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
.
1
,
Feb
r
u
ar
y
2020
,
p
p
.
538
~
548
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
1
0
i
1
.
pp
5
3
8
-
548
538
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
The ups
urg
e of
d
eep learning
for c
o
m
put
er vis
io
n a
pplica
tions
P
riy
a
nk
a
P
a
t
el
1
,
A
m
it
T
ha
kk
a
r
2
1
U &
P
U
P
a
tel
De
p
a
rtm
e
n
t
o
f
Co
m
p
u
ter E
n
g
in
e
e
rin
g
,
Ch
a
n
d
u
b
h
a
i
S
P
a
tel
In
sti
tu
te
o
f
T
e
c
h
n
o
lo
g
y
,
F
a
c
u
lt
y
o
f
T
e
c
h
n
o
lo
g
y
&
En
g
in
e
e
rin
g
,
CHA
RUS
AT
Un
iv
e
rsit
y
,
I
n
d
ia
2
S
m
t.
Ku
n
d
a
n
b
e
n
Din
s
h
a
P
a
tel
De
p
a
rtm
e
n
t
o
f
In
f
o
r
m
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
,
Ch
a
n
d
u
b
h
a
i
S
P
a
tel
In
stit
u
te
o
f
T
e
c
h
n
o
lo
g
y
,
F
a
c
u
lt
y
o
f
T
e
c
h
n
o
lo
g
y
&
En
g
in
e
e
rin
g
,
CHA
RUS
AT
Un
iv
e
rsit
y
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Feb
26
,
2
0
1
9
R
ev
i
s
ed
A
u
g
5
,
20
19
A
cc
ep
ted
A
u
g
29
,
2
0
19
A
rti
f
icia
l
in
telli
g
e
n
c
e
(
A
I)
is
a
d
d
it
io
n
a
ll
y
se
r
v
in
g
to
a
b
ra
n
d
n
e
w
b
re
e
d
o
f
c
o
rp
o
ra
ti
o
n
s
d
isru
p
t
in
d
u
stries
f
ro
m
re
sto
ra
ti
v
e
e
x
a
m
in
a
ti
o
n
to
h
o
rti
c
u
lt
u
re
.
Co
m
p
u
ters
c
a
n
’t
n
e
v
e
rth
e
les
s
r
e
p
lac
e
h
u
m
a
n
s,
h
o
w
e
v
e
r,
th
e
y
w
il
l
w
o
rk
su
p
e
rb
ly
tak
in
g
c
a
re
o
f
th
e
e
v
e
r
y
d
a
y
tan
g
le
o
f
o
u
r
li
v
e
s.
T
h
e
e
ra
is
re
c
o
n
stru
c
ti
n
g
b
ig
b
u
si
n
e
ss
a
n
d
h
a
s
b
e
e
n
o
n
th
e
rise
in
re
c
e
n
t
y
e
a
r
s
w
h
ich
h
a
s
g
ro
u
n
d
e
d
w
it
h
t
h
e
su
c
c
e
ss
o
f
d
e
e
p
lea
rn
in
g
(DL
).
Cy
b
e
r
-
se
c
u
rit
y
,
A
u
to
a
n
d
h
e
a
lt
h
i
n
d
u
stry
a
re
th
re
e
in
d
u
stri
e
s
in
n
o
v
a
ti
n
g
w
it
h
A
I
a
n
d
DL
te
c
h
n
o
l
o
g
ies
a
n
d
a
lso
Ba
n
k
in
g
,
re
tail
,
f
in
a
n
c
e
,
ro
b
o
ti
c
s,
m
a
n
u
f
a
c
tu
rin
g
.
T
h
e
h
e
a
lt
h
c
a
re
in
d
u
stry
is
o
n
e
o
f
th
e
e
a
rli
e
st
a
d
o
p
ters
o
f
A
I
a
n
d
DL
.
D
L
a
c
c
o
m
p
li
sh
in
g
e
x
c
e
p
ti
o
n
a
l
d
im
e
n
sio
n
s
lev
e
ls
o
f
a
c
c
u
ra
ten
e
ss
to
th
e
p
o
in
t
w
h
e
re
D
L
a
lg
o
rit
h
m
s
c
a
n
o
u
tp
e
rf
o
rm
h
u
m
a
n
s
a
t
c
la
ss
i
fy
in
g
v
id
e
o
s
&
i
m
a
g
e
s.
T
h
e
m
a
jo
r
d
riv
e
rs
th
a
t
c
a
u
se
d
th
e
b
re
a
k
th
ro
u
g
h
o
f
d
e
e
p
n
e
u
ra
l
n
e
tw
o
r
k
s
a
r
e
th
e
p
ro
v
isio
n
o
f
g
ian
t
a
m
o
u
n
ts
o
f
c
o
a
c
h
in
g
in
f
o
rm
a
ti
o
n
,
p
o
w
e
r
fu
l
m
a
c
h
in
e
in
f
ra
stru
c
tu
re
,
a
n
d
a
d
v
a
n
c
e
s
in
a
c
a
d
e
m
ia.
D
L
is
h
e
a
v
il
y
e
m
p
lo
y
e
d
in
e
a
c
h
a
c
a
d
e
m
e
to
re
v
ie
w
in
telli
g
e
n
c
e
a
n
d
w
it
h
in
t
h
e
trad
e
-
in
b
u
il
d
i
n
g
in
telli
g
e
n
t
s
y
ste
m
s
to
h
e
l
p
h
u
m
a
n
s
in
v
a
ried
tas
k
s.
T
h
e
re
b
y
D
L
s
y
ste
m
s
b
e
g
i
n
to
c
ru
sh
n
o
t
so
lely
c
las
sic
a
l
w
a
y
s,
b
u
t
a
d
d
it
io
n
a
ll
y
,
h
u
m
a
n
b
e
n
c
h
m
a
rk
s
in
n
u
m
e
ro
u
s
tas
k
s
li
k
e
i
m
a
g
e
c
las
sif
i
c
a
ti
o
n
,
a
c
ti
o
n
d
e
tec
ti
o
n
,
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
,
sig
n
a
l
p
ro
c
e
ss
,
a
n
d
li
n
g
u
isti
c
c
o
m
m
u
n
ica
ti
o
n
p
r
o
c
e
ss
.
K
ey
w
o
r
d
s
:
A
l
g
o
r
ith
m
s
A
p
p
licatio
n
s
A
r
ti
f
icial
i
n
tel
lig
e
n
ce
Dee
p
l
ea
r
n
in
g
His
to
r
y
Ma
ch
i
n
e
l
ea
r
n
i
n
g
Mo
d
els
Neu
r
al
n
et
w
o
r
k
Neu
r
al
n
et
w
o
r
k
r
u
le
s
R
ev
o
l
u
tio
n
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
:
P
r
iy
a
n
k
a
P
atel,
U
&
P
U
P
atel
Dep
ar
tm
e
n
t o
f
C
o
m
p
u
ter
E
n
g
in
ee
r
i
n
g
,
C
h
a
n
d
u
b
h
ai
S P
atel
I
n
s
tit
u
te
o
f
T
ec
h
n
o
lo
g
y
,
C
H
AR
US
A
T
Un
i
v
er
s
it
y
,
C
h
a
n
g
a,
3
8
8
4
2
1
,
Gu
j
ar
at,
I
n
d
ia
.
E
m
ail:
p
r
i
y
a
n
k
ap
atel.
it@
c
h
ar
u
s
at.
ac
.
i
n
1.
I
NT
RO
D
UCT
I
O
N
An
i
n
ter
esti
n
g
asp
ec
t
r
e
g
ar
d
in
g
n
e
u
r
al
s
y
s
te
m
s
i
s
to
w
h
at
e
x
ten
t
t
h
e
y
h
a
v
e
ta
k
en
to
b
e
an
o
v
er
n
i
g
h
t
ac
h
iev
e
m
e
n
t.
An
d
r
e
y
Ku
r
e
n
k
o
v
e
ill
u
s
tr
ate
a
v
er
y
n
ice
h
i
s
t
o
r
y
c
h
ar
t
i
n
Fi
g
u
r
e
1
.
Hi
s
to
r
y
r
et
u
r
n
s
t
h
e
w
h
o
l
e
d
is
tan
ce
to
t
h
e
1
9
4
0
s
.
Dee
p
l
ea
r
n
in
g
h
as
e
x
tr
e
m
el
y
j
u
s
t
ta
k
en
o
f
f
i
n
t
h
e
las
t
f
iv
e
y
ea
r
s
f
r
o
m
t
h
e
j
o
u
r
n
e
y
o
f
Neu
r
al
Net
w
o
r
k
.
T
h
e
r
ea
s
o
n
is
t
h
e
ex
p
a
n
d
ed
ac
ce
s
s
ib
ili
t
y
o
f
n
a
m
e
i
n
f
o
r
m
atio
n
alo
n
g
s
id
e
th
e
s
ig
n
i
f
ica
n
tl
y
ex
p
an
d
ed
co
m
p
u
tatio
n
al
t
h
r
o
u
g
h
p
u
t o
f
c
u
r
r
e
n
t p
r
o
ce
s
s
o
r
s
.
Dee
p
lear
n
in
g
al
g
o
r
ith
m
s
ar
e
co
m
p
o
s
ed
o
f
alg
o
r
ith
m
s
w
h
i
ch
allo
w
s
s
o
f
t
w
ar
e
to
in
s
tr
u
ct
an
d
tr
ain
its
el
f
to
p
er
f
o
r
m
tas
k
s
ad
ep
tl
y
w
ith
t
h
e
h
u
g
e
d
ataset
o
f
co
n
t
en
t
o
r
te
x
t,
i
m
ag
e
s
,
s
o
u
n
d
s
,
s
p
ee
ch
es,
v
id
eo
s
o
r
ti
m
e
s
er
ies
f
r
a
m
es
b
y
i
n
tr
o
d
u
cin
g
w
it
h
m
u
l
ti
-
la
y
er
ed
n
eu
r
al
n
et
w
o
r
k
s
w
it
h
h
u
g
e
a
m
o
u
n
t
s
o
f
d
ata.
Dee
p
lear
n
in
g
(
D
L
)
is
al
s
o
k
n
o
w
n
l
i
k
e
it i
s
th
e
s
u
b
s
et
o
f
m
ac
h
in
e
l
ea
r
n
in
g
(
M
L
)
m
ad
e
o
u
t o
f
ca
l
cu
latio
n
s
t
h
at
allo
w
p
r
o
g
r
am
m
i
n
g
to
p
r
ep
ar
e
its
elf
to
p
er
f
o
r
m
e
v
er
y
d
a
y
j
o
b
s
,
s
i
m
ilar
to
d
is
co
u
r
s
e
a
n
d
p
ictu
r
e
ac
k
n
o
w
led
g
m
e
n
t,
b
y
u
n
co
v
er
in
g
m
u
lti
-
la
y
er
ed
n
eu
r
al
s
y
s
te
m
s
to
tr
e
m
en
d
o
u
s
m
ea
s
u
r
es o
f
i
n
f
o
r
m
atio
n
.
A
r
ti
f
icial
I
n
tel
lig
e
n
ce
(
A
I
)
c
o
u
ld
b
e
a
m
ac
h
i
n
e
t
h
at
s
o
l
v
es
y
o
u
r
r
ea
l
-
li
f
e
p
r
o
b
le
m
s
w
h
ile
n
o
t
o
b
tain
in
g
a
n
y
h
u
m
a
n
s
co
n
ce
r
n
ed
in
i
t,
o
r
to
b
e
s
p
ec
i
f
ic
w
e
ar
e
ab
le
to
s
a
y
t
h
at
a
p
ac
k
a
g
e
th
at
i
s
s
o
r
t
o
f
a
litt
le
k
id
t
h
at
w
a
n
ts
s
o
m
e
i
n
s
tr
u
ctio
n
a
n
d
tr
ain
in
g
a
n
d
a
f
ter
w
ar
d
it's
ca
p
ab
le
o
f
r
eso
l
u
tio
n
y
o
u
r
r
ea
l
-
li
f
e
i
s
s
u
es.
An
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
Th
e
u
p
s
u
r
g
e
o
f d
ee
p
lea
r
n
in
g
fo
r
co
mp
u
ter visi
o
n
a
p
p
lica
tio
n
s
(
P
r
iya
n
ka
P
a
tel)
539
ass
es
s
m
en
t
o
f
th
e
i
m
p
ac
t
o
f
A
I
r
eq
u
ir
es
a
co
m
p
r
eh
e
n
s
io
n
o
f
b
o
th
t
h
e
p
r
ac
tice
o
f
in
n
o
v
a
tio
n
tec
h
n
o
lo
g
y
an
d
la
w
.
W
h
ile
t
h
e
r
ef
er
en
ce
d
ex
a
m
i
n
atio
n
s
ass
e
s
s
ed
ex
er
cise
s
o
f
atto
r
n
e
y
s
an
d
cu
r
r
en
t
i
n
n
o
v
atio
n
s
,
t
h
e
y
d
o
n
'
t
ad
eq
u
atel
y
co
n
n
ec
t
w
ith
f
u
tu
r
e
co
n
d
itio
n
s
o
f
in
n
o
v
at
io
n
an
d
if
y
o
u
w
o
u
ld
li
k
e
a
p
ar
ticu
lar
d
ef
in
itio
n
b
y
A
la
n
T
u
r
in
g
,
A
I
d
e
f
in
ed
as
"
th
e
s
c
ien
ce
o
f
m
a
k
i
n
g
co
m
p
u
ter
s
d
o
th
in
g
s
t
h
at
r
eq
u
ir
e
in
tel
lig
e
n
ce
w
h
e
n
d
o
n
e
b
y
h
u
m
a
n
s
.
No
w
,
A
r
ti
f
icial
I
n
tel
lig
e
n
ce
co
u
ld
b
e
a
b
ig
tr
ee
th
at
h
a
s
s
e
v
er
al
b
r
an
ch
e
s
to
r
ev
ie
w
an
d
s
p
ec
ial
ize
w
it
h
it"
[
1
,
2]
.
Fig
u
r
e
1
.
B
r
ief
h
is
to
r
y
o
f
n
e
u
r
al
n
et
w
o
r
k
a
n
d
d
ee
p
lea
r
n
in
g
Fig
u
r
e
2
is
b
r
ief
o
f
A
r
ti
f
icia
l
I
n
telli
g
e
n
ce
tr
ee
an
d
d
if
f
er
e
n
ce
b
et
w
ee
n
ar
ti
f
icial
i
n
telli
g
e
n
ce
,
Ma
ch
in
e
L
ea
r
n
i
n
g
,
Neu
r
al
Net
w
o
r
k
,
Na
tu
r
al
L
a
n
g
u
a
g
e
P
r
o
ce
s
s
in
g
,
an
d
Dee
p
L
ea
r
n
in
g
:
AI:
B
u
ild
in
g
s
y
s
te
m
s
B
u
ild
in
g
f
r
a
m
e
w
o
r
k
s
t
h
at
ca
n
d
o
s
a
v
v
y
th
i
n
g
s
i
m
p
lie
s
P
C
s
w
it
h
t
h
e
ca
p
ac
it
y
to
r
ea
s
o
n
lik
e
h
u
m
a
n
s
.
M
L
:
I
t
is
lik
e
w
i
s
e
a
s
u
b
ca
te
g
o
r
y
o
f
A
r
ti
f
icial
I
n
telli
g
e
n
c
e.
Stru
ct
u
r
e
a
f
r
a
m
e
w
o
r
k
s
wh
ich
w
ill
g
ai
n
a
s
a
m
atter
o
f
f
ac
t i
m
p
l
ies
m
ac
h
in
e
w
it
h
t
h
e
ca
p
ac
it
y
to
lear
n
w
it
h
o
u
t b
ein
g
u
n
eq
u
iv
o
ca
ll
y
c
u
s
t
o
m
ized
.
NL
P
:
I
t
i
s
a
s
u
b
ca
teg
o
r
y
o
f
Ar
tif
icial
I
n
tel
lig
e
n
ce
.
St
r
u
ct
u
r
e
a
f
r
a
m
e
w
o
r
k
s
th
at
ca
n
co
m
p
r
eh
en
d
lan
g
u
ag
e.
I
t
is
a
s
u
b
ca
te
g
o
r
y
o
f
A
I
.
DL
:
I
t
is
a
s
u
b
ca
te
g
o
r
y
o
f
M
ac
h
in
e
L
ea
r
n
i
n
g
.
Stru
ct
u
r
e
f
r
a
m
e
w
o
r
k
s
th
a
t
u
t
ilizatio
n
DN
N
o
n
a
s
u
b
s
ta
n
tial
ar
r
an
g
e
m
en
t
o
f
in
f
o
r
m
atio
n
im
p
lies
co
m
p
u
ter
s
w
it
h
th
e
ab
ilit
y
to
lear
n
b
y
u
s
i
n
g
ar
ti
f
ici
al
n
eu
r
al
n
et
w
o
r
k
s
,
w
h
ic
h
w
er
e
r
o
u
s
ed
b
y
th
e
m
o
d
el
an
d
ca
p
ac
ity
o
f
th
e
m
o
d
el
ce
r
eb
r
u
m
as s
h
w
o
n
i
n
Fi
g
u
r
e
2
.
NN:
A
b
io
lo
g
icall
y
en
liv
e
n
e
d
n
et
w
o
r
k
o
f
A
r
ti
f
icial
Ne
u
r
o
n
s
.
A
r
ti
f
icial
I
n
telli
g
e
n
ce
is
en
co
m
p
as
s
ed
w
it
h
n
u
m
er
o
u
s
d
i
v
er
s
e
s
u
b
-
s
p
ec
ial
t
ies.
E
x
ten
s
i
v
el
y
,
th
e
y
ca
n
b
e
g
ath
er
ed
:
R
ea
s
o
n
i
n
g
to
o
ls
-
to
g
en
er
ate
co
n
cl
u
s
io
n
f
r
o
m
a
v
ailab
le
d
ata
o
r
k
n
o
w
l
ed
g
e
lik
e
p
ict
u
r
e,
co
n
ten
t,
v
i
d
eo
s
.
Un
s
u
p
er
v
is
ed
d
ee
p
lear
n
in
g
to
o
ls
-
cr
ea
te
g
en
er
al
s
y
s
te
m
s
t
h
at
ca
n
b
e
t
r
ain
ed
w
it
h
ti
n
y
a
m
o
u
n
t
o
f
d
ata.
T
h
e
m
ai
n
o
b
j
ec
tiv
e
o
f
th
is
k
in
d
o
f
lear
n
in
g
r
esear
ch
is
to
p
r
e
-
tr
ain
a
m
o
d
el
lik
e
d
is
cr
i
m
i
n
ato
r
o
r
en
co
d
er
n
et
w
o
r
k
to
b
e
u
s
ed
f
o
r
o
th
er
task
s
.
Su
p
er
v
is
ed
to
o
ls
-
T
o
u
n
d
er
s
tan
d
co
m
p
r
e
h
e
n
d
i
m
p
o
r
tan
ce
a
n
d
s
etti
n
g
t
h
r
o
u
g
h
clas
s
i
f
icatio
n
a
n
d
r
eg
r
es
s
io
n
.
Nee
d
to
g
u
id
e
to
t
ea
ch
th
e
al
g
o
r
ith
m
w
h
at
co
n
clu
s
io
n
s
it
s
h
o
u
ld
co
m
e
u
p
w
it
h
.
L
ea
r
n
i
n
g
to
o
ls
-
T
o
p
u
t
o
n
a
n
d
b
r
o
ad
en
u
n
d
er
s
ta
n
d
in
g
th
r
o
u
g
h
m
ac
h
i
n
e
lear
n
in
g
a
n
d
p
r
ed
ictiv
e
a
n
al
y
s
i
s
.
Op
ti
m
izatio
n
to
o
ls
-
T
o
p
u
t
o
n
in
f
o
r
m
a
tio
n
to
ex
p
r
ess
in
q
u
ir
ies t
h
r
o
u
g
h
e
x
p
er
t f
r
a
m
e
w
o
r
k
s
.
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
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
538
-
548
540
Fig
u
r
e
2
.
T
y
p
e
o
f
lear
n
in
g
an
d
its
ap
p
licatio
n
s
s
tr
u
ctu
r
e
s
y
s
te
m
s
t
h
at
u
s
e
d
n
n
o
n
a
lar
g
e
s
e
t o
f
d
ata
[
3
,
4
]
2.
M
ACH
I
NE
L
E
AR
NIN
G
AP
P
RO
ACH
ML
is
a
k
i
n
d
o
f
ar
tif
icial
b
r
ain
p
o
w
er
w
h
ic
h
allo
w
s
s
o
f
t
w
ar
e
ap
p
licatio
n
s
to
co
n
cl
u
s
io
n
w
it
h
p
r
o
g
r
ess
iv
el
y
p
r
ec
is
e
i
n
ex
p
e
ctin
g
r
esu
lts
w
i
th
o
u
t
b
ein
g
e
x
p
r
ess
l
y
r
ev
i
s
ed
.
T
h
e
ess
en
t
i
al
s
tar
t
o
f
M
L
i
s
to
co
n
s
tr
u
ct
ca
lcu
latio
n
s
w
h
ic
h
w
il
l
g
e
t
i
n
p
u
t
d
ata
a
n
d
u
ti
li
ze
m
ea
s
u
r
ab
le
ex
a
m
i
n
atio
n
t
o
p
r
ed
ict
ass
o
ciate
o
u
tp
u
t
v
al
u
e
w
it
h
in
an
ad
eq
u
ate
r
an
g
e.
Fi
g
u
r
e
3
s
h
o
w
s
t
h
e
s
tr
ai
g
h
t
f
o
r
w
ar
d
p
r
o
ce
s
s
o
f
m
ac
h
in
e
lear
n
in
g
.
I
n
m
ac
h
in
e
lear
n
i
n
g
f
ea
t
u
r
es
ar
e
m
a
n
u
all
y
e
x
tr
ac
ted
an
d
t
h
e
m
o
d
el
cr
ea
tes
a
f
ter
e
x
tr
ac
tio
n
o
f
f
ea
t
u
r
es.
T
h
ese
ap
p
r
o
ac
h
f
o
llo
w
s
s
h
allo
w
n
e
t
w
o
r
k
a
s
s
h
o
w
n
in
Fig
u
r
e
4
.
W
e
ca
n
i
m
p
r
o
v
e
n
et
w
o
r
k
t
h
r
o
u
g
h
p
r
o
v
id
in
g
m
o
r
e
ex
a
m
p
le
s
an
d
m
o
r
e
tr
ain
i
n
g
d
ata
to
th
e
n
et
w
o
r
k
.
I
n
d
ee
p
lear
n
in
g
,
t
h
e
f
ea
t
u
r
es
f
r
o
m
th
e
d
ataset
ar
e
au
to
m
at
icall
y
e
x
tr
ac
ted
.
T
h
e
P
er
f
o
r
m
s
o
f
th
e
m
o
d
el
is
―
e
n
d
-
to
-
e
n
d
lear
n
i
n
g
‖
m
ea
n
s
d
ee
p
n
et
w
o
r
k
,
n
o
t
li
k
e
a
s
h
allo
w
n
e
t
w
o
r
k
.
W
e
ca
n
i
m
p
r
o
v
e
th
e
n
et
w
o
r
k
b
y
p
r
o
v
id
in
g
h
u
g
er
d
ataset
s
o
,
a
d
ee
p
lea
r
n
in
g
alg
o
r
i
t
h
m
i
s
a
s
ca
le
w
it
h
d
ata.
Fig
u
r
e
5
is
th
e
ex
a
m
p
le
o
f
D
L
.
Fig
u
r
e
3
.
B
asic
m
ac
h
i
n
e
lear
n
i
n
g
p
r
o
ce
s
s
f
lo
w
Fig
u
r
e
4
.
Ma
ch
in
e
lear
n
i
n
g
ap
p
r
o
ac
h
to
h
an
d
cr
af
t f
ea
t
u
r
es
an
d
class
i
f
icat
io
n
Fig
u
r
e
5
.
Dee
p
lear
n
in
g
ap
p
r
o
ac
h
to
au
to
ex
tr
ac
ti
n
g
f
ea
t
u
r
es
an
d
class
i
f
icatio
n
[
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
Th
e
u
p
s
u
r
g
e
o
f d
ee
p
lea
r
n
in
g
fo
r
co
mp
u
ter visi
o
n
a
p
p
lica
tio
n
s
(
P
r
iya
n
ka
P
a
tel)
541
3.
NE
URA
L
N
E
T
WO
RK
S (
NN)
AND
DE
E
P
L
E
ARNI
NG
NN
an
d
D
L
p
r
esen
tl
y
g
i
v
e
t
h
e
m
o
s
t
e
f
f
ec
ti
v
e
s
o
l
u
tio
n
s
t
o
s
ev
er
al
i
s
s
u
es
in
i
m
ag
e
r
ec
o
g
n
itio
n
,
s
p
ee
ch
r
ec
o
g
n
it
io
n
,
tex
t
r
ec
o
g
n
i
tio
n
,
ti
m
e
s
er
ies
p
r
o
b
lem
s
,
v
id
eo
p
r
o
ce
s
s
in
g
,
an
d
n
at
u
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
.
Dee
p
lear
n
in
g
is
n
o
t
h
i
n
g
b
u
t
an
alg
o
r
it
h
m
w
h
ic
h
is
w
a
y
s
m
ar
ter
th
a
n
s
h
allo
w
m
ac
h
i
n
e
lear
n
in
g
al
g
o
r
ith
m
s
b
u
t
it
is
n
o
t
an
ea
s
y
tas
k
.
Dee
p
n
eu
r
al
n
et
w
o
r
k
s
ar
e
h
ar
d
er
to
tr
ain
an
d
r
eq
u
ir
ed
g
r
ap
h
ical
p
r
o
ce
s
s
in
g
s
u
p
p
o
r
t
f
o
r
b
etter
an
d
f
ast
r
esu
lts
.
No
w
ad
a
y
s
d
ee
p
lear
n
in
g
ap
p
licatio
n
s
ar
e
u
s
ed
ev
er
y
w
h
er
e
m
a
n
y
e
x
a
m
p
les
o
f
d
ee
p
lear
n
in
g
ar
e
a
v
ailab
le
i
n
c
u
r
r
en
t
ti
m
e
s
s
u
ch
as
Go
o
g
le
’
s
s
el
f
-
d
r
iv
e
n
ca
r
,
ap
p
les
f
ac
e
r
ec
o
g
n
itio
n
s
y
s
te
m
,
SI
R
I
,
C
o
r
tan
a
p
er
s
o
n
al
ass
i
s
tan
t
all
th
o
s
e
ar
e
d
ev
elo
p
ed
u
s
in
g
d
ee
p
lear
n
in
g
alg
o
r
it
h
m
al
s
o
n
e
w
l
y
i
n
tr
o
d
u
ce
d
Am
az
o
n
Go
s
to
r
es
also
in
clu
d
e
s
d
ee
p
lear
n
in
g
.
T
h
r
ee
C
ap
ab
ilit
ies
o
f
Dee
p
L
ea
r
n
i
n
g
f
ir
s
t
is
Gen
er
aliza
b
ilit
y
—
h
o
w
p
r
ec
is
el
y
t
h
e
m
ac
h
i
n
e
w
i
ll
e
x
ec
u
te
esti
m
atio
n
o
n
g
iv
e
n
d
ata
wh
ich
h
a
s
n
o
t
b
ee
n
f
o
r
m
u
lated
y
et?
Seco
n
d
is
T
r
ain
ab
ilit
y
h
o
w
r
ap
id
l
y
d
ec
ilit
r
e
a
DL
f
r
a
m
e
w
o
r
k
w
ill
g
et
ac
cu
s
to
m
ed
to
it
s
co
n
ce
r
n
.
An
d
t
h
ir
d
is
E
x
p
r
es
s
ib
ilit
y
—
t
h
i
s
f
ea
t
u
r
e
d
elin
e
ates
h
o
w
w
ell
a
m
ac
h
i
n
e
w
i
ll
ev
a
lu
ate
g
e
n
er
a
l
esti
m
atio
n
s
.
Dee
p
lear
n
i
n
g
,
f
o
r
in
s
ta
n
ce
,
in
ter
p
r
etab
ilit
y
,
m
ea
s
u
r
ed
q
u
alit
y
,
tr
an
s
f
er
ab
i
lit
y
,
i
n
er
tn
e
s
s
,
il
l
-
d
is
p
o
s
ed
s
o
u
n
d
n
e
s
s
,
an
d
s
ec
u
r
it
y.
4.
CURR
E
NT
CH
A
L
L
E
N
G
E
S
I
n
cu
r
r
en
t
ti
m
e
Dee
p
L
ea
r
n
i
n
g
h
as
t
u
r
n
ed
o
u
t
to
b
e
o
n
e
in
ev
er
y
th
i
n
g
ab
o
u
t
f
ir
s
t
in
v
esti
g
atio
n
r
eg
io
n
s
i
n
cr
ea
tin
g
astu
te
m
ac
h
in
e
s
.
T
h
e
m
aj
o
r
it
y
o
f
n
o
tab
le
ap
p
licatio
n
s
lik
e
I
m
a
g
e
an
d
Sp
ee
ch
R
ec
o
g
n
itio
n
,
T
ex
t
p
r
o
ce
s
s
in
g
,
a
n
d
NP
L
-
Na
t
u
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
o
f
A
I
ar
e
d
r
iv
e
n
b
y
Dee
p
L
ea
r
n
in
g
n
o
w
ad
a
y
s
v
er
y
ea
s
il
y
.
Dee
p
L
ea
r
n
i
n
g
alg
o
r
it
h
m
s
co
p
y
h
u
m
a
n
ce
r
eb
r
u
m
s
ex
p
lo
itatio
n
ar
tif
icia
l n
e
u
r
al
n
et
w
o
r
k
s
an
d
m
o
r
e
a
n
d
in
cr
ea
s
i
n
g
l
y
m
o
r
e
p
r
o
f
o
u
n
d
l
y
f
i
g
u
r
e
o
u
t
h
o
w
to
p
r
ec
is
el
y
ta
k
e
ca
r
e
o
f
a
g
iv
e
n
is
s
u
e.
Ho
wev
er
th
er
e
ar
e
v
ital
ch
alle
n
g
e
s
i
n
Dee
p
L
ea
r
n
i
n
g
s
y
s
te
m
s
t
h
at
w
e
'v
e
g
o
t
to
s
e
e
m
o
u
t
f
o
r
lo
ts
a
n
d
lo
ts
o
f
i
n
f
o
r
m
atio
n
,
a
lar
g
e
a
m
o
u
n
t
o
f
d
ata
a
n
d
p
len
t
y
o
f
i
n
f
o
r
m
atio
n
,
O
v
er
f
itti
n
g
i
n
n
eu
r
al
n
et
w
o
r
k
s
,
H
y
p
er
p
ar
a
m
eter
Op
ti
m
izatio
n
,
R
eq
u
ir
es
h
i
g
h
-
p
er
f
o
r
m
a
n
ce
h
ar
d
w
ar
e,
Neu
r
al
n
e
t
w
o
r
k
s
ar
e
in
es
s
en
ce
a
B
lack
-
b
o
x
,
L
ac
k
o
f
Fle
x
ib
ili
t
y
a
n
d
Mu
ltit
a
s
k
in
g
.
Dee
p
lear
n
in
g
is
a
m
e
th
o
d
o
lo
g
y
t
h
at
m
o
d
els
h
u
m
a
n
th
eo
r
etica
l
r
ea
s
o
n
in
g
o
r
p
o
s
s
ib
l
y
s
p
ea
k
s
to
en
d
ea
v
o
r
to
ap
p
r
o
ac
h
it
r
ath
e
r
u
tili
zi
n
g
.
T
h
o
u
g
h
,
t
h
is
i
n
n
o
v
atio
n
h
as
an
al
s
o
s
o
m
e
s
et
o
f
d
is
ad
v
a
n
ta
g
es
to
o
.
Ma
n
ag
e
m
e
n
t
o
f
C
o
n
s
ta
n
t
I
n
p
u
t
Data
:
I
n
DL
,
a
tr
ain
i
n
g
p
r
o
ce
s
s
d
ep
en
d
s
o
n
ex
a
m
i
n
i
n
g
a
lo
t
o
f
in
f
o
r
m
at
io
n
.
A
lb
eit,
s
p
ee
d
il
y
f
lo
w
i
n
g
i
n
p
u
t
d
ata
an
d
g
iv
e
a
litt
le
p
er
io
d
o
f
ti
m
e
to
w
ar
d
g
u
ar
a
n
teei
n
g
a
p
r
o
d
u
ct
iv
e
tr
ain
i
n
g
p
r
o
ce
s
s
.
W
h
ich
is
th
e
m
ain
ca
u
s
e
th
at
d
ata
r
esear
c
h
er
s
n
ee
d
to
ad
j
u
s
t
t
h
eir
al
g
o
r
ith
m
in
th
e
m
a
n
n
er
in
w
h
ic
h
n
eu
r
al
n
et
w
o
r
k
s
w
i
ll
d
ea
l
w
i
t
h
a
lo
t
o
f
co
n
s
ta
n
t
i
n
p
u
t
d
ata
?
Gu
ar
an
teein
g
C
o
n
cl
u
s
io
n
T
r
an
s
p
ar
en
c
y
:
O
n
e
m
o
r
e
cr
itical
d
is
ad
v
a
n
ta
g
e
o
f
d
ee
p
lear
n
in
g
p
r
o
g
r
a
m
m
i
n
g
is
,
DL
is
u
n
eq
u
ip
p
ed
f
o
r
f
u
r
n
is
h
ed
o
p
in
io
n
s
b
ec
au
s
e
it
h
as
alr
ea
d
y
ac
h
iev
e
d
a
co
n
v
i
n
ci
n
g
co
n
clu
s
io
n
.
Un
lik
e
i
n
ca
s
e
o
f
a
n
cie
n
t
m
ac
h
i
n
e
lear
n
in
g
,
w
e
ca
n
't
tr
ac
k
an
as
s
o
ciate
al
g
o
r
ith
m
f
o
r
m
u
la
to
s
ea
r
c
h
w
h
y
,
o
u
r
s
y
s
te
m
h
as
f
i
x
ed
th
a
t
it
'
s
a
ca
t
o
n
an
i
m
ag
e,
n
o
t
a
tig
er
.
T
o
ad
d
r
ess
b
lu
n
d
er
s
i
n
d
ee
p
lear
n
in
g
p
r
o
g
r
a
m
o
r
alg
o
r
ith
m
,
w
e
n
ee
d
to
u
p
d
ate
t
h
e
en
tire
ca
lc
u
lat
io
n
an
d
alg
o
r
it
h
m
to
o
.
R
eso
u
r
ce
-
De
m
a
n
d
in
g
T
ec
h
n
o
lo
g
y
:
I
t
is
an
alto
g
et
h
er
r
eso
u
r
ce
-
d
e
m
a
n
d
in
g
tec
h
n
o
lo
g
y
.
I
t
n
ee
d
s
ad
d
itio
n
al
p
o
w
er
f
u
l
h
ig
h
-
p
er
f
o
r
m
a
n
ce
GP
Us,
m
e
an
s
its
w
o
r
k
o
n
C
P
Us
b
u
t
it
g
iv
es
tr
e
m
en
d
o
u
s
r
esu
lt
s
o
n
GP
Us,
it
also
r
eq
u
ir
ed
lar
g
e
a
m
o
u
n
t
s
o
f
s
to
r
ag
e
s
p
ac
e
to
tr
ain
i
n
g
a
m
o
d
el,
etc.
Mo
r
eo
v
er
,
th
e
tech
n
o
lo
g
y
ta
k
es
a
lo
n
g
er
tr
ain
in
g
ti
m
e
as
co
m
p
ar
ed
to
an
cien
t
ML
.
D
e
s
p
ite
all
its
ch
alle
n
g
e
s
,
d
ee
p
lear
n
in
g
ex
p
lo
r
er
s
ad
v
a
n
ce
d
tech
n
iq
u
e
s
f
o
r
th
o
s
e
r
esear
ch
er
s
w
h
o
h
av
e
t
h
e
i
n
te
n
tio
n
to
u
s
e
u
n
s
tr
u
ct
u
r
ed
b
ig
d
ata
an
al
y
tics
.
I
n
d
ee
d
,
DL
ta
s
k
s
o
f
d
ata
p
r
o
ce
s
s
in
g
g
iv
e
s
ig
n
i
f
ica
n
t b
en
e
f
its
to
E
d
u
ca
to
r
s
,
a
n
al
y
s
t,
co
r
p
o
r
atio
n
s
.
5.
RE
C
E
NT
D
E
V
E
L
O
P
M
E
NT
S
T
h
er
e
ar
e
s
o
m
an
y
co
m
p
u
t
er
v
is
io
n
p
r
o
b
le
m
s
a
n
d
is
s
u
es
li
k
e
clas
s
i
f
icatio
n
r
eo
r
g
an
izatio
n
,
id
en
ti
f
icatio
n
,
la
n
g
u
a
g
e
p
r
o
ce
s
s
i
n
g
,
v
id
eo
p
r
o
ce
s
s
i
n
g
,
g
estu
r
e
d
etec
tio
n
,
r
o
b
o
tics
etc
.
w
i
ll
c
u
r
r
en
tl
y
i
n
th
e
w
o
r
ld
o
f
p
r
o
g
r
ess
i
n
g
d
ee
p
lear
n
in
g
ar
ea
all
C
o
m
p
u
ter
v
i
s
io
n
p
r
o
b
lem
s
b
e
m
ea
s
u
r
ed
as
r
eso
lv
ed
.
I
n
r
ec
en
t
m
o
d
el
-
b
ased
d
ev
elo
p
m
en
t,
m
o
d
els
b
ased
o
n
C
NN
-
C
o
n
v
o
lu
tio
n
al
Neu
r
al
Ne
t
w
o
r
k
s
h
av
e
r
ev
o
l
u
tio
n
ized
th
e
en
tire
f
ield
o
f
co
m
p
u
ter
v
is
io
n
.
I
t
is
n
o
w
v
er
y
ea
s
y
to
d
ev
elo
p
th
r
o
u
g
h
s
ev
er
al
Dee
p
L
ea
r
n
in
g
co
n
f
i
g
u
r
atio
n
s
u
tili
zi
n
g
ad
j
u
s
tin
g
w
ith
p
r
e
-
tr
ai
n
ed
w
e
ig
h
ts
.
Fo
r
ex
am
p
le,
clas
s
if
icatio
n
th
r
o
u
g
h
I
m
a
g
eNe
t
.
Nev
er
th
e
less
,
t
h
e
m
o
r
e
s
ec
u
r
it
y
is
s
u
e
s
o
f
o
b
j
ec
t
d
etec
tio
n
an
d
s
eg
m
e
n
tatio
n
n
ee
d
ad
v
a
n
ce
d
w
a
y
s
to
cr
ac
k
th
e
s
o
lu
tio
n
.
Ob
j
ec
t
d
etec
tio
n
co
n
s
i
s
ts
o
f
lear
n
i
n
g
t
h
e
o
b
j
ec
ts
an
d
d
r
a
w
i
n
g
a
r
ec
tan
g
le
b
o
u
n
d
in
g
b
o
x
,
w
h
er
ea
s
s
eg
m
e
n
tatio
n
tar
g
ets
to
s
p
o
t
p
r
ec
is
e
p
ix
els
t
h
at
f
it
e
v
er
y
o
b
ject.
On
e
a
m
o
n
g
s
t
m
o
s
t
v
ar
ia
ti
o
n
s
th
r
o
u
g
h
i
m
ag
e
class
i
f
icatio
n
is
t
h
at
a
s
i
m
ilar
i
m
a
g
e
co
u
ld
co
n
tai
n
m
an
y
o
b
j
ec
ts
an
d
p
eo
p
le
m
i
g
h
t
r
e
m
ai
n
in
to
tall
y
d
i
v
er
s
e
s
izes,
li
g
h
tin
g
,
p
r
o
p
o
r
tio
n
,
a
n
d
p
ar
tl
y
o
cc
lu
d
ed
.
P
r
o
f
ess
o
r
Nick
R
ee
d
,
ac
ad
e
m
y
d
ir
ec
to
r
at
th
e
T
r
an
s
p
o
r
t
R
esear
ch
L
ab
o
r
ato
r
y
(
T
R
L
)
,
ag
r
ee
s
th
a
t
d
ee
p
lear
n
in
g
i
s
a
v
er
y
i
m
p
o
r
tan
t
to
o
l,
b
u
t
o
n
e
th
at
r
aises
s
er
io
u
s
co
n
ce
r
n
s
.
"
Dee
p
lear
n
in
g
is
s
o
m
e
w
h
at
o
p
aq
u
e.
I
t c
an
b
e
h
ar
d
to
u
n
d
er
s
ta
n
d
th
e
r
u
les o
r
k
n
o
w
led
g
e
lear
n
ed
b
y
th
e
s
y
s
te
m
.
I
n
t
h
e
ca
s
e
o
f
s
elf
-
d
r
iv
i
n
g
ca
r
s
,
t
h
i
s
m
a
y
b
e
co
m
e
i
m
p
o
r
tan
t
in
th
e
e
v
en
t
o
f
a
co
llis
io
n
.
Fo
r
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
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
538
-
548
542
ex
a
m
p
le,
i
f
a
v
e
h
icle
i
s
d
ep
en
d
en
t
o
n
a
d
ee
p
lear
n
i
n
g
a
lg
o
r
ith
m
,
it
m
a
y
b
e
d
i
f
f
ic
u
lt
to
u
n
d
er
s
ta
n
d
h
o
w
th
e
v
eh
ic
le
u
s
ed
t
h
e
a
v
ailab
le
i
n
f
o
r
m
atio
n
to
d
eter
m
i
n
e
it
s
ac
t
io
n
s
,
s
u
b
s
eq
u
en
tl
y
r
es
u
lti
n
g
i
n
a
lack
o
f
clar
it
y
o
v
er
liab
ilit
y
.
"
6.
T
O
WARD
S DE
E
P
L
E
ARN
I
NG
Dee
p
lear
n
in
g
a
p
ar
ad
ig
m
o
f
Ma
ch
i
n
e
lear
n
i
n
g
w
h
ich
h
a
s
s
h
o
w
n
i
n
cr
ed
ib
le
p
r
o
m
is
e
s
in
r
ec
en
t
ti
m
e.
T
h
is
is
b
ec
au
s
e
o
f
th
e
f
ac
t t
h
at
Dee
p
L
ea
r
n
i
n
g
s
h
o
w
s
a
g
r
ea
t
an
alo
g
y
w
it
h
t
h
e
f
u
n
ctio
n
i
n
g
o
f
th
e
h
u
m
a
n
b
r
ain
.
I
n
b
elo
w
Fig
u
r
e
6
s
h
o
w
s
t
h
e
T
r
ad
itio
n
al
Ma
ch
in
e
L
ea
r
n
i
n
g
ap
p
r
o
ac
h
es.
I
t
s
ee
m
s
in
a
p
ictu
r
e
o
f
m
ac
h
in
e
lear
n
in
g
th
a
t
f
ea
t
u
r
es
ex
tr
ac
ti
o
n
an
d
class
i
f
icatio
n
ar
e
d
o
n
e
s
ep
ar
ately
a
n
d
b
o
th
co
n
ce
r
n
ed
w
ith
t
h
e
co
m
p
le
x
d
esig
n
an
d
lo
ts
o
f
s
ig
n
i
f
ican
t
m
at
h
e
m
a
tics
,
af
ter
d
o
in
g
it
v
e
r
y
p
r
ec
is
el
y
s
o
m
eti
m
e
s
it
w
a
s
n
o
t
ev
en
ef
f
icie
n
t,
an
d
d
id
n
’
t p
er
f
o
r
m
w
e
ll
m
ea
n
s
in
co
n
ce
r
n
w
it
h
r
ea
l
-
w
o
r
ld
ap
p
licatio
n
s
th
e
ac
c
u
r
ac
y
le
v
el
w
a
s
n
’
t s
u
itab
le.
Fig
u
r
e
6
.
T
r
ad
itio
n
al
m
ac
h
i
n
e
lear
n
in
g
f
lo
w
7.
AT
P
RE
SE
N
T
,
AB
O
UT
D
E
E
P
L
E
ARNI
N
G
DL
n
et
w
o
r
k
s
w
i
ll
p
er
f
o
r
m
f
e
atu
r
e
ex
tr
ac
tio
n
a
n
d
clas
s
i
f
ica
tio
n
i
n
o
n
e
s
h
o
t,
th
at
’
s
w
h
y
s
o
m
e
ti
m
es
p
eo
p
le
u
s
e
to
ca
ll
o
n
e
s
h
o
r
t
lear
n
in
g
in
s
tead
o
f
d
ee
p
lear
n
in
g
.
W
h
ich
i
m
p
lie
s
it
j
u
s
t
n
ee
d
to
p
lan
an
d
d
esig
n
o
n
e
m
o
d
el.
T
h
e
f
u
n
d
a
m
e
n
tal
ad
v
an
ta
g
e
o
f
d
ee
p
lear
n
in
g
o
v
er
ML
al
g
o
r
ith
m
is
,
D
L
h
a
s
th
e
ab
ilit
y
to
cr
ea
te
n
e
w
f
ea
t
u
r
es
f
r
o
m
p
ar
tial
s
er
i
es
o
f
f
ea
tu
r
es
lo
ca
ted
i
n
t
h
e
t
r
ain
in
g
d
ataset.
Als
o
as
s
ee
n
in
Fig
u
r
e
7
Feat
u
r
e
E
x
tr
ac
tio
n
a
n
d
clas
s
i
f
icatio
n
w
il
l
b
e
d
o
n
e
i
n
o
n
e
s
h
o
r
t.
D
L
m
o
d
el
h
a
s
t
h
e
co
n
v
e
n
ien
ce
o
f
GP
Us
a
n
d
GP
U
w
il
l
p
r
o
ce
s
s
h
u
g
e
a
m
o
u
n
ts
o
f
lab
eled
d
ata
in
p
ar
allel
at
h
i
g
h
s
p
ee
d
s
e
m
p
o
w
er
s
.
D
L
m
o
d
el
s
to
b
e
m
u
ch
f
as
ter
th
an
M
L
m
et
h
o
d
s
.
I
n
D
L
Mo
d
el
w
i
th
b
ac
k
p
r
o
p
ag
atio
n
al
g
o
r
ith
m
,
ef
f
ec
ti
v
e
lo
s
s
f
u
n
c
tio
n
li
k
e
R
e
L
u
,
an
d
a
b
ig
a
m
o
u
n
t
o
f
p
ar
a
m
eter
s
,
th
e
s
e
D
L
n
et
w
o
r
k
s
ar
e
ca
p
ab
le
to
le
ar
n
ex
tr
e
m
el
y
co
m
p
le
x
a
n
d
a
b
s
tr
ac
t
f
ea
t
u
r
es
f
r
o
m
th
e
s
et,
tr
ad
itio
n
al
M
L
n
ee
d
to
h
an
d
cr
af
t
t
h
e
f
ea
t
u
r
es
an
d
n
o
ab
s
tr
ac
t
f
ea
t
u
r
es
ar
e
th
er
e.
So
m
e
P
y
to
r
c
h
,
T
en
s
o
r
Flo
w
,
an
d
Ker
as
ar
e
h
i
g
h
-
le
v
el
o
p
en
s
o
u
r
ce
lib
r
ar
ies
ar
e
th
er
e
w
i
th
D
L
f
r
am
e
w
o
r
k
s
f
o
r
ea
s
y
i
m
p
le
m
en
ta
tio
n
.
No
w
ad
a
y
s
t
h
e
n
e
w
e
s
t
tr
en
d
i
n
r
esear
ch
an
d
d
ev
elo
p
m
e
n
t
in
Ma
c
h
i
n
e
L
ea
r
n
i
n
g
is
Dee
p
lear
n
in
g
,
w
h
y
b
ec
a
u
s
e
D
L
m
e
th
o
d
s
h
av
e
ca
r
r
ied
g
r
o
u
n
d
-
b
r
ea
k
in
g
ad
v
a
n
ce
s
in
co
m
p
u
ter
v
is
io
n
an
d
m
ac
h
i
n
e
lear
n
in
g
.
Dee
p
L
ea
r
n
i
n
g
is
o
p
en
,
o
u
tp
er
f
o
r
m
,
s
tate
-
of
-
ar
t
t
ec
h
n
o
lo
g
y
.
I
n
r
ec
en
t
ti
m
e
m
a
n
y
n
e
w
ap
p
licatio
n
lik
e
s
p
ee
c
h
,
v
id
eo
,
r
o
b
o
tics
,
s
ec
u
r
it
y
m
a
k
es p
r
ac
ticall
y
f
ea
s
ib
le.
Fig
u
r
e
7
.
Dee
p
lear
n
in
g
f
lo
w
8.
T
YP
E
O
F
ARCH
I
T
E
CT
UR
E
S
R
esear
ch
er
s
h
a
v
e
cr
ea
ted
a
v
ar
iet
y
o
f
ar
ch
itect
u
r
es
w
h
ich
w
e
ca
n
s
ee
i
n
b
elo
w
c
h
ar
t
th
at
t
h
o
s
e
ex
is
ted
f
r
o
m
th
e
late
1
9
4
0
s
,
ar
ch
itect
u
r
es
a
n
d
ap
p
r
o
ac
h
es
a
r
e
m
o
d
if
ied
i
n
d
a
y
b
y
d
a
y
d
e
v
elo
p
m
e
n
t
a
n
d
n
e
w
ar
ch
itect
u
r
es
ar
e
cr
ea
ted
b
u
t
a
f
ter
2
0
0
6
th
e
n
e
w
ar
ch
i
tectu
r
e
s
an
d
g
r
ap
h
ical
p
r
o
ce
s
s
in
g
u
n
i
ts
(
GP
Us)
b
r
o
u
g
h
t
th
e
m
to
t
h
e
p
o
le
o
f
A
I
.
Fro
m
th
e
la
s
t
t
w
o
d
ec
ad
es,
d
ee
p
lear
n
in
g
ar
c
h
itect
u
r
es
h
a
v
e
g
r
ea
tl
y
ex
p
a
n
d
ed
th
e
p
r
o
b
le
m
s
n
e
u
r
al
n
et
w
o
r
k
s
ca
n
ad
d
r
ess
.
Dee
p
lear
n
i
n
g
is
m
u
c
h
m
o
r
e
th
a
n
t
h
e
n
e
u
r
al
n
et
w
o
r
k
f
o
llo
w
s
th
e
Fi
g
u
r
e
8
in
tr
o
d
u
ce
f
r
o
m
m
an
y
r
esear
c
h
p
ap
er
s
b
y
Fa
v
io
Váz
q
u
ez
.
T
ab
le
1
s
h
o
w
s
t
h
e
Neu
r
al
n
e
t
w
o
r
k
to
Dee
p
lear
n
i
n
g
m
o
d
els,
r
u
les
a
n
d
s
o
m
e
f
u
n
ctio
n
s
w
it
h
t
h
eir
ap
p
licatio
n
s
.
Mc
C
u
l
lo
ch
–
P
itts
m
o
d
el,
Heb
b
R
u
le,
P
er
ce
p
tr
o
n
,
A
d
ap
ti
v
e
li
n
ea
r
ele
m
en
t,
M
u
lti
-
L
a
y
e
r
P
er
ce
p
tr
o
n
(
ML
P
)
,
Ho
p
f
iel
d
Net
w
o
r
k
C
ir
cu
it,
B
ac
k
-
P
r
o
p
ag
atio
n
,
S
u
p
p
o
r
t
v
ec
to
r
Ma
ch
i
n
e
(
SV
M)
,
B
o
ltzm
a
n
n
Ma
c
h
in
e,
R
estric
ted
B
o
ltzm
a
n
n
Ma
ch
in
e
(
R
B
M)
,
Dee
p
B
o
ltzm
an
n
Ma
c
h
in
e
(
DB
M)
,
Au
to
en
co
d
er
,
R
ec
u
r
r
en
t
Neu
r
al
n
e
t
w
o
r
k
(
R
NN)
,
C
o
n
v
o
lu
t
io
n
n
eu
r
al
Net
w
o
r
k
o
r
C
NN
o
r
C
o
n
v
Net,
L
o
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
L
ST
M)
,
Dee
p
b
elief
n
et
w
o
r
k
(
DM
N)
,
Dee
p
A
u
to
-
en
co
d
er
,
Sp
ar
s
e
Au
to
en
co
d
er
,
A
tten
tio
n
b
ased
L
ST
M
(
AL
ST
M)
,
Gen
er
ativ
e
A
d
v
er
s
ar
ial
Net
w
o
r
k
(
G
A
N)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
Th
e
u
p
s
u
r
g
e
o
f d
ee
p
lea
r
n
in
g
fo
r
co
mp
u
ter visi
o
n
a
p
p
lica
tio
n
s
(
P
r
iya
n
ka
P
a
tel)
543
Fig
u
r
e
8
.
A
r
ti
f
icial
i
n
telli
g
e
n
t,
m
ac
h
i
n
e
lear
n
in
g
,
d
ee
p
lear
n
i
n
g
ap
p
r
o
ac
h
'
s
r
ev
o
l
u
tio
n
t
i
m
e
lin
e
T
ab
le
1
.
Mo
d
els an
d
r
u
le
w
ith
th
eir
g
r
ap
h
ical
r
ep
r
ese
n
tatio
n
an
d
ap
p
licatio
n
s
P
r
o
p
o
se
d
mo
d
e
l
s
G
r
a
p
h
i
c
a
l
r
e
p
r
e
se
n
t
a
t
i
o
n
A
p
p
l
i
c
a
t
i
o
n
s
T
i
me
l
i
n
e
M
c
C
u
l
l
o
c
h
–
P
i
t
t
s
mo
d
e
l
-
F
i
r
st
e
v
e
r
mat
h
e
ma
t
i
c
a
l
mo
d
e
l
-
L
i
n
e
a
r
t
h
r
e
sh
o
l
d
g
a
t
e
-
W
e
i
g
h
t
s
a
n
d
b
i
a
s
a
r
e
f
i
x
e
d
-
G
e
n
e
r
a
t
e
s b
i
n
a
r
y
o
u
t
p
u
t
1
9
4
3
[
6
]
H
e
b
b
R
u
l
e
-
I
t
i
s a
l
e
a
r
n
i
n
g
R
u
l
e
.
-
I
t
i
s u
se
d
t
o
s
p
e
c
i
f
y
t
h
e
w
e
i
g
h
t
i
n
p
r
o
p
o
r
t
i
o
n
t
o
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
.
-
I
t
w
o
r
k
s we
l
l
i
f
t
h
e
i
n
p
u
t
p
a
t
t
e
r
n
s
a
r
e
o
r
t
h
o
g
o
n
a
l
o
r
u
n
c
o
r
r
e
l
a
t
e
d
.
1
9
4
9
[
7
]
P
e
r
c
e
p
t
r
o
n
-
U
se
d
f
o
r
su
p
e
r
v
i
se
d
l
e
a
r
n
i
n
g
o
f
b
i
n
a
r
y
c
l
a
ssi
f
i
e
r
s
-
C
l
a
ssi
f
i
c
a
t
i
o
n
a
n
d
-
R
e
g
r
e
ssi
o
n
1
9
5
8
[
8
]
A
d
a
p
t
i
v
e
l
i
n
e
a
r
e
l
e
me
n
t
-
S
i
n
g
l
e
l
a
y
e
r
A
N
N
-
A
d
a
p
t
i
v
e
l
i
n
e
a
r
u
n
i
t
/
n
e
u
r
o
n
1
9
6
0
[
9
,
1
0
]
M
u
l
t
i
-
L
a
y
e
r
p
e
r
c
e
p
t
r
o
n
-
F
e
e
d
f
o
r
w
a
r
d
a
r
t
i
f
i
c
i
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
.
-
S
p
e
e
c
h
R
e
c
o
g
n
i
t
i
o
n
-
I
mag
e
R
e
c
o
g
n
i
t
i
o
n
-
M
a
c
h
i
n
e
T
r
a
n
sl
a
t
i
o
n
-
C
l
a
ssi
f
i
c
a
t
i
o
n
1
9
6
9
[
11
,
1
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
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
538
-
548
544
T
ab
le
1
.
Mo
d
els an
d
r
u
le
w
ith
th
eir
g
r
ap
h
ical
r
ep
r
ese
n
tatio
n
an
d
ap
p
licatio
n
s
(
co
n
tin
u
e
)
P
r
o
p
o
se
d
mo
d
e
l
s
G
r
a
p
h
i
c
a
l
r
e
p
r
e
se
n
t
a
t
i
o
n
A
p
p
l
i
c
a
t
i
o
n
s
T
i
me
l
i
n
e
H
o
p
f
i
e
l
d
N
e
t
w
o
r
k
C
i
r
c
u
i
t
-
R
e
c
u
r
r
e
n
t
A
r
t
i
f
i
c
i
a
l
N
e
u
r
a
l
N
e
t
w
o
r
k
-
R
e
c
o
g
n
i
t
i
o
n
1
9
8
2
[
13
]
B
a
c
k
-
P
r
o
p
a
g
a
t
i
o
n
-
C
l
a
ssi
f
i
c
a
t
i
o
n
-
T
i
me
-
seri
e
s Pr
e
d
i
c
t
i
o
n
-
F
u
n
c
t
i
o
n
A
p
p
r
o
x
i
m
a
t
i
o
n
1
9
7
0
[1
4
]
S
u
p
p
o
r
t
v
e
c
t
o
r
M
a
c
h
i
n
e
-
B
i
o
i
n
f
o
r
mat
i
c
s
-
G
e
n
e
r
a
l
i
z
e
d
P
r
e
d
i
c
t
i
v
e
C
o
n
t
r
o
l
-
O
b
j
e
c
t
D
e
t
e
c
t
i
o
n
-
H
a
n
d
w
r
i
t
i
n
g
R
e
c
o
g
n
i
t
i
o
n
-
T
e
x
t
A
n
d
H
y
p
e
r
t
e
x
t
C
a
t
e
g
o
r
i
z
a
t
i
o
n
-
I
mag
e
s C
l
a
ssi
f
i
c
a
t
i
o
n
-
P
r
o
t
e
i
n
F
o
l
d
A
n
d
R
e
mo
t
e
H
o
mo
l
o
g
y
D
e
t
e
c
t
i
o
n
1
9
6
3
[1
5
]
1
9
9
2
[1
6
]
B
o
l
t
z
man
M
a
c
h
i
n
e
-
P
a
t
t
e
r
n
r
e
c
o
g
n
i
t
i
o
n
-
C
o
mb
i
n
a
t
o
r
i
a
l
o
p
t
i
m
i
z
a
t
i
o
n
-
S
p
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
1
9
8
6
[
1
7
]
R
e
st
r
i
c
t
e
d
B
o
l
t
z
man
M
a
c
h
i
n
e
&
D
e
e
p
B
o
l
t
z
man
M
a
c
h
i
n
e
-
P
a
t
t
e
r
n
R
e
c
o
g
n
i
t
i
o
n
-
C
l
a
ssi
f
i
c
a
t
i
o
n
-
C
o
mb
i
n
a
t
o
r
i
a
l
O
p
t
i
mi
z
a
t
i
o
n
-
S
p
e
e
c
h
R
e
c
o
g
n
i
t
i
o
n
1
9
8
6
[1
8
]
A
u
t
o
e
n
c
o
d
e
r
-
U
se
d
f
o
r
u
n
s
u
p
e
r
v
i
se
d
l
e
a
r
n
i
n
g
a
n
d
d
i
me
n
s
i
o
n
a
l
i
t
y
r
e
d
u
c
t
i
o
n
-
I
t
’
s fo
r
r
e
p
r
e
se
n
t
a
t
i
o
n
l
e
a
r
n
i
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
-
I
mag
e
g
e
n
e
r
a
t
i
o
n
-
D
a
t
a
v
i
su
a
l
i
z
a
t
i
o
n
-
N
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ssi
n
g
1
9
8
6
[1
9
]
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
Th
e
u
p
s
u
r
g
e
o
f d
ee
p
lea
r
n
in
g
fo
r
co
mp
u
ter visi
o
n
a
p
p
lica
tio
n
s
(
P
r
iya
n
ka
P
a
tel)
545
T
ab
le
1
.
Mo
d
els an
d
r
u
le
w
ith
th
eir
g
r
ap
h
ical
r
ep
r
ese
n
tatio
n
an
d
ap
p
licatio
n
s
(
co
n
tin
u
e
)
P
r
o
p
o
se
d
mo
d
e
l
s
G
r
a
p
h
i
c
a
l
r
e
p
r
e
se
n
t
a
t
i
o
n
A
p
p
l
i
c
a
t
i
o
n
s
T
i
me
l
i
n
e
R
e
c
u
r
r
e
n
t
N
e
u
r
a
l
n
e
t
w
o
r
k
-
S
p
e
e
c
h
R
e
c
o
g
n
i
t
i
o
n
-
H
a
n
d
w
r
i
t
i
n
g
R
e
c
o
g
n
i
t
i
o
n
1
9
8
6
[
2
0
-
2
3
]
C
o
n
v
o
l
u
t
i
o
n
n
e
u
r
a
l
N
e
t
w
o
r
k
or
C
N
N
or
C
o
n
v
N
e
t
-
I
mag
e
A
n
d
V
i
d
e
o
R
e
c
o
g
n
i
t
i
o
n
-
I
mag
e
C
l
a
ssi
f
i
c
a
t
i
o
n
-
V
i
d
e
o
A
n
a
l
y
si
s
-
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
P
r
o
c
e
ssi
n
g
(
N
L
P
)
-
M
e
d
i
c
a
l
I
mag
e
A
n
a
l
y
si
s
1
9
8
3
[
2
4
]
1
9
9
9
[
2
5
]
L
o
n
g
sh
o
r
t
-
t
e
r
m
me
mo
r
y
-
S
p
e
e
c
h
R
e
c
o
g
n
i
t
i
o
n
-
G
e
st
u
r
e
R
e
c
o
g
n
i
t
i
o
n
-
H
a
n
d
w
r
i
t
i
n
g
R
e
c
o
g
n
i
t
i
o
n
-
N
L
P
&Te
x
t
C
o
mp
r
e
ssi
o
n
-
I
mag
e
C
a
p
t
i
o
n
i
n
g
1
9
9
7
[
26
]
D
e
e
p
b
e
l
i
e
f
n
e
t
w
o
r
k
-
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
U
n
d
e
r
st
a
n
d
i
n
g
-
I
mag
e
R
e
c
o
g
n
i
t
i
o
n
-
F
a
i
l
u
r
e
P
r
e
d
i
c
t
i
o
n
-
I
n
f
o
r
mat
i
o
n
R
e
t
r
i
e
v
a
l
2
0
0
7
2
0
0
9
[
2
7
]
-
D
e
e
p
A
u
t
o
e
n
c
o
d
e
r
-
U
se
d
f
o
r
u
n
s
u
p
e
r
v
i
se
d
l
e
a
r
n
i
n
g
a
n
d
d
i
me
n
s
i
o
n
a
l
i
t
y
r
e
d
u
c
t
i
o
n
-
I
t
’
s fo
r
r
e
p
r
e
se
n
t
a
t
i
o
n
l
e
a
r
n
i
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
-
I
mag
e
g
e
n
e
r
a
t
i
o
n
-
D
a
t
a
v
i
su
a
l
i
z
a
t
i
o
n
-
N
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ssi
n
g
2
0
0
0
s
[2
8
]
-
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
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
538
-
548
546
T
ab
le
1
.
Mo
d
els an
d
r
u
le
w
ith
th
eir
g
r
ap
h
ical
r
ep
r
ese
n
tatio
n
an
d
ap
p
licatio
n
s
(
co
n
tin
u
e
)
P
r
o
p
o
se
d
mo
d
e
l
s
G
r
a
p
h
i
c
a
l
r
e
p
r
e
se
n
t
a
t
i
o
n
A
p
p
l
i
c
a
t
i
o
n
s
T
i
me
l
i
n
e
S
p
a
r
se
A
u
t
o
e
n
c
o
d
e
r
-
U
se
d
f
o
r
u
n
s
u
p
e
r
v
i
se
d
l
e
a
r
n
i
n
g
a
n
d
d
i
me
n
s
i
o
n
a
l
i
t
y
r
e
d
u
c
t
i
o
n
-
O
b
j
e
c
t
R
e
c
o
g
n
i
t
i
o
n
-
M
e
d
i
c
a
l
D
i
a
g
n
o
si
s
2
0
0
0
s
[
2
9
-
3
1
]
A
t
t
e
n
t
i
o
n
b
a
se
d
L
S
T
M
-
U
se
d
t
o
p
r
e
d
i
c
t
t
a
r
g
e
t
w
o
r
d
-
D
e
e
p
N
L
P
2
0
1
0
s
[
3
2
-
3
4
]
G
e
n
e
r
a
t
i
v
e
A
d
v
e
r
sari
a
l
N
e
t
w
o
r
k
-
u
n
s
u
p
e
r
v
i
se
d
mac
h
i
n
e
l
e
a
r
n
i
n
g
-
R
e
c
o
n
st
r
u
c
t
3
D
mo
d
e
l
s
f
r
o
m i
mag
e
.
2
0
1
4
[3
5
-
3
7
]
9.
H
O
W
DE
E
P
L
E
AR
NIN
G
WO
RK
S
Dee
p
lear
n
in
g
w
o
r
k
i
n
g
a
n
d
th
e
n
e
u
r
al
n
et
w
o
r
k
w
o
r
k
i
n
g
i
s
al
m
o
s
t
t
h
e
s
a
m
e.
D
L
w
o
r
k
s
f
o
r
b
o
th
s
u
p
er
v
i
s
ed
lear
n
in
g
an
d
u
n
s
u
p
er
v
is
ed
lear
n
i
n
g
.
An
d
as
ea
r
lier
d
is
cu
s
s
ed
in
p
ap
er
D
L
ca
n
s
o
l
v
e
m
a
n
y
co
m
p
le
x
p
r
o
b
le
m
s
o
f
co
m
p
u
t
er
v
is
io
n
w
h
ic
h
w
er
e
n
o
t
ea
s
il
y
s
o
lv
ed
b
y
Ma
c
h
in
e
lear
n
i
n
g
.
I
t
m
ad
e
u
p
o
f
t
w
o
p
r
in
cip
al
s
tag
e
s
:
tr
ain
i
n
g
an
d
T
esti
n
g
.
a)
T
h
e
tr
ain
in
g
s
ta
g
es
-
lab
eli
n
g
p
r
o
ce
s
s
o
f
lar
g
e
a
m
o
u
n
t
s
o
f
d
ata,
an
d
d
ef
in
e
t
h
e
id
en
tica
l
f
ea
t
u
r
es.
T
h
e
tr
ain
i
n
g
m
o
d
el
co
m
p
ar
es
t
h
es
e
f
ea
t
u
r
es
a
n
d
k
ee
p
s
t
h
e
m
to
m
ak
e
p
r
ec
is
e
r
e
aso
n
i
n
g
an
d
c
o
n
clu
s
io
n
w
h
e
n
i
t
en
co
u
n
ter
s
r
elate
d
d
ata
n
ex
t t
i
m
e.
T
h
e
DL
tr
ai
n
in
g
p
r
o
ce
s
s
e
n
co
m
p
as
s
es t
h
e
f
o
llo
w
i
n
g
s
ta
g
es:
1)
A
N
Ns ex
p
lo
r
e
a
s
et
o
f
b
i
n
ar
y
f
alse/
tr
u
e
q
u
e
s
tio
n
s
o
r
.
2)
E
x
tr
ac
t n
u
m
er
al
v
al
u
es
f
r
o
m
d
ata
b
ar
.
3)
C
las
s
i
f
y
d
ata
co
r
r
esp
o
n
d
to
th
e
p
r
o
p
e
r
r
esp
o
n
s
es g
o
t.
4)
L
ab
elli
n
g
Da
ta.
b)
T
h
e
test
i
n
g
s
ta
g
es
-
lab
el
n
e
w
u
n
e
x
p
o
s
ed
d
ata
u
s
i
n
g
t
h
eir
p
r
ev
io
u
s
k
n
o
w
led
g
e
an
d
th
e
n
m
a
k
e
a
co
n
clu
s
io
n
.
I
n
an
c
ien
t
m
ac
h
in
e
lear
n
in
g
,
n
e
w
f
ea
tu
r
e
e
x
tr
ac
t
io
n
w
ill
b
e
ex
tr
ac
ted
f
r
o
m
a
f
u
n
d
a
m
e
n
tal
d
ata
s
tack
ed
i
n
to
t
h
e
m
ac
h
i
n
e.
An
al
y
s
t
f
o
r
m
u
late
s
M
L
i
n
s
tr
u
cti
o
n
s
a
n
d
co
r
r
ec
ts
th
e
e
r
r
o
r
s
e
n
co
u
n
ter
ed
b
y
a
m
ac
h
in
e.
T
h
is
m
e
th
o
d
o
lo
g
y
w
ip
es
o
u
t
a
n
e
g
ati
v
e
o
v
er
tr
ai
n
i
n
g
i
m
p
ac
t
m
u
c
h
o
f
t
h
e
ti
m
e
b
e
s
ee
n
i
n
d
ee
p
lear
n
in
g
.
I
n
ML
,
t
h
e
an
al
y
s
t
s
u
p
p
lie
s
b
o
th
ex
a
m
p
les
a
n
d
tr
ain
i
n
g
d
ata
to
ass
is
t
t
h
e
s
y
s
te
m
to
m
a
k
e
c
o
r
r
ec
t
d
ec
is
io
n
s
b
y
t
h
e
m
ac
h
in
e
th
is
n
o
r
m
is
k
n
o
w
n
as
s
u
p
er
v
is
ed
lear
n
i
n
g
.
I
n
o
t
h
er
w
o
r
d
s
,
i
n
a
n
a
n
cien
t
m
ac
h
in
e
lear
n
in
g
,
a
co
m
p
u
ter
s
o
l
v
es
a
h
u
g
e
n
u
m
b
er
o
f
ta
s
k
s
,
h
o
w
e
v
er
,
it
ca
n
'
t
m
o
u
n
t
s
u
ch
u
n
d
er
tak
i
n
g
s
w
it
h
o
u
t
h
u
m
a
n
co
n
tr
o
l.
Div
er
s
it
y
b
et
wee
n
m
ac
h
i
n
e
lear
n
i
n
g
an
d
d
ee
p
lear
n
in
g
:
1)
Dee
p
lear
n
in
g
r
eq
u
ir
es
u
n
lab
elled
tr
ain
in
g
d
ata
i
n
h
u
g
e
q
u
an
tit
y
to
m
a
k
e
co
n
ci
s
e
co
n
cl
u
s
io
n
s
w
h
il
e
Ma
ch
i
n
e
L
ea
r
n
i
n
g
w
i
ll u
t
ilize
a
s
m
all
a
m
o
u
n
t o
f
d
ata
f
r
o
m
t
h
e
an
al
y
s
t
.
2)
Ma
ch
i
n
e
lear
n
i
n
g
e
x
p
ec
ts
f
ea
t
u
r
es
to
b
e
p
r
ec
is
el
y
r
ec
o
g
n
ize
d
b
y
a
n
al
y
s
t
w
h
ile
d
ee
p
lear
n
i
n
g
g
e
n
er
ates
n
e
w
f
ea
tu
r
es i
n
d
ep
en
d
en
tl
y
.
3)
Un
li
k
e
m
ac
h
i
n
e
lear
n
i
n
g
,
d
ee
p
lear
n
in
g
n
ee
d
s
GP
U
-
h
ig
h
-
p
er
f
o
r
m
a
n
ce
h
ar
d
w
ar
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:
2
0
8
8
-
8708
Th
e
u
p
s
u
r
g
e
o
f d
ee
p
lea
r
n
in
g
fo
r
co
mp
u
ter visi
o
n
a
p
p
lica
tio
n
s
(
P
r
iya
n
ka
P
a
tel)
547
4)
Ma
ch
i
n
e
lear
n
i
n
g
is
o
late
s
u
n
d
er
tak
in
g
s
i
n
to
li
ttle
p
ar
ts
a
n
d
af
ter
w
ar
d
j
o
in
an
d
co
n
cl
u
d
e
in
to
o
n
e
o
u
tp
u
t
w
h
ile
d
ee
p
lear
n
in
g
s
o
l
v
es t
h
e
p
r
o
b
lem
o
n
t
h
e
e
n
d
-
to
-
en
d
b
asis
.
5)
I
n
co
m
p
ar
is
o
n
w
it
h
Ma
c
h
in
e
l
ea
r
n
in
g
,
d
ee
p
lear
n
in
g
n
ee
d
s
c
o
n
s
id
er
ab
l
y
m
o
r
e
ti
m
e
to
tr
ain
.
6)
Ma
ch
i
n
e
lear
n
i
n
g
g
i
v
es e
n
o
u
g
h
tr
an
s
p
ar
en
c
y
f
o
r
its
d
ec
is
io
n
s
th
a
n
d
ee
p
lear
n
in
g
.
T
h
e
id
ea
o
f
d
ee
p
lear
n
in
g
i
n
f
er
s
th
at
t
h
e
m
ac
h
i
n
e
m
ak
e
s
it
s
u
s
e
f
u
l
n
es
s
in
d
ep
en
d
e
n
t
d
ec
is
io
n
s
f
r
o
m
an
y
o
n
e
else
a
s
lo
n
g
as
i
t
is
co
n
ce
iv
ab
le
at
th
e
p
r
ese
n
t
ti
m
e.
T
o
s
u
r
m
i
s
e,
ap
p
lic
atio
n
s
o
f
d
ee
p
lear
n
in
g
u
tili
ze
a
v
ar
io
u
s
le
v
elled
m
e
th
o
d
o
lo
g
y
in
cl
u
d
i
n
g
d
ec
id
in
g
t
h
e
m
o
s
t
i
m
p
er
ativ
e
attr
ib
u
tes to
lo
o
k
at
.
10.
CO
NCLU
SI
O
N
An
o
b
s
er
v
at
io
n
s
o
f
t
h
i
s
p
ap
er
is
U
n
s
u
p
er
v
is
ed
lear
n
i
n
g
h
ad
b
ee
n
a
m
aj
o
r
ef
f
ec
t
o
n
i
n
r
e
v
i
v
in
g
d
ee
p
lear
n
in
g
,
b
u
t
t
h
e
s
u
p
er
v
is
ed
lear
n
in
g
h
a
s
g
r
ea
t
s
u
cc
es
s
i
n
th
e
1
9
6
0
s
to
2
0
0
0
s
s
o
,
d
ee
p
lear
n
in
g
h
as
s
in
ce
b
ee
n
o
v
er
s
h
ad
o
w
ed
,
b
u
t
t
h
e
n
a
g
ai
n
a
f
ter
2
0
0
0
s
d
ee
p
lear
n
in
g
is
o
n
to
p
to
s
o
lv
e
s
u
p
er
v
is
e
d
an
d
u
n
s
u
p
er
v
i
s
ed
co
m
p
le
x
p
r
o
b
le
m
s
.
A
lt
h
o
u
g
h
w
e
h
av
e
n
o
t
f
o
cu
s
ed
to
r
ev
ie
w
t
h
ese
k
i
n
d
s
o
f
lear
n
i
n
g
.
T
h
e
f
u
tu
r
e
e
x
p
ec
ta
tio
n
f
o
r
d
ee
p
lear
n
in
g
to
b
ec
o
m
e
f
ar
m
o
r
e
i
m
p
o
r
tan
t
i
n
t
h
e
lo
n
g
er
ter
m
.
Sp
ee
ch
r
ec
o
g
n
i
tio
n
,
n
at
u
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
in
g
,
a
n
d
v
id
eo
an
al
y
s
i
s
ar
e
lar
g
el
y
u
n
s
u
p
er
v
i
s
ed
:
w
e
r
ev
ie
w
ed
th
e
d
if
f
er
en
t
m
o
d
els
o
f
d
ee
p
lear
n
i
n
g
to
s
o
lv
e
t
h
ese
k
in
d
s
o
f
p
r
o
b
le
m
s
.
Dee
p
l
ea
r
n
i
n
g
al
g
o
r
ith
m
s
an
d
m
o
d
els
ar
e
tr
ain
ed
e
n
d
-
to
-
en
d
a
n
d
th
a
t
u
s
es
r
ein
f
o
r
ce
m
en
t le
ar
n
i
n
g
to
d
ec
i
d
e
w
h
er
e
to
lo
o
k
.
Dee
p
lear
n
i
n
g
a
n
d
r
ein
f
o
r
ce
m
e
n
t le
ar
n
in
g
alr
ea
d
y
o
u
tp
er
f
o
r
m
in
o
n
e
s
h
o
r
t
clas
s
i
f
icatio
n
an
d
f
ea
t
u
r
e
ex
tr
ac
tio
n
a
n
d
p
r
o
d
u
ce
i
m
p
r
es
s
iv
e
r
es
u
lts
.
Natu
r
al
la
n
g
u
a
g
e
p
r
o
ce
s
s
i
n
g
is
an
o
t
h
er
ar
ea
in
w
h
ich
d
ee
p
lear
n
in
g
i
s
d
o
in
g
lar
g
e
i
m
p
ac
t
o
f
ex
p
er
i
m
e
n
ts
o
v
er
th
e
n
ex
t
f
e
w
y
ea
r
s
.
Sp
ee
c
h
r
ec
o
g
n
itio
n
an
d
v
id
eo
an
a
l
y
s
is
ar
e
also
m
aj
o
r
co
m
p
lex
is
s
u
e
s
i
n
co
m
p
u
ter
v
is
io
n
.
Fo
r
ex
a
m
p
le
N
L
P
p
r
o
b
lem
,
w
e
ex
p
ec
t
a
m
o
d
el
w
h
ic
h
u
s
e
R
ec
u
r
r
en
t
Neu
r
al
Net
w
o
r
k
to
u
n
d
er
s
ta
n
d
d
o
cu
m
en
t
s
c
o
n
ten
t
o
r
s
e
n
t
en
ce
s
w
ill
b
ec
o
m
e
th
e
b
etter
r
u
le
f
o
r
s
elec
tiv
el
y
at
ten
d
i
n
g
to
o
n
e
p
ar
t
at
a
ti
m
e.
E
v
en
t
u
all
y
,
m
aj
o
r
p
r
o
g
r
ess
in
ar
ti
f
icia
l
in
telli
g
e
n
ce
w
ill
co
m
e
ab
o
u
t
t
h
r
o
u
g
h
s
y
s
te
m
s
d
ee
p
l
ea
r
n
i
n
g
.
I
n
s
p
ite
o
f
th
e
f
ac
t
t
h
at
to
p
r
o
d
u
ce
g
o
o
d
r
esu
lt
d
ee
p
lear
n
in
g
h
as b
ee
n
u
s
ed
f
o
r
im
a
g
e,
s
p
ee
ch
,
v
id
eo
an
d
h
an
d
w
r
i
tin
g
r
ec
o
g
n
itio
n
.
RE
F
E
R
E
NC
E
S
[1
]
H.
A
.
S
im
o
n
,
―
A
rti
f
icia
l
in
telli
g
e
n
c
e
:
a
n
e
m
p
iri
c
a
l
sc
ien
c
e
,‖
Arti
fi
c
ia
l
I
n
telli
g
e
n
c
e
,
v
o
l.
77
,
p
p
.
95
-
1
2
7
,
1
9
9
5
.
[2
]
T
.
De
tt
m
e
rs,
T
.
D
e
tt
m
e
rs,
T
.
D
e
t
tm
e
r
s,
E.
S
h
e
lh
a
m
e
r
a
n
d
T
.
De
tt
m
e
rs,
"
De
e
p
L
e
a
rn
in
g
in
a
Nu
tsh
e
ll
:
Histo
ry
a
n
d
T
ra
in
in
g
|
NV
IDIA
De
v
e
lo
p
e
r
Blo
g
"
,
NV
IDIA
De
v
e
lo
p
e
r
Blo
g
,
2
0
1
9
.
[
On
li
n
e
]
.
A
v
a
il
a
b
le:
h
tt
p
s:/
/d
e
v
b
lo
g
s.
n
v
id
ia.co
m
/d
e
e
p
-
lea
rn
in
g
-
nut
s
h
e
ll
-
h
ist
o
ry
-
train
in
g
.
[3
]
h
tt
p
s:/
/www
.
x
e
n
o
n
sta
c
k
.
c
o
m
/b
lo
g
/d
a
ta
-
sc
ien
c
e
/o
v
e
r
v
ie
w
-
of
-
a
rti
f
ic
ial
-
in
telli
g
e
n
c
e
-
a
n
d
-
ro
le
-
of
-
n
a
tu
r
a
l
-
lan
g
u
a
g
e
-
p
ro
c
e
ss
in
g
-
in
-
b
ig
-
d
a
ta
[4
]
h
ttp
s
:/
/i2
.
w
p
.
co
m
/v
i
n
ce
j
ef
f
s
.
c
o
m
/
w
p
-
co
n
ten
t/u
p
lo
ad
s
/2
0
1
7
/0
3
/A
I
_
Au
to
m
ated
_
I
n
telli
g
en
c
e.
p
n
g
[5
]
h
tt
p
s:/
/
ww
w
.
m
a
th
w
o
rk
s.co
m
/d
isc
o
v
e
r
y
/d
e
e
p
-
lea
rn
in
g
.
h
tm
l
.
[6
]
W
.
S
.
M
c
Cu
ll
o
c
h
a
n
d
W
.
P
it
t
s,
―
A
lo
g
ica
l
c
a
lcu
lu
s
o
f
th
e
id
e
a
s
imm
a
n
e
n
t
in
n
e
rv
o
u
s
a
c
ti
v
it
y
,‖
T
h
e
b
u
ll
e
ti
n
o
f
ma
th
e
ma
ti
c
a
l
b
io
p
h
y
sic
s
,
v
o
l.
5
,
p
p
.
1
1
5
-
133
,
1
9
4
3
.
[7
]
D.
O.
He
b
b
,
―
T
h
e
o
rg
a
n
iza
ti
o
n
o
f
b
e
h
a
v
io
r.
A
n
e
u
r
o
p
sy
c
h
o
lo
g
ica
l
th
e
o
ry
,‖
1949
.
[8
]
F.
Ro
se
n
b
latt,
―
T
h
e
p
e
rc
e
p
tro
n
:
a
p
ro
b
a
b
il
isti
c
m
o
d
e
l
f
o
r
in
f
o
r
m
a
ti
o
n
sto
ra
g
e
a
n
d
o
rg
a
n
iza
ti
o
n
in
t
h
e
b
ra
i
n
,‖
Psy
c
h
o
lo
g
ica
l
re
v
iew
,
v
o
l.
65
,
p
p
.
3
8
6
,
1
9
5
8
.
[9
]
B.
W
id
ro
w
,
―
A
d
a
p
ti
v
e
a
d
a
li
n
e
Ne
u
ro
n
Us
in
g
Ch
e
m
ica
l
,‖
me
misto
rs
,
1
9
6
0
.
[1
0
]
Eri
k
ss
o
n
,
K.,
&
Jo
h
n
so
n
,
C.
(1
9
8
8
).
A
n
a
d
a
p
ti
v
e
f
in
it
e
e
le
m
e
n
t
m
e
th
o
d
f
o
r
li
n
e
a
r
e
ll
ip
ti
c
p
r
o
b
lem
s.
M
a
th
e
m
a
ti
c
s
o
f
Co
m
p
u
tatio
n
,
5
0
(1
8
2
)
,
3
6
1
-
3
8
3
.
[1
1
]
M
.
L
.
M
in
sk
i
a
n
d
S
.
A
.
P
a
p
e
rt
,
―
P
e
rc
e
p
tro
n
s:
a
n
in
tro
d
u
c
ti
o
n
to
c
o
m
p
u
tatio
n
a
l
g
e
o
m
e
tr
y
,‖
MA
,
M
I
T
P
re
ss
,
Ca
m
b
rid
g
e
,
1
9
6
9
.
[1
2
]
Ru
m
e
lh
a
rt,
D.
E.
,
Hi
n
t
o
n
,
G
.
E.
,
&
W
il
li
a
m
s,
R.
J.
(1
9
8
8
).
L
e
a
rn
in
g
re
p
re
se
n
tatio
n
s
b
y
b
a
c
k
-
p
ro
p
a
g
a
ti
n
g
e
rro
rs.
Co
g
n
it
iv
e
m
o
d
e
li
n
g
,
5
(3
),
1
.
[1
3
]
J.
J.
Ho
p
f
ield
,
―N
e
u
ra
l
n
e
tw
o
rk
s
a
n
d
p
h
y
sic
a
l
s
y
ste
m
s
w
it
h
e
m
e
rg
e
n
t
c
o
ll
e
c
ti
v
e
c
o
m
p
u
tatio
n
a
l
a
b
il
it
ies
,‖
Pro
c
e
e
d
in
g
s
o
f
t
h
e
n
a
ti
o
n
a
l
a
c
a
d
e
my
o
f
sc
ien
c
e
s
,
v
o
l.
79
,
p
p
.
2
5
5
4
-
2
5
5
8
,
1
9
8
2
.
[1
4
]
D.
E.
Ru
m
e
lh
a
rt,
e
t
a
l
.,
―
L
e
a
rn
in
g
re
p
re
se
n
tatio
n
s
b
y
b
a
c
k
-
p
ro
p
a
g
a
ti
n
g
e
rro
rs
,‖
n
a
tu
re
,
v
o
l.
3
2
3
,
p
p
.
533
,
1
9
8
6
.
[1
5
]
V
a
p
n
ik
a
n
d
L
e
rn
e
r
,
―
In
tr
o
d
u
c
e
t
h
e
G
e
n
e
ra
li
z
e
d
P
o
rtra
it
a
lg
o
rit
h
m
(th
e
a
lg
o
rit
h
m
im
p
le
m
e
n
ted
b
y
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
s is
a
n
o
n
li
n
e
a
r
g
e
n
e
ra
li
z
a
ti
o
n
o
f
th
e
G
e
n
e
ra
li
z
e
d
P
o
r
trait
a
lg
o
rit
h
m
)
,
‖
1
9
6
3
.
[1
6
]
V
.
V
a
p
n
ik
a
n
d
A
.
L
e
rn
e
r
,
―
P
a
tt
e
rn
re
c
o
g
n
it
io
n
u
sin
g
g
e
n
e
ra
li
z
e
d
p
o
rtrait
m
e
th
o
d
,‖
Au
to
ma
ti
o
n
a
n
d
Rem
o
te
Co
n
tro
l
,
v
o
l.
2
4
,
pp.
7
7
4
-
7
8
0
,
1
9
6
3
.
[1
7
]
D.
H.
A
c
k
le
y
,
e
t
a
l
.,
―
A
lea
rn
in
g
a
lg
o
rit
h
m
f
o
r
Bo
lt
z
m
a
n
n
m
a
c
h
in
e
s
,‖
Co
g
n
it
ive
sc
ien
c
e
,
v
o
l.
9
,
p
p
.
147
-
1
6
9
,
1
9
8
5
.
[1
8
]
P.
S
m
o
len
sk
y
,
―
In
f
o
r
m
a
ti
o
n
p
ro
c
e
ss
in
g
in
d
y
n
a
m
ica
l
s
y
ste
m
s
:
F
o
u
n
d
a
t
io
n
s
o
f
h
a
r
m
o
n
y
th
e
o
ry
,‖
C
o
lo
ra
d
o
Un
iv
a
t
Bo
u
l
d
e
r
De
p
t
o
f
Co
m
p
u
ter S
c
ien
c
e
,
1
9
8
6
.
[1
9
]
D.
E.
Ru
m
e
lh
a
rt,
e
t
a
l
.,
―
L
e
a
rn
in
g
in
tern
a
l
re
p
re
se
n
tati
o
n
s
b
y
e
rro
r
p
ro
p
a
g
a
ti
o
n
,‖
i
n
R
u
m
e
lh
a
rt
D
.
,
e
t
a
l
.
,
―
P
a
ra
ll
e
l
d
istri
b
u
ted
p
ro
c
e
ss
in
g
:
Ex
p
l
o
ra
ti
o
n
s in
th
e
m
icro
stru
c
tu
re
o
f
c
o
g
n
it
i
o
n
,
‖
F
o
u
n
d
a
ti
o
n
s
,
v
o
l.
1
,
1
9
8
6
.
[2
0
]
R.
J.
W
il
li
a
m
s,
e
t
a
l
.,
―
L
e
a
rn
in
g
re
p
re
se
n
tat
io
n
s
b
y
b
a
c
k
-
p
ro
p
a
g
a
ti
n
g
e
rro
rs
,‖
Na
t
u
re
,
v
o
l.
3
2
3
,
p
p
.
5
3
3
-
5
3
6
,
1
9
8
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.