I
nte
rna
t
io
na
l J
o
urna
l o
f
Ro
bo
t
ics a
nd
Aut
o
m
a
t
io
n
(
I
J
RA
)
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
2
0
1
5
,
p
p
.
2
92
~3
10
I
SS
N:
2089
-
4856
292
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
J
RA
Using
Deep
Lea
r
nin
g
f
o
r
H
u
m
a
n
Co
m
pu
ter
I
nter
face
v
ia
Electroencep
ha
lo
g
ra
phy
Va
m
s
i M
a
ncha
la
,
Sa
ng
ra
m
Red
k
a
r
,
T
o
m
Su
g
a
r
T
h
e
P
o
ly
tec
h
n
ic S
c
h
o
o
l
,
A
rizo
n
a
S
tate
Un
iv
e
rsity
,
US
A
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l
20
,
2
0
1
5
R
ev
i
s
ed
Oct
2
7
,
2
0
1
5
A
cc
ep
ted
No
v
20
,
2
0
1
6
In
th
is
p
a
p
e
r,
se
v
e
ra
l
tec
h
n
iq
u
e
s
u
se
d
to
p
e
rf
o
rm
EE
G
si
g
n
a
l
p
re
-
p
ro
c
e
ss
in
g
,
f
e
a
tu
re
e
x
trac
ti
o
n
a
n
d
sig
n
a
l
c
las
sif
ic
a
t
io
n
h
a
v
e
b
e
e
n
d
isc
u
ss
e
d
,
im
p
le
m
e
n
ted
,
v
a
li
d
a
ted
a
n
d
v
e
rif
ied
;
e
ff
icie
n
t
su
p
e
rv
ise
d
a
n
d
u
n
su
p
e
rv
ise
d
m
a
c
h
in
e
lea
rn
in
g
m
o
d
e
ls,
f
o
r
th
e
EE
G
m
o
to
r
im
a
g
e
r
y
c
las
si
f
i
c
a
ti
o
n
a
re
id
e
n
ti
f
ied
.
Bra
in
Co
m
p
u
ter
In
ter
f
a
c
e
s
a
re
b
e
c
o
m
in
g
th
e
n
e
x
t
g
e
n
e
ra
ti
o
n
c
o
n
tro
ll
e
rs
n
o
t
o
n
ly
in
t
h
e
m
e
d
ica
l
d
e
v
ice
s
f
o
r
d
isa
b
led
i
n
d
iv
id
u
a
l
s
b
u
t
a
lso
in
t
h
e
g
a
m
in
g
a
n
d
e
n
terta
in
m
e
n
t
in
d
u
stries
.
I
n
o
rd
e
r
t
o
b
u
i
ld
a
n
e
f
fe
c
ti
v
e
Bra
in
Co
m
p
u
ter
In
terf
a
c
e
,
it
is
im
p
o
rtan
t
to
h
a
v
e
ro
b
u
st
sig
n
a
l
p
ro
c
e
ss
in
g
a
n
d
m
a
c
h
in
e
lea
rn
in
g
m
o
d
u
les
w
h
ich
o
p
e
ra
te
o
n
th
e
EE
G
sig
n
a
ls
a
n
d
e
sti
m
a
te
th
e
c
u
rre
n
t
th
o
u
g
h
t
o
r
i
n
ten
t
o
f
th
e
u
se
r.
M
o
t
o
r
Im
a
g
e
r
y
(
im
a
g
in
a
r
y
h
a
n
d
a
n
d
leg
m
o
v
e
m
e
n
ts)
sig
n
a
ls
a
re
a
c
q
u
ire
d
u
si
n
g
th
e
Emo
ti
v
EE
G
h
e
a
d
se
t.
T
h
e
sig
n
a
l
h
a
v
e
b
e
e
n
e
x
t
ra
c
ted
a
n
d
su
p
p
li
e
d
to
t
h
e
m
a
c
h
in
e
lea
rn
in
g
(M
L
)
sta
g
e
,
w
h
e
re
in
,
se
v
e
ra
l
M
L
tec
h
n
iq
u
e
s
a
re
a
p
p
li
e
d
a
n
d
v
a
li
d
a
ted
.
T
h
e
p
e
rf
o
r
m
a
n
c
e
s
o
f
v
a
rio
u
s
M
L
tec
h
n
i
q
u
e
s
a
re
c
o
m
p
a
re
d
a
n
d
so
m
e
im
p
o
rtan
t
o
b
se
rv
a
ti
o
n
s
a
re
re
p
o
rted
.
F
u
rth
e
r,
De
e
p
L
e
a
rn
in
g
tec
h
n
i
q
u
e
s
li
k
e
a
u
to
e
n
c
o
d
in
g
h
a
v
e
b
e
e
n
u
se
d
to
p
e
rf
o
r
m
u
n
su
p
e
rv
ise
d
f
e
a
tu
re
lea
rn
in
g
.
T
h
e
re
li
a
b
il
it
y
o
f
th
e
fe
a
tu
re
s
is
p
re
se
n
ted
a
n
d
a
n
a
ly
z
e
d
b
y
p
e
rf
o
r
m
in
g
c
las
si
f
ica
ti
o
n
b
y
u
sin
g
th
e
M
L
te
c
h
n
iq
u
e
s.
It
is
sh
o
w
n
th
a
t
h
a
n
d
e
n
g
in
e
e
re
d
‘a
d
-
h
o
c
’
f
e
a
tu
re
e
x
trac
ti
o
n
tec
h
n
iq
u
e
s
a
re
les
s
re
li
a
b
le
th
a
n
t
h
e
a
u
to
m
a
ted
(‘De
e
p
L
e
a
rn
in
g
’)
f
e
a
tu
re
lea
rn
in
g
tec
h
n
iq
u
e
s.
A
ll
th
e
f
in
d
i
n
g
s
in
th
is
re
se
a
rc
h
,
c
a
n
b
e
u
se
d
b
y
th
e
BCI
re
se
a
rc
h
c
o
m
m
u
n
it
y
f
o
r
b
u
il
d
in
g
m
o
to
r
im
a
g
e
r
y
b
a
s
e
d
BCI
a
p
p
li
c
a
ti
o
n
s
su
c
h
a
s
Ga
m
in
g
,
Ro
b
o
t
c
o
n
tr
o
l
a
n
d
a
u
to
n
o
m
o
u
s v
e
h
icle
s.
K
ey
w
o
r
d
:
B
eh
av
io
r
C
las
s
i
f
icatio
n
Dee
p
L
ea
r
n
i
n
g
E
lectr
o
en
ce
p
h
alo
g
r
ap
h
y
(
E
E
G)
Mo
to
r
I
m
ag
er
y
Neu
tr
al
Net
w
o
r
k
Co
p
y
rig
h
t
©
201
5
In
s
t
it
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
:
San
g
r
a
m
R
ed
k
ar
,
Dep
ar
te
m
en
t o
f
E
n
g
in
ee
r
i
n
g
,
T
h
e
Po
ly
tec
h
n
ic
Sc
h
o
o
l,
Me
s
a,
AZ
-
8
5
2
1
2
,
USA
.
E
m
ail: s
s
r
ed
k
ar
@
g
m
ail.
co
m
1.
I
NT
RO
DUC
T
I
O
N
R
ec
en
t
ad
v
a
n
ce
m
e
n
t
s
a
n
d
d
is
co
v
er
ies
i
n
t
h
e
ar
ea
s
o
f
b
r
ain
i
m
ag
in
g
a
n
d
co
g
n
iti
v
e
n
eu
r
o
s
c
ien
ce
h
av
e
en
ab
led
r
esear
ch
er
s
to
in
ter
ac
t
d
ir
ec
tly
w
it
h
th
e
h
u
m
an
b
r
ain
.
W
ith
th
e
aid
o
f
th
ese
tech
n
o
lo
g
ie
s
an
d
s
o
p
h
is
ticated
s
e
n
s
o
r
s
,
r
esear
c
h
er
s
a
r
e
ab
le
to
o
b
s
er
v
e
an
d
m
o
n
ito
r
t
h
e
ch
a
n
g
in
g
th
o
u
g
h
t
p
r
o
ce
s
s
i
n
t
h
e
f
o
r
m
o
f
lo
w
p
o
w
er
elec
tr
ical
s
i
g
n
als
.
T
h
ese
s
ig
n
als ar
e
u
s
ed
to
m
a
k
e
b
r
ain
-
co
m
p
u
ter
in
ter
f
ac
es (
B
C
I
s
)
p
o
s
s
ib
le
an
d
d
ev
elo
p
co
m
m
u
n
icat
io
n
s
y
s
te
m
s
i
n
w
h
ich
u
s
er
s
e
x
p
licitl
y
m
an
ip
u
late
t
h
eir
t
h
o
u
g
h
t
p
r
o
ce
s
s
,
in
s
tead
o
f
m
o
to
r
m
o
v
e
m
e
n
ts
,
to
co
n
tr
o
l c
o
m
p
u
t
er
s
o
r
co
m
m
u
n
icatio
n
d
ev
ices.
A
cq
u
ir
in
g
t
h
e
b
r
ain
s
i
g
n
a
ls
,
a
cc
u
r
atel
y
,
is
t
h
e
f
ir
s
t
s
tep
i
n
v
o
lv
ed
in
B
C
I
s
.
I
t
i
s
i
m
p
o
r
ta
n
t
to
h
a
v
e
a
co
m
p
lete
k
n
o
w
led
g
e
o
f
t
h
e
p
h
y
s
io
lo
g
y
a
n
d
an
ato
m
y
o
f
h
u
m
an
b
r
a
in
.
T
h
is
w
o
u
ld
b
e
h
e
lp
f
u
l
in
id
en
ti
f
y
i
n
g
th
e
co
r
r
ec
t lo
ca
tio
n
s
o
f
t
h
e
s
e
n
s
o
r
y
n
o
d
es a
n
d
m
ea
s
u
r
e
th
e
r
e
q
u
ir
ed
s
ig
n
als.
E
lectr
o
en
ce
p
h
alo
g
r
ap
h
y
r
ef
er
s
to
th
e
p
h
en
o
m
e
n
o
n
o
f
r
ec
o
r
d
i
n
g
t
h
e
elec
tr
ical
ac
ti
v
it
y
a
lo
n
g
th
e
s
ca
lp
an
d
E
lectr
o
en
ce
p
h
alo
g
r
a
m
(
E
E
G)
is
r
e
f
er
r
ed
to
th
e
r
e
co
r
d
ed
s
ig
n
al
s
an
d
is
t
h
e
m
ea
s
u
r
e
o
f
v
o
lta
g
e
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
I
SS
N:
2089
-
4856
Usi
n
g
Dee
p
Lea
r
n
in
g
fo
r
Hu
m
a
n
C
o
mp
u
ter I
n
terf
a
ce
v
ia
E
le
ctro
en
ce
p
h
a
lo
g
r
a
p
h
y
(
V
a
msi
Ma
n
ch
a
la
)
293
f
l
u
ctu
a
tio
n
s
/v
ar
iatio
n
s
o
cc
u
r
r
ed
d
u
e
to
th
e
f
lo
w
o
f
elec
tr
o
ch
e
m
ical
c
u
r
r
en
ts
i
n
th
e
n
e
u
r
o
n
s
o
f
t
h
e
b
r
ain
.
Du
r
in
g
t
h
e
s
i
g
n
al
r
ec
o
r
d
in
g
p
r
o
ce
d
u
r
e,
elec
tr
o
d
es
c
o
n
s
is
ti
n
g
o
f
s
m
all
m
etal
d
is
c
s
ar
e
p
as
ted
o
v
er
th
e
s
ca
lp
.
T
o
m
ain
tain
p
r
o
p
er
co
n
n
ec
tiv
i
t
y
w
i
th
th
e
ac
tu
al
elec
tr
ical
s
ig
n
als,
t
h
e
s
e
elec
tr
o
d
es a
r
e
w
ett
ed
b
y
a
co
n
d
u
ct
in
g
j
ell
o
r
liq
u
id
.
Ho
w
e
v
er
,
t
h
e
B
C
I
w
o
r
ld
is
n
o
w
s
ee
i
n
g
s
o
m
e
co
m
m
er
cial
d
r
y
E
E
G
h
ea
d
s
e
ts
,
w
h
ic
h
w
o
u
ld
s
er
v
e
th
e
p
u
r
p
o
s
e
o
f
ca
p
t
u
r
in
g
t
h
e
d
ata
an
d
tr
an
s
f
er
r
in
g
to
th
e
C
o
m
p
u
ter
t
h
r
o
u
g
h
w
ir
ele
s
s
m
ed
iu
m
.
P
atter
n
s
o
f
th
e
E
E
G
s
ig
n
al
s
,
d
etec
ted
b
y
th
e
elec
tr
o
d
es,
r
ep
r
esen
t
th
at
t
h
er
e
is
co
n
ti
n
u
o
u
s
ac
tiv
it
y
p
r
ese
n
t
in
t
h
e
h
u
m
a
n
b
r
ain
an
d
th
e
v
ar
y
in
g
i
n
te
n
s
it
ies
o
f
th
e
s
ig
n
al
ar
e
d
eter
m
in
e
d
b
y
th
e
c
h
an
g
i
n
g
m
e
n
tal
a
n
d
p
h
y
s
ical
s
tate
s
o
f
t
h
e
b
o
d
y
.
T
h
ese
in
ten
s
itie
s
o
f
th
e
E
E
G
Sig
n
a
ls
r
ec
o
r
d
ed
o
v
er
th
e
s
u
r
f
ac
e
o
f
t
h
e
b
r
ain
r
an
g
e
f
r
o
m
0
m
icr
o
v
o
lt
s
to
2
0
0
m
icr
o
v
o
lts
.
T
h
e
r
h
y
th
m
ic
ac
ti
v
it
y
o
f
th
e
b
r
ain
s
i
g
n
al
s
is
o
f
te
n
d
iv
id
ed
t
o
d
if
f
er
en
t
b
an
d
s
i
n
ter
m
s
o
f
f
r
eq
u
en
c
y
.
A
lt
h
o
u
g
h
t
h
ese
f
r
eq
u
en
c
y
b
a
n
d
s
ar
e
a
m
atter
o
f
n
o
m
e
n
clat
u
r
e,
th
ese
d
esig
n
atio
n
s
ar
e
u
s
u
all
y
u
s
ed
to
i
m
p
l
y
th
e
f
ac
t
th
at
th
e
r
h
y
t
h
m
ic
a
ctiv
it
y
i
n
a
ce
r
tain
f
r
eq
u
en
c
y
r
an
g
e
is
o
b
s
er
v
ed
d
u
e
to
ce
r
tain
b
io
lo
g
ical
s
ig
n
i
f
ica
n
ce
a
n
d
ar
e
o
f
te
n
n
o
ted
to
h
a
v
e
ce
r
tai
n
d
is
tr
ib
u
ti
o
n
o
v
er
t
h
e
s
ca
lp
.
Fi
g
u
r
e
1
s
h
o
w
s
t
h
e
d
i
f
f
er
en
t
f
r
eq
u
en
c
y
b
a
n
d
s
t
h
e
E
E
G
d
ata
is
d
iv
id
ed
i
n
to
,
an
d
T
ab
le
1
s
h
o
w
s
th
e
s
i
g
n
i
f
ica
n
ce
o
f
th
e
s
e
f
r
eq
u
e
n
c
y
b
an
d
s
an
d
r
elate
d
co
g
n
iti
v
e
tas
k
s
t
h
e
s
e
b
an
d
s
co
r
r
esp
o
n
d
to
.
Fig
u
r
e
1.
Fre
q
u
en
c
y
P
lo
ts
o
f
E
E
G
in
d
if
f
er
en
t
Fre
q
u
en
c
y
R
a
n
g
e
s
.
[
1
]
T
ab
le
1
.
Sig
n
i
f
ican
ce
o
f
E
E
G
in
d
if
f
er
en
t f
r
eq
u
e
n
c
y
b
a
n
d
s
.
Ty
p
e
F
r
e
q
u
e
n
c
y
(
H
z
)
L
o
c
a
t
i
o
n
U
se
D
e
l
t
a
u
p
t
o
4
Ev
e
r
y
w
h
e
r
e
o
c
c
u
r
d
u
r
i
n
g
sl
e
e
p
,
c
o
ma
T
h
e
t
a
4
–
7
H
z
t
e
mp
o
r
a
l
a
n
d
p
a
r
i
e
t
a
l
c
o
r
r
e
l
a
t
e
d
w
i
t
h
e
mo
t
i
o
n
a
l
st
r
e
ss
(
f
r
u
st
r
a
t
i
o
n
&
d
i
sa
p
p
o
i
n
t
me
n
t
)
A
l
p
h
a
8
–
1
2
H
z
o
c
c
i
p
i
t
a
l
a
n
d
p
a
r
i
e
t
a
l
r
e
d
u
c
e
a
m
p
l
i
t
u
d
e
w
i
t
h
se
n
so
r
y
st
i
m
u
l
a
t
i
o
n
o
r
me
n
t
a
l
i
mag
e
r
y
B
e
t
a
1
2
–
3
0
H
z
p
a
r
i
e
t
a
l
a
n
d
f
r
o
n
t
a
l
c
a
n
i
n
c
r
e
a
se
a
m
p
l
i
t
u
d
e
d
u
r
i
n
g
i
n
t
e
n
se
me
n
t
a
l
a
c
t
i
v
i
t
y
Mu
9
-
1
1
H
z
f
r
o
n
t
a
l
(
mo
t
o
r
c
o
r
t
e
x
)
d
i
m
i
n
i
s
h
e
s w
i
t
h
mo
v
e
me
n
t
o
r
i
n
t
e
n
t
i
o
n
o
f
mo
v
e
m
e
n
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4856
IJ
RA
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
2
0
1
5
:
2
92
–
3
1
0
294
10
-
2
0
s
y
s
te
m
i
s
an
i
n
ter
n
at
i
o
n
all
y
ac
ce
p
ted
an
d
p
r
ac
ticed
s
ch
e
m
e
o
f
elec
tr
o
d
e
p
lace
m
en
t
o
n
th
e
h
u
m
a
n
s
ca
lp
.
T
h
e
1
0
an
d
2
0
in
t
h
e
n
a
m
e
r
ef
er
to
th
e
p
e
r
ce
n
tag
e
d
i
s
tan
ce
o
f
n
o
d
es
f
r
o
m
ea
ch
o
th
er
i
n
p
r
o
p
o
r
tio
n
to
th
e
h
ea
d
s
ize.
T
h
e
elec
tr
o
d
e
lo
ca
tio
n
s
s
u
g
g
e
s
t
ed
b
y
th
is
m
et
h
o
d
b
elo
n
g
to
l
o
ca
tio
n
s
o
n
ce
r
eb
r
al
co
r
tex
an
d
th
e
letter
s
F,
T
,
C
,
P
an
d
O
d
en
o
te
th
e
f
r
o
n
tal,
te
m
p
o
r
al,
ce
n
tr
al,
p
ar
ietal
an
d
o
cc
ip
ital
r
esp
ec
tiv
el
y
.
E
x
ce
p
t
f
o
r
th
e
ce
n
tr
al
lo
ca
tio
n
th
e
r
e
m
ai
n
i
n
g
ar
e
all
lo
b
es
o
f
th
e
b
r
ain
.
T
h
e
n
u
m
b
er
s
in
d
icate
th
e
p
o
s
itio
n
o
f
t
h
e
n
o
d
e
o
n
t
h
e
s
c
alp
,
ev
en
n
u
m
b
er
d
en
o
te
r
ig
h
t
s
id
e
o
f
th
e
h
ea
d
,
o
d
d
n
u
m
b
er
d
en
o
te
th
e
le
f
t
s
i
d
e
an
d
Z
i
n
d
icate
s
t
h
at
t
h
e
n
o
d
e
is
lo
ca
ted
o
n
th
e
ce
n
tr
al
l
in
e
o
f
th
e
h
ea
d
.
Fi
g
u
r
e
2
illu
s
tr
ates
th
e
s
e
s
ta
n
d
ar
d
elec
tr
o
d
e
p
o
s
itio
n
s
an
d
Fig
u
r
e
3
illu
s
tr
ated
th
e
k
i
n
d
.
Fig
u
r
e
2
.
Stan
d
ar
d
elec
tr
o
d
e
p
o
s
itio
n
s
a
n
d
p
lace
m
e
n
t o
n
t
h
e
h
u
m
a
n
s
ca
lp
.
[
2
]
B
C
I
is
a
b
r
an
c
h
o
f
H
u
m
a
n
C
o
m
p
u
ter
I
n
ter
f
ac
e
,
w
h
ich
in
v
o
lv
e
s
o
b
tain
i
n
g
th
e
b
r
ai
n
s
i
g
n
als,
co
r
r
esp
o
n
d
in
g
to
s
p
ec
if
ic
f
o
r
m
o
f
th
o
u
g
h
t
s
,
an
d
tr
an
s
lati
n
g
t
h
e
m
to
m
ac
h
i
n
e
c
o
m
m
a
n
d
s
.
I
t
is
a
co
m
m
u
n
icatio
n
s
y
s
te
m
w
h
ic
h
p
er
f
o
r
m
s
t
h
e
tr
an
s
f
er
o
f
m
es
s
ag
e
s
o
r
co
m
m
a
n
d
s
b
y
t
h
e
m
ea
n
s
o
f
h
u
m
a
n
th
o
u
g
h
ts
a
n
d
n
o
t c
o
n
v
e
n
tio
n
al
l
y
b
y
p
er
ip
h
er
al
n
er
v
es.
A
h
ig
h
l
y
a
n
ticip
ated
ap
p
licatio
n
a
m
o
n
g
s
t
th
e
B
C
I
co
m
m
u
n
itie
s
is
th
at
t
h
e
f
u
t
u
r
e
u
s
er
-
co
m
m
u
n
icatio
n
s
y
s
te
m
s
w
o
u
l
d
r
eq
u
ir
e
a
p
ar
allel
f
ee
d
b
ac
k
o
f
t
h
e
u
s
er
m
e
n
tal
s
tate
o
r
in
t
en
tio
n
s
alo
n
g
w
it
h
h
is
p
h
y
s
ical
s
tate.
Fo
r
ex
a
m
p
le,
it
is
i
m
p
o
r
tan
t
f
o
r
t
h
e
au
t
o
m
o
b
ile
to
r
ea
ct
to
t
h
e
u
s
er
’
s
d
r
o
w
s
i
n
es
s
.
T
h
ese
f
u
tu
r
e
ap
p
licatio
n
s
ar
e
ca
lled
s
y
s
te
m
-
s
y
m
b
io
s
i
s
o
r
ef
f
ec
tiv
e
co
m
p
u
ti
n
g
a
n
d
r
eq
u
ir
e
th
e
s
y
s
te
m
s
to
g
at
h
e
r
d
etails
r
eg
ar
d
in
g
m
e
n
tal
s
tates
lik
e
e
m
o
tio
n
s
,
atte
n
tio
n
,
w
o
r
k
lo
ad
,
s
tr
ess
,
f
ati
g
u
e,
etc.
a
n
d
in
ter
p
r
et
th
e
m
.
[
3
]
Fig
u
r
e
3
.
Sh
o
w
in
g
t
h
e
p
h
y
s
io
l
o
g
ical
s
i
g
n
a
ls
ex
p
ec
ted
f
r
o
m
e
ac
h
n
o
d
e
o
f
th
e
1
0
-
2
0
s
y
s
te
m
.
[
4
]
On
li
n
e
an
d
/o
r
o
f
f
li
n
e
ev
a
lu
at
i
o
n
o
f
ap
p
licatio
n
s
u
s
in
g
t
h
e
p
h
y
s
io
lo
g
ical
d
ata
m
ig
h
t
lead
to
s
ev
er
al
in
s
i
g
h
ts
r
eg
ar
d
i
n
g
t
h
e
u
s
er
s
s
t
ate
an
d
h
elp
in
co
m
p
ar
in
g
d
i
f
f
er
en
t
u
s
e
ca
s
e
s
.
Fo
r
in
s
tan
ce
,
a
r
ec
en
t
r
esear
ch
o
n
an
al
y
z
in
g
t
h
e
b
r
ain
i
m
ag
in
g
r
es
u
lt
s
o
f
ce
ll
p
h
o
n
e
u
s
e
d
u
r
i
n
g
d
r
iv
in
g
h
as
p
r
o
v
ed
th
at
e
v
en
h
a
n
d
s
f
r
ee
a
n
d
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
I
SS
N:
2089
-
4856
Usi
n
g
Dee
p
Lea
r
n
in
g
fo
r
Hu
m
a
n
C
o
mp
u
ter I
n
terf
a
ce
v
ia
E
le
ctro
en
ce
p
h
a
lo
g
r
a
p
h
y
(
V
a
msi
Ma
n
ch
a
la
)
295
v
o
ice
ac
tiv
a
ted
u
s
e
o
f
m
o
b
ile
p
h
o
n
e
is
a
s
d
an
g
er
o
u
s
a
s
d
r
u
n
k
en
d
r
i
v
in
g
.
An
o
th
er
r
ec
e
n
t
r
e
s
ea
r
ch
i
n
t
h
is
E
E
G
d
ata
ev
alu
a
tio
n
h
as
b
ee
n
co
n
d
u
cted
b
y
A
r
izo
n
a
Sta
te
Un
i
v
er
s
it
y
,
w
h
ich
f
o
c
u
s
es
t
o
f
in
d
o
u
t
h
o
w
to
lev
er
ag
e
s
o
cial
m
ed
ia
to
i
m
p
r
o
v
e
ed
u
ca
tio
n
al
an
d
tr
ain
in
g
en
v
ir
o
n
m
e
n
t
s
.
T
h
e
g
o
al
o
f
th
is
r
esear
ch
w
as
to
an
al
y
ze
th
e
E
E
G
d
ata
ca
p
t
u
r
e
d
f
r
o
m
s
t
u
d
en
ts
w
h
ile
t
h
e
y
w
e
r
e
u
s
i
n
g
Face
b
o
o
k
an
d
tr
y
m
a
k
e
a
r
ec
o
r
d
o
f
w
h
at
th
e
y
w
er
e
lo
o
k
in
g
at
a
n
d
als
o
th
eir
af
f
ec
ted
s
ta
te,
an
d
u
lt
i
m
atel
y
f
o
r
w
ar
d
t
h
eir
f
i
n
d
i
n
g
s
to
u
s
e
in
o
n
li
n
e
lear
n
in
g
co
m
m
u
n
ities
a
n
d
m
a
k
e
o
n
li
n
e
lear
n
i
n
g
m
o
r
e
in
ter
esti
n
g
f
o
r
th
e
s
t
u
d
en
t
s
.
[
5
]
T
h
e
g
a
m
i
n
g
i
n
d
u
s
tr
y
is
ea
r
n
in
g
m
o
s
t
o
f
its
m
ar
k
et
s
h
ar
e
b
y
m
ak
in
g
u
s
e
o
f
th
e
w
ea
r
ab
le
te
ch
n
o
lo
g
y
.
P
ar
ticu
lar
l
y
,
o
v
er
t
h
e
p
ast
f
e
w
y
ea
r
s
,
n
e
w
g
a
m
e
h
a
v
e
b
ee
n
d
ev
elo
p
ed
b
ased
o
n
th
e
co
m
m
er
ciall
y
av
ai
lab
le
E
E
G
h
ea
d
s
ets
b
y
t
h
e
co
m
p
an
ies
lik
e
Ne
u
r
o
Sk
y
,
E
m
o
ti
v
,
U
n
cle
Milto
n
,
Ma
ttel
an
d
Mi
n
d
Ga
m
es.
T
h
e
u
s
u
a
l
g
a
m
in
g
e
x
p
er
i
en
ce
h
as
b
ee
n
e
n
h
a
n
ce
d
a
n
d
en
r
ic
h
ed
b
y
t
h
e
u
s
e
o
f
B
C
I
s
in
th
e
g
a
m
i
n
g
in
d
u
s
tr
y
.
Fo
r
ex
a
m
p
le,
a
t
y
p
ical
B
C
I
b
ased
g
a
m
e
w
o
u
ld
n
o
lo
n
g
er
b
e
co
n
tr
o
lled
b
y
t
h
e
k
e
y
b
o
ar
d
b
u
t
w
o
u
ld
f
u
n
ctio
n
b
ased
o
n
t
h
e
m
en
tal
s
tate
s
,
i
m
m
er
s
io
n
,
f
lo
w
,
s
u
r
p
r
is
e,
an
d
f
r
u
s
tr
atio
n
etc
.
o
f
th
e
p
la
y
er
.
[
3
]
B
r
ain
C
o
m
p
u
ter
I
n
ter
f
ac
e
s
ar
e
alr
ea
d
y
b
ein
g
u
s
ed
i
n
co
n
tr
o
llin
g
m
an
y
d
e
v
ices
li
k
e
m
o
to
r
i
ze
d
w
h
ee
l
ch
air
s
,
p
r
o
s
th
et
ic
li
m
b
s
,
s
i
m
u
late
m
u
s
c
u
lar
m
o
v
e
m
en
t
,
co
n
tr
o
llin
g
h
o
m
e
ap
p
lian
ce
s
,
lig
h
t
s
,
r
o
o
m
te
m
p
er
atu
r
e,
tele
v
is
io
n
,
o
p
er
atin
g
d
o
o
r
s
,
etc.
T
h
e
n
ee
d
f
o
r
B
r
ain
C
o
m
p
u
ter
I
n
ter
f
ac
e
s
in
t
h
e
e
m
b
ed
d
ed
m
ar
k
et
i
s
b
ein
g
ex
p
lo
r
ed
,
r
ec
en
t
ad
v
a
n
ce
s
i
n
B
C
I
h
av
e
s
ee
n
p
r
o
j
ec
ts
u
s
in
g
o
f
f
th
e
s
h
el
f
E
E
G
h
ea
d
s
ets
a
n
d
e
m
b
ed
d
ed
s
in
g
le
b
o
ar
d
co
m
p
u
ter
s
lik
e
B
ea
g
le
B
o
n
e
B
lack
a
n
d
R
a
s
p
b
er
r
y
P
i.
[
3
]
2.
RE
SE
ARCH
M
E
T
H
O
D
As
d
is
c
u
s
s
ed
ea
r
lier
,
E
E
G
is
a
r
ec
o
r
d
in
g
o
f
t
h
e
b
io
-
p
o
ten
ti
als
f
r
o
m
th
e
s
u
r
f
ac
e
o
f
th
e
s
c
alp
.
Mo
r
e
s
p
ec
if
icall
y
,
t
h
e
s
e
r
ec
o
r
d
in
g
s
ar
e
th
e
elec
tr
o
ch
e
m
ical
p
o
ten
t
ials
m
ea
s
u
r
ed
f
r
o
m
t
h
e
n
e
u
r
o
n
s
a
t
t
h
e
ce
r
eb
r
u
m
o
f
th
e
h
u
m
a
n
b
r
ain
.
Si
n
ce
t
h
ese
s
i
g
n
al
s
ar
e
r
ec
o
r
d
ed
f
r
o
m
t
h
e
s
u
r
f
ac
e
o
f
t
h
e
s
ca
lp
,
it
is
m
o
s
t
lik
e
l
y
t
h
at
p
o
ten
tials
f
r
o
m
m
a
n
y
ce
lls
ar
e
b
ein
g
m
ea
s
u
r
ed
at
t
h
e
s
a
m
e
ti
m
e.
At
f
ir
s
t
g
la
n
ce
,
E
E
G
d
ata
m
a
y
lo
o
k
l
ik
e
a
n
u
n
s
tr
u
ct
u
r
ed
,
n
o
n
-
s
tatio
n
a
r
y
,
n
o
is
y
s
i
g
n
al.
Ho
w
e
v
er
,
ad
v
an
ce
d
s
ig
n
al
p
r
o
ce
s
s
i
n
g
tech
n
iq
u
es
ca
n
b
e
u
s
ed
to
s
ep
ar
ate
d
if
f
er
en
t
co
m
p
o
n
e
n
ts
o
f
t
h
e
b
r
ain
w
a
v
es.
T
h
ese
s
e
p
ar
ate
co
m
p
o
n
e
n
t
s
ca
n
t
h
en
b
e
ass
o
ciate
d
w
it
h
d
if
f
er
e
n
t b
r
ain
ar
ea
s
an
d
f
u
n
ct
io
n
s
.
I
n
o
r
d
er
to
ca
r
r
y
o
n
w
it
h
t
h
e
s
i
g
n
a
l
ac
q
u
is
itio
n
s
ta
g
e,
it
is
i
m
p
o
r
tan
t
to
id
en
ti
f
y
w
h
et
h
e
r
th
e
B
C
I
s
ig
n
al
s
ar
e
g
o
in
g
to
b
e
d
e
p
en
d
en
t
o
r
in
d
ep
en
d
en
t;
h
a
v
e
ev
o
k
ed
o
r
s
p
o
n
tan
eo
u
s
i
n
p
u
t
s
.
I
n
ad
d
itio
n
to
th
ese,
it
is
also
i
m
p
o
r
tan
t
to
d
ec
id
e
o
n
w
h
ic
h
m
et
h
o
d
to
u
s
e
i
n
o
b
tain
i
n
g
t
h
e
s
i
g
n
als
;
a
n
o
n
-
i
n
v
asi
v
e
o
r
i
n
v
a
s
i
v
e.
Ulti
m
a
tel
y
,
i
n
t
h
e
s
i
g
n
al
ac
q
u
is
itio
n
s
ta
g
e,
th
e
s
i
g
n
als
ar
e
o
b
tain
ed
f
r
o
m
t
h
e
elec
tr
o
d
es,
a
m
p
li
f
ied
,
d
ig
i
tized
an
d
m
ad
e
av
ailab
le
f
o
r
th
e
f
u
r
th
er
s
ta
g
es.
I
t
is
al
w
a
y
s
p
o
s
s
ib
le
th
at
t
h
e
ac
q
u
ir
ed
E
E
G
d
ata
i
s
co
m
b
in
ed
w
it
h
a
lo
t
o
f
a
r
ti
f
ac
ts
d
u
e
to
t
h
e
elec
tr
ical
ac
ti
v
it
y
o
f
e
y
e
s
(
E
O
G:
E
lectr
o
o
cc
u
lo
g
r
a
m
)
o
r
m
u
s
cles
(
E
MG
:
E
lectr
o
m
y
o
g
r
a
m
)
.
T
h
e
b
est
w
a
y
to
av
o
id
th
e
s
e
u
n
w
a
n
ted
co
m
p
o
n
en
t
s
is
to
m
ain
tain
id
ea
l
co
n
d
itio
n
s
d
u
r
i
n
g
th
e
s
i
g
n
al
ac
q
u
is
i
tio
n
,
l
ik
e
m
ai
n
tai
n
in
g
a
r
elax
ed
p
o
s
itio
n
w
h
ic
h
w
o
u
ld
in
v
o
l
v
e
m
in
i
m
u
m
o
r
n
o
p
h
y
s
ical
m
o
v
e
m
e
n
ts
.
Ho
w
e
v
er
,
o
n
a
p
r
ac
tical
n
o
te,
m
ai
n
tai
n
i
n
g
s
u
ch
lab
o
r
ato
r
y
co
n
d
it
io
n
s
i
n
e
v
er
y
d
a
y
B
C
I
s
is
n
o
t
r
ea
lizab
l
e
an
d
s
u
c
h
s
y
s
te
m
s
w
h
e
n
u
s
ed
o
u
td
o
o
r
s
to
o
p
e
r
ate
em
b
ed
d
ed
ap
p
licatio
n
s
lik
e
U
GV
o
r
w
h
ee
lc
h
air
,
is
n
o
t
co
n
s
i
d
er
ed
t
o
b
e
r
o
b
u
s
t
an
d
r
eliab
le.
T
h
is
p
r
o
b
lem
ca
n
g
en
er
all
y
b
e
s
o
lv
ed
b
y
ad
o
p
tin
g
e
f
f
ec
ti
v
e
p
r
e
-
p
r
o
ce
s
s
in
g
te
ch
n
iq
u
es
w
h
ic
h
ar
e
r
esp
o
n
s
ib
le
to
c
lean
th
e
s
ig
n
a
l
f
r
o
m
u
n
w
an
ted
ar
tif
ac
ts
a
n
d
/o
r
en
h
a
n
ce
th
e
i
n
f
o
r
m
atio
n
e
m
b
ed
d
ed
i
n
t
h
e
s
e
s
ig
n
al
s
.
I
t
is
o
b
s
er
v
ed
th
a
t
t
h
e
a
m
p
lit
u
d
e
o
f
t
h
ese
m
u
s
c
le
ar
ti
f
ac
ts
is
m
u
c
h
h
i
g
h
er
t
h
an
t
h
e
u
s
u
al
E
E
G
s
i
g
n
a
ls
an
d
d
u
r
in
g
m
o
s
t
o
f
f
l
in
e
a
n
al
y
s
is
th
e
s
e
ca
n
b
e
r
em
o
v
ed
b
y
v
is
u
al
in
s
p
ec
tio
n
.
B
u
t
to
eli
m
i
n
at
e
th
ese
ar
tif
ac
ts
i
n
a
m
o
r
e
ef
f
ec
tiv
e
m
a
n
n
er
it i
s
i
m
p
o
r
ta
n
t to
ap
p
l
y
v
a
r
io
u
s
s
p
at
io
-
s
p
ec
tr
o
-
te
m
p
o
r
al
f
ilter
i
n
g
t
ec
h
n
iq
u
es.
Featu
r
e
ex
tr
ac
tio
n
is
p
h
en
o
m
en
o
n
o
f
b
u
ild
i
n
g
a
f
ea
t
u
r
e
v
e
cto
r
o
f
f
ea
tu
r
es
w
h
ic
h
ar
e
co
n
s
id
er
ed
as
s
u
b
s
et
o
f
d
ata,
d
er
i
v
ed
f
r
o
m
t
h
e
m
ai
n
s
ig
n
als
an
d
,
w
h
ic
h
b
est
d
ef
i
n
e
s
t
h
e
s
ig
n
al
o
f
i
n
ter
est
a
n
d
r
ef
lect
s
t
h
e
s
i
m
ilar
ities
a
n
d
d
if
f
er
en
ce
s
b
e
t
w
ee
n
s
i
g
n
als o
f
s
a
m
e
an
d
d
if
f
er
en
t c
lass
e
s
r
esp
ec
tiv
e
l
y
.
I
d
en
tify
i
n
g
a
n
d
ex
tr
ac
ti
n
g
r
el
ev
an
t
f
ea
t
u
r
es
i
s
o
n
e
o
f
t
h
e
m
o
s
t
i
m
p
o
r
tan
t
s
tep
s
in
a
B
C
I
as
it
is
p
r
o
v
ed
to
b
e
c
r
u
cial
f
o
r
an
ef
f
ec
tiv
e
class
if
icatio
n
s
ta
g
e.
I
f
t
h
e
f
ea
t
u
r
es
ex
tr
ac
ted
f
r
o
m
E
E
G
ar
e
n
o
t
r
elev
an
t
to
th
e
co
r
r
esp
o
n
d
in
g
n
eu
r
o
p
h
y
s
io
lo
g
ical
ac
tio
n
,
it
w
o
u
ld
b
e
v
er
y
d
i
f
f
ic
u
lt
f
o
r
th
e
B
C
I
to
class
i
f
y
t
h
e
tr
ain
in
g
s
ig
n
al
s
i
n
to
th
e
ir
r
esp
ec
tiv
e
c
l
ass
es
a
n
d
h
en
ce
th
e
s
y
s
te
m
wo
u
ld
n
o
t
b
e
p
er
f
o
r
m
i
n
g
e
f
f
ec
ti
v
el
y
d
u
r
i
n
g
th
e
test
p
h
ase.
T
h
u
s
,
e
v
en
i
f
ap
p
l
y
i
n
g
class
if
icatio
n
s
tep
s
o
n
th
e
r
aw
s
i
g
n
als
m
ig
h
t
g
i
v
e
r
e
s
u
l
ts
,
it
w
o
u
ld
b
e
a
s
lo
w
p
r
o
ce
s
s
an
d
it is
r
ec
o
m
m
e
n
d
e
d
to
u
s
e
an
e
f
f
ec
ti
v
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
tec
h
n
iq
u
e
i
n
o
r
d
er
to
m
a
x
i
m
ize
th
e
s
p
ee
d
an
d
ef
f
icie
n
c
y
o
f
th
e
B
C
I
.
Of
te
n
ti
m
es,
i
f
a
lear
n
i
n
g
a
lg
o
r
ith
m
d
o
es
n
o
t
b
eh
a
v
e
as
d
esi
r
ed
it
is
m
o
s
t
li
k
el
y
d
u
e
to
t
h
e
h
i
g
h
b
ia
s
o
r
h
ig
h
v
ar
ian
ce
p
r
o
b
le
m
i
n
t
h
e
s
y
s
te
m
.
H
ig
h
b
ias
is
o
cc
u
r
r
ed
d
u
e
to
u
n
d
er
f
itt
in
g
o
f
th
e
alg
o
r
ith
m
.
T
h
e
b
ias
er
r
o
r
o
f
th
e
s
y
s
te
m
is
attr
ib
u
t
ed
to
i
ts
i
n
ab
ilit
y
to
ap
p
r
o
p
r
iatel
y
c
h
o
o
s
e
th
e
f
u
n
ctio
n
f
,
to
esti
m
ate
lab
el
s
y
o
f
an
in
p
u
t
f
ea
tu
r
e
v
ec
to
r
,
f
r
o
m
all
th
e
p
o
s
s
ib
le
s
et
o
f
m
ap
p
in
g
f
u
n
ctio
n
s
.
On
th
e
o
t
h
er
h
a
n
d
,
a
h
ig
h
v
ar
ia
n
ce
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4856
IJ
RA
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
2
0
1
5
:
2
92
–
3
1
0
296
p
r
o
b
lem
is
ca
u
s
ed
d
u
e
to
o
v
er
f
itti
n
g
o
f
t
h
e
m
ap
p
in
g
f
u
n
ctio
n
.
T
h
is
m
ig
h
t
r
ed
u
ce
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
s
y
s
te
m
w
h
e
n
p
r
o
v
id
ed
w
it
h
n
e
w
te
s
ti
n
g
d
ata.
[
6
]
A
lo
n
g
w
it
h
b
ias
a
n
d
v
ar
ia
n
ce
p
r
o
b
lem
s
,
i
t
is
also
i
m
p
o
r
tan
t
to
u
n
d
er
s
ta
n
d
t
h
e
s
ig
n
i
f
ica
n
c
e
o
f
u
s
i
n
g
cr
o
s
s
-
v
alid
atio
n
i
n
th
e
s
elec
t
io
n
p
r
o
ce
d
u
r
e
o
f
a
Ma
ch
in
e
L
ea
r
n
in
g
m
o
d
el,
to
v
alid
ate
th
e
ex
p
er
im
e
n
tal
r
es
u
lts
.
Valid
atio
n
tech
n
iq
u
es
ar
e
m
o
t
iv
ated
b
y
t
w
o
f
u
n
d
a
m
e
n
tal
a
n
d
m
o
s
t
i
m
p
o
r
tan
t
p
r
o
b
le
m
s
i
n
Ma
ch
i
n
e
L
ea
r
n
i
n
g
:
Mo
d
el
Selectio
n
an
d
P
er
f
o
r
m
a
n
ce
E
s
ti
m
atio
n
.
a.
Mo
d
el
Selectio
n
:
A
l
m
o
s
t
al
w
a
y
s
,
t
h
e
p
er
f
o
r
m
an
ce
o
f
p
atter
n
r
ec
o
g
n
itio
n
a
n
d
th
e
clas
s
i
f
ica
tio
n
tech
n
iq
u
e
s
d
ep
en
d
s
o
n
s
i
n
g
le
/
m
u
lt
ip
le
p
ar
a
m
eter
s
.
Fo
r
in
s
ta
n
ce
,
en
li
s
te
d
b
elo
w
ar
e
s
o
m
e
o
f
th
e
p
ar
a
m
eter
s
u
s
ed
f
o
r
m
o
d
el
s
elec
tio
n
i
n
d
if
f
er
e
n
t c
l
ass
i
f
icatio
n
tech
n
iq
u
e
s
[
7
]
.
b.
No
n
li
n
ea
r
R
eg
r
es
s
io
n
: P
o
ly
n
o
m
ial
s
w
it
h
d
if
f
er
en
t d
eg
r
ee
s
.
c.
K
-
Nea
r
est Ne
ig
h
b
o
r
s
: D
i
f
f
er
e
n
t c
h
o
ice
o
f
K.
d.
Dec
is
io
n
T
r
ee
s
: D
if
f
er
en
t c
h
o
i
ce
s
o
f
n
u
m
b
er
o
f
le
v
el
s
.
e.
SVM:
Di
f
f
er
e
n
t c
h
o
ices o
f
t
h
e
m
is
cla
s
s
i
f
icat
io
n
p
en
alt
y
h
y
p
er
p
ar
am
eter
C.
f.
R
eg
u
lar
ized
Mo
d
els:
Di
f
f
er
e
n
t
ch
o
ices o
f
t
h
e
r
eg
u
lar
izatio
n
p
ar
am
eter
.
g.
Ker
n
el
b
ased
Me
th
o
d
s
: D
i
f
f
er
en
t c
h
o
ice
s
o
f
k
er
n
els.
h.
P
er
f
o
r
m
a
n
ce
E
s
ti
m
atio
n
:
O
n
c
e
th
e
m
o
d
el
is
ch
o
s
e
n
it
is
i
m
p
o
r
tan
t
to
est
i
m
a
te
it
s
p
er
f
o
r
m
an
ce
,
w
h
ich
i
s
t
y
p
icall
y
m
ea
s
u
r
ed
b
y
ev
a
lu
at
i
n
g
t
h
e
tr
u
e
er
r
o
r
r
ate
-
th
e
cla
s
s
if
ier
s
er
r
o
r
r
ate
o
n
th
e
en
tire
d
ata
s
et.
[
7
]
.
i.
Ma
ch
i
n
e
L
ea
r
n
i
n
g
C
la
s
s
i
f
icat
io
n
k
-
Nea
r
e
s
t
Nei
g
h
b
o
r
C
las
s
if
ier
:
k
-
Nea
r
es
t
Nei
g
h
b
o
r
(
k
-
NN)
i
s
s
i
m
p
l
e
an
d
ef
f
ec
ti
v
e
cla
s
s
i
f
ier
.
T
h
e
class
i
f
ier
co
m
p
ar
es
th
e
test
d
ata
w
it
h
t
h
e
tr
ai
n
i
n
g
d
ata.
I
t
ev
al
u
ates
t
h
e
d
is
tan
ce
s
o
f
ea
ch
v
ec
to
r
in
t
h
e
tr
ain
in
g
d
ata
f
o
r
m
t
h
e
test
v
e
cto
r
,
f
in
d
s
k
n
ea
r
est
n
ei
g
h
b
o
r
s
ar
o
u
n
d
th
e
te
s
t
s
a
m
p
le
a
n
d
ass
i
g
n
s
t
h
e
clas
s
l
ab
el
w
h
ich
i
s
f
o
u
n
d
a
m
o
n
g
s
t
m
aj
o
r
ity
o
f
th
e
k
n
ea
r
e
s
t
n
ei
g
h
b
o
r
s
.
T
h
e
b
ias
o
f
th
e
k
-
N
N
alg
o
r
ith
m
i
s
v
er
y
lo
w
s
i
n
ce
it
is
d
ec
id
in
g
b
ased
o
n
th
e
n
ea
r
b
y
p
o
in
t
s
.
Ho
w
e
v
er
,
it
h
as
a
v
er
y
h
ig
h
v
ar
ian
ce
.
So
m
e
o
f
t
h
e
d
is
ta
n
ce
f
u
n
ct
io
n
s
u
s
ed
i
n
t
h
e
k
-
NN
al
g
o
r
ith
m
ar
e
E
au
clid
ea
n
,
Stan
d
ar
d
ized
E
u
clid
ea
n
,
C
it
y
b
lo
ck
,
C
h
eb
y
c
h
ev
,
C
o
s
i
n
e
d
is
ta
n
ce
,
Ma
n
h
at
tan
,
Min
k
o
w
s
k
i,
Ha
m
m
in
g
,
co
r
r
elatio
n
d
is
tan
ce
,
etc.
F
ig
u
r
e
4
s
h
o
w
s
r
eg
io
n
co
n
s
i
s
ti
n
g
o
f
t
h
e
test
s
a
m
p
le
a
n
d
its
n
ea
r
e
s
t
n
eig
h
b
o
r
s
.
Fig
u
r
e
4
.
Sh
o
w
in
g
t
h
e
t
y
p
ical
s
ch
e
m
a
o
f
K
-
NN
[
8
]
j.
L
i
n
ea
r
Dis
cr
i
m
i
n
an
t
A
n
a
l
y
s
is
:
T
h
e
w
o
r
k
i
n
g
p
r
in
cip
le
o
f
L
DA
is
to
m
ak
e
u
s
e
o
f
a
h
y
p
er
-
p
lan
e
w
h
ic
h
s
ep
ar
ates
th
e
s
i
g
n
als
b
elo
n
g
i
n
g
to
d
if
f
er
e
n
t
clas
s
es.
I
n
a
t
w
o
-
clas
s
p
r
o
b
lem
,
t
h
e
t
w
o
cla
s
s
e
s
ar
e
s
ep
ar
ated
b
y
a
h
y
p
er
-
p
la
n
e
an
d
th
e
s
ig
n
als
b
elo
n
g
i
n
g
to
d
if
f
er
en
t
class
es
ar
e
o
n
eit
h
er
s
id
es
o
f
t
h
e
h
y
p
er
-
p
lan
e
.
Si
m
i
lar
to
a
tw
o
-
clas
s
p
r
o
b
le
m
,
d
if
f
er
en
t
s
i
g
n
als
b
elo
n
g
in
g
to
d
if
f
er
en
t
class
e
s
in
a
m
u
lti
-
clas
s
p
r
o
b
lem
ar
e
s
ep
ar
ated
b
y
m
u
ltip
l
e
h
y
p
e
r
-
p
lan
es.
[
9
]
Fig
u
r
e
5
.
L
D
A
h
y
p
er
-
p
la
n
e
[
9
]
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
I
SS
N:
2089
-
4856
Usi
n
g
Dee
p
Lea
r
n
in
g
fo
r
Hu
m
a
n
C
o
mp
u
ter I
n
terf
a
ce
v
ia
E
le
ctro
en
ce
p
h
a
lo
g
r
a
p
h
y
(
V
a
msi
Ma
n
ch
a
la
)
297
L
D
A
g
e
n
er
all
y
a
s
s
u
m
es
a
n
o
r
m
al
d
is
tr
ib
u
tio
n
o
f
t
h
e
d
ata
with
s
a
m
e
co
v
ar
ian
ce
m
atr
ices
f
o
r
all
th
e
s
ig
n
al
s
.
E
ac
h
h
y
p
er
-
p
lan
e
s
ep
ar
atin
g
o
n
e
cla
s
s
f
r
o
m
th
e
o
t
h
er
class
es
is
o
b
tain
ed
b
y
ev
a
lu
atio
n
th
e
p
r
o
j
ec
tio
n
th
at
m
a
x
i
m
ize
s
th
e
d
is
ta
n
ce
b
et
w
ee
n
th
e
m
ea
n
o
f
o
n
e
class
f
r
o
m
th
e
m
ea
n
s
o
f
all
o
th
er
cl
ass
es
a
n
d
m
i
n
i
m
izes
th
e
i
n
ter
clas
s
co
v
ar
ian
ce
.
Fi
g
u
r
e
5
s
h
o
w
s
th
e
s
ep
ar
ati
n
g
p
lan
e
b
et
w
ee
n
t
w
o
c
lass
e
s
.
T
h
e
m
ai
n
ad
v
an
tag
e
s
o
f
th
is
m
et
h
o
d
is
th
at
it
h
a
s
a
v
e
r
y
lo
w
co
m
p
u
tatio
n
a
l
r
eq
u
ir
em
en
ts
an
d
co
m
p
le
x
itie
s
,
w
h
ic
h
m
a
k
es
it
s
u
itab
le
f
o
r
r
ea
l
ti
m
e
e
m
b
ed
d
ed
ap
p
li
ca
tio
n
s
.
Ho
w
e
v
er
th
e
m
ai
n
d
r
a
w
b
ac
k
o
f
t
h
is
m
et
h
o
d
is
th
a
t
it
w
o
u
ld
n
o
t
w
o
r
k
ef
f
ec
tiv
e
l
y
o
n
n
o
n
-
li
n
ea
r
co
m
p
lex
E
E
G
d
ata.
k.
Su
p
p
o
r
t
Vec
to
r
Ma
ch
in
es
(
SV
M)
:
L
ik
e
L
D
A
,
S
VM
is
also
u
s
ed
to
class
i
f
y
s
i
g
n
al
s
in
to
d
if
f
er
en
t
clas
s
es
an
d
id
en
tify
th
e
m
w
h
en
r
eq
u
ir
ed
,
w
it
h
th
e
aid
o
f
a
h
y
p
er
-
p
lan
e.
Ho
w
e
v
er
,
SVM
tr
ies
to
s
o
lv
e
th
e
p
r
o
b
lem
o
f
n
o
n
-
l
in
ea
r
co
m
p
le
x
s
i
g
n
al
s
.
I
n
SVM,
th
e
s
elec
ti
o
n
o
f
th
e
h
y
p
er
-
p
la
n
e
is
m
ad
e
to
m
a
x
i
m
ize
th
e
w
id
th
o
f
t
h
e
b
an
d
w
h
i
ch
s
ep
ar
ates
th
e
n
ea
r
est
tr
ai
n
in
g
p
o
in
ts
to
in
cr
ea
s
e
t
h
e
g
en
er
aliza
tio
n
ca
p
ab
ilit
ies.
[
9
,
1
0
]
T
h
e
h
y
p
er
-
p
lan
e,
also
ca
lled
as
d
ec
is
io
n
b
o
r
d
er
,
s
eg
m
e
n
ts
th
e
f
ea
t
u
r
e
s
p
ac
e
in
to
p
ar
ts
e
q
u
al
to
th
e
n
u
m
b
er
o
f
class
e
s
o
f
t
h
e
s
i
g
n
als.
T
h
e
r
esu
lt
o
f
th
e
clas
s
i
f
ic
atio
n
s
ta
g
e
w
o
u
ld
d
ep
en
d
s
o
n
w
h
ich
p
ar
t
o
f
th
e
p
lan
e
is
t
h
e
test
s
ig
n
al
lo
ca
ted
.
Fig
u
r
e
6
s
h
o
w
s
t
h
e
o
p
ti
m
al
h
y
p
er
-
p
lan
e
s
ep
ar
atin
g
t
w
o
p
la
n
es i
n
SVM.
Fig
u
r
e
6
.
H
y
p
er
-
p
lan
e
a
n
d
s
u
p
p
o
r
t v
ec
to
r
s
[
1
1
]
Dep
en
d
in
g
u
p
o
n
w
h
et
h
er
o
r
n
o
t
th
e
ti
m
e
s
er
ies
s
i
g
n
als
i
s
li
n
ea
r
l
y
s
ep
ar
ab
le,
th
e
SVM
m
et
h
o
d
w
o
u
ld
b
e
ab
le
to
co
n
v
er
t
th
e
d
ata
in
t
o
lin
ea
r
l
y
s
ep
ar
ab
le
an
d
cr
ea
te
n
o
n
li
n
ea
r
d
ec
is
io
n
b
o
u
n
d
ar
i
es
to
class
i
f
y
t
h
e
m
(
Fig
u
r
e
6
)
.
T
h
is
p
h
en
o
m
e
n
o
n
o
f
b
u
ild
i
n
g
n
o
n
-
li
n
ea
r
d
ec
is
io
n
b
o
u
n
d
ar
ies
is
n
o
t
m
u
c
h
co
m
p
lex
a
s
is
m
a
k
i
n
g
th
e
u
s
e
o
f
a
k
er
n
el
tr
ick
to
i
m
p
licitl
y
m
ap
th
e
d
ata
to
an
o
th
er
s
p
ac
e
o
f
h
ig
h
er
d
i
m
e
n
s
io
n
a
lit
y
,
w
h
er
e
th
e
d
ata
is
lin
ea
r
l
y
s
ep
ar
ab
le
an
d
th
e
r
eg
u
lar
li
n
ea
r
clas
s
i
f
ier
s
ar
e
s
t
ill
ap
p
licab
le.
T
h
e
k
er
n
el
g
en
er
all
y
u
s
ed
in
B
C
I
r
esear
ch
is
t
h
e
Gau
s
s
ia
n
k
er
n
e
l:
2
2
,
2
xy
K
x
y
e
x
p
(
4
)
l.
Naïv
e
B
a
y
es c
la
s
s
i
f
ier
:
Naï
v
e
B
ay
e
s
p
r
o
b
ab
ilit
y
f
u
n
ctio
n
is
a
s
f
o
llo
w
s
-
1
12
11
|
|
,
,
.
.
.
,
|
m
l
i
l
i
lm
Nm
q
i
q
qi
p
c
p
x
c
p
c
x
x
x
p
c
p
x
c
(
5
)
W
h
er
e
N
is
th
e
to
tal
n
u
m
b
er
o
f
class
e
s
.
T
h
e
in
d
iv
id
u
al
p
r
o
b
ab
ilit
ies
o
n
th
e
r
ig
h
t
-
h
an
d
s
i
d
e
o
f
th
e
eq
u
atio
n
ar
e
ev
al
u
ated
f
r
o
m
t
h
e
tr
ain
in
g
d
ata
[
1
0
]
.
m.
L
o
g
i
s
tic
r
eg
r
es
s
io
n
u
s
ed
f
o
r
C
las
s
i
f
icatio
n
:
U
n
li
k
e
in
t
h
e
r
eg
r
ess
io
n
p
r
o
b
lem
,
th
e
o
u
tp
u
t
v
alu
e
s
y
o
f
th
e
m
o
d
el
ta
k
e
a
li
m
ited
n
u
m
b
er
o
f
d
is
cr
ete
v
alu
e
s
in
th
e
clas
s
if
icatio
n
p
r
o
b
lem
.
Fo
r
ex
a
m
p
l
e
in
a
b
in
ar
y
class
i
f
icatio
n
th
e
o
u
tp
u
t
y
m
i
g
h
t
e
ith
er
ta
k
e
a
v
al
u
e
o
f
1
o
r
0
d
ep
en
d
in
g
o
n
w
h
et
h
er
o
r
n
o
t
th
e
i
n
p
u
t
f
ea
t
u
r
e
v
ec
to
r
b
elo
n
g
s
to
t
h
e
d
esire
d
class
?
[
1
2
]
Fo
r
lo
g
is
t
ic
r
eg
r
e
s
s
io
n
u
s
ed
f
o
r
cl
ass
i
f
icatio
n
,
a
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4856
IJ
RA
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
2
0
1
5
:
2
92
–
3
1
0
298
s
ig
m
o
id
f
u
n
ct
io
n
is
u
s
ed
as
a
h
y
p
o
th
e
s
is
to
p
r
ed
ict
th
e
o
u
tp
u
t
class
as
t
h
e
o
u
tp
u
t
o
f
a
s
ig
m
o
id
w
o
u
l
d
r
an
g
e
b
et
w
ee
n
0
an
d
1
.
Vec
to
r
s
w
h
ic
h
p
r
o
d
u
ce
o
u
tp
u
t
lo
w
e
r
th
an
0
.
5
w
o
u
ld
b
e
ass
i
g
n
ed
a
0
class
a
n
d
th
e
o
n
es
w
i
th
a
n
o
u
tp
u
t v
al
u
e
m
o
r
e
t
h
a
n
0
.
5
w
o
u
ld
b
e
ass
i
g
n
ed
a
1
,
as sh
o
w
n
in
F
ig
u
r
e
7
.
[
1
2
]
1
1
T
T
x
h
x
g
x
e
(
6
)
()
(
)
1
/
(
1
)
z
g
z
e
(
7
)
Her
e,
(
)
is
ca
lled
t
h
e
lo
g
is
tic
f
u
n
ctio
n
o
r
th
e
s
i
g
m
o
id
f
u
n
ctio
n
a
n
d
ar
e
t
h
e
p
ar
a
m
eter
s
(
als
o
ca
lled
as
w
ei
g
h
ts
)
p
ar
a
m
eter
iz
in
g
t
h
e
s
p
ac
e
o
f
lo
g
is
t
ic
f
u
n
c
ti
o
n
m
ap
p
in
g
X
an
d
Y.
Fig
u
r
e
7
.
Sh
o
w
in
g
t
h
e
lo
g
i
s
tic
f
u
n
ctio
n
.
[
1
2
]
T
h
e
m
ain
f
o
c
u
s
o
f
t
h
e
lo
g
is
tic
r
eg
r
ess
io
n
cla
s
s
i
f
ier
i
s
to
e
v
a
lu
ate
t
h
e
v
al
u
es
o
f
t
h
e
w
e
ig
h
t
s
,
in
an
iter
ativ
e
f
a
s
h
io
n
,
s
o
as
to
r
ed
u
ce
th
e
d
if
f
er
en
ce
b
et
w
ee
n
t
h
e
h
y
p
o
t
h
esi
s
o
f
an
in
p
u
t
f
ea
tu
r
e
v
ec
to
r
an
d
th
e
co
r
r
esp
o
n
d
in
g
o
u
tp
u
t.
T
h
is
is
ac
h
iev
ed
b
y
co
m
p
u
tin
g
th
e
co
s
t
f
u
n
ctio
n
(
)
f
o
r
ev
er
y
s
et
o
f
w
eig
h
t
s
an
d
co
m
p
ar
i
n
g
i
t
w
it
h
th
at
o
b
tai
n
e
d
f
r
o
m
ea
r
lie
r
s
ets o
f
t
h
eta.
[
1
2
]
2
1
1
l
o
g
(
1
)
l
o
g
1
m
i
i
i
i
i
J
y
h
x
y
h
x
m
(
8
)
T
h
e
ef
f
icie
n
c
y
o
f
t
h
e
L
o
g
i
s
ti
c
R
eg
r
ess
io
n
f
u
n
ctio
n
i
n
clas
s
if
y
i
n
g
th
e
co
r
r
ec
t
class
d
ep
e
n
d
s
o
n
th
e
s
elec
tio
n
o
f
t
h
e
d
ata
f
i
tti
n
g
f
u
n
ctio
n
.
T
h
e
f
u
n
c
tio
n
m
i
g
h
t e
i
t
h
er
u
n
d
er
-
f
it o
r
o
v
er
-
f
it th
e
d
a
ta
(
Fig
u
r
e
8
)
.
Fig
u
r
e
8
.
Sh
o
w
in
g
d
i
f
f
er
e
n
t k
i
n
d
s
o
f
f
u
n
ctio
n
s
u
s
ed
to
f
it t
h
e
d
ata
[
1
3
]
n.
A
r
ti
f
icial
Ne
u
r
al
Net
w
o
r
k
s
:
A
N
N
is
an
as
s
e
m
b
l
y
o
f
s
e
v
er
al
ar
tif
icial
n
e
u
r
o
n
s
w
h
ic
h
h
a
v
e
ca
p
ab
ilit
y
to
p
r
o
d
u
ce
n
o
n
-
li
n
ea
r
d
ec
is
io
n
b
o
u
n
d
ar
ies
an
d
w
h
en
co
m
b
i
n
e
d
w
it
h
class
if
ier
s
ar
e
ca
p
ab
le
o
f
s
o
lv
i
n
g
th
e
m
u
lti
clas
s
p
r
o
b
le
m
.
A
t
y
p
ic
al
A
N
N
is
co
m
p
o
s
ed
o
f
s
e
v
e
r
al
la
y
er
s
o
f
n
e
u
r
o
n
s
:
a
n
i
n
p
u
t
la
y
er
,
o
n
e
o
r
s
ev
er
al
h
id
d
en
la
y
er
s
an
d
a
n
o
u
tp
u
t
la
y
er
,
t
h
e
n
u
m
b
er
o
f
n
eu
r
o
n
s
in
w
h
ich
ar
e
b
ased
o
n
th
e
n
u
m
b
er
o
f
class
es i
n
t
h
e
p
r
o
b
lem
.
[
1
0
]
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
I
SS
N:
2089
-
4856
Usi
n
g
Dee
p
Lea
r
n
in
g
fo
r
Hu
m
a
n
C
o
mp
u
ter I
n
terf
a
ce
v
ia
E
le
ctro
en
ce
p
h
a
lo
g
r
a
p
h
y
(
V
a
msi
Ma
n
ch
a
la
)
299
Neu
r
al
Net
w
o
r
k
s
b
eh
a
v
e
as
u
n
iv
er
s
al
ap
p
r
o
x
i
m
ate
s
w
h
en
b
u
ilt
o
f
e
n
o
u
g
h
n
e
u
r
o
n
s
an
d
la
y
er
s
as
th
e
y
ca
n
ap
p
r
o
x
i
m
ate
an
y
co
n
ti
n
u
o
u
s
f
u
n
ctio
n
.
An
o
th
er
ad
v
an
t
ag
e
th
at
m
ak
e
t
h
e
Neu
r
al
Net
w
o
r
k
s
f
le
x
ib
le
f
o
r
a
g
r
ea
t n
u
m
b
er
o
f
p
r
o
b
le
m
s
i
s
t
h
at
th
e
y
ca
n
class
if
y
a
n
y
n
u
m
b
er
o
f
class
es.
T
h
e
in
tu
i
tio
n
f
o
r
Neu
r
al
Ne
t
w
o
r
k
s
ca
n
b
e
b
u
ilt
o
v
er
t
h
e
u
n
d
er
s
ta
n
d
i
n
g
d
e
v
elo
p
ed
o
n
L
o
g
i
s
tic
R
eg
r
es
s
io
n
,
i
n
t
h
e
p
r
ev
io
u
s
s
e
ctio
n
.
C
o
n
s
id
er
a
s
u
p
er
v
is
ed
l
ea
r
n
in
g
p
r
o
b
le
m
,
p
r
o
v
id
ed
w
i
th
lab
eled
tr
ain
i
n
g
d
ata
,
x
i
y
i
.
Neu
r
al
Net
w
o
r
k
s
g
i
v
e
a
w
a
y
o
f
d
ef
i
n
in
g
a
co
m
p
le
x
,
n
o
n
-
li
n
ea
r
f
o
r
m
o
f
h
y
p
o
t
h
eses
,
Wb
hx
,
w
it
h
p
ar
a
m
eter
s
,
Wb
th
at
o
n
e
ca
n
f
i
t to
o
u
r
d
ata.
Fig
u
r
e
9
.
Sin
g
le
Neu
r
o
n
u
s
ed
i
n
a
NN
[
1
4
]
A
s
i
n
g
le
n
e
u
r
o
n
(
Fi
g
u
r
e
9
)
is
a
b
asic
co
m
p
u
tatio
n
al
u
n
i
t
i
n
a
co
m
p
le
x
NN,
ta
k
es
in
p
u
ts
1
,
2
,
3
.
.
.
.
.
.
x
x
x
an
d
o
u
tp
u
ts
3
1
,
)(
()
T
Wb
i
ii
h
x
W
x
f
x
f
Wb
,
w
h
er
e
:
f
R
R
is
ca
lled
th
e
ac
tiv
a
tio
n
f
u
n
ctio
n
.
Mo
s
t c
ases
i
t is eit
h
e
r
a
s
ig
m
o
id
f
u
n
ctio
n
o
r
a
tan
h
f
u
n
ctio
n
.
()
(
)
1
/
(
1
)
z
f
z
e
(
9
)
(
)
(
)
zz
zz
f
ee
z
t
a
n
h
z
ee
(
1
0
)
A
Neu
r
al
Net
w
o
r
k
is
b
u
i
lt
b
y
co
n
n
ec
ti
n
g
m
u
ltip
le
s
i
m
p
le
n
e
u
r
o
n
s
to
g
et
h
er
to
f
o
r
m
a
co
m
p
lex
n
et
w
o
r
k
.
[
1
4
]
.
Fo
r
ex
a
m
p
le,
Fig
u
r
e
1
0
s
h
o
w
s
a
NN
w
h
ich
i
s
b
u
ilt
w
i
th
o
n
e
i
n
p
u
t
la
y
er
,
o
n
e
h
id
d
en
la
y
er
an
d
an
o
u
tp
u
t la
y
er
,
ca
p
ab
le
o
f
c
lass
i
f
y
in
g
t
w
o
d
if
f
er
en
t a
ctio
n
s
.
Fig
u
r
e
1
0
.
T
y
p
ical
NN.
[
1
4
]
T
h
e
co
m
p
u
ta
tio
n
f
o
r
n
eu
r
al
n
e
t
w
o
r
k
i
n
Fi
g
u
r
e
1
1
is
g
iv
e
n
b
y
(
u
s
i
n
g
t
h
e
n
o
tatio
n
s
as p
er
T
a
b
le
2
)
:
(
2
)
(
1
)
(
1
)
(
1
)
(
1
)
1
1
1
1
1
2
2
1
3
3
1
(
2
)
(
1
)
(
1
)
(
1
)
(
1
)
2
2
1
1
2
2
2
2
3
3
2
(
2
)
(
1
)
(
1
)
(
1
)
(
1
)
1
3
1
1
3
2
2
3
3
3
3
(
3
)
(
2
)
(
2
)
(
2
)
(
2
)
(
2
)
(
2
)
(
2
)
,
1
1
1
1
1
2
2
1
3
3
1
()
()
()
(
)
(
)
wb
a
f
W
x
W
x
W
x
b
a
f
W
x
W
x
W
x
b
a
f
W
x
W
x
W
x
b
h
x
a
f
W
a
W
a
W
a
b
(
1
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4856
IJ
RA
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
2
0
1
5
:
2
92
–
3
1
0
300
Fo
r
f
ix
ed
tr
ain
in
g
s
et
(
1
)
(
1
)
(
)
(
)
{
(
,
y
)
,
.
.
.
.
.
.
(
,
y
)
}
mm
xx
o
f
m
t
r
ain
in
g
e
x
a
m
p
le
s
.
Fo
r
a
s
in
g
le
tr
ain
i
n
g
ex
a
m
p
le
(
,
)
xy
,
th
e
co
s
t f
u
n
ctio
n
i
s
d
ef
i
n
ed
as
-
2
,
1
(
,
;
,
)
(
)
y
2
Wb
J
W
b
x
y
h
x
(
1
2
)
An
d
f
o
r
a
tr
ain
i
n
g
s
et
o
f
m
s
a
m
p
les,
th
e
o
v
er
all
co
s
t
f
u
n
ct
io
n
wo
u
ld
b
e
-
11
(
)
(
)
(
)
2
1
1
1
1
11
2
(
)
(
)
(
)
2
,
1
1
1
1
1
(
,
)
(
,
;
,
)
(
)
2
11
(
)
y
(
)
22
l
l
l
l
l
l
n
s
s
m
i
i
l
ji
i
l
i
j
n
s
s
m
i
i
l
W
b
ji
i
l
i
j
J
W
b
J
W
b
x
y
W
m
h
x
W
m
(
1
3
)
T
h
e
f
ir
s
t
ter
m
i
n
t
h
e
ab
o
v
e
eq
u
atio
n
is
a
n
a
v
er
ag
e
s
u
m
-
of
-
s
q
u
ar
es
er
r
o
r
ter
m
.
T
h
e
s
ec
o
n
d
ter
m
i
s
a
r
eg
u
lar
izatio
n
ter
m
(
also
ca
lle
d
a
w
ei
g
h
t
d
ec
a
y
ter
m
)
t
h
at
te
n
d
s
to
d
ec
r
ea
s
e
th
e
m
a
g
n
i
tu
d
e
o
f
th
e
w
eig
h
t
s
,
an
d
h
elp
s
p
r
ev
en
t
o
v
er
f
i
tti
n
g
.
[
1
4
]
T
h
e
u
lti
m
ate
g
o
al
in
th
e
Ne
u
r
al
Net
w
o
r
k
s
is
to
co
m
e
u
p
w
it
h
t
h
e
b
est
s
et
o
f
p
ar
am
eter
s
(
1
)
(
1
)
(
2
)
(
2
)
(
W
,
b
,
W
,
b
)
,
Wb
,
w
h
ic
h
w
o
u
ld
m
i
n
i
m
ize
th
e
(
,
)
J
W
b
.
T
o
tr
ain
th
e
n
et
w
o
r
k
,
we
w
il
l
in
itia
lize
ea
ch
p
ar
a
m
eter
()
l
ij
W
an
d
ea
ch
()
l
i
b
to
r
an
d
o
m
n
o
n
-
ze
r
o
v
alu
es,
an
d
u
p
d
ate
th
e
()
l
ij
W
an
d
()
l
i
b
f
o
r
ev
er
y
i
ter
atio
n
b
y
ap
p
l
y
in
g
tech
n
iq
u
es
li
k
e
g
r
ad
ien
t
d
es
ce
n
t.
On
e
iter
atio
n
o
f
g
r
ad
ien
t
d
ec
en
t
u
p
d
ated
th
e
p
ar
am
eter
s
as f
o
llo
w
s
:
(
)
(
)
()
(
)
(
)
()
:
(
,
)
:
(
,
)
ll
ij
ij
l
ij
ll
ii
l
i
W
W
J
W
b
W
b
b
J
W
b
b
(
1
4
)
T
ab
le
2
.
No
tatio
n
s
u
s
ed
in
Ne
u
r
al
Net
w
o
r
k
s
(
i
)
(
i
)
(
,
y
)
x
i
t
h
T
r
a
i
n
i
n
g
e
x
a
mp
l
e
,
()
wb
hx
O
u
t
p
u
t
o
f
h
y
p
o
t
h
e
si
s o
n
i
n
p
u
t
x
,
u
s
i
n
g
p
a
r
a
me
t
e
r
s
,
Wb
.
()
l
ij
W
T
h
e
p
a
r
a
me
t
e
r
a
sso
c
i
a
t
e
d
w
i
t
h
t
h
e
c
o
n
n
e
c
t
i
o
n
b
e
t
w
e
e
n
u
n
i
t
j
i
n
l
a
y
e
r
l
,
a
n
d
u
n
i
t
i
i
n
l
a
y
e
r
1
l
.
()
l
i
b
T
h
e
b
i
a
s
t
e
r
m a
sso
c
i
a
t
e
d
w
i
t
h
u
n
i
t
i
i
n
l
a
y
e
r
1
l
.
()
l
i
a
A
c
t
i
v
a
t
i
o
n
o
f
u
n
i
t
i
i
n
l
a
y
e
r
l
o
f
t
h
e
n
e
t
w
o
r
k
.
Fig
u
r
e
1
1
.
Neu
r
al
Net
w
o
r
k
s
h
o
w
i
n
g
i
n
p
u
t,
h
id
d
en
an
d
o
u
tp
u
t la
y
er
s
f
o
r
m
u
l
ti
-
clas
s
class
if
i
ca
tio
n
[
1
4
]
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
I
SS
N:
2089
-
4856
Usi
n
g
Dee
p
Lea
r
n
in
g
fo
r
Hu
m
a
n
C
o
mp
u
ter I
n
terf
a
ce
v
ia
E
le
ctro
en
ce
p
h
a
lo
g
r
a
p
h
y
(
V
a
msi
Ma
n
ch
a
la
)
301
o.
Dee
p
L
ea
r
n
i
n
g
:
Ma
n
y
o
f
t
h
e
f
ea
t
u
r
es
ar
e
d
is
co
v
er
ed
b
y
o
b
s
er
v
atio
n
o
f
r
a
w
d
ata
b
y
m
a
n
y
r
esear
ch
er
s
,
o
v
er
s
ev
er
al
y
ea
r
s
.
T
h
e
ar
ea
o
f
f
ea
t
u
r
e
ex
tr
ac
t
io
n
is
co
n
s
id
er
ed
to
b
e
alm
o
s
t
s
at
u
r
ated
an
d
t
h
e
r
esear
ch
er
s
ar
e
s
ee
in
g
t
h
e
m
s
el
v
es
g
et
tin
g
in
ter
ested
to
ex
p
lo
r
e
m
o
r
e
s
o
p
h
is
ticated
an
d
au
to
m
ated
f
ea
t
u
r
e
ex
tr
ac
tio
n
tech
n
iq
u
es.
Dee
p
L
ea
r
n
i
n
g
i
s
o
n
e
ar
ea
c
u
r
r
en
t
l
y
b
ei
n
g
e
x
p
lo
r
ed
b
y
th
e
Ma
ch
in
e
L
ea
r
n
in
g
r
esear
c
h
co
m
m
u
n
itie
s
to
e
m
u
late
t
h
e
f
ea
tu
r
e
lear
n
i
n
g
an
d
cla
s
s
i
f
ica
tio
n
m
ec
h
a
n
is
m
ta
k
i
n
g
p
lace
in
t
h
e
h
u
m
a
n
b
r
ain
to
u
n
d
er
s
tan
d
th
e
i
n
f
o
r
m
atio
n
it
g
ets
f
r
o
m
d
if
f
er
en
t
n
at
u
r
al
s
en
s
o
r
s
,
b
y
b
r
ea
k
i
n
g
d
o
wn
th
e
co
m
p
le
x
in
f
o
r
m
atio
n
i
n
to
n
e
w
a
n
d
s
i
m
p
le
r
ep
r
esen
tatio
n
s
.
On
e
p
o
ten
tial
u
s
e
o
f
Dee
p
L
ea
r
n
i
n
g
is
u
n
s
u
p
er
v
is
ed
f
ea
t
u
r
e
lear
n
i
n
g
,
w
h
ic
h
tr
ies to
u
n
d
er
s
ta
n
d
th
e
co
m
p
le
x
d
ata
an
d
r
ep
r
esen
t it
in
m
u
ch
le
s
s
c
o
m
p
le
x
it
y
.
Dee
p
L
ea
r
n
in
g
r
ef
er
s
to
a
r
at
h
er
w
id
e
cla
s
s
o
f
m
ac
h
i
n
e
lear
n
in
g
tech
n
iq
u
e
s
a
n
d
ar
ch
itect
u
r
e.
B
ased
o
n
h
o
w
t
h
e
ar
ch
itect
u
r
e
h
a
s
b
ee
n
d
esi
g
n
ed
an
d
it
s
i
n
ten
d
ed
u
s
e,
Dee
p
L
ea
r
n
i
n
g
tech
n
iq
u
es
ca
n
b
e
clas
s
i
f
ied
in
to
th
r
ee
m
aj
o
r
ar
ea
s
.
p.
Dee
p
n
et
w
o
r
k
s
f
o
r
u
n
s
u
p
er
v
i
s
ed
o
r
g
en
er
ati
v
e
lear
n
i
n
g
:
U
s
ed
to
ca
p
tu
r
e
h
i
g
h
-
o
r
d
er
co
r
r
elatio
n
o
f
t
h
e
d
ata
to
an
al
y
ze
p
atter
n
s
a
n
d
s
y
n
th
e
s
ize
t
h
e
m
w
h
e
n
n
o
i
n
f
o
r
m
atio
n
ab
o
u
t th
e
tar
g
et
clas
s
is
a
v
ailab
le.
q.
Dee
p
n
et
w
o
r
k
s
f
o
r
s
u
p
er
v
is
ed
lear
n
in
g
:
T
ar
g
et
lab
el
d
ata
a
r
e
m
ad
e
av
a
i
lab
le
f
o
r
s
u
ch
k
in
d
s
o
f
tech
n
iq
u
e
s
to
d
ir
ec
tly
p
r
o
v
id
e
d
is
cr
i
m
in
at
iv
e
in
f
o
r
m
a
tio
n
f
o
r
p
atter
n
class
i
f
icatio
n
p
u
r
p
o
s
es.
r.
H
y
b
r
id
d
ee
p
n
et
w
o
r
k
s
:
I
t
is
a
b
len
d
o
f
b
o
th
Su
p
er
v
is
ed
an
d
Un
s
u
p
er
v
is
ed
tech
n
iq
u
e
s
to
p
r
o
d
u
ce
h
ig
h
er
class
i
f
icatio
n
r
ates.
Her
ein
,
t
h
e
n
et
w
o
r
k
w
o
r
k
s
w
it
h
u
n
s
u
p
e
r
v
is
ed
an
d
lar
g
e
l
y
g
en
er
at
iv
e
p
r
e
-
tr
ain
i
n
g
to
b
o
o
s
t
th
e
ef
f
ec
ti
v
e
n
ess
o
f
s
u
p
er
v
is
ed
tr
ain
in
g
.
T
h
is
is
p
r
o
ce
d
u
r
e
is
f
o
u
n
d
cr
itical
w
h
e
n
t
h
e
tr
ain
i
n
g
d
ata
ar
e
li
m
ited
.
s.
T
r
ain
in
g
a
Dee
p
Mo
d
el:
Dee
p
Mo
d
els
ar
e
tr
ain
ed
in
a
g
r
ee
d
y
la
y
er
-
w
i
s
e
u
n
s
u
p
er
v
i
s
ed
m
a
n
n
er
.
T
h
i
s
g
r
ee
d
y
la
y
er
-
w
i
s
e
u
n
s
u
p
er
v
is
ed
lear
n
in
g
al
g
o
r
ith
m
f
ir
s
t
s
ta
r
ts
w
it
h
t
h
e
tr
ai
n
i
n
g
o
f
t
h
e
f
ir
s
t
la
y
er
o
f
t
h
e
m
o
d
el
i
n
an
u
n
s
u
p
er
v
is
ed
f
a
s
h
io
n
to
y
ield
an
i
n
itia
l
s
et
o
f
p
ar
am
eter
s
f
o
r
th
e
f
ir
s
t
la
y
er
o
f
th
e
n
et
w
o
r
k
[
1
5
]
.
T
h
e
o
u
tp
u
t
f
o
r
m
th
e
f
ir
s
t
lay
er
is
a
r
ed
u
ce
d
r
ep
r
esen
tatio
n
o
f
th
e
in
p
u
t
an
d
is
s
u
p
p
lied
as
an
in
p
u
t
to
th
e
s
ec
o
n
d
la
y
er
w
h
ich
i
s
s
i
m
ilar
l
y
tr
ai
n
ed
u
s
in
g
th
e
s
a
m
e
u
n
s
u
p
er
v
i
s
ed
alg
o
r
ith
m
,
to
y
ield
th
e
i
n
itial
p
ar
am
eter
s
o
f
t
h
at
la
y
er
.
Ag
ai
n
,
th
e
o
u
tp
u
t
f
o
r
m
t
h
e
s
ec
o
n
d
la
y
er
is
u
s
ed
as
a
n
in
p
u
t
to
tr
a
in
t
h
e
th
ir
d
an
d
th
is
p
r
o
ce
s
s
co
n
tin
u
es
u
n
ti
l
a
ll
th
e
p
ar
a
m
eter
s
o
f
ea
ch
la
y
er
h
av
e
an
i
n
itial
v
al
u
es
w
h
i
ch
ar
e
r
ed
u
ce
d
r
ep
r
esen
tatio
n
s
o
f
th
e
p
r
ev
io
u
s
la
y
er
[
15]
.
Fo
llo
w
i
n
g
t
h
is
u
n
s
u
p
er
v
i
s
ed
p
r
e
-
tr
ain
i
n
g
p
h
ase,
o
f
o
b
tai
n
i
n
g
th
e
i
n
itial
p
ar
a
m
eter
s
o
f
t
h
e
s
tack
ed
n
eu
r
al
n
et
w
o
r
k
,
th
e
co
m
p
le
te
n
et
w
o
r
k
ca
n
t
h
e
n
b
e
f
in
e
-
t
u
n
e
d
b
y
ap
p
l
y
in
g
s
u
p
er
v
i
s
ed
b
ac
k
p
r
o
p
ag
atio
n
in
t
h
e
r
ev
er
s
e
d
ir
ec
tio
n
.
B
ac
k
p
r
o
p
ag
atio
n
is
r
esp
o
n
s
ib
le
to
r
ea
d
j
u
s
t
th
e
w
eig
h
t
s
in
a
n
iter
ativ
e
f
ash
io
n
b
y
tr
y
i
n
g
to
r
ed
u
ce
th
e
er
r
o
r
(
co
s
t
f
u
n
ctio
n
)
b
et
w
ee
n
th
e
tr
u
e
lab
els
a
n
d
t
h
e
lab
els
o
b
tain
ed
f
r
o
m
t
h
e
n
e
t
w
o
r
k
,
d
u
r
in
g
ea
c
h
iter
atio
n
.
A
s
th
e
w
ei
g
h
t
s
ar
e
a
d
j
u
s
ted
to
o
b
tain
th
e
c
lo
s
est
o
u
tp
u
t
lab
els,
t
h
e
i
n
ter
n
al
h
id
d
e
n
u
n
it
s
b
ec
o
m
e
th
e
b
est r
ep
r
esen
tatio
n
s
o
f
th
e
i
n
p
u
t f
ea
tu
r
es
[
1
5
]
.
t.
Au
to
en
co
d
er
:
Au
to
en
co
d
er
s
o
f
f
er
a
m
et
h
o
d
o
f
au
to
m
atic
all
y
lear
n
in
g
f
ea
t
u
r
es
f
r
o
m
u
n
lab
eled
d
ata,
allo
w
i
n
g
f
o
r
u
n
s
u
p
er
v
i
s
ed
lea
r
n
in
g
.
I
t
p
er
f
o
r
m
s
b
ac
k
p
r
o
p
ag
atio
n
w
it
h
o
u
t
an
y
k
n
o
w
led
g
e
o
f
t
h
e
lab
el
s
[
1
6
]
.
A
n
a
u
to
en
co
d
er
is
an
ar
tif
ic
ial
n
e
u
r
al
n
et
w
o
r
k
t
h
at
is
ab
le
to
b
e
tr
ain
ed
in
a
co
m
p
letel
y
u
n
s
u
p
er
v
is
ed
m
a
n
n
er
.
I
n
th
e
u
s
u
al
n
eu
r
al
n
et
w
o
r
k
s
,
lab
eled
d
ata
w
er
e
r
eq
u
ir
ed
to
tr
ain
th
e
n
et
w
o
r
k
u
s
i
n
g
th
e
b
ac
k
p
r
o
p
ag
atio
n
p
h
ase
b
y
f
i
n
e
-
tu
n
i
n
g
t
h
e
i
n
itiall
y
a
s
s
i
g
n
ed
w
ei
g
h
ts
.
W
h
er
ea
s
,
t
h
e
au
to
en
co
d
er
s
p
r
o
v
id
e
th
e
ab
il
it
y
to
lear
n
t
h
e
i
n
f
o
r
m
atio
n
w
i
th
o
u
t
t
h
e
n
ee
d
f
o
r
lab
eled
d
ata.
An
a
u
to
en
co
d
er
n
e
u
r
al
n
et
w
o
r
k
p
er
f
o
r
m
s
b
ac
k
p
r
o
p
ag
atio
n
b
y
s
etti
n
g
t
h
e
tar
g
et
v
alu
es
to
t
h
e
i
n
p
u
t
v
al
u
es.
I
n
o
th
er
w
o
r
d
s
,
a
n
au
to
en
co
d
er
n
eu
r
al
n
et
w
o
r
k
(
s
h
o
w
n
i
n
Fig
u
r
e
1
2
)
,
an
u
n
s
u
p
er
v
is
ed
f
ea
t
u
r
e
lear
n
i
n
g
al
g
o
r
ith
m
t
h
at
tr
ai
n
s
th
e
,
()
wb
hx
s
ettin
g
th
e
tar
g
et
v
a
lu
e
s
to
b
e
eq
u
al
to
th
e
in
p
u
t
s
i.e
.
it u
s
es
(
)
(
)
ii
yx
.
T
h
is
s
tr
u
ct
u
r
e
h
as b
ee
n
p
r
o
v
e
d
to
b
e
u
s
ed
ef
f
ec
ti
v
el
y
i
n
d
i
f
f
er
en
t k
i
n
d
s
o
f
ap
p
licatio
n
s
,
o
n
e
b
ein
g
t
h
e
s
o
lu
tio
n
to
th
e
d
i
m
e
n
s
io
n
alit
y
p
r
o
b
lem
o
f
th
e
E
E
G
d
ata,
wh
er
ein
th
e
i
n
ter
m
ed
iar
y
ac
t
iv
atio
n
v
a
lu
e
s
o
f
th
e
h
id
d
en
la
y
er
ca
n
b
e
p
ass
ed
a
s
f
ea
t
u
r
es
(
w
i
th
r
ed
u
ce
d
d
i
m
e
n
s
io
n
alit
y
)
to
a
s
u
p
er
v
is
ed
lear
n
in
g
a
lg
o
r
it
h
m
.
Fo
r
ex
a
m
p
le,
co
n
s
id
er
an
E
E
G
m
o
to
r
i
m
a
g
er
y
d
ata
s
et
co
n
s
is
ti
n
g
o
f
s
i
n
g
le
tr
ial
s
w
it
h
d
ata
s
p
an
n
ed
o
v
er
5
s
ec
o
n
d
s
w
it
h
1
2
8
Hz
f
r
eq
u
e
n
c
y
.
T
h
e
to
tal
n
u
m
b
er
o
f
f
ea
t
u
r
es
in
a
s
i
n
g
le
tr
ial
ar
e
1
2
8
*
5
6
4
0
w
h
ic
h
is
h
u
g
e
an
d
co
m
p
u
tatio
n
al
l
y
i
n
t
en
s
e
f
o
r
a
n
o
r
m
al
cla
s
s
i
f
icati
o
n
tech
n
iq
u
e
li
k
e
L
D
A
,
S
VM
.
Ho
w
ev
er
,
o
n
e
ca
n
u
s
e
a
h
id
d
en
la
y
er
o
f
2
0
0
n
o
d
es
to
co
n
s
tr
u
c
t
an
a
u
to
en
co
d
er
an
d
th
e
ac
ti
v
atio
n
s
()
l
i
a
f
o
r
ea
ch
t
r
ain
in
g
s
a
m
p
le
ar
e
u
n
iq
u
e
a
n
d
is
to
tall
y
b
ase
d
o
n
th
e
w
ei
g
h
ts
o
f
th
e
n
e
t
wo
r
k
o
b
tain
ed
b
y
tr
ain
in
g
it
u
s
in
g
al
l
th
e
tr
ai
n
in
g
s
a
m
p
les.
B
y
li
m
it
in
g
th
e
n
u
m
b
er
o
f
h
id
d
en
u
n
it
s
,
an
d
p
er
f
o
r
m
i
n
g
th
e
tr
ai
n
in
g
,
au
to
e
n
co
d
er
w
i
ll
r
esu
l
t
in
a
co
m
p
r
es
s
ed
r
ep
r
esen
tatio
n
o
f
t
h
e
d
ata.
T
h
e
ab
o
v
e
d
is
cu
s
s
io
n
o
f
b
ein
g
ab
le
to
co
m
e
u
p
w
it
h
a
n
e
w
r
ep
r
esen
tatio
n
o
f
t
h
e
i
n
p
u
t
f
ea
t
u
r
es,
w
it
h
r
ed
u
ce
d
d
im
e
n
s
io
n
alit
y
is
r
ea
l
izab
le
o
n
l
y
i
f
th
e
h
id
d
en
la
y
er
h
as
a
lo
w
er
n
u
m
b
er
o
f
n
o
d
es.
B
u
t
ev
en
w
h
e
n
th
e
n
u
m
b
er
o
f
h
id
d
en
u
n
its
i
s
la
r
g
e,
m
a
y
b
e
g
r
ea
ter
t
h
an
t
h
e
n
u
m
b
er
o
f
in
p
u
t,
o
n
e
ca
n
s
till
co
m
e
u
p
w
it
h
in
ter
esti
n
g
f
ea
tu
r
es
b
y
i
m
p
o
s
i
n
g
o
t
h
er
co
n
s
tr
ai
n
t
s
o
n
t
h
e
n
e
t
w
o
r
k
[
1
4
]
.
On
e
w
a
y
to
ac
h
ie
v
e
th
is
i
s
to
i
m
p
o
s
e
s
p
ar
s
it
y
co
n
s
tr
ai
n
t
o
n
t
h
e
h
i
d
d
en
u
n
i
ts
.
“S
p
a
r
s
ity
is
a
v
ery
u
s
efu
l
p
r
o
p
ert
y
o
f
s
o
me
Ma
ch
in
e
Lea
r
n
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.