
y
,
Fi
g
u
r
e
9
s
h
o
w
s
t
h
e
r
esu
lt
o
f
th
e
p
r
ed
ictio
n
o
f
P
M2
.
5
p
ar
ticu
late
m
ater
ial
r
e
m
o
v
i
n
g
o
u
tlier
s
.
T
h
is
r
e
s
u
l
t
s
h
o
w
s
t
h
at
th
e
p
r
ed
ictio
n
is
ac
h
iev
ed
f
o
r
a
li
m
ited
r
an
g
e
o
f
d
ata,
w
h
ic
h
m
a
y
b
e
r
elate
d
to
th
e
lo
w
a
m
o
u
n
t o
f
d
ata
av
a
il
ab
le
f
o
r
tr
ain
in
g
.
Fig
u
r
e
7
.
P
er
f
o
r
m
a
n
ce
tr
ain
i
n
g
f
o
r
P
M2
.
5
d
ata
u
s
i
n
g
5
0
0
0
ep
o
ch
Fig
u
r
e
8
.
P
er
f
o
r
m
a
n
ce
tr
ain
i
n
g
f
o
r
P
M2
.
5
d
ata
u
s
i
n
g
ea
r
l
y
s
to
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
6
,
Decem
b
er
2020
:
6
5
7
4
-
6
5
8
1
6580
Fig
u
r
e
9
.
P
r
ed
ictio
n
r
esu
lts
4.
CO
NCLU
SI
O
N
C
u
r
r
en
tl
y
,
th
er
e
is
a
s
u
itab
le
a
m
o
u
n
t
o
f
l
ib
r
ar
ies
th
a
t
allo
w
th
e
d
esig
n
a
n
d
m
o
d
elin
g
o
f
ar
ti
f
icial
n
eu
r
al
n
et
w
o
r
k
s
,
w
h
er
e
it
h
as
b
ee
n
s
h
o
w
n
t
h
at
Ma
c
h
in
e
L
ea
r
n
in
g
tech
n
iq
u
es
ar
e
a
s
u
itab
l
e
to
o
l
f
o
r
m
o
d
elin
g
p
r
ed
ictio
n
an
d
class
if
ica
tio
n
p
r
o
b
lem
s
.
I
n
th
e
f
ir
s
t
s
ta
g
e
o
f
th
e
p
r
o
j
ec
t,
f
r
o
m
t
h
e
in
f
o
r
m
atio
n
co
llected
,
th
e
e
m
p
t
y
f
ield
s
w
er
e
clea
n
ed
to
av
o
id
er
r
o
r
s
an
d
th
e
n
n
o
r
m
alize
d
.
T
h
is
a
s
p
ec
t
w
a
s
o
f
i
m
p
o
r
ta
n
ce
to
ac
h
ie
v
e
a
s
atis
f
ac
to
r
y
r
esu
lt.
I
n
th
is
r
eg
ar
d
,
at
th
e
d
ev
elo
p
m
e
n
t
s
ta
g
e,
th
e
s
ta
n
d
ar
d
n
o
r
m
al
d
is
tr
i
b
u
tio
n
m
et
h
o
d
w
a
s
u
s
ed
to
n
o
r
m
alize
t
h
e
i
n
p
u
t d
a
ta.
As
m
e
n
tio
n
ed
ab
o
v
e,
th
e
d
ata
w
a
s
d
iv
id
ed
in
s
u
c
h
a
w
a
y
th
a
t
r
ec
o
r
d
s
w
er
e
k
ep
t
f
o
r
b
o
th
tr
ain
i
n
g
a
n
d
test
i
n
g
;
af
ter
m
ak
i
n
g
t
h
e
ad
j
u
s
t
m
e
n
t
s
o
n
t
h
e
tr
ai
n
in
g
o
f
t
h
e
m
o
d
el
i
t
w
as
p
o
s
s
ib
le
to
r
ed
u
ce
t
h
e
f
i
n
al
M
SE
f
r
o
m
3
0
to
3
.
5
6
in
th
e
p
r
ed
ictio
n
o
f
th
e
a
v
er
ag
e
co
n
ce
n
tr
ati
o
n
o
f
P
M2
.
5
p
o
llu
tan
t
s
.
T
h
e
d
esig
n
o
f
t
h
e
n
e
u
r
a
l
n
et
w
o
r
k
w
as
b
ased
o
n
t
h
e
r
u
l
e
o
f
th
e
g
eo
m
etr
ic
p
y
r
a
m
id
o
f
th
e
h
id
d
en
la
y
er
s
o
b
tain
i
n
g
a
s
u
itab
le
r
es
u
lt
f
o
r
MSE
.
W
h
en
t
h
er
e
is
n
o
t
m
u
c
h
tr
ai
n
i
n
g
d
ata,
ac
co
r
d
in
g
to
l
iter
atu
r
e
a
n
alter
n
ati
v
e
is
to
u
s
e
a
s
m
al
l
n
et
w
o
r
k
w
it
h
f
e
w
h
id
d
en
la
y
er
s
to
av
o
id
o
v
er
f
itti
n
g
.
An
o
th
er
u
s
ef
u
l
alter
n
ati
v
e
to
av
o
id
o
v
er
f
itti
n
g
is
th
e
ea
r
l
y
s
to
p
,
w
h
ic
h
w
as
co
n
s
id
er
ed
in
th
i
s
w
o
r
k
.
T
h
e
p
r
o
t
o
ty
p
e
o
f
t
h
e
n
eu
r
al
n
e
t
w
o
r
k
ap
p
lied
to
p
r
ed
ict
th
e
P
M2
.
5
p
o
llu
tan
t
s
,
is
a
n
e
x
a
m
p
le
o
f
t
h
e
ap
p
licatio
n
o
f
Ma
ch
in
e
L
e
ar
n
in
g
m
et
h
o
d
s
in
th
e
d
i
f
f
er
en
t
asp
ec
t
s
th
a
t
af
f
ec
t
h
u
m
a
n
b
ein
g
s
.
T
h
e
r
esu
lt
o
f
th
is
w
o
r
k
ca
n
b
e
i
m
p
r
o
v
ed
to
ac
h
iev
e
a
b
ett
er
p
r
ed
icti
o
n
f
o
r
a
lar
g
e
v
al
u
e
o
f
d
ata.
ACK
NO
WL
E
D
G
E
M
E
NT
S
T
h
e
au
th
o
r
s
t
h
an
k
t
h
e
Fac
u
lt
y
o
f
E
n
g
in
ee
r
i
n
g
o
f
t
h
e
U
n
i
v
er
s
id
ad
Dis
tr
ital
Fra
n
cisco
J
o
s
é
d
e
C
ald
as,
esp
ec
iall
y
t
h
e
Do
cto
r
ate
in
E
n
g
in
ee
r
i
n
g
an
d
t
h
e
s
p
ec
ializatio
n
in
So
f
t
w
ar
e
E
n
g
i
n
ee
r
in
g
.
RE
F
E
R
E
NC
E
S
[1
]
M.
G
.
C
.
Ja
n
u
c
h
s
,
“
A
p
li
c
a
c
ió
n
d
e
téc
n
ica
s d
e
In
telig
e
n
c
ia A
rti
f
icia
l
a
la p
re
d
icc
ió
n
d
e
c
o
n
tam
in
a
n
tes
a
tm
o
s
f
é
ri
c
o
s
,
”
P
h
.
D.
d
isse
rtatio
n
,
Un
iv
e
rsid
a
d
P
o
li
téc
n
ica
d
e
M
a
d
rid
,
M
a
d
r
id
,
S
p
a
in
,
p
p
.
1
-
2
2
5
,
2
0
1
2
.
[2
]
J.
A
.
Z
.
P
e
n
a
,
“
Estrate
g
ias
p
a
ra
m
it
ig
a
r
la
c
o
n
tam
in
a
c
i
ó
n
d
e
l
a
ire
e
n
z
o
n
a
s
a
led
a
ñ
a
s
a
g
ra
n
d
e
s
a
v
e
n
id
a
s
d
e
Bo
g
o
tá
,
”
M
.
S
c
.
d
isse
rtatio
n
,
Un
iv
e
rsid
a
d
N
a
c
io
n
a
l
d
e
C
o
lo
m
b
ia,
Bo
g
o
tá,
Co
l
o
m
b
ia,
p
p
.
1
-
1
2
9
,
2
0
1
6
.
[3
]
S
p
id
e
r
U
rb
a
n
M
a
n
a
g
e
m
e
n
t
P
latf
o
rm
,
“
S
o
lu
c
io
n
e
s
p
a
ra
la
c
o
n
ta
m
in
a
c
ió
n
a
tm
o
s
f
é
ric
a
e
n
las
sm
a
rtciti
e
s
,
”
2
0
1
9
.
Av
a
il
a
b
le:
h
tt
p
s://
w
ww
.
u
rb
a
n
so
l
u
ti
o
n
s.e
s/e
s/b
lo
g
/
1
1
9
-
so
l
u
c
io
n
e
s
-
p
a
ra
-
la
-
c
o
n
tam
in
a
c
io
n
-
a
tm
o
s
f
e
ric
a
-
en
-
las
-
s
m
a
rt
-
c
it
ies
.
[4
]
R.
Ku
m
a
r
V
.
a
n
d
P
.
Dix
it
,
“
Da
il
y
p
e
a
k
lo
a
d
f
o
re
c
a
st
u
sin
g
a
rti
ficia
l
n
e
u
ra
l
n
e
tw
o
rk
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
9
,
n
o
.
4
,
p
p
.
2
2
5
6
-
2
2
6
3
,
2
0
1
9
.
[5
]
S
.
Ba
rh
m
i
a
n
d
O.
El
F
a
tn
i,
“
Ho
u
rly
w
in
d
sp
e
e
d
f
o
re
c
a
stin
g
b
a
s
e
d
o
n
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
a
n
d
a
rti
f
icia
l
n
e
u
ra
l
n
e
tw
o
rk
s
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
Arti
f
icia
l
I
n
telli
g
e
n
c
e
,
v
o
l.
8
,
n
o
.
3
,
p
p
.
2
8
6
-
2
9
1
,
2
0
1
9
.
[6
]
N.
H.
A
b
d
-
Ra
h
m
a
n
a
n
d
M.
H
.
L
e
e
,
“
A
rti
f
i
c
ial
n
e
u
ra
l
n
e
tw
o
r
k
f
o
re
c
a
stin
g
p
e
rf
o
rm
a
n
c
e
w
it
h
m
issin
g
v
a
lu
e
im
p
u
tatio
n
s
,
”
I
n
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
i
a
l
I
n
telli
g
e
n
c
e
,
v
o
l.
9
,
n
o
.
1
,
p
p
.
3
3
-
3
9
,
2
0
2
0
.
[7
]
S
e
c
re
taría
Distrit
a
l
d
e
Am
b
ien
te
d
e
Bo
g
o
tá,
“
Re
d
d
e
M
o
n
it
o
re
o
d
e
Ca
li
d
a
d
d
e
l
A
ire
d
e
Bo
g
o
tá
–
RM
CA
B
,
”
2019.
Av
a
il
a
b
le:
h
tt
p
:/
/am
b
ien
teb
o
g
o
ta
.
g
o
v
.
c
o
/red
-
de
-
c
a
li
d
a
d
-
d
e
l
-
a
ire.
[8
]
I
.
N
u
rh
a
id
a
,
e
t
a
l
.
,
“
Im
p
le
m
e
n
tatio
n
o
f
d
e
e
p
n
e
u
ra
l
n
e
tw
o
rk
s
(DN
N)
w
it
h
b
a
tch
n
o
rm
a
li
z
a
ti
o
n
f
o
r
b
a
t
ik
p
a
tt
e
rn
re
c
o
g
n
it
io
n
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
1
0
,
n
o
.
2
,
p
p
.
2
0
4
5
-
2
0
5
3
,
2
0
2
0
.
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
-
8708
P
r
ed
ictio
n
o
f a
tmo
s
p
h
eric p
o
llu
tio
n
u
s
in
g
n
eu
r
a
l n
etw
o
r
ks mo
d
el
o
f fin
e
p
a
r
ticles
.
..
(
Ju
a
n
C
a
milo
P
ed
r
a
z
a
)
6581
[9
]
C.
S
.
P
in
t
o
a
n
d
D
.
C
.
G
a
larz
a
,
“
F
u
n
d
a
m
e
n
to
s Bá
sic
o
s d
e
Esta
d
ísti
c
a
,
”
Qu
it
o
,
S
i
n
e
d
it
o
rial,
p
p
.
1
-
2
2
4
,
2
0
1
7
.
[1
0
]
L.
L
.
C
.
P
e
ra
lt
a
,
e
t
a
l.
,
“
A
p
li
c
a
c
ió
n
d
e
Re
d
e
s
n
e
u
ro
n
a
les
a
rti
f
icia
les
p
a
ra
la
p
re
d
icc
ió
n
d
e
c
a
li
d
a
d
d
e
a
ire
,
”
M
e
c
á
n
ica
Co
mp
u
t
a
c
io
n
a
l
,
v
o
l.
XX
V
II,
p
p
.
3
6
0
7
-
3
6
2
5
,
2
0
0
8
.
[1
1
]
B.
B.
Be
z
a
b
e
h
a
n
d
A.
D.
M
e
n
g
ist
u
,
“
T
h
e
e
ff
e
c
ts o
f
m
u
lt
ip
le l
a
y
e
rs
f
e
e
d
-
f
o
r
w
a
rd
n
e
u
ra
l
n
e
tw
o
rk
tran
sf
e
r
f
u
n
c
ti
o
n
i
n
d
ig
it
a
l
b
a
se
d
e
th
i
o
p
ian
s
o
il
c
la
ss
if
ic
a
ti
o
n
a
n
d
m
o
istu
re
p
re
d
ict
io
n
,
”
I
n
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
Co
mp
u
ter
E
n
g
in
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
1
0
,
n
o
.
4
,
p
p
.
4
0
7
3
-
4
0
7
9
,
2
0
2
0
.
[1
2
]
R.
L
ip
p
m
a
n
n
,
“
A
n
in
tro
d
u
c
ti
o
n
to
c
o
m
p
u
ti
n
g
w
it
h
n
e
u
ra
l
n
e
ts
,
”
IE
EE
AS
S
P
M
a
g
a
zi
n
e
,
v
o
l.
4
,
n
o
.
2
,
p
p
.
4
-
2
2
,
1
9
8
7
.
[1
3
]
R.
He
c
h
t
-
Nie
lse
n
,
“
Co
u
n
terp
r
o
p
a
g
a
ti
o
n
n
e
tw
o
rk
s
,
”
Ap
p
l
ied
Op
t
ics
,
v
o
l.
2
6
,
n
o
.
2
3
,
p
p
.
4
9
7
9
-
4
9
8
4
,
1
9
8
7
.
[1
4
]
T
.
M
a
ste
rs,
“
P
ra
c
ti
c
a
l
Ne
u
ra
l
Ne
tw
o
r
k
Re
c
ip
ies
in
C+
+
,
”
M
o
rg
a
n
Ka
u
fm
a
n
n
,
p
p
.
1
-
4
9
3
,
2
0
1
4
.
[1
5
]
R.
Ru
ss
e
l,
“
Re
d
e
s Ne
u
ro
n
a
les
,
”
Cre
a
teSp
a
c
e
In
d
e
p
e
n
d
e
n
t
P
u
b
l
ish
i
n
g
P
latf
o
rm
,
p
p
.
1
-
9
0
,
2
0
1
8
.
[1
6
]
R.
F
ló
re
z
a
n
d
J.
F
e
rn
á
n
d
e
z
,
“
Las
Re
d
e
s
Ne
u
ro
n
a
les
A
rti
f
icia
l
e
s
,
”
Ne
ti
b
lo
,
p
p
.
1
5
2
,
2
0
0
8
.
[
O
n
l
in
e
].
A
v
a
il
a
b
le:
h
tt
p
s:/
/d
ial
n
e
t.
u
n
iri
o
ja.es
/se
rv
let/l
ib
ro
?
c
o
d
ig
o
=
3
9
5
2
4
1
.
[1
7
]
Ke
ra
s,
“
Ke
ra
s Do
c
u
m
e
n
tatio
n
,
”
2
0
1
9
.
A
v
a
il
a
b
le:
h
tt
p
s://
k
e
ra
s.io
/
.
[1
8
]
T
e
n
so
rF
lo
w
,
“
Ba
sic
Re
g
r
e
ss
io
n
:
p
re
d
ict
f
u
e
l
e
ff
icie
n
c
y
,
”
2
0
1
9
.
[
On
li
n
e
]
.
A
v
a
il
a
b
le:
h
tt
p
s:/
/www
.
ten
so
rf
lo
w
.
o
rg
/t
u
to
rials/
k
e
ra
s/b
a
sic
_
re
g
re
ss
io
n
.
[1
9
]
Un
ip
y
th
o
n
,
“
Có
m
o
d
e
sa
rro
ll
a
r
m
o
d
e
lo
s
d
e
De
e
p
L
e
a
rn
in
g
c
o
n
Ke
ra
s
,
”
2018
.
[
On
li
n
e
]
.
A
v
a
il
a
b
le:
h
tt
p
s:/
/u
n
ip
y
th
o
n
.
c
o
m
/co
m
o
-
d
e
sa
rro
ll
a
r
-
m
o
d
e
lo
s
-
de
-
d
e
e
p
-
lea
rn
i
n
g
-
c
o
n
-
k
e
ra
s/
.
[2
0
]
C.
Ig
e
l
a
n
d
M
.
Hü
sk
e
n
,
“
Im
p
ro
v
in
g
th
e
Rp
r
o
p
L
e
a
rn
in
g
A
lg
o
rit
h
m
,
”
Pro
c
e
e
d
in
g
s
o
f
th
e
S
e
c
o
n
d
In
ter
n
a
ti
o
n
a
l
S
y
mp
o
si
u
m o
n
Ne
u
ra
l
Co
m
p
u
t
a
ti
o
n
,
p
p
.
1
1
5
-
1
2
1
,
2
0
0
0
.
[2
1
]
R.
V
.
K.
Re
d
d
y
,
e
t
a
l.
,
“
Ha
n
d
w
rit
ten
Hin
d
i
Dig
it
s
Re
c
o
g
n
it
i
o
n
Us
in
g
Co
n
v
o
lu
t
io
n
a
l
Ne
u
ra
l
Ne
tw
o
rk
w
it
h
RM
S
p
r
o
p
Op
ti
m
iza
ti
o
n
,
”
S
e
c
o
n
d
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
In
telli
g
e
n
t
C
o
mp
u
ti
n
g
a
n
d
C
o
n
tro
l
S
y
ste
ms
(
ICICCS
)
,
p
p
.
4
5
-
5
1
,
2
0
1
8
.
[2
2
]
X
.
Ya
o
,
“
Ev
o
lv
in
g
a
rti
f
icia
l
n
e
u
ra
l
n
e
tw
o
rk
s
,
”
Pro
c
e
e
d
in
g
s o
f
t
h
e
I
EE
E
,
v
o
l.
8
7
,
n
o
.
9
,
p
p
.
1
4
2
3
-
1
4
4
7
,
1
9
9
9
.
[2
3
]
D.
T.
V
.
Dh
a
rm
a
jee
-
Ra
o
a
n
d
K.
V
.
Ra
m
a
n
a
,
“
A
No
v
e
l
A
p
p
ro
a
c
h
f
o
r
Eff
icie
n
t
T
r
a
in
in
g
o
f
De
e
p
Ne
u
ra
l
Ne
tw
o
rk
s
,
”
In
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
i
n
e
e
rin
g
a
n
d
C
o
mp
u
ter
S
c
ien
c
e
(
I
J
EE
CS
)
,
v
o
l.
1
1
,
n
o
.
3
,
p
p
.
9
5
4
-
9
6
1
,
2
0
1
8
.
[2
4
]
L
.
F
.
Be
rto
n
a
,
“
En
tren
a
m
ien
to
d
e
Re
d
e
s
Ne
u
ro
n
a
les
b
a
sa
d
o
e
n
A
lg
o
rit
m
o
s
Ev
o
lu
ti
v
o
s
,
”
T
ra
b
a
jo
d
e
G
r
a
d
o
,
Un
iv
e
rsid
a
d
d
e
B
u
e
n
o
s A
ires
,
A
r
g
e
n
ti
n
a
,
p
p
.
1
-
2
5
3
,
2
0
0
5
.
[2
5
]
J.
A
.
V
.
T
o
rre
s
a
n
d
J.
A
.
D.
Riv
e
ra
,
“
En
tren
a
m
ien
to
d
e
u
n
a
re
d
n
e
u
ro
n
a
l
m
u
lt
ica
p
a
p
a
ra
la
tas
a
d
e
c
a
m
b
io
e
u
ro
-
d
ó
lar
(EUR/US
D)
T
ra
in
i
n
g
a
m
u
l
ti
lay
e
r
n
e
u
ra
l
n
e
tw
o
rk
f
o
r
th
e
Eu
r
o
-
d
o
ll
a
r
(EUR/
USD)
e
x
c
h
a
n
g
e
ra
te
,
”
In
g
e
n
ier
í
a
e
In
v
e
stig
a
c
ió
n
,
v
o
l.
2
7
,
n
o
.
3
,
p
p
.
1
0
6
-
1
1
7
,
2
0
0
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.