T
E
L
KO
M
NIK
A
, V
ol
.
17
,
No.
5,
O
c
tob
er
20
1
9,
p
p.2
57
2
~
25
8
6
IS
S
N: 1
69
3
-
6
93
0
,
accr
ed
ited
F
irst
Gr
ad
e b
y K
em
en
r
istekdikti,
Decr
ee
No: 2
1/E/
K
P
T
/20
18
DOI:
10.12928/TE
LK
OM
N
IK
A
.v
1
7
i
5
.
10202
◼
25
72
Rec
ei
v
ed
J
un
e
5
, 2
01
8
; Re
v
i
s
ed
F
eb
r
u
ary
10
,
20
1
9
;
A
c
c
ep
ted
Ma
r
c
h 1
2
, 2
0
1
9
Und
ers
tan
din
g
user in
ten
tio
n i
n im
a
ge retrie
v
a
l
:
g
eneraliz
a
tio
n s
electio
n
u
sing
mu
lt
iple c
on
c
ep
t h
ierar
c
hies
A
b
d
e
lmad
jid Y
o
u
ce
f
a
*
1
,
M
o
h
amm
ed L
amin
e Kh
er
f
i
2
, Bel
al
Kh
a
ldi
3
, O
u
ss
am
a
A
iadi
4
1
L
AG
E L
a
b
o
r
a
to
ry
,
Uni
v
e
r
s
i
t
é
Ka
s
d
i
M
a
rb
a
h
O
u
a
rg
l
a
,
Al
g
e
ri
a
2
L
AM
IA
L
a
b
o
ra
to
ry
,
Uni
v
e
r
s
i
t
é
d
u
Qué
b
e
c
à
T
ro
i
s
-
Riv
i
è
r
e
s
,
3
3
5
1
,
b
o
u
l
.
d
e
s
F
o
r
g
e
s
,
C.P.
5
0
0
,
T
ro
i
s
-
Riv
i
è
re
s
,
Ca
n
a
d
a
2
,3
,4
L
INAT
I
L
a
b
o
ra
to
ry
,
Un
i
v
e
rs
i
té
Ka
s
d
i
M
a
rb
a
h
Ou
a
rg
l
a
,
Al
g
e
ri
a
*C
o
rre
s
p
o
n
d
i
n
g
a
u
th
o
r,
e
-
m
a
i
l
:
a
b
d
e
l
m
a
d
j
i
d
y
o
u
c
e
fa
@g
m
a
i
l
.c
o
m
Ab
strac
t
Im
a
g
e
r
e
tri
e
v
a
l
i
s
t
h
e
te
c
h
n
i
q
u
e
th
a
t
h
e
l
p
s
u
s
e
r
s
t
o
f
i
n
d
a
n
d
re
tr
i
e
v
e
d
e
s
i
re
d
i
m
a
g
e
s
fro
m
a
h
u
g
e
i
m
a
g
e
d
a
ta
b
a
s
e
.
Th
e
u
s
e
r
h
a
s
f
i
rs
t
l
y
to
fo
rm
u
l
a
te
a
q
u
e
ry
th
a
t
e
x
p
r
e
s
s
e
s
h
i
s
/
h
e
r
n
e
e
d
s
.
T
h
i
s
q
u
e
ry
m
a
y
a
p
p
e
a
r
i
n
te
x
tu
a
l
fo
rm
a
s
i
n
s
e
m
a
n
ti
c
re
tri
e
v
a
l
(SR),
i
n
v
i
s
u
a
l
e
x
a
m
p
l
e
f
o
rm
a
s
i
n
q
u
e
ry
b
y
v
i
s
u
a
l
e
x
a
m
p
l
e
(Q
BVE),
o
r
a
s
a
c
o
m
b
i
n
a
ti
o
n
o
f
t
h
e
s
e
two
fo
rm
s
n
a
m
e
d
q
u
e
r
y
b
y
s
e
m
a
n
ti
c
e
x
a
m
p
l
e
(Q
BSE).
Th
e
fo
c
u
s
o
f
th
i
s
p
a
p
e
r
l
i
e
s
i
n
th
e
te
c
h
n
i
q
u
e
s
o
f
a
n
a
l
y
z
i
n
g
q
u
e
r
i
e
s
c
o
m
p
o
s
e
d
o
f
m
u
l
ti
p
l
e
s
e
m
a
n
ti
c
e
x
a
m
p
l
e
s
.
Th
i
s
i
s
a
v
e
r
y
c
h
a
l
l
e
n
g
i
n
g
ta
s
k
d
u
e
to
th
e
d
i
ff
e
re
n
t
i
n
te
r
p
re
ta
t
i
o
n
s
th
a
t
c
a
n
b
e
d
ra
wn
fro
m
th
e
s
a
m
e
q
u
e
r
y
.
T
o
s
o
l
v
e
s
u
c
h
a
p
ro
b
l
e
m
,
we
i
n
tro
d
u
c
e
a
m
o
d
e
l
b
a
s
e
d
o
n
B
a
y
e
s
i
a
n
g
e
n
e
ra
l
i
z
a
ti
o
n
.
In
c
o
g
n
i
t
i
v
e
s
c
i
e
n
c
e
,
B
a
y
e
s
i
a
n
g
e
n
e
r
a
l
i
z
a
ti
o
n
,
whi
c
h
i
s
th
e
b
a
s
e
o
f
m
o
s
t
wor
k
s
i
n
l
i
te
r
a
t
u
re
,
i
s
a
m
e
th
o
d
th
a
t
tr
i
e
s
to
fi
n
d
,
i
n
o
n
e
h
i
e
ra
r
c
h
y
o
f
c
o
n
c
e
p
ts
,
th
e
p
a
r
e
n
t
c
o
n
c
e
p
t
o
f
a
g
i
v
e
n
s
e
t
o
f
c
o
n
c
e
p
ts
.
In
a
d
d
i
ti
o
n
a
n
d
i
n
s
te
a
d
o
f
u
s
i
n
g
o
n
e
s
i
n
g
l
e
c
o
n
c
e
p
t
h
i
e
ra
r
c
h
y
,
we
p
ro
p
o
s
e
a
g
e
n
e
r
a
l
i
z
a
ti
o
n
s
o
i
t
c
a
n
b
e
u
s
e
d
wit
h
m
u
l
ti
p
l
e
h
i
e
ra
r
c
h
i
e
s
wh
e
re
e
a
c
h
o
n
e
h
a
s
a
d
i
ff
e
r
e
n
t
s
e
m
a
n
ti
c
c
o
n
te
x
t
a
n
d
c
o
n
t
a
i
n
s
s
e
v
e
ra
l
a
b
s
tr
a
c
t
i
o
n
l
e
v
e
l
s
.
O
u
r
m
e
t
h
o
d
c
o
n
s
i
s
t
s
i
n
fi
n
d
i
n
g
t
h
e
o
p
t
i
m
a
l
g
e
n
e
r
a
l
i
z
a
ti
o
n
b
y
,
fi
r
s
tl
y
,
d
e
t
e
rm
i
n
i
n
g
th
e
a
p
p
ro
p
ri
a
te
c
o
n
c
e
p
t
h
i
e
r
a
rc
h
y
,
a
n
d
t
h
e
n
d
e
te
rm
i
n
i
n
g
th
e
a
p
p
ro
p
r
i
a
te
l
e
v
e
l
o
f
g
e
n
e
ra
l
i
z
a
ti
o
n
.
Ex
p
e
ri
m
e
n
ta
l
e
v
a
l
u
a
t
i
o
n
s
d
e
m
o
n
s
tr
a
te
t
h
a
t
o
u
r
m
e
th
o
d
,
whi
c
h
u
s
e
s
m
u
l
t
i
p
l
e
h
i
e
ra
r
c
h
i
e
s
,
y
i
e
l
d
s
b
e
tt
e
r
re
s
u
l
t
s
th
a
n
t
h
o
s
e
u
s
i
n
g
o
n
l
y
o
n
e
s
i
n
g
l
e
h
i
e
ra
r
c
h
y
.
Key
w
ords
:
b
a
y
e
s
i
a
n
m
o
d
e
l
s
o
f
g
e
n
e
r
a
l
i
z
a
ti
o
n
,
c
o
n
c
e
p
t
h
i
e
ra
r
c
h
y
,
g
e
n
e
r
a
l
i
z
a
ti
o
n
o
f
c
o
n
c
e
p
t
s
,
i
m
a
g
e
re
tri
e
v
a
l
,
q
u
e
r
y
e
x
p
a
n
s
i
o
n
,
u
s
e
r
i
n
te
n
ti
o
n
Copy
righ
t
©
2
0
1
9
Uni
v
e
rsi
t
a
s
Ahm
a
d
D
a
hl
a
n.
All
rig
ht
s
r
e
s
e
rve
d
.
1.
Int
r
o
d
u
ctio
n
Due
to
the
ex
p
l
os
i
v
e
gro
wt
h
i
n
di
gi
t
al
i
m
ag
es
,
t
he
r
e
h
as
be
en
an
i
nc
r
ea
s
i
n
g
i
nt
e
r
es
t
i
n
de
v
el
op
i
ng
t
ec
hn
i
qu
es
to
he
l
p
us
ers
r
etri
e
v
i
ng
th
ei
r
de
s
i
r
ed
i
m
ag
es
.
T
he
s
e
t
ec
hn
i
qu
es
ar
e
c
al
l
e
d
“
i
m
ag
e
r
etri
ev
al
”
an
d
the
y
c
a
n
be
c
l
as
s
i
f
i
ed
i
n
t
o
t
w
o
m
ai
n
c
at
eg
or
i
es
w
h
i
c
h
are,
c
on
ten
t
ba
s
ed
i
m
ag
e
r
etri
ev
al
C
B
I
R
[
1
-
4
]
an
d
tex
t
-
ba
s
ed
i
m
ag
e
r
etri
ev
al
T
B
IR
[
5
-
7
]
.C
B
IR
tec
h
ni
qu
es
us
e
th
e
v
i
s
u
al
c
on
t
en
t
i
n
ord
er
t
o
r
etr
i
e
v
e,
f
or
a
gi
v
en
qu
er
y
(
e.
g.,
i
m
ag
e
ex
a
m
pl
e,
s
k
etc
h,
f
ea
ture
v
ec
t
or,
etc
.)
[
8
]
,
t
h
e
s
i
m
i
l
ar
on
es
.
T
hi
s
v
i
s
u
al
c
on
ten
t
c
a
n
be
r
ep
r
es
en
t
e
d
i
n
term
s
o
f
gl
o
ba
l
f
ea
tures
[
9
,
10
]
(
c
ol
or
, s
h
ap
e
, a
n
d t
ex
t
ure)
or l
oc
al
f
ea
tures
[
11
]
(
S
IFT
k
e
y
p
oi
nts
…)
.
Q
ue
r
y
b
y
v
i
s
ua
l
ex
am
pl
e
Q
B
V
E
i
s
o
ne
of
the
m
os
t
us
ed
ap
pro
ac
he
s
i
n
CB
IR.
Ho
wev
er,
the
s
em
an
ti
c
ga
p
b
et
w
e
en
the
l
o
w
-
l
ev
el
v
i
s
ua
l
f
ea
ture
s
an
d
t
he
h
i
g
h
-
l
e
v
e
l
s
em
an
ti
c
m
ea
ni
ng
of
i
m
ag
es
c
au
s
es
a
hi
gh
l
i
m
i
tat
i
o
n
i
n
CB
I
R
pe
r
f
or
m
an
c
e.
T
he
s
em
an
ti
c
ga
p
c
o
ul
d
be
de
f
i
ne
d
as
the
c
on
tr
ad
i
c
ti
o
n
b
et
w
e
en
t
he
h
um
an
j
ud
gm
en
t
an
d
C
B
IR
r
es
u
l
ts
.
In
ot
he
r
w
or
ds
,
the
s
em
an
ti
c
ga
p
i
s
th
e
d
i
s
c
r
ep
an
c
y
be
t
w
e
en
t
w
o
i
nt
erpr
eta
ti
o
ns
,
on
e
of
th
e
us
er
an
d
the
oth
er
of
the
m
ac
hi
ne
[
12
]
.
T
B
IR
tec
hn
i
c
s
us
e
t
ex
t
(
e.g
.,
i
m
ag
e
a
nn
o
tat
i
on
or
tex
t
s
urr
ou
nd
i
ng
i
t)
as
i
m
ag
e
de
s
c
r
i
pto
r
.
Due
to
i
ts
s
i
m
pl
i
c
i
t
y
an
d
r
ap
i
di
t
y
,
T
B
IR
s
ee
m
s
to
be
m
ore
de
s
i
r
a
bl
e
a
nd
prac
t
i
c
al
f
or
us
ers
.
Ho
wev
er,
t
he
qu
al
i
t
y
of
T
B
IR
d
ep
e
nd
s
on
the
qu
al
i
t
y
of
the
an
no
t
ati
on
s
tha
t
are
of
ten
am
bi
gu
ou
s
an
d
i
nc
om
pl
et
e.
F
or
ex
am
pl
e,
the
s
am
e
i
m
ag
e
m
a
y
be
an
no
t
ate
d
wi
th
t
wo
v
er
y
di
f
f
erent
an
no
t
ati
on
s
ba
s
e
d
o
n
t
he
i
n
teres
ts
or
th
e
ps
y
c
h
ol
o
gi
c
a
l
s
t
ate
of
t
he
an
n
ota
t
or.
A
dd
i
ti
on
al
l
y
t
he
a
nn
ot
at
i
o
ns
m
a
y
be
i
nc
om
pl
ete
a
nd
do
no
t
f
ul
l
y
de
s
c
r
i
be
the
c
on
t
en
t
of
the
i
m
ag
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NIK
A
IS
S
N: 1
69
3
-
6
93
0
◼
Unde
r
s
ta
nd
i
ng
us
er i
n
ten
t
i
o
n i
n
i
m
ag
e retr
i
ev
al
:
ge
n
era
l
i
z
at
i
on
… (
A
bd
el
ma
d
j
i
d Y
ou
c
efa
)
2573
In
order
to
el
i
m
i
na
te
the
l
i
m
i
tat
i
on
s
of
Q
B
V
E
an
d
T
B
IR,
a
n
al
ternat
i
v
e
p
arad
i
gm
ha
s
been
prop
os
ed
a
nd
de
n
o
ted
as
q
ue
r
y
b
y
s
em
an
ti
c
ex
am
pl
e
Q
B
S
E
tha
t
c
om
bi
ne
s
bo
th
tec
hn
i
qu
es
[
13
]
.
In
c
urr
en
t
wor
k
,
w
e
are
c
o
nc
erne
d
w
i
t
h
Q
B
S
E
.
I
n
Q
B
S
E
pa
r
a
di
g
m
,
the
qu
er
y
i
s
c
o
m
po
s
ed
of
m
ul
ti
pl
e
i
m
ag
es
,
w
he
r
e
ea
c
h
i
m
ag
e
i
s
l
ab
e
l
ed
wi
th
d
i
f
f
erent
k
e
y
wor
ds
tha
t
de
s
c
r
i
be
t
he
d
i
f
f
erent
v
i
s
u
al
c
on
c
e
pts
w
i
thi
n
th
e
i
m
ag
e
(
e.g
.,
h
ou
s
e,
r
a
i
n,
s
u
ns
et,
etc
.)
A
s
a
qu
er
y
,
t
he
s
y
s
tem
us
es
th
e
k
e
y
w
ords
an
no
ta
ti
ng
the
i
m
ag
es
r
a
the
r
th
a
n
the
i
m
ag
e
s
th
em
s
el
v
es
.
F
urtherm
ore,
an
d
i
n
ord
er
to
ob
t
ai
n
a
be
tte
r
pe
r
f
orm
an
c
e,
the
s
y
s
t
em
s
ho
ul
d
no
t
us
e
the
s
e
c
on
c
ep
ts
as
the
y
ar
e,
i
ns
t
ea
d,
i
t
ha
s
to
ge
n
eral
i
z
e
t
he
m
to
s
o
m
e
c
o
m
m
on
or
m
ore
ge
ne
r
al
c
on
c
ep
ts
(
e.g
.,
t
he
us
er
i
s
l
oo
k
i
ng
f
or
an
i
m
al
s
,
l
an
ds
c
ap
es
,
etc
.)
.
T
he
proc
es
s
of
m
ov
i
ng
f
r
om
a
s
et
of
c
on
c
ep
ts
to
a m
ore c
om
m
on
or gen
eral
c
o
nc
ep
t
i
s
c
al
l
ed
“
g
en
era
l
i
z
ati
on
”
.
Ind
e
ed
,
us
i
ng
qu
erie
s
t
ha
t
are
c
om
po
s
ed
o
f
m
ul
ti
pl
e
i
m
ag
es
(
i
.e.
,
m
ul
ti
pl
e
s
e
m
an
ti
c
ex
am
pl
es
)
c
ou
l
d
s
i
g
ni
f
i
c
an
t
l
y
i
m
prov
e
r
es
ul
ts
.
H
o
w
e
v
er,
f
i
nd
i
ng
t
he
ap
propr
i
ate
ge
ne
r
al
i
z
ati
on
f
or
the
s
e
s
em
an
ti
c
ex
am
pl
es
i
s
a
v
er
y
c
om
pl
i
c
ate
d
t
as
k
.
Rec
en
tl
y
,
m
an
y
s
t
ud
i
es
ha
v
e
be
en
do
ne
tr
y
i
n
g
t
o
u
nd
ers
ta
nd
an
d
s
i
m
ul
ate
ho
w
hu
m
an
s
ge
n
eral
i
z
e.
S
om
e
of
tho
s
e
w
ork
s
ha
v
e
us
ed
m
ac
hi
ne
v
i
s
i
on
te
c
hn
i
q
ue
s
[
14
-
16
]
,
ot
he
r
s
ha
v
e
op
t
ed
f
or
B
a
y
es
i
an
m
od
el
s
of
ge
ne
r
al
i
z
ati
on
[
17
-
21
]
.
T
hu
s
,
a
grea
t
pro
gres
s
ha
s
b
e
en
ac
h
i
e
v
e
d
an
d
g
en
era
l
i
z
ati
o
n
m
eth
od
s
ha
v
e
be
en
pro
po
s
ed
.
S
tart
i
ng
f
r
o
m
on
e
c
on
c
ep
t
h
i
erar
c
h
y
an
d
a
s
et
of
gi
v
en
p
os
i
ti
v
e
c
on
c
ep
ts
,
the
k
e
y
i
de
a
i
s
to
f
i
n
d
th
e
a
pp
r
op
r
i
ate
l
e
v
e
l
t
he
s
e
c
o
nc
ep
ts
n
ee
d
to
be
ge
n
eral
i
z
e
d
to.
A
c
o
nc
ep
t
hi
erar
c
h
y
i
s
m
ad
e
up
of
s
ev
era
l
ab
s
tr
ac
t
i
on
l
ev
el
s
where
ea
c
h
l
ev
el
h
ol
ds
a
s
et
of
c
on
c
ep
ts
,
whi
c
h
are
r
e
pres
en
t
ed
b
y
l
e
af
no
d
es
,
as
F
i
gu
r
e
1
s
ho
w
s
.
Ho
we
v
er,
on
e
s
ho
ul
d
k
no
w
tha
t
the
s
am
e
s
et
of
c
on
c
ep
ts
c
ou
l
d
be
r
ep
r
es
en
ted
b
y
di
f
f
erent
c
o
nc
ep
t
hi
erar
c
hi
es
ba
s
ed
on
the
s
e
l
ec
te
d
c
on
t
ex
t.
F
or
e
x
a
m
pl
e,
a
ni
m
al
s
c
ou
l
d
b
e
c
ate
go
r
i
z
e
d
i
n
a
c
on
c
e
pt
h
i
erar
c
h
y
ba
s
ed
on
th
ei
r
c
l
as
s
es
, regi
on
of
l
ea
v
i
n
g a
n
d
d
i
et,
et
c
.
F
i
gu
r
e
1.
A
n e
x
am
pl
e
of
a
c
on
c
ep
t
hi
erar
c
h
y
an
d
the
c
orr
es
po
nd
i
ng
i
m
ag
es
of
s
o
m
e l
ea
f
no
de
s
(
Im
ag
eNe
t a
s
i
ns
t
an
c
e)
Evaluation Warning : The document was created with Spire.PDF for Python.
◼
IS
S
N: 1
69
3
-
6
93
0
T
E
L
KO
M
NIK
A
V
ol
.
17
,
No
.
5,
O
c
tob
er 20
19
:
257
2
-
25
8
6
2574
Des
pi
t
e t
h
e g
r
ea
t
progr
es
s
ac
hi
e
v
e
d b
y
l
i
teratur
e
w
ork
s
tr
y
i
n
g t
o
i
m
prov
e g
en
era
l
i
z
at
i
on
,
us
i
ng
on
l
y
on
e
c
o
nc
ep
t
h
i
erar
c
h
y
r
es
tr
i
c
ts
m
ac
hi
ne
to
on
l
y
on
e
s
em
an
ti
c
c
on
tex
t.
Ins
t
ea
d,
m
a
c
hi
ne
s
h
ou
l
d
b
e
a
bl
e
t
o
ge
ne
r
a
l
i
z
e
i
n
m
ul
ti
pl
e
c
ate
g
orie
s
(
i
.
e.,
c
o
nte
x
ts
)
a
s
hu
m
an
do
es
.
T
he
ge
n
eral
i
z
ati
on
s
ho
u
l
d
,
the
r
ef
ore,
be
c
arr
i
ed
o
ut
us
i
ng
m
ul
ti
pl
e
c
on
c
ep
t
hi
erar
c
hi
es
.
T
o
m
a
k
e
thi
n
gs
c
l
e
arer
,
l
et
us
tak
e
the
ex
am
pl
e
i
l
l
us
tr
ate
d
i
n
F
i
gu
r
e
2.
H
um
an
c
an
ge
ne
r
al
i
z
e
th
e
c
on
c
ep
ts
E
l
ep
ha
nt
,
Z
eb
r
a
an
d
G
i
r
af
f
e
to
on
e
h
y
p
oth
es
i
s
f
r
o
m
the
f
ol
l
o
wi
ng
h
y
po
th
es
i
s
s
pa
c
e
H
=
{
M
am
m
al
,
A
f
r
i
c
a
an
i
m
al
s
,
Her
bi
v
ores
}
.
A
h
y
p
oth
es
i
s
s
pa
c
e
i
s
a
s
et
of
al
l
the
po
s
s
i
b
l
e
ge
ne
r
al
i
z
ati
on
s
ob
tai
ne
d f
r
om
th
e c
on
c
ep
ts
th
at
c
om
po
s
e t
he
gi
v
e
n q
ue
r
y
.
A
s
s
ho
w
n
i
n
F
i
gu
r
e
2
(
a),
an
i
m
al
s
ha
v
e
b
ee
n
c
at
eg
ori
z
e
d
ac
c
ordi
ng
to
th
ei
r
f
am
i
l
y
,
whereas
i
n
F
i
gu
r
e
2
(
b)
ac
c
ordi
ng
t
o
th
ei
r
d
i
et
an
d
i
n
F
i
gu
r
e
2
(
c
)
ac
c
ordi
ng
to
the
i
r
r
e
gi
o
n
of
l
i
v
i
ng
.
T
hi
s
m
ea
ns
tha
t
the
ge
ne
r
al
i
z
ati
on
i
n
ea
c
h
c
as
e
wi
l
l
be
p
erf
or
m
ed
us
i
ng
d
i
f
ferent
c
on
c
e
pt
hi
erar
c
h
y
.
T
he
r
ef
ore,
f
utu
r
e
wor
k
s
s
ho
ul
d
f
oc
us
on
h
o
w
to
c
om
bi
ne
m
ul
ti
pl
e
c
on
c
ep
t
h
i
erar
c
h
i
es
to
gras
p
hu
m
an
i
nt
en
t
i
on
b
y
d
ete
r
m
i
ni
ng
the
ap
pr
op
r
i
a
te
c
on
t
ex
t a
n
d l
ev
el
of
th
e g
en
era
l
i
z
at
i
o
n.
In
c
l
as
s
i
c
al
t
ec
hn
i
qu
es
,
s
i
m
i
l
ar
i
t
y
be
t
ween
i
m
ag
es
w
as
c
al
c
ul
at
ed
b
as
ed
o
n
the
n
um
be
r
of
c
o
m
m
on
c
on
c
ep
ts
t
ha
t
an
no
tat
e
th
os
e
i
m
ag
es
.
Ho
wev
er,
o
ur
a
pp
r
oa
c
h
i
s
no
t
l
i
m
i
ted
to
th
i
s
na
ï
v
e
tec
h
ni
qu
e.
Ins
tea
d,
i
t
al
s
o
a
na
l
y
s
es
th
e
s
em
an
ti
c
r
el
at
i
o
ns
hi
p
be
t
wee
n
di
ff
erent
i
m
ag
e
c
on
c
ep
ts
.
I
n
c
o
nc
ep
t
hi
erar
c
hi
es
,
a
s
em
an
ti
c
r
e
l
at
i
on
s
hi
p
c
ou
l
d
be
de
f
i
n
ed
as
a
l
i
nk
tha
t
bi
nd
s
two
c
on
c
ep
ts
(
i
.
e.,
f
ath
er/s
on
no
de
s
)
.
Lo
c
at
i
ng
t
he
f
at
he
r
no
d
e
f
or
a
gi
v
en
s
et
of
no
de
s
i
s
c
al
l
e
d
ge
ne
r
al
i
z
ati
on
.
I
t
i
s
a
v
er
y
c
ha
l
l
en
g
i
n
g
tas
k
to
de
t
erm
i
ne
w
h
at
r
e
l
at
i
on
s
h
i
p
as
s
em
bl
es
a
s
et
of
gi
v
en
c
on
c
e
pts
.
F
or
ex
am
pl
e,
are
El
ep
ha
n
t
,
G
i
r
af
f
e
an
d
Z
eb
r
a
Ma
m
m
al
s
or
A
fr
i
c
an
an
i
m
al
s
?
O
ur
ap
pro
ac
h
tr
i
es
t
o
g
en
eral
i
z
e
th
e
qu
er
y
c
on
c
e
pts
b
y
f
i
nd
i
ng
th
e
m
os
t
proba
bl
e
r
el
ati
on
s
hi
p
tha
t
as
s
em
bl
es
the
m
.
In
ad
di
t
i
on
,
w
e
ex
tr
ac
t
th
e
c
on
c
e
pts
tha
t
are
r
e
l
at
ed
to
t
ho
s
e
of
the
qu
er
y
,
whi
c
h
are
c
al
l
e
d “hi
dd
e
n c
o
nc
ep
ts
”
.
T
o
be
tte
r
gras
p
us
er
i
nte
nti
on
,
i
n
th
i
s
pa
p
er,
w
e
pro
po
s
e
a
m
eth
od
tha
t
ge
n
eral
i
z
es
us
er
qu
eri
es
us
i
n
g m
ul
ti
p
l
e c
o
nc
ep
t h
i
erar
c
h
i
es
. I
n o
ur ap
proac
h,
we, f
i
r
s
tl
y
,
tr
y
to
de
t
erm
i
ne
th
e m
os
t
proba
bl
e
c
o
nte
x
t
whi
c
h
c
orr
es
po
nd
s
to
s
o
m
e
c
on
c
ep
t
h
i
erar
c
h
y
.
A
f
t
er
de
term
i
ni
n
g
the
a
pp
r
op
r
i
at
e
c
on
c
ep
t
h
i
erar
c
h
y
,
we
g
en
era
l
i
z
e
the
qu
er
y
c
on
c
ep
ts
an
d
ex
tr
a
c
t
the
h
i
dd
en
c
on
c
ep
ts
i
n
order
t
o
be
u
s
ed
i
n
l
at
ter
r
etri
ev
al
proc
es
s
.
In
ad
di
t
i
on
,
w
e
i
ntro
d
uc
e
t
w
o
ne
w
c
on
c
ep
t
h
i
erar
c
h
i
es
to
be
us
ed
i
n
ou
r
m
eth
od
al
on
g
wi
th
Im
ag
eNe
t.
O
ur
p
ap
er
i
s
organi
z
e
d
as
f
ol
l
o
w
s
.
In
s
ec
ti
on
2,
w
e
o
v
erv
i
e
w
t
he
r
el
ate
d
w
ork
.
In
s
ec
ti
on
3
,
we
de
s
c
r
i
b
e
ou
r
propos
e
d
s
ol
ut
i
on
.
S
ec
t
i
on
4
s
ho
w
s
de
ta
i
l
s
of
the
ex
p
erim
en
tat
i
on
a
nd
t
he
o
bta
i
ne
d
r
es
ul
ts
.
F
i
na
l
l
y
,
w
e
dra
w
s
om
e
c
on
c
l
us
i
o
ns
.
F
i
gu
r
e
2
. C
on
c
ep
ts
m
a
y
b
e
c
ate
go
r
i
z
ed
i
n
di
f
f
erent
w
a
y
s
ba
s
e
d o
n
th
e
s
el
ec
t
ed
c
on
tex
t:
(
a)
a
ni
m
al
s
c
ate
g
ori
z
ed
ac
c
ordi
n
g t
o
the
i
r
f
am
i
l
y
, (b)
a
n
i
m
al
s
c
ate
go
r
i
z
ed
ac
c
ordi
n
g t
o t
he
i
r
di
et,
(c
)
a
ni
m
al
s
c
ate
go
r
i
z
e
d a
c
c
ordi
n
g t
o
the
i
r
l
i
v
i
n
g reg
i
on
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NIK
A
IS
S
N: 1
69
3
-
6
93
0
◼
Unde
r
s
ta
nd
i
ng
us
er i
n
ten
t
i
o
n i
n
i
m
ag
e retr
i
ev
al
:
ge
n
era
l
i
z
at
i
on
… (
A
bd
el
ma
d
j
i
d Y
ou
c
efa
)
2575
2.
Rel
ated
W
o
r
k
Ma
n
y
l
i
t
erature
r
es
ea
r
c
he
s
i
n c
o
gn
i
ti
v
e s
c
i
en
c
e
ha
v
e
at
tem
pte
d t
o
de
v
e
l
op
m
eth
od
s
th
at
are
a
bl
e
t
o
s
i
m
ul
ate
t
he
pe
r
f
or
m
an
c
e
of
the
h
um
an
to
l
ea
r
n
n
ov
el
v
i
s
ua
l
c
o
nc
ep
ts
f
r
o
m
po
s
i
ti
v
e
ex
am
pl
es
.
F
or
ex
am
pl
e,
i
n
[
22
]
,
t
he
a
uth
ors
h
av
e
tr
i
e
d
to
ex
pl
ai
n
ho
w
a
hu
m
an
c
hi
l
d
l
ea
r
ns
ne
w
wor
ds
f
r
o
m
a
s
et
of
pre
-
pr
ov
i
de
d
p
os
i
ti
v
e
ex
am
pl
es
.
Hum
an
s
are
ab
l
e
t
o
g
en
er
al
i
z
e
c
om
pl
ex
s
ets
of
i
m
ag
es
tha
t
c
on
ta
i
n
di
f
f
erent
ob
j
ec
ts
(
e.
g.,
n
atu
r
al
s
c
en
e,
an
i
m
al
s
,
etc
.)
v
er
y
qu
i
c
k
l
y
.
B
es
i
d
es
, h
um
an
s
are ab
l
e t
o e
x
tr
ac
t th
e re
l
at
i
on
s
h
i
p
b
et
w
ee
n a
gi
v
e
n s
et
of
c
on
c
ep
ts
i
n d
i
f
f
erent
c
on
tex
ts
.
B
y
ex
p
l
oi
ti
n
g
c
on
c
ep
t
h
i
erar
c
hi
es
,
nu
m
erous
att
e
m
pts
ha
v
e
be
en
m
ad
e
i
n
the
l
i
t
erature,
att
em
pti
ng
to
r
ea
c
h
hu
m
an
-
l
i
k
e
o
bj
ec
t
ge
n
eral
i
z
ati
on
or
c
ate
go
r
i
z
a
ti
on
.
Deng
et
a
l
.
[
14
]
,
ha
v
e
i
ntr
o
du
c
ed
ne
w
c
l
as
s
i
f
i
ers
tha
t
ex
pl
o
i
t
a
c
on
c
ep
t
h
i
erar
c
h
y
c
on
s
i
s
ti
n
g
of
m
an
y
l
e
v
e
l
s
of
ab
s
tr
ac
ti
on
.
T
he
y
h
av
e
propos
e
d
a
Dual
A
c
c
urac
y
Re
war
d
T
r
ad
e
-
of
f
S
ea
r
c
h
(
DA
RT
S
)
al
go
r
i
thm
tha
t
ai
m
s
to
s
el
ec
t
the
ap
pro
pria
te
l
e
v
e
l
of
c
ate
go
r
i
z
at
i
o
n
i
n
thi
s
c
on
c
ep
t
hi
erar
c
h
y
.
Ho
wev
er,
the
s
e
c
l
as
s
i
f
i
ers
are
no
t
c
om
pl
ete
l
y
ac
c
urate
i
n
i
de
nt
i
f
y
i
n
g
l
e
af
no
de
c
l
as
s
es
.
S
om
e
oth
er
w
ork
s
[
15
,
16
]
ha
v
e
tr
i
ed
t
o
ha
nd
l
e
the
pro
bl
em
o
f
f
i
nd
i
n
g
ne
w
c
ate
go
r
i
es
b
as
ed
on
pre
-
pr
ov
i
de
d
s
ets
of
l
a
be
l
ed
ex
a
m
pl
es
.
T
he
m
ai
n
ai
m
i
n
[
15
]
f
or
ex
am
pl
e
was
ho
w
to
l
e
arn
a
n
e
w
v
i
s
ua
l
c
ate
go
r
y
(
i
.
e.,
ge
n
eral
i
z
a
ti
on
)
f
r
om
f
ew
p
os
i
t
i
v
e
ex
am
pl
es
.
S
al
ak
hu
tdi
no
v
et
al
[
16
]
,
ha
v
e
pres
e
nte
d
a
hi
erar
c
hi
c
al
c
l
as
s
i
f
i
c
ati
o
n
m
od
el
t
h
at
a
l
l
o
w
s
r
are
ob
j
ec
ts
to
bo
r
r
o
w
s
tat
i
s
ti
c
al
s
tr
e
ng
t
h
f
r
om
r
el
ate
d
o
bj
ec
ts
tha
t
ha
v
e
m
an
y
tr
ai
ni
n
g
ex
am
pl
es
.
Ho
w
e
v
er,
t
he
f
orm
er
tw
o
wor
k
s
ha
v
e
tr
i
e
d
to
i
m
prov
e
the
ge
n
era
l
i
z
at
i
on
us
i
n
g
on
l
y
t
he
l
ea
v
e
no
de
s
of
the
c
on
c
e
pt
h
i
era
r
c
h
y
.
T
hu
s
,
the
y
d
i
d
n
ot
a
d
dres
s
the
i
s
s
ue
of
di
s
c
ov
erin
g
t
he
h
i
dd
en
c
on
c
ep
ts
be
t
w
ee
n t
h
e l
ea
v
e n
o
de
s
,
w
h
i
c
h
i
s
a
k
e
y
i
de
a f
or v
i
s
ua
l
c
o
nc
ep
t
l
ea
r
ni
n
g.
B
a
y
es
i
a
n
m
od
el
s
of
ge
ne
r
al
i
z
at
i
o
n
[
17
-
20
]
ha
v
e
be
en
ex
te
ns
i
v
e
l
y
us
ed
i
n
c
o
gn
i
t
i
v
e
sc
i
en
c
e
i
n
or
de
r
to
r
es
o
l
v
e
the
i
s
s
ue
of
l
e
arni
n
g
ne
w
w
ords
or
c
on
c
e
pts
f
r
o
m
a
n
i
n
i
ti
al
s
et
of
wor
ds
or
c
on
c
ep
ts
.
G
i
v
en
a
c
on
c
ep
t
h
i
erar
c
h
y
,
B
a
y
e
s
i
an
m
od
el
s
of
ge
ne
r
a
l
i
z
at
i
on
ba
s
i
c
i
de
a
r
ev
o
l
v
es
aroun
d
f
i
nd
i
ng
th
e
op
t
i
m
al
de
gree
of
ge
n
eral
i
z
a
ti
o
n,
i
n
thi
s
h
i
erar
c
h
y
,
f
or
an
y
s
et
of
c
on
c
ep
ts
[
21
]
.
T
en
en
ba
um
an
d
G
r
i
f
f
i
ths
,
[
18
]
ha
v
e
r
e
f
err
ed
to
s
uc
h
an
ap
pro
ac
h
as
‘
the
s
i
z
e
prin
c
i
pl
e
’
an
d
t
he
y
ha
v
e
s
h
o
w
n
ho
w
i
t
c
o
ul
d
p
ote
nti
al
l
y
ex
pl
ai
n
a
wi
de
r
an
ge
of
ph
en
om
en
a
i
n
c
ate
go
r
y
l
ea
r
n
i
ng
, g
en
era
l
i
z
at
i
on
, a
n
d s
i
m
i
l
arit
y
j
u
dg
m
en
t.
S
uc
h
ph
en
om
en
a
w
er
e
no
t
prev
i
o
us
l
y
un
i
f
i
e
d
un
de
r
o
ne
s
i
ng
l
e
r
a
ti
on
al
-
i
nf
erenc
e.
I
n
m
ore
r
ec
en
t
w
ork
,
X
u
an
d
T
en
e
nb
au
m
[
19
]
ha
v
e
de
v
el
op
e
d
a
n
e
w
B
a
y
es
i
a
n
w
or
d
-
l
ea
r
ni
n
g
m
od
el
.
T
he
i
r
m
od
el
a
pp
e
ared
to
be
c
a
pa
b
l
e
of
m
i
m
i
c
k
i
ng
hu
m
an
ge
ne
r
a
l
i
z
at
i
on
j
ud
gm
en
ts
to
c
r
ea
t
e
a
h
y
po
the
s
i
s
s
pa
c
e
f
or
th
r
ee
c
ate
go
r
i
es
(
an
i
m
al
s
,
v
e
hi
c
l
es
,
an
d
v
eg
eta
b
l
es
)
w
i
t
h
f
e
w
po
s
i
ti
v
e
e
x
a
m
pl
es
.
Ho
wev
er,
th
ei
r
wor
k
i
s
too
ha
r
d
to
be
ex
ten
de
d
to
oth
er c
a
t
eg
ori
es
.
A
bb
ott
e
t
al
,
[
20
]
ha
v
e
pro
po
s
ed
a
B
a
y
es
i
an
-
b
as
ed
m
od
el
f
or
au
tom
ati
c
al
l
y
ge
ne
r
ati
n
g
h
y
p
oth
es
i
s
s
pa
c
es
t
ha
t
are
us
ed
f
or
ge
ne
r
a
l
i
z
at
i
o
n.
I
n
t
he
i
r
m
od
el
,
W
ordNet
da
ta
b
as
e
ha
s
be
e
n
us
ed
t
o
g
en
era
te
the
tr
ee
-
s
tr
uc
tured
h
y
po
the
s
i
s
s
pa
c
e
f
or
di
f
f
erent
c
on
c
ep
ts
.
W
o
r
dNet
i
s
a
da
ta
ba
s
e
t
ha
t
en
c
o
de
s
th
e
s
e
m
an
ti
c
r
el
ati
on
s
h
i
ps
b
et
ween
c
on
c
ep
ts
as
a
ne
t
wor
k
.
O
n
the
ot
he
r
ha
nd
, Im
ag
eNet
ha
s
b
ee
n
us
ed
to
i
n
di
c
a
te
t
he
i
m
ag
es
c
orr
es
po
nd
i
ng
to
ea
c
h
of
th
es
e c
on
c
e
pts
.
Unl
i
k
e
the
prev
i
o
us
wor
k
s
,
A
bb
ott
’
s
a
uto
m
ati
c
al
l
y
ge
n
erated
h
y
p
oth
es
i
s
s
pa
c
e
th
at
c
an
be
us
e
d
i
n a
n
y
c
ate
g
or
y
.
Rec
en
t
w
ork
s
i
n
v
i
s
u
al
r
ec
og
ni
t
i
o
n
[
23
-
26
]
an
d
i
m
ag
e
r
etri
e
v
al
[
27
]
ha
v
e
us
e
d
hi
erar
c
h
i
c
al
s
tr
uc
tur
es
tha
t
c
on
ta
i
n
a
h
i
gh
nu
m
be
r
of
c
l
as
s
es
.
N.
V
erm
a
et
al
.
[
24
]
pro
po
s
ed
a
no
v
el
f
r
am
ew
ork
to
de
term
i
ne
th
e
s
i
m
i
l
arit
y
r
ate
b
et
w
e
en
i
m
ag
es
.
In
th
e
h
i
erar
c
h
y
,
t
w
o
i
m
ag
es
are
c
on
s
i
de
r
ed
to
be
s
i
m
i
l
ar
i
f
the
di
s
t
an
c
e
be
t
we
en
t
he
i
r
an
no
ta
ti
o
ns
(
i
.e
.,
c
on
c
ep
ts
)
i
s
m
i
ni
m
u
m
an
d
v
i
c
e
-
v
ers
a.
J
i
a
et
al
.
[
23
]
ha
v
e
pro
po
s
e
d
a
s
y
s
tem
tha
t
i
nte
gr
ate
s
bo
th
B
a
y
es
i
a
n
m
od
el
s
of
ge
ne
r
al
i
z
at
i
o
n
an
d
m
ac
hi
ne
v
i
s
i
o
n
tec
hn
i
qu
es
.
T
he
i
r
m
ai
n
ai
m
w
as
to
de
term
i
ne
whet
h
er
a
qu
er
y
i
m
ag
e
i
s
r
el
a
te
d
to
a
c
on
c
ep
t
ge
n
erated
f
r
om
s
o
m
e
gi
v
en
s
et
of
i
m
ag
es
.
Li
k
ewi
s
e,
th
e
y
ha
v
e
us
ed
I
m
ag
eNet
da
t
ab
as
e
to
b
ui
l
d
the
i
r
h
y
p
ot
he
s
i
s
s
pa
c
e
.
In
ad
d
i
ti
on
t
o
th
e
hi
g
h
pe
r
f
or
m
an
c
e
the
i
r
s
y
s
t
em
s
ho
w
s
,
i
t
s
ee
m
s
to
be
s
i
m
i
l
ar
to
hu
m
an
r
ea
s
on
i
n
g
i
n
ge
ne
r
al
i
z
ati
on
.
H
o
w
e
v
er
,
al
l
w
ork
s
i
n
the
c
o
nte
x
t
o
f
c
on
c
ep
t
ge
n
eral
i
z
at
i
on
s
uf
f
er
f
r
o
m
on
e
m
a
j
or
probl
em
whi
c
h
i
s
p
erf
or
m
i
ng
ge
n
eral
i
z
ati
on
us
i
n
g
on
l
y
o
ne
c
o
nc
ep
t
hi
erar
c
h
y
.
T
he
r
ef
ore
,
the
y
are
r
es
tr
i
c
ted
to
on
l
y
o
ne
c
on
tex
t
of
ge
n
eral
i
z
ati
o
n
un
l
i
k
e
hu
m
an
s
.
T
o
m
a
k
e
thi
s
l
at
te
r
po
i
nt
c
l
ea
r
er
l
et
’
s
ta
k
e t
he
ex
am
pl
e
i
l
l
us
tr
at
ed
i
n
Fi
gu
r
e
3.
In
F
i
gu
r
e
3
(
a), t
he
r
el
a
ti
o
ns
hi
p b
et
w
e
en
the
three
i
m
ag
es
c
om
es
i
n
term
s
of
F
a
m
i
l
y
(
i
.
e.,
B
i
r
ds
)
,
whereas
i
n
F
i
g
ure
3
(
b)
a
no
th
er
k
i
nd
of
r
el
at
i
on
s
h
i
p
ga
th
ers
the
t
hree
i
m
ag
es
,
w
h
i
c
h
i
s
th
e
d
i
e
t
(
i
.e.
,
om
ni
v
ores
)
.
F
i
n
al
l
y
,
i
n
F
i
gu
r
e
3
(
c
)
,
the
r
e
l
at
i
on
s
h
i
p
i
s
l
i
v
i
n
g reg
i
on
(
i
.e.
,
A
s
i
a A
ni
m
al
s
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
◼
IS
S
N: 1
69
3
-
6
93
0
T
E
L
KO
M
NIK
A
V
ol
.
17
,
No
.
5,
O
c
tob
er 20
19
:
257
2
-
25
8
6
2576
F
i
gu
r
e
3
.
E
x
am
pl
es
of
s
o
m
e g
e
ne
r
a
l
i
z
at
i
on
s
i
n
di
f
f
erent c
on
tex
ts
(
a) g
en
era
l
i
z
a
ti
o
n b
y
f
am
i
l
y
,
(
b)
ge
ne
r
al
i
z
ati
on
b
y
l
i
v
i
ng
r
eg
i
on
, (c
)
ge
ne
r
a
l
i
z
at
i
on
b
y
di
e
t
S
up
po
s
i
ng
th
at
w
e
h
av
e
a
qu
er
y
t
ha
t
c
on
t
ai
ns
three
c
on
c
ep
ts
:
E
l
ep
h
an
t,
Z
eb
r
a,
an
d
G
i
r
af
f
e
as
w
e
ha
v
e
s
h
o
w
n
prev
i
ou
s
l
y
i
n
F
i
g
ure
2.
Co
n
v
en
ti
o
na
l
s
y
s
tem
s
[
2
3
]
i
nte
r
pret,
or
r
a
the
r
ge
ne
r
al
i
z
e,
thi
s
q
ue
r
y
t
o
th
e
c
on
c
ep
t
M
am
m
al
,
w
hi
c
h
i
s
tot
al
l
y
c
orr
ec
t.
H
o
w
e
v
er,
s
ev
eral
o
the
r
m
ea
ni
ng
f
ul
c
on
c
e
pts
c
an
b
e
i
nf
err
ed
.
T
he
s
e
c
on
c
ep
ts
,
s
uc
h
as
A
f
r
i
c
an
an
i
m
al
,
m
ay
be
c
l
os
es
t
to
the
us
er
i
nt
en
t
i
on
t
ha
n
t
he
c
on
c
ep
t
Ma
m
m
al
.
In
or
de
r
to
r
em
ov
e
th
i
s
c
on
f
us
i
o
n
an
d
prec
i
s
el
y
de
tec
ts
t
he
us
er
i
nte
nti
on
,
w
e
prop
os
e
t
o
e
nric
h
the
ex
i
s
ti
ng
hi
erar
c
hi
es
w
i
t
h
oth
er
on
es
;
f
or
ex
am
pl
e,
ad
di
ng
th
e
hi
er
arc
h
y
tha
t
as
s
em
bl
es
c
o
nc
ep
ts
ac
c
ordi
n
g
t
o
t
he
i
r
di
et
an
d
al
s
o
ac
c
ordi
ng
to
th
ei
r
r
e
g
i
on
of
l
i
v
i
ng
.
F
i
g
ure
4
s
h
o
w
s
t
he
d
i
f
f
erenc
e
be
t
ween
o
ur
ge
ne
r
al
i
z
at
i
o
n
a
nd
tha
t
of
a
c
o
nv
en
t
i
on
al
s
y
s
t
em
.
In
the
ne
x
t
s
ec
t
i
on
,
w
e
i
ntr
od
uc
e
ou
r
m
eth
od
where
w
e
tr
y
t
o
i
m
prov
e t
he
ge
ne
r
a
l
i
z
at
i
on
tas
k
b
y
m
ak
i
ng
i
t a
bl
e
to
de
al
wi
th
m
ul
ti
pl
e c
on
c
ep
t h
i
e
r
arc
hi
es
.
F
i
gu
r
e
4.
I
l
l
us
tr
ate
s
th
e d
i
f
ferenc
e b
et
w
e
en
ou
r
g
en
era
l
i
z
at
i
on
s
c
he
m
e
an
d
a c
on
v
en
ti
o
na
l
s
y
s
tem
i
n
[
23
]
3.
P
r
o
p
o
se
d
S
o
lut
ion
In
th
i
s
s
ec
ti
o
n,
we p
r
es
en
t
t
he
d
eta
i
l
s
of
the
pr
op
os
e
d
m
eth
od
.
W
e s
tart b
y
gi
v
i
ng
de
ta
i
l
s
ab
ou
t
the
B
a
y
es
i
an
Co
nc
e
pt
Le
arn
i
n
g,
af
ter
tha
t
we
de
f
i
ne
th
e
c
on
c
ep
t
hi
erar
c
hi
es
CHs
i
n
ou
r
f
r
a
m
ew
ork
.
T
he
n,
w
e
ex
pl
a
i
n
ou
r
ge
ne
r
al
i
z
at
i
on
s
c
he
m
e,
an
d
f
i
n
al
l
y
we
pres
e
nt
the
an
a
l
y
s
i
s
of
s
o
m
e e
x
am
pl
e q
ue
r
i
es
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NIK
A
IS
S
N: 1
69
3
-
6
93
0
◼
Unde
r
s
ta
nd
i
ng
us
er i
n
ten
t
i
o
n i
n
i
m
ag
e retr
i
ev
al
:
ge
n
era
l
i
z
at
i
on
… (
A
bd
el
ma
d
j
i
d Y
ou
c
efa
)
2577
3.1
. B
a
y
e
s
ian Co
n
ce
p
t
Le
ar
n
ing
B
a
y
es
i
a
n
f
r
a
m
ew
ork
f
or
c
on
c
ep
t
l
e
arni
ng
a
nd
ge
ne
r
al
i
z
at
i
on
tec
h
ni
qu
es
[
19
]
,
ar
e
pa
r
ti
c
u
l
arl
y
us
ef
ul
i
n
the
c
a
s
e
w
h
ere
l
e
arni
ng
i
s
p
erf
orm
ed
us
i
ng
on
l
y
a
s
m
al
l
nu
m
be
r
o
f
po
s
i
ti
v
e
ex
am
pl
es
. In
pa
r
t
i
c
ul
ar, t
h
e
prob
l
em
c
an
be
l
oo
k
ed
to
as
f
ol
l
o
w
s
:
G
i
v
en
a s
et
of
n
ex
am
pl
es
(
i
.e.
,
i
m
ag
es
i
n
ou
r
c
as
e)
X
={
x
1
….
x
n
}
whi
c
h
c
a
n
be
group
e
d
un
d
er
a
s
pe
c
i
f
i
c
c
on
c
ep
t
C
as
s
ho
w
n
i
n
F
i
gu
r
e
3
.
G
i
v
en
a
n
e
w
ex
a
m
pl
e
y
,
t
he
qu
es
ti
o
n
i
s
:
Is
y
a
m
e
m
be
r
of
X
or
no
t.
T
o
an
s
w
er
th
i
s
qu
es
ti
on
,
B
a
y
es
i
an
c
on
c
e
p
t
l
ea
r
n
i
n
g
as
s
um
es
the
ex
i
s
ten
c
e
of
a
h
y
p
oth
es
i
s
s
pa
c
e
H
s
uc
h
tha
t
H=
{
h
1
….
h
n
}
where
th
e
m
os
t
ap
propr
i
ate
h
y
po
t
he
s
i
s
hi
c
an
b
e
c
on
s
i
de
r
e
d
as
C.
E
ac
h
h
y
p
oth
es
i
s
hi
(
e
.g.
,
A
ni
m
al
,
M
am
m
al
,
B
i
r
d)
c
orr
es
po
nd
s
t
o
o
ne
c
l
us
ter
i
n
the
c
on
c
ep
t
hi
erar
c
hi
es
.
A
n
i
l
l
us
tr
ati
o
n
i
s
gi
v
e
n
i
n
F
i
gu
r
e
2.
T
he
B
a
y
es
i
an
l
e
ar
ne
r
ev
al
u
ate
s
al
l
th
e
h
y
p
ot
he
s
e
s
hi
us
i
ng
B
a
y
es
r
ul
e a
s
f
ol
l
o
w
s
:
(
ℎ
|
)
∝
(
ℎ
)
(
|
ℎ
)
(
1
)
s
uc
h
tha
t
P
(
h|
X
)
i
s
th
e
po
s
teri
or
proba
bi
l
i
t
y
,
P
(
h)
the
pr
i
or
prob
ab
i
l
i
t
y
an
d
P
(
X
|
h)
the
l
i
k
el
i
h
oo
d
. T
he
pri
or
P
(
h
)
of
th
e h
y
po
t
he
s
i
s
i
s
de
f
i
n
e
d a
c
c
ordi
ng
t
o t
h
e E
r
l
an
g d
i
s
tr
i
bu
ti
on
:
(
ℎ
)
(
|
ℎ
|
/
2
)
e
xp
{
−
|
ℎ
|
/
}
(
2)
w
he
r
e
|
h
|
i
s
the
s
i
z
e
of
th
e
h
y
po
t
he
s
i
s
h
(
nu
m
be
r
of
l
ea
f
no
d
es
)
an
d
σ
pa
r
am
ete
r
i
s
the
m
ea
n
s
i
z
e
of
th
e
b
as
i
c
l
e
v
e
l
h
y
p
o
the
s
es
.
T
he
l
i
k
el
i
ho
o
d
i
s
d
e
term
i
ne
d
b
y
t
he
as
s
um
pti
o
n
of
r
an
do
m
l
y
s
a
m
pl
ed
po
s
i
t
i
v
e e
x
am
pl
es
.
In
the
s
i
m
pl
es
t
c
as
e,
e
ac
h
ex
am
pl
e
i
n
X
i
s
as
s
um
ed
to
be
i
nd
e
pe
nd
e
ntl
y
s
am
pl
e
d
f
r
o
m
a u
n
i
f
or
m
de
ns
i
t
y
o
v
er t
he
c
on
c
ep
t
C
.
For
n
ex
am
pl
es
we t
h
en
h
av
e:
(
|
ℎ
)
=
{
[
1
|
ℎ
|
]
if
1
,
…
,
∈
ℎ
0
if
a
n
y
∉
ℎ
(
3
)
p
r
i
or
wor
k
[
19
]
f
oc
us
ed
on
c
al
c
u
l
at
i
ng
th
e
pr
ob
a
bi
l
i
t
y
tha
t
a
ne
w
ob
j
ec
t
y
i
s
al
s
o
a
m
em
be
r
of
the
c
o
nc
ep
t
C
b
y
a
v
e
r
ag
i
n
g
t
he
pred
i
c
ti
o
ns
of
al
l
h
y
po
the
s
es
wei
g
hte
d
b
y
t
he
i
r
po
s
teri
or
prob
ab
i
l
i
ti
es
:
(
∈
|
)
=
∑
(
|
ℎ
)
ℎ
∈
(
ℎ
|
)
(
4
)
In
ou
r
m
eth
od
,
ho
wev
er,
we
f
oc
us
o
n
f
i
nd
i
n
g
the
h
y
p
o
the
s
i
s
h
t
ha
t
c
orr
es
po
nd
s
to
the
c
on
c
e
pt
C.
I
n
p
arti
c
ul
ar,
w
e
h
av
en
’
t
a
n
e
w
ex
am
pl
e
y
,
b
ut
r
ath
er
a
q
u
er
y
X
.
A
no
th
er
s
ub
s
tan
t
i
al
d
i
f
f
erenc
e
i
s
t
h
at
i
n
t
he
pre
v
i
ou
s
wor
k
s
[
19
]
t
he
h
y
po
the
s
i
s
s
pa
c
e
H
i
s
ge
n
erate
d
ac
c
ordi
ng
to
on
l
y
on
e
c
on
c
ep
t
h
i
erar
c
h
y
CH,
w
h
i
c
h
i
s
no
t
t
he
c
as
e
i
n
ou
r
wor
k
be
c
au
s
e
we
c
on
s
i
de
r
3 d
i
f
f
erent CHs
i
n
ge
ne
r
ati
ng
H.
T
o
de
term
i
ne
th
e
m
os
t
ap
propr
i
a
te
h
f
r
om
H,
w
e
c
a
l
c
ul
ate
th
e
po
s
teri
or
pr
ob
a
b
i
l
i
t
y
f
or
ea
c
h
h
,
th
e
a
pp
r
o
pria
t
e
h
tha
t
c
orr
es
po
n
ds
to
the
c
on
c
ep
t
C
i
s
t
he
on
e
h
a
v
i
ng
ob
ta
i
n
ed
the
h
i
g
he
s
t p
r
o
ba
b
i
l
i
t
y
s
c
ore (
i
.e.
, M
ax
i
m
u
m
a
P
os
teri
o
r
i
h
y
p
ot
he
s
i
s
h
M
AP
).
T
he
h
MA
P
i
s
gi
v
en
b
y
:
ℎ
=
a
r
gma
x
ℎ
∈
(
|
ℎ
)
(
ℎ
)
(
5
)
a
f
ter
ha
v
i
ng
d
ete
r
m
i
ne
d
th
e
m
os
t
ap
propr
i
at
e
h,
w
e
c
o
ns
i
de
r
th
e
CH
to
w
h
i
c
h
h
b
el
o
ng
s
,
an
d
w
e
om
i
t
the
2
oth
ers
.
A
f
terwar
ds
,
w
e
ex
tr
ac
t
th
e
r
em
ai
ni
n
g
c
on
c
ep
ts
tha
t
b
el
on
g
t
o
C
(
i
.e.
,
h
i
dd
en
c
on
c
ep
ts
)
.
F
i
na
l
l
y
,
we
de
t
e
c
t
i
m
ag
es
an
no
tat
ed
b
y
c
o
nc
ep
ts
c
on
ta
i
ne
d
i
n
C
an
d
di
s
pl
a
y
t
he
m
to
the
us
er.
3.2
.
P
r
es
ent
atio
n
of
O
u
r
C
o
n
ce
p
t
Hier
ar
chies
In
o
ur
f
r
am
ew
ork
,
w
e
us
e
three
k
i
nd
s
of
c
on
c
ep
t
h
i
e
r
arc
hi
es
to
ex
p
an
d
th
e
s
c
op
e
of
us
er
un
d
ers
tan
d
i
n
g,
w
h
ere
ea
c
h
h
i
erar
c
h
y
gro
up
s
c
on
c
ep
ts
ac
c
ordi
n
g
to
a
s
p
ec
i
f
i
c
r
el
ati
on
s
h
i
p
.
T
he
s
e rel
ati
on
s
h
i
ps
are:
f
am
i
l
y
,
di
et
a
nd
l
i
v
i
ng
pl
ac
e
.
Nex
t,
w
e
gi
v
e
de
t
ai
l
s
ab
ou
t
ea
c
h o
f
th
em
.
Evaluation Warning : The document was created with Spire.PDF for Python.
◼
IS
S
N: 1
69
3
-
6
93
0
T
E
L
KO
M
NIK
A
V
ol
.
17
,
No
.
5,
O
c
tob
er 20
19
:
257
2
-
25
8
6
2578
3.2
.1
. Con
ce
p
t
H
ier
ar
ch
y
A
c
cordin
g
t
o
F
am
il
y
CH
a
(
Imag
eNet h
ie
r
a
r
ch
y
)
W
e
us
e
I
m
ag
e
Net
hi
erar
c
h
y
as
the
f
i
r
s
t
CH,
w
e
de
no
te
i
t
b
y
CH
a
.
Im
ag
e
Net
i
s
a
l
arge
i
m
ag
e
da
tab
as
e
w
h
i
c
h
i
s
ba
s
ed
o
n
the
W
ordNet
hi
erar
c
h
y
.
E
ac
h
c
on
c
ep
t
i
n
W
ordNet
i
s
de
s
c
r
i
be
d
b
y
m
ul
t
i
pl
e
wor
ds
w
h
i
c
h
are
c
a
l
l
ed
a
"
s
y
no
n
y
m
s
et
"
or
"
s
y
ns
et
"
.
W
e
h
av
e
c
h
os
en
Im
ag
e
Net
hi
erar
c
h
y
be
c
a
us
e
i
t
ha
s
a
r
i
c
h
h
i
erar
c
h
y
of
c
on
c
ep
ts
an
d
i
t
as
s
em
bl
es
m
i
l
l
i
on
s
of
i
m
ag
es
(
ab
ou
t
ten
m
i
l
l
i
on
i
m
ag
es
th
at
ha
v
e
be
en
m
an
ua
l
l
y
a
nn
ot
ate
d).
In
ou
r
w
ork
,
w
e
are
i
nte
r
es
t
ed
b
y
t
he
pa
r
t
whi
c
h
c
ate
go
r
i
z
es
th
e
an
i
m
al
s
as
s
ho
w
n
i
n
F
i
gu
r
e
5
.
3.2
.2
. Con
ce
p
t
H
ier
ar
ch
y
A
c
cordin
g
to
Di
et
CH
b
W
e
bu
i
l
d
t
hi
s
t
y
p
e
of
r
el
ati
on
s
hi
p
ba
s
ed
o
n
W
i
k
i
pe
di
a
.
O
ur
hi
er
arc
h
y
i
s
bu
i
l
t
ba
s
ed
o
n
the
f
oo
d
na
t
ure of
ea
c
h “s
y
ns
et”
as
s
ho
wn i
n
F
i
g
ure
6
.
W
e d
en
ote
the
c
urr
e
nt
CH
b
y
C
H
b
.
3.2
.3
. Con
ce
p
t
H
ier
ar
ch
y
A
c
cordin
g
to
R
egio
n
of
L
i
v
ing
CH
c
T
he
r
eg
i
o
n
of
l
i
v
i
n
g
i
s
a
s
y
n
on
y
m
s
et
w
i
t
hi
n
Im
ag
eNet
,
thu
s
,
we
a
do
pt
a
CH
t
ha
t
g
r
ou
ps
c
on
c
ep
ts
ac
c
ordi
ng
t
o reg
i
o
n o
f
l
i
v
i
ng
as
s
ho
w
n
i
n F
i
g
u
r
e
7
.
W
e d
en
ote
i
t b
y
C
H
c
.
F
i
gu
r
e
5
. I
l
l
us
tr
ati
on
of
th
e
c
on
c
ep
t
hi
erar
c
h
y
CH
a
F
i
gu
r
e
6
. I
l
l
us
tr
ati
on
of
th
e
c
on
c
ep
t
hi
erar
c
h
y
CH
b
F
i
gu
r
e
7
. I
l
l
us
tr
ati
on
of
th
e
c
on
c
ep
t h
i
erar
c
h
y
CH
c
3.3
. O
u
r
G
ener
ali
z
atio
n
S
cheme
P
r
ev
i
o
us
s
tud
i
es
[
19
-
21
,
23
]
ha
v
e
at
tem
pte
d
to
l
ea
r
n
c
on
c
ep
ts
us
i
n
g
a
f
ew
n
u
m
be
r
o
f
po
s
i
t
i
v
e
ex
am
pl
es
.
Ho
wev
er,
the
s
e
s
tu
di
es
ha
v
e
f
o
c
us
ed
on
l
y
on
c
ho
os
i
n
g
t
he
ap
propr
i
ate
gene
r
al
i
z
ati
on
l
e
v
el
i
n
a
s
i
ng
l
e
c
on
c
ep
t
h
i
erar
c
h
y
.
T
hi
s
,
i
n
f
ac
t,
c
an
y
i
e
l
d
m
i
no
r
or
c
om
pl
ete
l
y
i
r
r
el
e
v
a
nt
r
es
ul
ts
.
T
o
ov
erc
om
e
thi
s
probl
em
an
d
i
m
pr
ov
e
the
r
es
ul
ts
of
the
en
gi
n
e
w
e
us
e
three
k
i
nd
s
of
CH
i
n
t
he
ge
n
er
al
i
z
a
ti
o
n.
T
he
d
eta
i
l
s
of
ou
r
ge
n
eral
i
z
at
i
o
n
s
c
he
m
e
are
i
l
l
us
tr
ate
d
i
n Fi
gu
r
e
8
.
T
he
r
e a
r
e
s
i
x
m
ai
n s
t
ep
s
to
ge
n
eral
i
z
e t
he
qu
er
y
wh
i
c
h a
r
e
de
s
c
r
i
be
d
as
f
ol
l
o
w
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NIK
A
IS
S
N: 1
69
3
-
6
93
0
◼
Unde
r
s
ta
nd
i
ng
us
er i
n
ten
t
i
o
n i
n
i
m
ag
e retr
i
ev
al
:
ge
n
era
l
i
z
at
i
on
… (
A
bd
el
ma
d
j
i
d Y
ou
c
efa
)
2579
F
i
gu
r
e
8
.
I
l
l
us
tr
ate
s
th
e m
ai
n
s
tep
s
of
ou
r
g
en
eral
i
z
ati
o
n s
c
he
m
e
3.3
.1
.
Inp
u
t
Imag
e
s (F
o
r
m
u
latio
n
Q
u
er
y
)
O
ur
s
y
s
tem
s
ho
w
s
the
us
er
s
o
m
e
i
m
ag
es
fr
o
m
da
tas
et
(
u
s
er
i
nte
r
f
ac
e).
T
he
us
er
ha
s
to
s
el
ec
t
s
om
e
i
m
ag
es
ex
am
pl
e
(
2
-
5
i
m
ag
es
)
r
ep
r
es
en
t
hi
s
n
ee
ds
t
o
f
orm
ul
ate
qu
er
y
,
as
s
ho
w
n
i
n Fi
gu
r
e
9
.
F
i
gu
r
e
9
.
Us
er
i
nt
erf
ac
e
of
ou
r
s
y
s
tem
3.3
.2
.
Co
n
ce
p
t
s
of
I
mag
e
s
In
th
e
da
tas
et
ea
c
h
i
m
ag
e
an
n
ota
t
ed
wi
th
a
c
on
c
e
pt,
we
us
e
Im
ag
e
Net
da
tas
et
th
i
s
c
ol
l
ec
t
i
on
a
nn
ot
ate
d
f
r
om
W
o
r
d
Net.
A
f
ter
the
us
er
f
orm
ul
ate
s
hi
s
qu
er
y
,
ou
r
s
y
s
t
em
ha
s
t
o
ex
tr
ac
t
the
c
o
nc
ep
ts
of
ea
c
h
i
m
ag
e
i
n
the
qu
er
y
of
t
ho
s
e
c
on
c
e
pts
w
h
i
c
h
w
e
c
al
l
ed
c
o
nc
ep
t
qu
er
y
as
s
ho
wn i
n
F
i
gu
r
e
1
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.
◼
IS
S
N: 1
69
3
-
6
93
0
T
E
L
KO
M
NIK
A
V
ol
.
17
,
No
.
5,
O
c
tob
er 20
19
:
257
2
-
25
8
6
2580
F
i
gu
r
e
1
0
.
Im
ag
e
s
an
d t
he
i
r
c
orr
es
po
nd
i
ng
c
o
nc
ep
ts
3.3
.3
.
H
y
p
o
t
h
es
is
S
p
ac
e (Relatio
n
ship
s)
A
f
ter
f
i
nd
i
n
g
c
on
c
e
pts
of
th
e
qu
er
y
ou
r
s
y
s
tem
be
gi
ns
s
ea
r
c
hi
ng
f
or
al
l
the
r
el
a
ti
o
ns
hi
ps
be
t
w
e
en
c
on
c
ep
ts
qu
er
y
.
A
l
l
r
el
a
ti
o
ns
hi
ps
i
n
a
l
l
k
i
nd
s
of
c
on
c
ep
ts
h
i
erar
c
h
y
are
c
a
l
l
ed
h
y
p
oth
es
i
s
s
pa
c
e.
3.3
.4
.
F
ind
ing
t
h
e
A
p
p
r
o
p
r
iate R
ela
t
ion
ship
(
h
M
A
P
)
A
f
ter
c
r
ea
ti
n
g
th
e
h
y
po
the
s
i
s
s
pa
c
e
ou
r
s
y
s
tem
ha
s
to
f
i
nd
the
ap
propr
i
ate
r
e
l
at
i
o
ns
hi
p
ga
th
erin
g t
h
os
e c
on
c
ep
ts
.
T
he
m
ax
a
po
s
teri
or r
e
pres
en
ts
th
i
s
r
el
a
ti
on
s
hi
p.
3.3
.5
.
Hid
d
en
C
o
n
ce
p
t
s
Hi
dd
en
c
on
c
e
pts
are
th
e
c
on
c
ep
ts
tha
t
ar
e
l
i
n
k
ed
w
i
th
c
on
c
e
pts
qu
er
y
b
y
the
a
pp
r
o
pria
t
e rel
ati
on
s
h
i
p
s
s
el
ec
ted
i
n
the
c
o
nc
ep
t
hi
erar
c
h
y
.
3.3
.6
.
Re
sult
s
F
i
na
l
l
y
ou
r
s
y
s
tem
s
ea
r
c
he
s
al
l
i
m
ag
es
an
no
tat
e
d
wi
th
c
on
c
ep
t
qu
er
y
an
d
h
i
dd
e
n
c
on
c
ep
ts
,
an
d
s
ho
w
s
r
es
ul
t
s
to
t
he
us
er.
F
or
t
he
s
ak
e
of
c
l
arit
y
,
l
et
us
i
l
l
us
tr
ate
th
i
s
b
y
a
s
i
m
pl
e
ex
am
pl
e.
S
up
po
s
e
tha
t
we
ha
v
e
3
po
s
i
ti
v
e
ex
a
m
pl
es
i
.e.
,
X
=
{
Li
on
,
G
i
r
aff
e,
Z
eb
r
a},
the
h
y
p
oth
es
es
we
c
an
c
o
n
s
i
de
r
as
c
an
d
i
d
ate
t
o
be
t
h
e
c
on
c
ep
t
C
are:
A
n
i
m
al
,
Ma
m
m
al
,
A
f
r
i
c
a
an
i
m
al
.
W
e
c
al
c
ul
at
e
t
h
e
p
os
teri
or
prob
ab
i
l
i
t
y
o
f
ea
c
h
h
y
p
oth
es
i
s
ac
c
o
r
di
ng
to
(
1).
T
he
h
y
po
the
s
i
s
th
at
ob
t
ai
ns
th
e
hi
g
he
s
t s
c
ore
i
s
c
on
s
i
d
ered a
s
C.
A
f
ter
de
t
erm
i
ni
ng
C,
w
e
g
i
v
e
b
ac
k
the
us
er
th
e
i
m
ag
es
an
n
ota
t
ed
w
i
t
h
al
l
the
c
o
nc
ep
ts
of
l
ea
f
no
de
s
un
de
r
C
i
.
e.,
we
c
on
s
i
de
r
th
e
c
on
c
ep
ts
c
on
ta
i
ne
d
i
n
th
e
qu
er
y
to
ge
t
he
r
wi
th
t
ho
s
e
whi
c
h
a
r
e
no
t
c
on
t
ai
ne
d
(
i
.
e.,
hi
dd
e
n
c
on
c
ep
ts
)
.
In
de
e
d,
thi
s
c
ou
l
d
h
el
p
i
n
i
m
prov
i
ng
t
he
qu
al
i
t
y
of
r
etri
ev
al
r
es
u
l
ts
. T
he
s
te
ps
of
ou
r
al
go
r
i
t
hm
are s
u
m
m
ariz
ed
i
n
A
l
go
r
i
thm
1.
A
l
g
orit
hm
1:
G
en
eral
i
z
at
i
on
of
qu
er
y
B
eg
i
n
1:
INP
UT
: X
=
{x
1,
x
2,
…x
n}
2:
Com
pu
te
po
s
t
erio
r
prob
a
bi
l
i
t
y
P
(
h
|
X
)
of
al
l
h
y
p
ot
he
s
es
h
i
n
CH
a,
C
Hb
an
d
CHc
ac
c
ordi
ng
to
(
1)
:
3:
F
i
nd
t
he
Ma
x
a
po
s
ter
i
ori
h
MA
P
ac
c
ordi
ng
t
o (5)
:
4:
S
el
ec
t a
p
propr
i
ate
CH a
nd
the
C.
5:
F
i
nd
H
i
d
de
n
Ci
(
T
he
c
on
c
e
pts
un
d
er C a
nd
whi
c
h
di
dn
’
t a
pp
e
ar i
n
the
qu
er
y
)
6:
O
UT
P
UT
: R
es
ul
t o
f
i
m
ag
es
Ii
a
nn
o
tat
e
d b
y
a
l
l
l
e
af
no
d
es
un
d
er the
c
on
c
ep
t
C.
E
nd
3.4.
E
xa
mp
les
of
G
en
er
a
li
z
ing
Q
u
er
ies
in
O
u
r
F
o
r
m
w
o
r
k
No
w
,
l
et
us
ex
pl
a
i
n,
b
y
e
x
a
m
pl
es
,
ho
w
o
ur
propos
ed
m
eth
od
wor
k
s
.
In
ou
r
c
as
e,
the
s
y
s
t
em
pres
en
ts
a
s
am
pl
e
of
i
m
ag
es
to
the
us
er,
t
he
n
the
us
er
s
e
l
ec
ts
i
m
ag
es
s
up
po
s
e
d
to
be
s
i
m
i
l
ar
to
what
h
e
i
s
l
o
ok
i
ng
f
or.
O
ur
pro
po
s
e
d
a
pp
r
oa
c
h
us
es
,
t
he
r
e
af
ter,
t
he
an
n
ota
t
i
on
s
as
s
i
gn
e
d
wi
th
tho
s
e
i
m
ag
e
s
i
n
order
t
o
d
i
s
c
ov
er
the
hi
d
de
n
r
el
ati
on
s
hi
p
be
t
wee
n
th
e
c
on
c
e
pts
c
on
tai
ne
d
i
n
the
qu
er
y
.
F
or
ex
am
pl
e,
the
ge
ne
r
al
i
z
i
ng
of
the
q
ue
r
i
es
Q
1
,
Q
2
a
n
d
Q
3
s
ho
wn
i
n
F
i
gu
r
es
1
1
-
1
3
are r
es
pe
c
t
i
v
el
y
de
m
on
s
tr
ate
d
in
T
ab
l
e
1,
T
ab
l
e
2,
a
nd
T
ab
l
e 3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NIK
A
IS
S
N: 1
69
3
-
6
93
0
◼
Unde
r
s
ta
nd
i
ng
us
er i
n
ten
t
i
o
n i
n
i
m
ag
e retr
i
ev
al
:
ge
n
era
l
i
z
at
i
on
… (
A
bd
el
ma
d
j
i
d Y
ou
c
efa
)
2581
F
i
gu
r
e
1
1
.
Il
l
us
tr
at
i
on
of
th
e
Q
ue
r
y
Q
1
F
i
gu
r
e
1
2
.
Il
l
us
tr
at
i
on
of
th
e
Q
ue
r
y
Q2
F
i
gu
r
e
1
3
.
Il
l
us
tr
at
i
on
of
th
e
Q
ue
r
y
Q3
3.4.
1
.
G
ene
r
ali
z
ing
Q
u
er
y
Q
1
Conc
ep
ts
Q
u
er
y
:
X
=
{l
eo
p
a
r
d,
oc
el
ot,
T
i
ge
r
, C
ou
ga
r
}
n=
3
H
y
p
oth
es
i
s
h
−
A
ni
m
al
−
Ma
m
m
al
−
F
el
i
ne
−
Car
ni
v
ores
S
i
z
e
of
ea
c
h h
y
p
oth
es
i
s
|
h
|
−
|
A
ni
m
al
|
=
10
0
−
|
Ma
m
m
al
|
=
86
−
|
Fe
l
i
n
e
|
=
5
−
|
Car
n
i
v
ores
|
=
3
4
Ma
x
i
m
u
m
a
P
os
teri
or
i
h
y
p
oth
es
i
s
h
MAP
:
T
hrough
the
r
es
ul
ts
gi
v
e
n
b
y
T
ab
l
e
1
th
e
Ma
x
i
m
um
a
P
os
teri
ori
h
y
p
oth
es
i
s
of
th
e
Q
ue
r
y
1
is
h
MAP
=
Fe
l
i
n
e
.
Hi
dd
en
c
o
nc
ep
ts
: L
i
o
n,
J
ag
ua
r
, P
an
t
he
r
, L
y
nx
.
T
ab
l
e
1
. G
e
ne
r
a
l
i
z
at
i
o
n o
f
t
he
Q
ue
r
y
Q
1
i
n
e
ac
h C
on
c
ep
t H
i
erar
c
h
y
C
o
n
c
e
p
t
h
ier
a
r
c
h
y
H
y
p
o
t
h
e
s
is
h
P
o
s
t
e
r
ior
p
r
o
b
a
b
il
it
y
P
(
h
|
X
)
CHa
M
a
mm
a
l
0
.
5
6
A
n
i
m
a
l
0
.
2
3
Feline
0
.
8
7
CHb
C
a
r
n
iv
o
r
e
s
0
.
5
5
CHc
N
o
r
e
lat
ion
m
a
t
c
h
e
d
/
Evaluation Warning : The document was created with Spire.PDF for Python.