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.
11
,
No
.
1
,
Feb
r
u
ar
y
2021
,
p
p
.
163
~
170
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
1
1
i
1
.
pp
1
6
3
-
170
163
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
A co
m
pa
riso
n of
a
ctiva
tion func
tions
in
m
ul
tilay
er n
eura
l
netw
o
rk
f
o
r pr
edi
cting th
e produc
ti
o
n and
cons
u
m
pti
o
n of
elect
rici
ty po
w
er
Abdu
lw
a
hed Sa
la
m
1
,
Ab
dela
a
ziz
E
l H
iba
o
ui
2
,
Abdu
lg
a
bb
a
r
Sa
if
3
1,
2
F
a
c
u
lt
y
o
f
S
c
ien
c
e
,
A
b
d
e
l
m
a
le
k
Essa
a
d
i
Un
iv
e
rsit
y
,
M
o
ro
c
c
o
3
Ce
n
ter f
o
r
A
rti
f
icia
l
In
telli
g
e
n
c
e
T
e
c
h
n
o
lo
g
y
,
F
a
c
u
lt
y
o
f
In
f
o
r
m
a
ti
o
n
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
Un
iv
e
rc
it
y
Ke
b
a
n
g
sa
a
n
M
a
lay
sia
,
M
a
lay
sia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
4
,
2
0
1
8
R
ev
i
s
ed
J
u
l 4
,
2
0
2
0
A
cc
ep
ted
J
u
l 2
9
,
2
0
2
0
P
re
d
ictin
g
e
lec
tri
c
it
y
p
o
we
r
is
a
n
im
p
o
rtan
t
tas
k
,
w
h
ich
h
e
lp
s
p
o
w
e
r
u
ti
li
ti
e
s
in
im
p
ro
v
in
g
t
h
e
ir
sy
ste
m
s’
p
e
rf
o
r
m
a
n
c
e
in
term
s
o
f
e
ffe
c
ti
v
e
n
e
ss
,
p
ro
d
u
c
ti
v
it
y
,
m
a
n
a
g
e
m
e
n
t
a
n
d
c
o
n
tr
o
l.
S
e
v
e
ra
l
re
se
a
r
c
h
e
s
h
a
d
i
n
tro
d
u
c
e
d
th
is
tas
k
u
sin
g
t
h
re
e
m
a
in
m
o
d
e
ls:
e
n
g
in
e
e
rin
g
,
sta
ti
stica
l
a
n
d
a
rti
f
icia
l
in
telli
g
e
n
c
e
.
Ba
se
d
o
n
t
h
e
e
x
p
e
rim
e
n
ts,
w
h
ich
u
se
d
a
rti
f
icia
l
in
telli
g
e
n
c
e
m
o
d
e
ls,
m
u
lt
il
a
y
e
r
n
e
u
ra
l
n
e
t
w
o
rk
s
m
o
d
e
l
h
a
s
p
ro
v
e
n
it
s
su
c
c
e
ss
in
p
re
d
ictin
g
m
a
n
y
e
v
a
lu
a
ti
o
n
d
a
tas
e
ts.
Ho
w
e
v
e
r,
th
e
p
e
rf
o
r
m
a
n
c
e
o
f
th
is m
o
d
e
l
d
e
p
e
n
d
s
m
a
in
l
y
o
n
t
h
e
ty
p
e
o
f
a
c
ti
v
a
ti
o
n
f
u
n
c
ti
o
n
.
T
h
e
re
f
o
re
,
th
is
p
a
p
e
r
in
tro
d
u
c
e
s
a
n
e
x
p
e
ri
m
e
n
tal
stu
d
y
f
o
r
in
v
e
sti
g
a
ti
n
g
th
e
p
e
r
f
o
rm
a
n
c
e
o
f
th
e
m
u
lt
il
a
y
e
r
n
e
u
ra
l
n
e
tw
o
rk
s
m
o
d
e
l
w
it
h
re
sp
e
c
t
to
d
if
f
e
re
n
t
a
c
ti
v
a
ti
o
n
f
u
n
c
ti
o
n
s a
n
d
d
if
f
e
re
n
t
d
e
p
th
s o
f
h
id
d
e
n
lay
e
rs.
T
h
e
e
x
p
e
rime
n
ts i
n
th
is p
a
p
e
r
c
o
v
e
r
th
e
c
o
m
p
a
riso
n
a
m
o
n
g
e
lev
e
n
a
c
ti
v
a
ti
o
n
f
u
n
c
ti
o
n
s
u
sin
g
f
o
u
r
b
e
n
c
h
m
a
rk
e
l
e
c
tri
c
it
y
d
a
tas
e
ts.
T
h
e
a
c
ti
v
a
ti
o
n
f
u
n
c
ti
o
n
s
u
n
d
e
r
e
x
a
m
in
a
ti
o
n
a
re
sig
m
o
id
,
h
y
p
e
rb
o
li
c
tan
g
e
n
t
,
S
o
f
tS
ig
n
,
S
o
f
tP
lu
s,
Re
L
U,
L
e
a
k
Re
L
U,
G
a
u
ss
ian
,
EL
U,
S
EL
U,
S
w
ish
a
n
d
A
d
ju
st
-
S
w
ish
.
Ex
p
e
rim
e
n
tal
re
su
lt
s
sh
o
w
th
a
t
Re
L
U
a
n
d
L
e
a
k
R
e
LU
a
c
ti
v
a
ti
o
n
f
u
n
c
ti
o
n
s o
u
tp
e
rf
o
rm
th
e
ir
c
o
u
n
terp
a
rts
in
a
ll
d
a
tas
e
ts.
K
ey
w
o
r
d
s
:
A
cti
v
atio
n
f
u
n
ctio
n
s
A
r
ti
f
icial
n
e
u
r
al
n
et
w
o
r
k
s
m
u
ltil
a
y
er
p
er
ce
p
tr
o
n
s
E
n
er
g
y
p
r
ed
ictio
n
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
:
A
b
d
u
l
w
ah
ed
Sala
m
,
Facu
lt
y
o
f
Sc
ien
ce
,
A
b
d
el
m
a
lek
E
s
s
aa
d
i U
n
iv
er
s
it
y
,
T
eto
u
an
,
Mo
r
o
cc
o
.
E
m
ail: sa
lla
m
.
y
e
@
y
a
h
o
o
.
co
m
1.
I
NT
RO
D
UCT
I
O
N
E
lectr
ical
p
o
w
er
i
s
p
r
o
d
u
ce
d
f
r
o
m
d
i
f
f
er
en
t
s
o
u
r
ce
s
,
s
o
m
e
o
f
t
h
e
m
ar
e
r
en
e
w
ab
les
a
n
d
o
th
er
s
ar
e
n
o
n
-
r
en
e
w
ab
le
s
.
No
w
ad
a
y
s
,
c
o
n
s
u
m
er
s
ca
n
b
ec
o
m
e
p
r
o
d
u
ce
r
s
i
f
t
h
e
y
h
av
e
o
th
er
p
o
w
er
r
eso
u
r
ce
s
,
e
s
p
ec
iall
y
r
en
e
w
ab
le
s
u
c
h
as
s
o
lar
o
r
w
in
d
p
o
w
er
,
an
d
th
i
s
t
u
r
n
ed
t
o
b
e
d
if
f
ic
u
lt
to
co
n
tr
o
l.
Du
e
t
o
th
at,
th
e
n
ee
d
f
o
r
a
s
tab
le
s
y
s
te
m
to
m
an
a
g
e
th
i
s
n
e
w
a
ttit
u
d
e
e
m
er
g
ed
.
T
h
is
s
y
s
te
m
i
s
s
tan
d
i
n
g
f
o
r
w
h
at
we
h
av
e
an
d
w
h
at
w
e
n
ee
d
.
Ma
n
y
o
r
g
an
izatio
n
s
ar
e
u
s
i
n
g
in
f
o
r
m
a
tio
n
s
y
s
te
m
to
h
elp
th
eir
j
o
b
s
b
y
s
to
r
i
n
g
in
f
o
r
m
atio
n
a
n
d
u
s
e
t
h
a
t
in
f
o
r
m
atio
n
to
h
elp
i
n
m
a
k
in
g
d
ec
is
io
n
s
in
th
e
f
u
t
u
r
e.
P
lan
n
in
g
b
y
f
o
r
ec
asti
n
g
is
to
a
n
al
y
z
e
h
is
to
r
ical
d
ata
to
s
av
e
e
n
er
g
y
a
u
s
a
g
e
[
1
]
.
T
h
is
d
ata
h
ad
b
ee
n
co
llected
o
v
er
th
e
ti
m
e
o
f
th
e
p
r
o
g
r
ess
.
T
h
i
s
d
ata
is
f
o
r
m
atted
m
i
n
u
te
l
y
,
h
o
u
r
l
y
,
d
ail
y
,
w
ee
k
l
y
,
m
o
n
th
l
y
,
q
u
ar
ter
l
y
,
o
r
y
ea
r
l
y
.
A
ls
o
,
it
co
n
s
is
t
s
o
f
u
n
i
v
ar
i
ate
o
r
m
u
lti
v
ar
iate
d
ep
en
d
in
g
o
n
t
h
e
i
n
p
u
t
v
ar
iab
les
th
at
a
f
f
ec
t
elec
tr
icit
y
.
D
u
e
to
th
e
co
m
p
licatio
n
o
f
p
r
ed
ictio
n
,
s
ev
er
al
m
o
d
el
s
h
av
e
b
ee
n
p
r
o
p
o
s
ed
.
T
h
ese
m
o
d
el
s
i
n
cl
u
d
e
en
g
i
n
ee
r
in
g
,
s
tatis
t
ical
a
n
d
ar
tific
ial
i
n
telli
g
en
ce
[
2
]
.
T
h
e
m
o
s
t
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.
11
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
:
16
3
-
1
7
0
164
u
s
ed
m
o
d
el
i
s
ar
ti
f
icial
n
e
u
r
al
n
et
w
o
r
k
(
A
N
N)
.
Fu
l
l
y
co
n
n
e
cted
s
u
p
er
v
i
s
ed
n
et
w
o
r
k
w
it
h
b
ac
k
p
r
o
p
ag
atio
n
lear
n
in
g
r
u
le
i
s
t
h
e
m
o
s
t o
f
te
n
u
s
ed
.
T
h
is
t
y
p
e
o
f
ANN
is
ex
ce
lle
n
t
at
p
r
ed
ictio
n
.
T
o
p
r
ed
ict
o
n
e
o
r
m
o
r
e
v
al
u
es
f
r
o
m
th
e
i
n
p
u
t
v
ar
iab
le
s
is
t
h
e
ai
m
o
f
s
u
p
er
v
is
ed
lear
n
in
g
.
R
e
g
r
es
s
io
n
th
at
d
ep
en
d
s
o
n
ex
a
m
p
le
p
air
s
o
f
d
ata
is
a
f
o
r
m
o
f
s
u
p
er
v
is
ed
lear
n
in
g
.
T
h
is
t
y
p
e
o
f
n
et
w
o
r
k
is
a
f
u
ll
y
i
n
ter
co
n
n
ec
ted
n
eu
r
o
n
s
in
d
if
f
er
en
t
la
y
er
s
,
th
e
i
n
p
u
t
la
y
er
,
th
e
o
u
tp
u
t
la
y
er
,
an
d
th
e
h
id
d
en
la
y
er
s
wh
ich
h
av
e
at
lea
s
t
o
n
h
id
d
en
l
a
y
er
[
3
]
.
T
h
e
h
id
d
en
la
y
er
co
n
n
ec
t
t
h
e
i
n
p
u
t
an
d
th
e
o
u
tp
u
t
la
y
er
s
.
T
h
e
in
p
u
t
la
y
er
r
ec
eiv
es
d
ata
t
h
at
t
h
e
n
eu
r
al
n
et
w
o
r
k
lear
n
s
f
r
o
m
a
n
d
th
e
o
u
tp
u
t
la
y
er
p
r
o
v
id
es
r
esp
o
n
s
e
to
th
e
in
p
u
t
d
ata.
T
h
er
e
is
n
o
ex
ac
t
r
o
le
to
d
eter
m
i
n
e
t
h
e
n
u
m
b
er
o
f
h
id
d
en
la
y
er
s
a
n
d
th
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
in
th
e
h
id
d
en
la
y
er
a
n
d
th
er
e
w
er
e
m
an
y
s
t
u
d
ies
t
h
at
tr
ied
to
f
i
g
u
r
e
th
e
s
e
n
u
m
b
er
s
.
T
h
e
n
eu
r
al
n
et
w
o
r
k
is
lear
n
ed
b
y
alg
o
r
it
h
m
s
.
T
h
er
e
ar
e
m
a
n
y
lear
n
in
g
alg
o
r
it
h
m
s
in
t
h
e
n
eu
r
al
n
e
t
w
o
r
k
b
u
t
th
e
m
o
s
t
co
m
m
o
n
l
y
u
s
ed
is
b
ac
k
p
r
o
p
ag
atio
n
o
f
er
r
o
r
.
T
h
e
er
r
o
r
in
p
r
e
d
ictio
n
is
f
ed
b
ac
k
w
ar
d
s
t
h
r
o
u
g
h
th
e
n
et
w
o
r
k
d
u
r
in
g
tr
ain
i
n
g
to
ad
j
u
s
t
th
e
w
ei
g
h
ts
a
n
d
m
i
n
i
m
ize
t
h
e
er
r
o
r
.
T
h
is
s
tep
is
r
ep
e
ated
u
n
til
ac
h
iev
in
g
th
e
m
i
n
i
m
u
m
er
r
o
r
o
r
r
ea
c
h
in
g
th
e
s
p
ec
i
f
ic
n
u
m
b
er
o
f
ep
o
ch
s
[
4
]
.
I
n
t
h
e
ce
n
ter
o
f
n
e
t
w
o
r
k
s
tan
d
s
th
e
ac
ti
v
atio
n
f
u
n
ctio
n
.
I
n
t
h
eo
r
y
,
an
y
f
u
n
ctio
n
ca
n
b
e
u
tili
ze
d
as
a
n
ac
ti
v
atio
n
f
u
n
ct
io
n
.
Ho
w
e
v
er
,
th
e
ac
ti
v
atio
n
f
u
n
c
tio
n
h
av
e
a
lin
ea
r
an
d
n
o
n
li
n
ea
r
ch
ar
ac
ter
.
No
n
lin
ea
r
ac
ti
v
atio
n
is
i
m
p
o
r
tan
t
i
n
o
r
d
er
to
b
e
ca
p
ab
le
to
d
is
tin
g
u
i
s
h
t
h
e
c
o
m
p
le
x
r
elatio
n
s
h
ip
s
w
h
ic
h
ex
is
t
in
th
e
f
ea
tu
r
e
s
p
ac
e
[
5
]
.
W
h
er
ea
s
lin
ea
r
f
u
n
ctio
n
s
ar
e
p
ar
ticu
lar
l
y
u
s
e
d
in
o
u
tp
u
t
la
y
er
s
e
s
p
ec
iall
y
f
o
r
r
eg
r
ess
io
n
,
n
o
n
-
li
n
ea
r
ac
tiv
atio
n
f
u
n
ct
io
n
s
ca
n
b
e
u
tili
ze
d
in
h
id
d
en
la
y
er
.
A
l
s
o
n
o
n
-
li
n
ea
r
ca
n
b
e
u
s
ed
in
o
u
tp
u
t
la
y
er
s
esp
ec
iall
y
f
o
r
class
i
f
icatio
n
[
6
]
.
Mo
r
eo
v
er
,
d
er
iv
ativ
e
o
f
ac
tiv
atio
n
f
u
n
ctio
n
i
s
i
n
v
o
lv
ed
in
t
h
e
ca
lc
u
latio
n
f
o
r
er
r
o
r
w
h
ic
h
ef
f
ec
t
s
o
n
w
ei
g
h
ts
o
f
n
e
u
r
o
n
co
n
n
ec
tio
n
s
.
W
e
o
r
g
an
ized
th
e
r
est
o
f
t
h
is
p
ap
er
as
f
o
llo
w
s
:
Sec
ti
o
n
2
p
r
esen
ts
ac
ti
v
ati
o
n
f
u
n
ctio
n
s
i
n
co
m
p
ar
is
o
n
.
I
n
s
ec
tio
n
3
w
e
p
r
esen
t
d
ata
s
et
s
d
escr
ip
tio
n
an
d
m
et
h
o
d
o
lo
g
y
t
h
at
u
s
e
d
in
t
h
is
w
o
r
k
i
n
s
ec
tio
n
4
.
E
x
p
er
i
m
e
n
ts
a
n
d
th
eir
r
esu
lts
co
v
er
ed
i
n
Sectio
n
5
Fin
all
y
,
Sectio
n
6
co
n
cl
u
d
es
th
e
p
ap
er
.
2.
ACTI
VAT
I
O
N
F
UNC
T
I
O
N
S IN CO
M
P
ARIS
O
N
T
h
e
ch
o
ice
o
f
ac
ti
v
atio
n
f
u
n
c
tio
n
s
s
tr
o
n
g
l
y
in
f
l
u
e
n
ce
s
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
n
e
u
r
al
n
et
w
o
r
k
s
.
I
n
th
is
s
ess
io
n
,
w
e
d
e
s
cr
ib
e
th
e
m
o
s
t
u
s
ed
ac
ti
v
atio
n
f
u
n
ctio
n
s
.
S
ig
m
o
id
[
7
]
i
s
a
w
id
el
y
u
s
ed
ac
tiv
atio
n
f
u
n
ctio
n
.
T
h
e
o
u
tp
u
t v
al
u
es a
r
e
b
et
w
ee
n
ze
r
o
an
d
o
n
e.
I
ts
eq
u
atio
n
is
:
(
1
)
Du
e
to
s
i
m
p
licit
y
o
f
s
i
g
m
o
id
d
er
iv
ativ
e,
Si
g
m
o
id
is
f
a
s
t
to
ex
ec
u
te.
B
u
t
its
p
r
o
b
le
m
is
th
at
g
r
ad
ien
t
ap
p
r
o
ac
h
es
to
ze
r
o
an
d
th
e
lea
r
n
in
g
o
f
n
e
t
w
o
r
k
b
ec
o
m
e
s
d
if
f
icu
lt.
T
h
e
h
y
p
er
b
o
lic
tan
g
e
n
t
f
u
n
ct
io
n
[
8
]
is
ze
r
o
ce
n
ter
ed
an
d
its
o
u
tp
u
t b
et
w
ee
n
-
1
an
d
1
,
(
2
)
So
f
tSi
g
n
[
8
]
f
u
n
c
tio
n
i
s
clo
s
el
y
r
elate
d
to
h
y
p
er
b
o
lic
tan
g
en
t.
So
f
t
Si
g
n
co
n
v
er
g
es
p
o
l
y
n
o
m
iall
y
w
h
er
ea
s
h
y
p
er
b
o
lic
tan
g
en
t c
o
n
v
er
g
es e
x
p
o
n
en
t
iall
y
.
So
f
t
Si
g
n
d
e
f
in
ed
as
(
3
)
R
ec
ti
f
ied
lin
ea
r
u
n
it
(
R
e
L
U)
[
9
]
d
ef
in
ed
as p
o
s
itiv
e
p
ar
t o
f
x
:
(
4
)
I
t
is
s
i
m
p
le,
an
d
f
ast.
I
n
ad
d
itio
n
,
it
r
ec
tif
ies
v
a
n
i
s
h
i
n
g
g
r
ad
ien
t
p
r
o
b
lem
.
T
h
e
m
o
s
t
asp
ec
t
o
f
R
e
L
U
is
d
ea
d
n
eu
r
o
n
s
w
h
ic
h
m
ea
n
it
is
n
ev
er
b
ee
n
ac
ti
v
ated
w
h
en
x
is
les
s
th
a
n
ze
r
o
.
In
[
1
0
]
,
au
th
o
r
s
d
ef
in
ed
a
leak
r
ec
tif
ied
li
n
ea
r
u
n
it
(
L
R
eL
U)
as
,
(
5
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
co
mp
a
r
is
o
n
o
f
a
ctiva
tio
n
fu
n
ctio
n
s
in
mu
ltil
a
ye
r
n
eu
r
a
l
…
(
A
b
d
u
lw
a
h
ed
S
a
l
a
m
)
165
I
t
is
s
i
m
i
lar
to
R
eL
U.
T
h
e
o
n
ly
d
i
f
f
er
en
ce
is
t
h
at
L
R
e
L
U
f
l
o
w
s
w
h
en
x
<0
an
d
th
at
s
o
l
v
e
s
th
e
d
ea
d
n
eu
r
o
n
s
o
f
R
eL
U
f
r
o
m
t
h
e
a
u
t
h
o
r
's p
o
in
t o
f
v
ie
w
.
So
f
tP
lu
s
[
1
1
]
:
(
6
)
So
f
tP
lu
s
is
j
u
s
t
a
s
m
o
o
th
ap
p
r
o
x
im
a
tio
n
to
th
e
L
R
e
L
U
b
u
t
it
is
p
o
s
itiv
e.
Fig
u
r
e
1
s
h
o
w
s
th
e
d
if
f
er
en
ce
b
et
w
ee
n
a
m
o
n
g
Sig
m
o
id
,
T
an
h
an
d
So
f
t
Sig
n
,
R
eL
U,
L
ea
k
R
e
L
U
a
n
d
So
f
tP
l
u
s
.
Fig
u
r
e.
1
.
Sig
m
o
id
,
T
an
h
,
So
f
t
Sig
n
,
L
ea
k
R
e
L
U,
R
e
L
U
an
d
So
f
tP
lu
s
ac
tiv
at
io
n
f
u
n
ctio
n
s
Gau
s
s
ia
n
ac
tiv
at
io
n
f
u
n
ctio
n
[
1
2
]
,
d
ef
in
ed
as
,
(
7
)
E
x
p
o
n
en
t
ial
li
n
ea
r
u
n
it
(
E
L
U)
[
1
3
]
is
lik
e
R
e
L
U
an
d
lea
k
y
R
eL
U
f
o
r
av
o
id
in
g
a
v
an
i
s
h
in
g
g
r
ad
ie
n
t
b
y
t
h
e
id
en
t
it
y
o
f
p
o
s
itiv
e
v
al
u
es,
(
8
)
w
h
er
e
α
=1
.
0
Scaled
ex
p
o
n
e
n
tial
l
in
ea
r
u
n
it
(
SEL
U)
[
1
4
]
w
h
ic
h
i
s
a
m
o
d
i
f
ied
t
y
p
e
o
f
E
L
U
w
it
h
t
w
o
f
i
x
ed
p
ar
am
eter
s
.
I
t is d
ef
i
n
ed
as
,
(
9
)
w
h
er
e
α
≈
1
.
6
7
3
3
an
d
γ
≈
1
.
0
5
0
7
Go
o
g
le
B
r
ain
in
tr
o
d
u
ce
s
a
n
e
w
ac
ti
v
atio
n
f
u
n
ctio
n
w
h
ic
h
is
ca
lled
S
w
i
s
h
[
1
5
]
(
1
0
)
A
d
j
u
s
ted
s
w
i
s
h
(E
-
S
w
i
s
h
)
[
1
6
]
is
a
m
o
d
if
ied
t
y
p
e
o
f
S
w
is
h
a
n
d
is
d
ef
i
n
ed
as
(
1
1
)
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.
11
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
:
16
3
-
1
7
0
166
T
h
e
d
if
f
er
en
ce
b
et
w
ee
n
S
w
i
s
h
a
n
d
E
-
S
w
i
s
h
is
t
h
e
p
o
s
iti
o
n
o
f
co
n
s
ta
n
t
β.
T
h
e
s
i
m
il
ar
ities
an
d
d
if
f
er
e
n
ce
s
b
et
w
ee
n
t
h
e
m
s
h
o
w
n
ab
o
v
e
i
n
Fig
u
r
e
2
.
S
w
i
s
h
an
d
E
-
S
w
is
h
ac
ti
v
atio
n
f
u
n
cti
o
n
s
ar
e
eq
u
al
w
h
e
n
β=1
.
A
cc
o
r
d
in
g
to
[
1
5
,
1
6
]
p
ap
er
s
,
th
e
y
p
er
f
o
r
m
ed
a
g
o
o
d
r
esu
lt
i
n
i
m
ag
e
clas
s
i
f
icatio
n
a
n
d
n
e
v
er
e
x
a
m
in
e
s
b
ef
o
r
e
in
r
e
g
r
ess
io
n
.
I
n
t
h
i
s
w
o
r
k
w
e
s
et
β=1
f
o
r
S
w
i
s
h
a
ctiv
atio
n
f
u
n
ctio
n
as
th
e
b
e
s
t
ac
h
ie
v
ed
r
esu
lt
i
n
th
e
o
r
ig
i
n
al
p
ap
er
an
d
β=1
.
5
i
n
th
e
E
-
S
w
i
s
h
ac
tiv
a
tio
n
f
u
n
ct
io
n
f
o
r
th
e
s
a
m
e
r
ea
s
o
n
.
Fig
u
r
e.
2
.
E
L
U,
SEL
U,
g
a
u
s
s
i
an
,
s
w
i
s
h
a
n
d
E
-
s
w
i
s
h
ac
ti
v
ati
o
n
f
u
n
ct
io
n
s
2
.
1
.
Da
t
a
s
et
des
c
ript
io
n
I
n
th
is
p
a
p
e
r
w
e
tr
y
to
c
o
v
e
r
d
i
f
f
er
en
t
ty
p
es o
f
d
at
a
s
et
s
w
h
ich
ar
e
r
ela
te
d
t
o
e
lec
tr
ici
ty
.
T
h
e
v
ar
i
ety
in
th
e
q
u
ality
o
f
th
e
d
ata
s
et
is
u
s
ed
.
C
o
n
s
u
m
p
ti
o
n
,
p
r
o
d
u
c
t
io
n
,
m
u
ltiv
ar
iat
e
an
d
u
n
iv
a
r
i
ate
ar
e
c
o
n
s
i
d
e
r
ed
.
A
ls
o
,
s
o
m
e
o
f
th
em
d
ep
en
d
s
o
n
d
ate
an
d
tim
e.
Mo
r
eo
v
e
r
,
w
e
co
n
s
i
d
e
r
th
e
r
ela
ti
o
n
s
h
ip
s
b
et
w
ee
n
th
e
in
p
u
t
an
d
ou
t
p
u
t v
a
r
i
ab
les
.
2
.
2
.
Co
m
bi
ned
cy
cle
po
w
er
pla
nt
C
o
m
b
in
ed
cy
cle
p
o
w
er
p
lan
t
d
at
a
s
et
c
o
n
s
is
t
s
o
f
f
o
u
r
in
p
u
t
v
ar
i
ab
les
tem
p
e
r
atu
r
e
,
am
b
ien
t
p
r
ess
u
r
e,
r
el
ativ
e
h
u
m
id
ity
an
d
ex
h
au
s
t
v
ac
u
u
m
.
T
h
o
s
e
in
p
u
t
v
ar
ia
b
l
es
a
r
e
u
s
e
d
to
p
r
ed
ict
e
lec
tr
i
ca
l
en
e
r
g
y
o
u
tp
u
t
.
T
h
e
d
a
ta
s
et
w
as
co
ll
ec
t
ed
o
v
e
r
s
ix
y
ea
r
s
(
2
0
0
6
-
2
0
1
1
)
an
d
c
o
n
tain
s
9
5
6
8
d
ata
p
o
in
ts
[
1
7
]
.
T
h
er
e
is
n
o
m
is
s
ed
d
at
a
in
th
is
d
at
a
s
et
.
Fig
u
r
e
3
b
elo
w
s
h
o
w
s
th
e
r
ela
ti
o
n
s
h
ip
b
e
t
w
ee
n
en
e
r
g
y
an
d
its
f
ac
t
o
r
s
.
Fig
u
r
e
3
.
Scatter
d
iag
r
a
m
o
f
el
ec
tr
ical
en
er
g
y
o
u
tp
u
t v
s
te
m
p
er
atu
r
e,
p
r
ess
u
r
e,
h
u
m
id
it
y
a
n
d
ex
h
au
s
t
v
ac
u
u
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
co
mp
a
r
is
o
n
o
f
a
ctiva
tio
n
fu
n
ctio
n
s
in
mu
ltil
a
ye
r
n
eu
r
a
l
…
(
A
b
d
u
lw
a
h
ed
S
a
l
a
m
)
167
2
.
3
.
E
nerg
y
ef
f
iciency
da
t
a
s
et
th
e
d
a
tase
t
c
o
n
s
is
t
o
f
eig
h
t
in
p
u
t
attr
ib
u
t
es
r
ela
tiv
e
c
o
m
p
ac
tn
ess
(
c
o
m
p
)
,
s
u
r
f
ac
e
ar
ea
(
s
a
)
,
w
all
A
r
ea
(
W
A
)
,
r
o
o
f
ar
ea
(
R
A
)
,
o
v
e
r
al
l
h
eig
h
t
(
OH
)
,
o
r
i
en
tat
io
n
(
O
r
ie
)
,
g
l
az
in
g
a
r
ea
(
GA
)
,
g
la
zin
g
ar
e
a
d
is
t
r
i
b
u
ti
o
n
(
GA
D)
w
h
ich
ar
e
th
e
b
u
il
d
in
g
d
es
ig
n
p
a
r
am
ete
r
s
.
T
h
e
r
es
p
o
n
s
es
o
u
tc
o
m
es
ar
e
h
ea
tin
g
l
o
a
d
(
HL
)
an
d
c
o
o
lin
g
lo
a
d
(
C
L
)
w
h
ich
ar
e
p
o
w
er
c
o
n
s
u
m
p
tio
n
[
1
8
]
.
2
.
4
.
Appl
ia
nces e
nerg
y
pred
ict
io
n da
t
a
s
et
It
co
n
s
is
ts
o
f
1
9
7
3
5
in
s
tan
c
es
o
f
2
8
att
r
i
b
u
t
es.
T
h
o
s
e
a
tt
r
i
b
u
t
es
a
r
e
r
ela
te
d
t
o
tem
p
er
atu
r
e,
h
u
m
id
ity
,
p
r
ess
u
r
e,
w
in
d
s
p
ee
d
,
d
ay
s
o
f
w
ee
k
an
d
d
ay
s
tatu
s
[
1
9
]
.
I
t
d
o
es n
o
t
c
o
n
tain
e
an
y
m
is
s
in
g
v
alu
es
.
2
.
5
.
I
nd
iv
idu
a
l ho
us
e
ho
ld el
ec
t
ric
po
w
er
c
o
ns
u
m
ptio
n d
a
t
a
s
et
T
h
is
d
at
ase
t
is
tim
e
s
e
r
i
es
w
h
ich
m
ea
s
u
r
es
an
ele
ct
r
ic
p
o
w
er
co
n
s
u
m
p
tio
n
in
o
n
e
h
o
u
s
eh
o
l
d
f
o
r
a
s
in
g
le
r
es
id
en
tia
l
cu
s
to
m
er
.
I
t
co
n
t
ain
s
th
e
c
o
n
s
u
m
p
tio
n
p
o
w
er
r
ec
o
r
d
th
a
t
is
tak
en
in
ea
ch
m
in
u
te
o
v
er
a
p
e
r
i
o
d
o
f
f
o
u
r
y
ea
r
s
b
etw
ee
n
D
ec
em
b
e
r
2
0
0
6
an
d
N
o
v
em
b
er
2
0
1
0
.
T
h
e
o
r
ig
in
al
d
a
taset
c
o
n
t
ain
s
n
in
e
att
r
i
b
u
tes
b
u
t
w
e
f
o
cu
s
e
d
o
n
t
h
e
d
a
te
t
im
e
as
a
tim
e
s
e
r
ies
a
n
d
th
e
h
o
u
s
eh
o
ld
g
l
o
b
a
l
m
in
u
te
-
av
e
r
ag
e
d
ac
tiv
e
p
o
w
er
.
T
h
e
r
e
a
r
e
m
is
s
in
g
v
alu
es
o
f
d
a
ta
an
d
w
e
f
ill
th
e
m
b
y
u
s
in
g
th
e
v
alu
es
th
at
co
m
e
b
ef
o
r
e
[
2
0
]
.
T
a
b
le
1
s
h
o
w
s
th
e
p
r
o
p
e
r
ti
es
o
f
d
a
ta
s
et
.
T
ab
le
1
.
I
n
d
iv
id
u
al
h
o
u
s
eh
o
ld
elec
tr
ic
p
o
w
er
co
n
s
u
m
p
tio
n
p
r
o
p
er
ties
P
r
o
p
e
r
t
i
y
V
a
l
u
e
C
o
u
n
t
2
,
0
7
5
,
2
5
9
M
a
x
1
1
.
1
2
0
Min
0
.
0
7
6
M
e
a
n
1
.
0
8
6
M
e
d
i
a
n
0
.
5
9
8
S
T
D
1
.
0
5
3
M
i
s
s
i
n
g
v
a
l
u
e
s
2
5
,
9
7
9
3.
M
E
T
H
O
DO
L
O
G
Y
W
e
u
s
e
n
e
u
r
al
n
et
w
o
r
k
w
it
h
d
if
f
er
en
t
d
ep
th
o
f
h
id
d
en
la
y
er
s
to
ex
a
m
i
n
e
th
e
m
o
s
t
ac
cu
r
ate
ac
tiv
atio
n
f
u
n
ctio
n
.
Ma
n
y
s
tu
d
ie
s
tr
ied
to
d
ef
in
e
th
e
n
u
m
b
er
o
f
h
id
d
en
lay
er
s
an
d
th
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
in
ea
ch
h
id
d
en
la
y
er
b
u
t
w
e
k
n
o
w
t
h
at
n
o
ex
a
ct
r
o
le
ca
n
d
ef
in
e
t
h
at.
I
n
th
i
s
p
ap
er
,
w
e
co
n
ce
n
tr
ate
o
n
t
h
e
ac
tiv
atio
n
f
u
n
ctio
n
an
d
w
e
d
o
n
o
t
f
o
cu
s
o
n
t
h
e
s
tr
u
ct
u
r
e
o
f
n
e
u
r
al
n
et
w
o
r
k
.
T
h
er
ef
o
r
e,
w
e
d
ef
i
n
ed
f
o
u
r
m
o
d
el
s
o
f
m
u
lti
la
y
er
n
eu
r
al
n
et
w
o
r
k
s
w
i
th
d
if
f
er
en
t
d
ep
th
.
T
h
e
f
ir
s
t
m
o
d
el
is
s
a
m
p
le
w
it
h
o
n
e
h
id
d
en
la
y
er
a
n
d
d
if
f
er
en
t
n
u
m
b
er
o
f
n
e
u
r
o
n
s
ac
co
r
d
in
g
to
th
e
f
o
u
r
d
if
f
er
en
t d
ataset
s
.
W
e
s
p
ec
if
ied
th
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
b
y
f
u
n
ctio
n
[
2
]
.
ℎ
=
+ 1
(
1
2
)
w
h
er
e
N
is
t
h
e
n
u
m
b
er
o
f
i
n
p
u
t d
ata
T
h
e
o
th
er
m
o
d
els
ar
e
d
ef
in
ed
w
ith
m
o
r
e
h
id
d
en
lay
er
s
.
Mo
d
el
2
co
n
s
is
ts
o
f
tw
o
h
id
d
en
lay
er
s
w
ith
3
0
an
d
2
0
n
eu
r
o
n
s
in
ea
ch
lay
er
.
T
h
e
th
ir
d
is
d
ee
p
er
th
an
p
r
ev
io
u
s
m
o
d
els
w
ith
9
h
id
d
en
lay
er
s
w
ith
2
4
0
,
2
0
0
,
1
6
0
,
1
2
0
,
8
0
,
6
0
,
4
0
,
3
0
,
2
0
n
eu
r
o
n
s
in
h
id
d
en
la
y
er
r
esp
ec
tiv
ely
.
T
h
e
last
m
o
d
el
is
d
esig
n
ed
w
ith
1
1
h
id
d
en
lay
er
s
w
ith
3
2
0
,
2
8
0
,
2
4
0
,
2
0
0
,
1
6
0
,
1
2
0
,
8
0
,
6
0
,
4
0
,
3
0
,
2
0
n
eu
r
o
n
s
.
W
e
tr
ied
to
g
o
d
ee
p
er
to
d
eter
m
in
e
th
e
ab
ilit
y
o
f
ac
tiv
atio
n
f
u
n
ctio
n
in
d
ee
p
n
eu
r
al
n
etw
o
r
k
.
W
e
im
p
lem
en
t
th
e
s
to
ch
asti
c
g
r
ad
ien
t
d
escen
t
(
SGD)
[
2
1
]
as
th
e
o
p
tim
izer
.
T
h
e
n
u
m
b
er
o
f
ep
o
ch
s
is
en
o
u
g
h
to
r
ea
ch
th
e
b
est
s
o
lu
tio
n
.
T
h
e
in
itial
lear
n
in
g
r
ate
is
s
et
to
0
.
1
an
d
w
e
r
ed
u
ce
d
it b
y
m
u
ltip
ley
in
g
it b
y
0
.
2
at
4
0
,
8
0
,
1
2
0
,
1
6
0
ep
o
ch
s
.
T
h
e
in
itializatio
n
o
f
tr
ain
in
g
is
Glo
r
o
t
u
n
if
o
r
m
in
itializatio
n
[
8
]
.
W
e
u
tili
ze
d
n
o
d
r
o
p
o
u
t
an
d
0
.
9
m
o
m
en
tu
m
.
T
h
e
p
ar
am
eter
s
th
at
u
s
ed
in
th
e
f
o
u
r
m
o
d
els
ar
e
eq
u
al
f
o
r
th
e
all
d
ata
s
ets
ex
ce
p
t
f
o
r
th
e
in
d
iv
id
u
al
h
o
u
s
eh
o
ld
elec
tr
ic
p
o
w
er
co
n
s
u
m
p
tio
n
d
ata
s
et
b
ec
au
s
e
it
h
as
h
u
g
e
n
u
m
b
er
o
f
d
ata
an
d
n
ee
d
s
lo
n
g
tim
e
o
f
ex
ec
u
tio
n
.
So
,
th
e
n
u
m
b
er
o
f
ep
o
ch
s
is
r
ed
u
ce
d
to
4
0
w
ith
A
d
am
o
p
tim
izer
[
2
2
]
as
it
co
n
s
id
er
s
f
aster
th
an
o
th
er
s
.
W
e
u
s
e
r
o
o
t
m
ea
n
s
q
u
ar
ed
er
r
o
r
(
R
MSE
)
[
32
]
w
h
ich
is
o
n
e
o
f
q
u
an
titi
v
e
ap
p
r
o
ac
h
es
o
f
m
o
d
el
p
er
f
o
r
m
ac
e
ev
alu
atio
n
.
I
t
co
m
p
ar
es
p
r
ed
icted
v
alu
es
to
th
e
ac
tu
al
v
alu
es
n
X
X
R
M
S
E
n
i
i
d
e
l
mo
i
obs
1
2
,
,
)
(
(
1
3
)
w
h
er
e
X
obs
is
p
r
ed
o
cted
v
alu
es a
n
d
X
m
odel
is
ac
tu
al
v
alu
es
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.
11
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
:
16
3
-
1
7
0
168
T
o
tr
ain
th
e
n
eu
r
al
n
etw
o
r
k
ef
f
icien
tly
[
2
4
]
,
w
e
p
er
f
o
r
m
d
ata
p
r
ep
r
o
ce
s
s
in
g
to
tr
an
s
f
o
r
m
in
p
u
t
d
ata
in
to
b
etter
f
o
r
m
.
T
h
e
tr
an
s
f
o
r
m
atio
n
w
as
u
s
ed
b
y
Min
-
Ma
x
n
o
r
m
aliza
tio
n
[
2
5
]
w
h
ich
is
o
n
e
o
f
th
e
m
o
s
t
u
s
ed
tech
n
iq
u
es
an
d
it
is
ac
co
m
p
lis
h
ed
b
y
:
=
x
−
−
(
1
4
)
4.
E
XP
E
R
I
M
E
NT
S AN
D
RE
S
UL
T
I
n
o
r
d
er
to
ev
alu
ate
th
e
ac
tiv
atio
n
f
u
n
ctio
n
w
ith
th
e
f
o
u
r
m
o
d
els
an
d
d
atasets
,
all
d
atasets
d
iv
id
e
in
to
tr
ain
s
et
an
d
test
s
et.
E
ac
h
m
o
d
el
tr
ain
s
b
y
u
s
in
g
7
5
%
o
f
th
e
d
ata
an
d
th
e
r
est
f
o
r
test
in
g
w
h
ich
ar
e
u
n
s
ee
n
v
alu
es.
T
h
e
test
s
et
is
u
s
u
ally
u
s
ed
to
g
a
g
e
th
e
m
o
d
els
b
u
t
also
tr
ain
in
g
s
et
s
h
o
w
s
th
e
ab
ilit
y
o
f
lear
n
in
g
.
W
e
co
m
p
ar
e
th
e
m
ed
ian
o
f
9
r
u
n
s
in
all
ex
ec
u
tio
n
m
o
d
els
o
f
d
atasets
ex
ce
p
t
f
o
r
th
e
I
n
d
iv
id
u
al
h
o
u
s
eh
o
ld
elec
tr
ic
p
o
w
er
co
n
s
u
m
p
tio
n
d
ataset.
As
s
h
o
w
n
i
n
T
ab
le
2
th
at
R
e
L
U
ac
ti
v
atio
n
g
et
s
b
est
r
es
u
lt
in
m
o
d
el
I
I
,
I
I
I
an
d
I
V
f
o
r
tr
ain
i
n
g
a
n
d
test
i
n
g
s
et
o
f
co
m
b
in
ed
c
y
cle
p
o
w
er
p
lan
t
d
ata
s
et.
I
n
ad
d
itio
n
,
SEL
U
g
et
s
b
est
p
er
f
o
r
m
a
n
ce
f
o
r
tr
ai
n
in
g
a
n
d
test
i
n
g
s
et
i
n
m
o
d
el
I
.
T
ab
le
3
s
h
o
w
s
t
h
at
R
e
L
U
a
n
d
L
ea
k
R
eL
U
g
e
t
th
e
b
est
p
er
f
o
r
m
a
n
ce
in
all
m
o
d
els
f
o
r
en
er
g
y
ef
f
icie
n
c
y
d
ata
s
et.
T
h
e
m
in
i
m
u
m
tr
ain
er
r
o
r
s
in
m
o
d
el
I
I
,
I
I
I
,
I
V
an
d
th
e
m
i
n
i
m
u
m
test
er
r
o
r
in
m
o
d
el
I
I
is
g
o
tten
b
y
R
e
L
U.
L
ea
k
R
e
L
U
ac
h
iev
e
s
th
e
m
i
n
i
m
u
m
er
r
o
r
s
in
th
e
r
est.
T
ab
le
2
.
R
MSE
co
m
p
ar
is
o
n
o
f
ac
tiv
atio
n
f
u
n
c
tio
n
i
n
4
m
o
d
els
fo
r
co
m
b
in
ed
cy
cle
p
o
w
er
p
lan
t
d
ataset
A
c
t
i
v
a
t
i
o
n
M
O
D
EL
I
M
O
D
EL
I
I
M
O
D
EL
I
I
I
M
O
D
EL
I
V
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
S
i
g
mo
i
d
4
.
4
2
1
4
.
6
1
9
4
.
5
0
1
4
.
7
0
7
4
.
1
8
2
4
.
3
6
5
4
.
1
2
5
4
.
3
1
0
T
a
n
h
4
.
2
5
1
4
.
4
4
2
4
.
1
3
0
4
.
3
2
6
3
.
9
7
8
4
.
2
2
8
3
.
9
2
7
4
.
2
0
4
S
o
f
t
S
i
g
n
4
.
2
5
1
4
.
4
5
1
4
.
1
3
2
4
.
3
2
5
4
.
0
3
9
4
.
2
6
5
3
.
7
4
2
4
.
1
0
2
S
o
f
t
P
l
u
s
4
.
3
0
3
4
.
4
0
4
4
.
2
3
6
4
.
4
1
9
4
.
1
5
0
4
.
3
2
3
4
.
1
5
5
4
.
3
4
0
R
e
L
U
4
.
2
1
8
4
.
4
1
3
3
.
8
5
6
4
.
1
4
1
3
.
2
1
4
3
.
8
4
5
3
.
1
2
6
3
.
8
6
0
L
R
e
L
U
4
.
2
1
2
4
.
4
0
9
3
.
9
1
5
4
.
1
7
3
3
.
4
3
8
3
.
9
3
9
3
.
3
8
3
3
.
9
4
3
G
a
u
ssi
a
n
4
.
2
2
3
4
.
3
9
4
4
.
1
5
3
4
.
3
2
8
1
7
.
0
7
1
7
.
0
4
1
7
.
0
7
1
7
.
0
4
EL
U
4
.
3
8
3
4
.
4
5
1
4
.
1
5
8
4
.
3
3
7
4
.
0
5
5
4
.
2
7
3
3
.
9
4
9
4
.
2
0
0
S
E
L
U
4
.
1
8
7
4
.
3
8
3
4
.
1
3
8
4
.
3
3
4
3
.
7
5
3
4
.
0
9
1
3
.
8
0
9
4
.
2
0
0
S
w
i
sh
4
.
2
3
9
4
.
4
2
4
4
.
2
3
3
4
.
4
1
3
4
.
0
7
1
4
.
2
7
3
3
.
9
4
5
4
.
1
8
7
E
-
S
w
i
sh
4
.
2
3
9
4
.
4
2
6
4
.
2
1
4
4
.
3
9
7
3
.
9
9
1
4
.
2
0
7
3
.
9
3
6
4
.
1
8
6
T
ab
le
3
.
R
MSE
co
m
p
ar
is
o
n
o
f
ac
tiv
atio
n
f
u
n
c
tio
n
i
n
4
m
o
d
els f
o
r
en
er
g
y
e
f
f
icien
c
y
(
h
ea
ti
n
g
lo
ad
)
d
ataset
A
c
t
i
v
a
t
i
o
n
M
O
D
EL
I
M
O
D
EL
I
I
M
O
D
EL
I
I
I
M
O
D
EL
I
V
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
S
i
g
mo
i
d
2
.
9
2
7
3
.
1
4
0
2
.
9
2
8
3
.
1
2
7
2
.
6
7
5
3
.
5
3
0
2
.
3
5
7
3
.
9
7
7
T
a
n
h
2
.
9
2
1
3
.
1
3
7
2
.
9
6
9
3
.
2
2
7
2
.
6
3
6
2
.
7
7
3
1
.
0
2
1
1
.
4
6
7
S
o
f
t
S
i
g
n
2
.
6
1
4
2
.
4
6
4
3
.
0
3
1
3
.
1
7
1
0
.
7
2
4
2
.
1
9
5
0
.
9
0
6
1
.
2
7
6
S
o
f
t
P
l
u
s
2
.
8
6
0
2
.
9
8
4
2
.
7
2
4
2
.
5
9
9
1
.
9
9
0
2
.
8
6
3
1
.
0
2
1
1
.
5
3
1
R
e
L
U
2
.
1
3
0
2
.
1
8
9
0
.
4
4
6
0
.
6
7
9
0
.
3
8
5
1
.
0
0
2
0
.
3
9
8
1
.
2
4
8
L
R
e
L
U
2
.
0
2
1
2
.
0
5
7
0
.
6
5
2
0
.
7
8
3
0
.
4
6
2
0
.
9
1
8
0
.
4
5
1
1
.
0
4
4
G
a
u
ssi
a
n
2
.
4
5
7
3
.
2
3
5
2
.
7
9
7
3
.
4
2
5
2
.
4
7
6
2
.
7
1
7
2
.
6
6
5
3
.
5
0
6
EL
U
2
.
7
5
2
2
.
6
6
9
2
.
7
6
6
2
.
8
0
9
1
.
7
9
7
2
.
0
3
5
1
.
6
1
2
2
.
1
9
0
S
E
L
U
2
.
2
6
5
2
.
4
0
0
2
.
8
5
7
2
.
9
9
1
0
.
4
7
7
1
.
1
6
5
0
.
4
3
9
1
.
0
6
2
S
w
i
sh
2
.
7
3
4
2
.
6
5
1
2
.
6
5
6
3
.
0
8
5
0
.
8
8
5
1
.
5
4
8
0
.
8
0
8
1
.
0
7
4
E
-
S
w
i
sh
2
.
7
2
5
2
.
6
3
9
2
.
7
0
4
2
.
7
2
9
1
.
3
7
0
1
.
7
5
9
0
.
8
3
3
1
.
1
3
5
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
R
e
L
U
a
n
d
L
ea
k
R
e
L
U
al
s
o
o
u
tp
er
f
o
r
m
t
h
e
o
th
er
ac
ti
v
atio
n
f
u
n
ctio
n
s
.
T
ab
le
3
an
d
4
p
r
esen
t
th
e
p
r
ed
ictio
n
o
f
th
e
s
a
m
e
d
ataset.
T
h
e
o
n
l
y
d
i
f
f
er
en
ce
is
t
h
at
T
ab
le
3
p
r
esen
ts
th
e
p
r
ed
ictio
n
o
f
h
ea
ti
n
g
lo
ad
r
esp
o
n
s
e
v
ar
iab
le
an
d
T
ab
le
4
s
h
o
w
s
t
h
e
p
r
ed
ictio
n
o
f
co
o
lin
g
lo
ad
r
esp
o
n
s
e
v
ar
iab
le.
T
h
e
co
r
r
elatio
n
a
m
o
n
g
v
ar
iab
les
in
ap
p
lia
n
ce
s
e
n
er
g
y
i
s
o
b
v
io
u
s
l
y
clea
r
t
h
at
t
h
e
r
el
atio
n
s
h
ip
s
b
et
w
ee
n
in
p
u
t a
n
d
o
u
tp
u
t o
f
t
h
e
ap
p
lian
ce
s
en
er
g
y
p
r
ed
ictio
n
d
ataset
a
r
e
w
ea
k
.
T
h
e
er
r
o
r
is
lar
g
e
co
m
p
ar
ed
to
o
th
er
d
atasets
.
I
n
t
h
e
o
r
ig
i
n
al
p
ap
er
[
1
9
]
,
m
u
ltip
le
l
in
ea
r
r
e
g
r
ess
io
n
m
o
d
el
g
ets
(
R
MSE
=
9
3
.
2
1
)
f
o
r
tr
ain
in
g
s
et
an
d
(
R
MSE
=9
3
.
1
8
)
f
o
r
tes
tin
g
s
et.
I
t
is
s
ee
n
i
n
T
ab
le
5
b
elo
w
th
at
SE
L
U
ac
h
ie
v
es
b
est
r
esu
lt
i
n
Mo
d
el
I
.
I
n
ad
d
itio
n
,
R
e
L
U
a
n
d
L
ea
k
R
e
L
U
s
till
o
u
tp
er
f
o
r
m
o
th
e
r
ac
tiv
atio
n
f
u
n
ctio
n
s
in
Mo
d
el
I
I
,
Mo
d
el
I
I
I
an
d
Mo
d
el
I
V.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
co
mp
a
r
is
o
n
o
f
a
ctiva
tio
n
fu
n
ctio
n
s
in
mu
ltil
a
ye
r
n
eu
r
a
l
…
(
A
b
d
u
lw
a
h
ed
S
a
l
a
m
)
169
T
ab
le
4
.
R
MSE
co
m
p
ar
is
o
n
o
f
ac
tiv
atio
n
f
u
n
c
tio
n
i
n
4
m
o
d
els f
o
r
en
er
g
y
e
f
f
icien
c
y
(
co
o
lin
g
lo
ad
)
d
ataset
A
c
t
i
v
a
t
i
o
n
M
O
D
EL
I
M
O
D
EL
I
I
M
O
D
EL
III
M
O
D
EL
I
V
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
S
i
g
mo
i
d
3
.
1
9
5
3
.
2
8
4
3
.
1
9
3
3
.
2
7
2
3
.
0
6
9
3
.
3
4
7
2
.
5
9
7
3
.
0
9
2
T
a
n
h
3
.
1
8
9
3
.
2
7
6
3
.
2
4
3
3
.
4
6
0
2
.
4
9
2
2
.
9
4
5
1
.
1
9
7
2
.
1
5
2
S
o
f
t
S
i
g
n
3
.
0
3
7
3
.
0
0
6
3
.
0
3
6
3
.
0
7
7
2
.
7
9
4
3
.
0
5
8
0
.
7
5
8
2
.
0
6
0
S
o
f
t
P
l
u
s
3
.
1
6
9
3
.
2
2
2
3
.
1
3
3
3
.
1
1
3
2
.
1
5
5
2
.
5
4
5
1
.
6
8
6
2
.
0
0
3
R
e
L
U
2
.
6
9
8
2
.
7
8
9
1
.
0
0
9
1
.
3
6
2
0
.
4
0
4
1
.
1
4
9
0
.
3
9
6
1
.
4
2
9
L
R
e
L
U
2
.
6
9
3
2
.
7
4
6
1
.
4
0
4
1
.
8
5
9
0
.
4
7
7
1
.
2
2
5
0
.
4
6
0
1
.
2
0
3
G
a
u
ssi
a
n
2
.
8
6
4
2
.
9
2
7
2
.
4
0
4
2
.
4
4
0
2
.
3
7
6
3
.
2
6
4
2
.
6
6
5
3
.
9
2
4
EL
U
3
.
1
3
3
3
.
1
5
0
3
.
1
1
8
3
.
0
3
6
2
.
9
8
9
3
.
0
3
2
2
.
4
2
0
2
.
5
2
1
S
E
L
U
2
.
8
5
5
2
.
9
0
6
3
.
0
8
1
3
.
1
7
1
0
.
7
5
5
1
.
4
2
9
0
.
6
3
2
1
.
4
6
9
S
w
i
sh
3
.
1
3
5
3
.
1
3
4
3
.
0
8
4
2
.
0
4
0
1
.
2
2
5
1
.
9
2
4
1
.
1
7
1
1
.
6
1
9
E
-
S
w
i
sh
3
.
1
2
6
3
.
1
3
8
2
.
9
2
9
3
.
0
0
4
1
.
3
0
0
1
.
7
8
8
1
.
1
0
2
1
.
4
6
4
T
ab
le
5
.
R
MSE
co
m
p
ar
is
o
n
o
f
ac
tiv
atio
n
f
u
n
c
tio
n
i
n
4
m
o
d
els f
o
r
ap
p
lian
ce
s
en
e
r
g
y
p
r
e
d
ic
tio
n
d
ataset
A
c
t
i
v
a
t
i
o
n
M
O
D
EL
I
M
O
D
EL
I
I
M
O
D
EL
I
I
I
M
O
D
EL
I
V
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
S
i
g
mo
i
d
9
3
.
4
7
9
3
.
7
1
9
3
.
4
6
9
3
.
7
0
8
1
.
6
4
8
5
.
1
0
7
3
.
9
9
8
1
.
8
2
T
a
n
h
9
3
.
4
6
9
3
.
6
9
8
6
.
9
2
8
8
.
8
1
5
9
.
6
0
7
6
.
5
2
5
4
.
7
5
7
4
.
6
0
S
o
f
t
S
i
g
n
8
9
.
2
7
9
0
.
1
1
8
1
.
6
6
8
5
.
2
0
5
0
.
5
4
7
3
.
4
3
4
4
.
3
6
7
2
.
7
1
S
o
f
t
P
l
u
s
9
3
.
4
3
9
3
.
6
7
9
2
.
2
0
9
2
.
6
0
6
9
.
9
2
8
1
.
4
3
6
3
.
8
1
7
8
.
6
3
R
e
L
U
8
4
.
7
7
8
6
.
8
2
7
7
.
4
4
8
2
.
4
2
3
7
.
7
7
7
2
.
8
6
3
6
.
5
6
7
2
.
2
6
L
R
e
L
U
8
4
.
5
7
8
6
.
6
2
7
4
.
9
3
8
1
.
6
6
4
3
.
4
7
7
2
.
2
7
3
9
.
2
7
7
1
.
3
1
G
a
u
ssi
a
n
8
5
.
3
7
8
7
.
2
0
7
8
.
7
1
8
3
.
0
9
5
7
.
4
1
7
7
.
0
8
7
0
.
9
7
8
0
.
8
8
EL
U
9
2
.
6
8
9
2
.
9
9
8
6
.
5
3
8
7
.
8
0
6
3
.
9
1
7
6
.
9
2
5
8
.
2
8
7
6
.
3
4
S
E
L
U
8
4
.
5
0
8
6
.
5
0
7
9
.
8
5
8
3
.
8
2
5
2
.
3
9
7
4
.
6
8
4
8
.
0
3
7
3
.
9
9
S
w
i
sh
9
1
.
1
0
9
1
.
6
7
8
8
.
0
1
8
8
.
9
5
5
7
.
5
3
7
5
.
6
9
5
0
.
2
2
7
3
.
8
2
E
-
S
w
i
sh
9
1
.
5
1
9
1
.
9
7
8
6
.
7
9
8
8
.
3
3
5
7
.
3
1
7
7
.
4
2
5
0
.
0
3
7
4
.
5
1
I
n
d
iv
id
u
a
l
h
o
u
s
e
h
o
ld
elec
tr
ic
p
o
w
er
co
n
s
u
m
p
tio
n
d
ataset
n
e
ed
s
lo
n
g
ti
m
e
to
e
x
ec
u
te
d
u
e
t
o
its
h
u
g
e
s
ize.
T
h
er
ef
o
r
e,
w
e
r
ed
u
ce
d
th
e
ep
o
ch
n
u
m
b
er
to
4
0
in
s
tead
o
f
2
0
0
.
I
n
ad
d
itio
n
,
th
e
n
u
m
b
er
o
f
i
m
p
le
m
en
ta
tio
n
s
is
r
ed
u
ce
d
to
th
r
ee
in
s
tead
o
f
n
i
n
e.
Mo
r
eo
v
er
,
w
e
j
u
s
t
te
s
ted
Mo
d
el
I
a
n
d
Mo
d
el
I
I
d
u
e
to
th
e
s
a
m
e
r
ea
s
o
n
.
T
ab
le
6
s
h
o
w
s
t
h
at
R
eL
U
A
cti
v
atio
n
f
u
n
ctio
n
o
u
tp
er
f
o
r
m
s
it
s
co
u
n
ter
p
ar
t
s.
T
ab
le
7
s
u
m
m
ar
izes
t
h
e
r
esu
l
ts
o
f
all
ex
p
er
m
in
ts
.
I
t
s
h
o
w
s
th
at
R
e
L
U
a
n
d
L
R
e
L
U
o
u
tp
er
f
o
r
m
t
h
e
o
th
er
ac
tiv
at
io
n
f
u
n
ctio
n
s
.
I
n
ad
d
itio
n
,
w
e
ca
n
o
b
s
er
v
e
th
at
R
e
L
U
lear
n
i
n
tr
ain
d
ata
b
etter
th
an
o
t
h
er
s
d
o
.
T
ab
le
6
.
R
MSE
co
m
p
ar
is
o
n
o
f
ac
tiv
atio
n
f
u
n
c
tio
n
i
n
4
m
o
d
els f
o
r
in
d
i
v
id
u
al
h
o
u
s
e
h
o
ld
elec
tr
ic
p
o
w
er
co
n
s
u
m
p
tio
n
d
ataset
A
c
t
i
v
a
t
i
o
n
M
O
D
EL
I
M
O
D
EL
I
I
T
ra
i
n
T
e
st
T
ra
i
n
T
e
st
S
i
g
mo
i
d
0
.
2
7
7
0
.
2
2
8
0
.
5
4
4
0
.
5
1
7
T
a
n
h
0
.
2
9
8
0
.
2
5
0
0
.
2
7
5
0
.
2
2
5
S
o
f
t
S
i
g
n
0
.
2
8
3
0
.
2
3
4
0
.
6
2
3
0
.
6
0
6
S
o
f
t
P
l
u
s
0
.
2
7
8
0
.
2
2
8
0
.
3
4
6
0
.
3
0
6
R
e
L
U
0
.
2
7
6
0
.
2
2
5
0
.
2
7
4
0
.
2
2
4
L
R
e
L
U
0
.
2
7
8
0
.
2
2
7
0
.
2
7
8
0
.
2
2
7
G
a
u
ssi
a
n
0
.
2
8
5
0
.
2
3
7
0
.
2
8
5
0
.
2
3
7
EL
U
0
.
2
7
9
0
.
2
3
0
0
.
2
7
9
0
.
2
3
1
S
E
L
U
0
.
2
8
7
0
.
2
3
9
0
.
2
8
5
0
.
2
3
6
S
w
i
sh
0
.
2
9
2
0
.
2
4
4
0
.
2
8
2
0
.
2
3
4
E
-
S
w
i
sh
0
.
3
2
1
0
.
2
7
8
0
.
3
3
4
0
.
2
9
6
T
ab
le
7
.
B
est p
er
f
o
r
m
a
n
ce
o
f
ac
tiv
atio
n
f
u
n
ctio
n
s
f
o
r
all
d
ata
s
ets
D
a
t
a
se
t
T
r
a
i
n
T
e
st
C
o
mb
i
n
e
d
c
y
c
l
e
p
o
w
e
r
p
l
a
n
t
R
e
L
U
R
e
L
U
En
e
r
g
y
e
f
f
i
c
i
e
n
c
y
(
H
e
a
t
i
n
g
L
o
a
d
)
R
e
L
U
R
e
L
U
En
e
r
g
y
e
f
f
i
c
i
e
n
c
y
(
C
o
o
l
i
n
g
L
o
a
d
)
R
e
L
U
R
e
L
U
A
p
p
l
i
a
n
c
e
s e
n
e
r
g
y
p
r
e
d
i
c
t
i
o
n
R
e
L
U
L
e
a
k
R
e
L
U
I
n
d
i
v
i
d
u
a
l
h
o
u
se
h
o
l
d
e
l
e
c
t
r
i
c
p
o
w
e
r
c
o
n
su
m
p
t
i
o
n
R
e
L
U
R
e
L
U
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.
11
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
:
16
3
-
1
7
0
170
5.
CO
NCLU
SI
O
N
P
r
ed
ictio
n
o
f
p
o
w
er
co
n
s
u
m
p
t
io
n
an
d
p
r
o
d
u
ctio
n
is
a
n
i
m
p
o
r
tan
t
s
tep
to
m
a
n
a
g
e
an
d
co
n
t
r
o
l
p
o
w
er
u
tili
t
ies.
T
h
is
p
r
ed
ictio
n
h
elp
s
th
e
u
tili
tie
s
to
f
ig
u
r
e
t
h
e
q
u
an
tit
y
o
f
co
n
s
u
m
ed
a
n
d
p
r
o
d
u
ce
d
elec
tr
icit
y
.
T
h
er
e
ar
e
m
an
y
m
o
d
els
d
es
ig
n
ed
f
o
r
th
at
p
u
r
p
o
s
e.
No
w
ad
a
y
s
,
ar
ti
f
icial
in
telli
g
en
ce
m
o
d
els
s
u
ch
as
n
e
u
r
al
n
et
w
o
r
k
s
(
Feed
f
o
r
w
o
r
d
n
e
u
r
al
n
et
w
o
r
k
)
p
r
o
v
ed
its
e
f
f
icien
c
y
.
B
u
ild
i
n
g
s
tr
o
n
g
m
o
d
el
f
o
r
p
r
ed
ictio
n
u
s
i
n
g
m
u
ltil
a
y
er
n
eu
r
al
n
et
w
o
r
k
n
ee
d
s
s
e
v
er
al
i
m
p
o
r
tan
t
t
h
in
g
s
o
n
e
o
f
th
e
m
i
s
ch
o
o
s
i
n
g
s
u
itab
le
a
ctiv
atio
n
f
u
n
ctio
n
.
W
e
co
m
p
ar
e
t
h
e
m
o
s
t
u
s
ed
ac
tiv
atio
n
f
u
n
ctio
n
s
o
n
d
ata
s
ets
r
elate
d
to
elec
tr
ical
p
o
wer
.
W
e
p
r
esen
t
t
h
at
R
eL
U
an
d
L
ea
k
R
e
L
U
o
u
tp
er
f
o
r
m
o
th
er
ac
ti
v
atio
n
f
u
n
ctio
n
.
R
eL
U
ac
tiv
at
io
n
ac
h
ie
v
es
1
3
o
v
er
1
8
o
f
tr
ain
s
et
an
d
8
o
v
er
1
8
o
f
test
in
g
s
ets
a
n
d
L
ea
k
R
e
L
U
ac
h
ie
v
es
3
o
v
e
r
1
8
in
tr
ain
in
g
s
et
s
an
d
8
o
v
e
r
1
8
in
test
in
g
s
et
s
.
R
eL
U
ac
h
ie
v
es
th
e
s
m
a
lles
t
er
r
o
r
in
tr
ain
in
g
s
e
t
in
al
l
d
atase
ts
.
R
e
L
U
i
n
co
m
b
i
n
ed
c
y
cle
p
o
w
er
p
lan
t,
e
n
er
g
y
ef
f
icien
c
y
(
Hea
ti
n
g
L
o
ad
)
an
d
en
er
g
y
e
f
f
icie
n
c
y
(
C
o
o
lin
g
L
o
ad
)
an
d
in
d
i
v
id
u
al
h
o
u
s
e
h
o
ld
elec
tr
ic
p
o
w
er
co
n
s
u
m
p
tio
n
d
atasets
p
er
f
o
r
m
s
t
h
e
m
in
i
m
u
m
R
M
SE
in
t
est
s
et.
I
n
t
h
e
o
th
er
h
a
n
d
,
L
ea
k
R
e
L
U
ac
h
iev
e
s
th
e
m
in
i
m
u
m
test
s
et
er
r
o
r
in
ap
p
lian
ce
s
en
er
g
y
p
r
ed
ictio
n
d
a
taset.
RE
F
E
R
E
NC
E
S
[1
]
P
.
Ch
u
jai,
N
.
Ke
rd
p
ra
so
p
,
a
n
d
K
.
Ke
rd
p
ra
so
p
,
“
T
i
m
e
S
e
ries
An
a
ly
sis
o
f
Ho
u
se
h
o
ld
El
e
c
tri
c
C
o
n
su
m
p
ti
o
n
w
it
h
A
RIM
A
a
n
d
A
RM
A
M
o
d
e
ls
,
”
Pro
c
e
e
d
in
g
s
o
f
th
e
I
n
ter
n
a
t
io
n
a
l
M
u
lt
iCo
n
fer
e
n
c
e
o
f
En
g
in
e
e
rs
a
n
d
C
o
mp
u
ter
S
c
ien
ti
sts,
v
o
l
.
1
,
p
p
.
2
9
5
-
3
0
0
,
2
0
1
3
.
[2
]
K.
G
.
S
h
e
e
la
a
n
d
S
.
N.
De
e
p
a
,
“
Re
v
ie
w
o
n
M
e
th
o
d
s
t
o
F
ix
Nu
m
b
e
r
o
f
Hid
d
e
n
Ne
u
ro
n
s
in
Ne
u
ra
l
Ne
tw
o
rk
s,”
M
a
th
e
ma
ti
c
a
l
Pro
b
lem
s in
E
n
g
in
e
e
rin
g
,
v
o
l.
2
0
1
3
,
p
p
.
1
-
1
1
,
2
0
1
3
.
[3
]
A
.
S
a
la
m
a
n
d
A
.
El
Hib
a
o
u
i,
“
Co
m
p
a
riso
n
o
f
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
rit
h
m
s
f
o
r
th
e
p
o
w
e
r
c
o
n
su
m
p
ti
o
n
p
re
d
icti
o
n
-
Ca
se
stu
d
y
o
f
teto
u
a
n
c
it
y
,
”
6
th
I
n
ter
n
a
ti
o
n
a
l
Ren
e
wa
b
le a
n
d
S
u
st
a
in
a
b
l
e
E
n
e
rg
y
Co
n
fer
e
n
c
e
,
p
p
.
1
-
5
,
2
0
1
8
.
[4
]
S
.
A
g
a
to
n
o
v
ic
-
Ku
strin
a
n
d
R.
B
e
re
s
f
o
rd
,
“
Ba
sic
c
o
n
c
e
p
ts
o
f
a
rti
f
icia
l
n
e
u
ra
l
n
e
tw
o
rk
(
A
NN
)
m
o
d
e
li
n
g
a
n
d
it
s
a
p
p
li
c
a
ti
o
n
i
n
p
h
a
rm
a
c
e
u
ti
c
a
l
re
se
a
rc
h
,
”
J
.
Ph
a
rm
.
Bi
o
me
d
.
An
a
l.
,
vo
l.
2
2
,
n
o
.
5
,
p
p
.
7
1
7
-
7
2
7
,
2
0
0
0
.
[5
]
B.
Din
g
,
H.
Qia
n
,
a
n
d
J.
Zh
o
u
,
“
A
c
ti
v
a
ti
o
n
f
u
n
c
ti
o
n
s
a
n
d
t
h
e
ir
c
h
a
ra
c
terist
ics
in
d
e
e
p
n
e
u
ra
l
n
e
tw
o
rk
s,”
Pro
c
e
e
d
in
g
s
o
f
t
h
e
3
0
t
h
Ch
i
n
e
se
Co
n
tro
l
a
n
d
De
c
isio
n
C
o
n
fer
e
n
c
e
,
p
p
.
1
8
3
6
-
1
8
4
1
,
2
0
1
8
.
[6
]
C.
Öz
k
a
n
a
n
d
F
.
S
u
n
a
r
Er
b
e
k
,
“
A
Co
m
p
a
riso
n
o
f
A
c
ti
v
a
ti
o
n
F
u
n
c
ti
o
n
s
f
o
r
M
u
l
ti
sp
e
c
tral
L
a
n
d
sa
t
T
M
I
m
a
g
e
Clas
sif
ic
a
ti
o
n
,
”
Ph
o
t
o
g
ra
mm
e
tric
En
g
in
e
e
rin
g
&
Rem
o
te S
e
n
sin
g
,
v
o
l.
6
9
,
n
o
.
1
1
,
p
p
.
1
2
2
5
-
1
2
3
4
,
2
0
0
3
.
[
7
]
A
.
A
.
M
i
n
a
i
a
n
d
R
.
D
.
W
i
l
l
i
a
m
s
,
“
O
n
t
h
e
d
e
r
i
v
a
t
i
v
e
s
o
f
t
h
e
s
i
g
m
o
i
d
,
”
N
e
u
r
a
l
N
e
t
w
o
r
k
s
,
v
o
l
.
6
,
n
o
.
6
,
p
p
.
8
4
5
-
8
5
3
,
1
9
9
3
.
[8
]
X
.
G
lo
ro
t
a
n
d
Y.
Be
n
g
io
,
“
U
n
d
e
rsta
n
d
in
g
th
e
d
if
f
icu
lt
y
o
f
train
in
g
d
e
e
p
f
e
e
d
f
o
rwa
rd
n
e
u
r
a
l
n
e
tw
o
rk
s,”
Pro
c
e
e
d
in
g
s
o
f
t
h
e
t
h
irtee
n
th
in
te
rn
a
ti
o
n
a
l
c
o
n
fer
e
n
c
e
o
n
a
rtif
ici
a
l
in
telli
g
e
n
c
e
a
n
d
st
a
ti
stics
,
p
p
.
2
4
9
-
2
5
6
,
2
0
1
0
.
[9
]
R.
H.
R.
Ha
h
n
l
o
se
r,
e
t
a
l.
,
“
Err
a
tu
m
:
Dig
it
a
l
se
lec
ti
o
n
a
n
d
a
n
a
l
o
g
u
e
a
m
p
li
f
ic
a
ti
o
n
c
o
e
x
ist
i
n
a
c
o
rtex
-
in
sp
ired
sili
c
o
n
c
ircu
it
,
”
Na
tu
re
,
v
o
l.
4
0
5
,
n
o
.
6
7
8
9
,
p
p
.
9
4
7
–
9
5
1
,
2
0
0
0
.
[1
0
]
B.
X
u
,
N.
W
a
n
g
,
T
.
Ch
e
n
,
a
n
d
M
.
L
i,
“
E
m
p
iri
c
a
l
Ev
a
lu
a
ti
o
n
o
f
Re
c
ti
f
ied
Ac
ti
v
a
ti
o
n
s
in
Co
n
v
o
lu
ti
o
n
a
l
Ne
tw
o
rk
,
”
a
rXiv p
re
p
ri
n
t
a
rX
iv:1
5
0
5
.
0
0
8
5
3
,
2
0
1
5
.
[1
1
]
V
.
Na
ir
a
n
d
G
.
E.
Hi
n
t
o
n
,
“
Re
c
ti
f
ied
L
in
e
a
r
Un
it
s Im
p
ro
v
e
Re
stricte
d
Bo
lt
z
m
a
n
n
M
a
c
h
in
e
s,
”
ICM
L
,
2
0
1
0
.
[1
2
]
S
.
L
e
e
a
n
d
R.
M
.
Kil,
“
A
Ga
u
ss
ia
n
p
o
ten
t
ial
f
u
n
c
ti
o
n
n
e
tw
o
rk
w
it
h
h
iera
rc
h
ica
ll
y
se
l
f
-
o
rg
a
n
izin
g
lea
rn
in
g
,
”
Ne
u
r
a
l
Ne
two
rk
s
,
v
o
l.
4
,
n
o
.
2
,
p
p
.
2
0
7
-
2
2
4
,
1
9
9
1
.
[1
3
]
D.
-
A
.
Cle
v
e
rt,
T
.
Un
terth
in
e
r,
a
n
d
S
.
Ho
c
h
re
it
e
r,
“
F
a
st
a
n
d
a
c
c
u
ra
te
d
e
e
p
n
e
tw
o
rk
le
a
rn
in
g
b
y
e
x
p
o
n
e
n
ti
a
l
li
n
e
a
r
u
n
it
s (
EL
US),
”
a
rXiv p
re
p
ri
n
t
a
rXiv:1
5
1
1
.
0
7
2
8
9
,
2
0
1
5
.
[1
4
]
G
.
Kla
m
b
a
u
e
r,
T
.
Un
terth
in
e
r,
A
.
M
a
y
r,
a
n
d
S
.
Ho
c
h
re
it
e
r,
“
S
e
lf
-
No
r
m
a
li
z
in
g
N
e
u
ra
l
Ne
t
w
o
rk
s,”
Ad
v
a
n
c
e
s
i
n
n
e
u
ra
l
in
f
o
rm
a
ti
o
n
p
ro
c
e
ss
in
g
sy
ste
ms
,
p
p
.
9
7
1
-
9
8
0
,
2
0
1
7
.
[
1
5
]
P
.
R
a
m
a
c
h
a
n
d
r
a
n
,
B
.
Z
o
p
h
,
a
n
d
Q
.
V
.
L
e
,
“
S
e
a
r
c
h
i
n
g
f
o
r
A
c
t
i
v
a
t
i
o
n
F
u
n
c
t
i
o
n
s
,
”
a
r
X
i
v
p
r
e
p
r
i
n
t
a
r
X
i
v
:
1
7
1
0
.
0
5
9
4
1
,
2017.
[1
6
]
E.
A
lc
a
id
e
a
n
d
Eri
c
,
“
E
-
s
w
i
sh
:
A
d
ju
stin
g
Ac
ti
v
a
ti
o
n
s
to
Diffe
re
n
t
Ne
t
w
o
rk
De
p
th
s,
”
a
rXiv
p
re
p
rin
t
a
rXiv:1
8
0
1
.
0
7
1
4
5
,
2
0
1
8
.
[1
7
]
L
.
X
.
Niu
a
n
d
X
.
J.
L
iu
,
“
M
u
lt
iv
a
riab
le
g
e
n
e
ra
li
z
e
d
p
re
d
ictiv
e
s
c
h
e
m
e
f
o
r
g
a
s
tu
rb
in
e
c
o
n
tro
l
i
n
c
o
m
b
in
e
d
c
y
c
le
p
o
w
e
r
p
lan
t,
”
IEE
E
Co
n
fer
e
n
c
e
o
n
Cy
b
e
rn
e
ti
c
s a
n
d
In
telli
g
e
n
t
S
y
st
e
ms
,
p
p
.
7
9
1
-
7
9
6
,
2
0
0
8
.
[1
8
]
A
.
T
sa
n
a
s
a
n
d
A
.
X
if
a
ra
,
“
Ac
c
u
ra
te
q
u
a
n
ti
tativ
e
e
stim
a
ti
o
n
o
f
e
n
e
rg
y
p
e
r
f
o
r
m
a
n
c
e
o
f
re
sid
e
n
ti
a
l
b
u
il
d
in
g
s
u
si
n
g
sta
ti
stica
l
m
a
c
h
in
e
lea
rn
in
g
to
o
ls,
”
En
e
rg
y
Bu
i
ld
.
,
v
o
l.
4
9
,
p
p
.
5
6
0
-
5
6
7
,
2
0
1
2
.
[1
9
]
L
.
M
.
Ca
n
d
a
n
e
d
o
,
V
.
F
e
ld
h
e
im
,
a
n
d
D.
De
ra
m
a
ix
,
“
Da
ta
d
riv
e
n
p
re
d
ictio
n
m
o
d
e
ls
o
f
e
n
e
rg
y
u
se
o
f
a
p
p
li
a
n
c
e
s
in
a
lo
w
-
e
n
e
rg
y
h
o
u
se
,
”
En
e
rg
y
Bu
i
ld
.
,
v
o
l.
1
4
0
,
p
p
.
8
1
-
9
7
,
2
0
1
7
.
[2
0
]
N.
F
a
ra
g
A
b
e
d
a
n
d
M
.
M
M
il
o
s
a
v
lj
e
v
ic,
“
S
in
g
le
Ho
m
e
El
e
c
tri
c
it
y
P
o
w
e
r
Co
n
su
m
p
ti
o
n
F
o
re
c
a
st
Us
in
g
Ne
u
ra
l
Ne
tw
o
rk
s M
o
d
e
l,
”
IJ
IS
E
T
-
In
ter
n
a
ti
o
n
a
l
J
.
In
n
o
v
.
S
c
i.
En
g
.
T
e
c
h
n
o
l.
,
v
o
l.
3
,
n
o
.
1
,
p
p
.
1
0
0
-
1
0
6
,
2
0
1
6
.
[2
1
]
T
.
Zh
a
n
g
a
n
d
T
o
n
g
,
“
S
o
lv
in
g
larg
e
sc
a
l
e
li
n
e
a
r
p
re
d
ictio
n
p
r
o
b
lem
s
u
sin
g
sto
c
h
a
stic
g
ra
d
ien
t
d
e
sc
e
n
t
a
lg
o
rit
h
m
s,”
T
we
n
ty
-
fi
rs
t
in
ter
n
a
ti
o
n
a
l
c
o
n
fer
e
n
c
e
o
n
M
a
c
h
in
e
le
a
rn
i
n
g
-
ICM
L
’0
4
,
2
0
0
4
.
[2
2
]
D.
P
.
Ki
n
g
m
a
a
n
d
J.
Ba
,
“
A
d
a
m
:
A
M
e
th
o
d
f
o
r
S
t
o
c
h
a
stic Op
ti
m
iza
ti
o
n
,
”
a
rXiv p
re
p
ri
n
t
a
rXiv:1
4
1
2
.
6
9
8
0
,
2
0
1
4
.
[2
3
]
T
.
Ch
a
i,
T
.
Ch
a
i,
a
n
d
R.
R.
Dra
x
ler,
“
Oz
o
n
e
h
e
a
lt
h
a
n
d
e
c
o
sy
ste
m
i
m
p
a
c
ts
V
ie
w
p
ro
jec
t
Ro
o
t
m
e
a
n
sq
u
a
re
e
rro
r
(RM
S
E)
o
r
m
e
a
n
a
b
so
lu
te
e
rro
r
(
M
A
E)?
-
Ar
g
u
m
e
n
ts
a
g
a
in
st
a
v
o
id
in
g
RM
S
E
i
n
t
h
e
li
tera
tu
re
,
”
Ge
o
sc
i.
M
o
d
e
l
De
v
,
v
o
l.
7
,
p
p
.
1
2
4
7
-
1
2
5
0
,
2
0
1
4
.
[2
4
]
A
.
F
a
m
il
i,
W
.
M
.
S
h
e
n
,
R.
W
e
b
e
r
,
a
n
d
E.
S
im
o
u
d
is,
“
Da
ta p
re
p
r
o
c
e
ss
in
g
a
n
d
in
telli
g
e
n
t
d
a
ta an
a
ly
sis,”
In
tell.
D
a
ta
An
a
l
.
,
v
o
l.
1
,
n
o
.
1
,
p
p
.
3
–
2
3
,
Ja
n
.
1
9
9
7
.
[2
5
]
.
T
.
Ja
y
a
la
k
sh
m
i
a
n
d
A
.
S
a
n
th
a
k
u
m
a
ra
n
,
“
S
tatisti
c
a
l
No
rm
a
li
z
a
ti
o
n
a
n
d
Ba
c
k
P
r
o
p
a
g
a
ti
o
n
f
o
r
Cl
a
ss
if
ic
a
ti
o
n
,
”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
C
o
mp
u
ter
T
h
e
o
ry
a
n
d
En
g
in
e
e
rin
g
,
v
o
l
.
3
,
n
o
.
1
,
p
p
.
1
7
9
3
-
8
2
0
1
,
2
0
1
1
Evaluation Warning : The document was created with Spire.PDF for Python.