I
AE
S
I
n
t
e
r
n
at
ion
al
Jou
r
n
al
of
Ar
t
if
icial
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
Vol.
14
,
No.
4
,
Augus
t
2025
,
pp.
2776
~
2787
I
S
S
N:
2252
-
8938
,
DO
I
:
10
.
11591/i
jai
.
v
14
.i
4
.
pp
27
76
-
2787
2776
Jou
r
n
al
h
omepage
:
ht
tp:
//
ij
ai
.
iaes
c
or
e
.
c
om
D
u
al
si
m
u
la
t
e
d
a
n
n
e
al
in
g sof
t
d
e
c
o
d
e
r
f
o
r
l
in
e
a
r
b
lo
c
k
c
od
e
s
Hi
c
h
am
T
ah
iri
Alao
u
i
1
,
A
h
m
e
d
Az
ou
ao
u
i
2
,
Ja
m
al
E
l
Kaf
i
1
1
L
a
R
oS
E
R
I
L
a
bor
a
to
r
y, D
e
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ie
nc
e
, F
a
c
ul
ty
of
S
c
ie
nc
e
s
, C
houa
ib
D
oukka
li
U
ni
ve
r
s
it
y, E
l
J
a
di
da
,
M
or
oc
c
o
2
C
omput
e
r
S
c
ie
nc
e
R
e
s
e
a
r
c
h L
a
bor
a
to
r
y (
L
a
R
I
)
, H
ig
he
r
S
c
hool
of
T
e
c
hnol
ogy, I
bn T
of
a
il
U
ni
ve
r
s
it
y, K
e
ni
tr
a
, M
or
oc
c
o
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
Apr
22
,
2024
R
e
vis
e
d
F
e
b
18
,
2025
Ac
c
e
pted
M
a
r
15
,
2025
T
h
i
s
p
ap
er
p
r
o
p
o
s
e
s
a
n
e
w
a
p
p
r
o
ach
t
o
s
o
ft
d
eco
d
i
n
g
f
o
r
l
i
n
ear
b
l
o
c
k
co
d
es
cal
l
e
d
d
u
a
l
s
i
mu
l
at
e
d
an
n
eal
i
n
g
s
o
f
t
d
ec
o
d
er
(
D
SA
S
D
)
w
h
i
c
h
u
t
i
l
i
ze
s
t
h
e
d
u
a
l
co
d
e
i
n
s
t
ead
o
f
t
h
e
o
ri
g
i
n
al
c
o
d
e,
u
s
i
n
g
t
h
e
s
i
mu
l
a
t
ed
a
n
n
ea
l
i
n
g
al
g
o
ri
t
h
m
as
p
res
e
n
t
e
d
i
n
a
p
rev
i
o
u
s
l
y
d
e
v
el
o
p
e
d
w
o
r
k
.
T
h
e
D
S
A
SD
al
g
o
ri
t
h
m
d
emo
n
s
t
rat
e
s
s
u
p
eri
o
r
d
ec
o
d
i
n
g
p
erfo
rma
n
ce
acro
s
s
a
w
i
d
e
ran
g
e
o
f
co
d
es
,
o
u
t
p
erfo
rm
i
n
g
c
l
as
s
i
ca
l
s
i
mu
l
at
e
d
an
n
eal
i
n
g
an
d
s
e
v
eral
o
t
h
er
t
es
t
ed
d
e
co
d
ers
.
W
e
c
o
n
d
u
c
t
a
co
m
p
reh
e
n
s
i
v
e
e
v
al
u
at
i
o
n
o
f
t
h
e
p
r
o
p
o
s
e
d
al
g
o
ri
t
h
m's
p
erfo
rma
n
ce,
o
p
t
i
mi
z
i
n
g
i
t
s
p
aramet
er
s
t
o
ach
i
ev
e
t
h
e
b
es
t
p
o
s
s
i
b
l
e
res
u
l
t
s
.
A
d
d
i
t
i
o
n
al
l
y
,
w
e
co
m
p
are
i
t
s
d
eco
d
i
n
g
p
erfo
rma
n
ce
an
d
al
g
o
ri
t
h
m
i
c
co
mp
l
ex
i
t
y
w
i
t
h
o
t
h
er
d
e
co
d
i
n
g
al
g
o
r
i
t
h
ms
i
n
i
t
s
ca
t
eg
o
ry
.
O
u
r
res
u
l
t
s
d
emo
n
s
t
r
at
e
a
g
ai
n
i
n
p
erfo
rma
n
ce
o
f
ap
p
r
o
x
i
m
at
el
y
2
.
5
d
B
at
a
b
i
t
erro
r
rat
e
(BE
R)
o
f
6
×
1
0
⁻
⁶
fo
r
t
h
e
L
D
PC
(6
0
,
3
0
)
co
d
e.
K
e
y
w
o
r
d
s
:
B
C
H
c
ode
s
De
c
oding
a
lgor
it
hms
Dua
l
c
ode
s
E
r
r
o
r
c
or
r
e
c
ti
on
c
ode
s
L
inea
r
block
c
ode
s
S
im
ulate
d
a
nne
a
li
ng
S
of
t
de
c
oding
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
Hic
ha
m
T
a
hir
i
Ala
oui
De
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ienc
e
,
F
a
c
ult
y
of
S
c
ien
c
e
s
,
C
houa
ib
Doukka
li
Unive
r
s
it
y
E
l
J
a
dida,
M
or
oc
c
o
E
mail:
tahi
r
i
-
a
laoui.
h@uc
d.
a
c
.
ma
1.
I
NT
RODU
C
T
I
ON
I
n
digi
tal
c
omm
un
ica
ti
on
s
ys
tems
,
e
ns
ur
ing
r
e
li
a
bl
e
inf
or
mation
tr
a
ns
mi
s
s
ion
ove
r
nois
y
c
ha
nne
ls
is
a
pa
r
a
mount
c
onc
e
r
n.
E
r
r
o
r
c
o
r
r
e
c
ti
on
tec
hniques
[
1
]
play
a
vit
a
l
r
ole
in
a
dd
r
e
s
s
ing
thi
s
c
ha
ll
e
nge
.
Among
thes
e
tec
hniques
,
li
ne
a
r
block
c
od
e
s
[
2
]
a
r
e
pa
r
ti
c
ular
ly
notew
or
thy
.
T
he
s
e
c
ode
s
a
dd
r
e
dunda
nt
bit
s
to
the
or
igi
na
l
mes
s
a
ge
,
f
or
m
ing
c
ode
wor
ds
,
whic
h
e
n
a
ble
e
r
r
or
de
tec
ti
on
a
nd
c
or
r
e
c
ti
on
dur
ing
tr
a
ns
mi
s
s
ion.
De
c
oding
a
lgor
it
hms
a
r
e
de
s
igned
to
de
ter
m
ine
t
he
mos
t
pr
oba
ble
tr
a
ns
mi
tt
e
d
c
ode
wor
d
f
r
o
m
the
r
e
c
e
ived
s
ignal,
a
s
s
hown
in
F
igu
r
e
1.
T
r
a
dit
ionally,
ha
r
d
-
de
c
is
ion
de
c
oding
a
lgor
it
hms
ha
ve
be
e
n
e
mpl
oye
d.
How
e
ve
r
,
thes
e
a
lgo
r
it
hms
c
a
n
be
c
omput
a
ti
ona
ll
y
int
e
ns
ive
a
nd
may
dis
c
a
r
d
va
luable
s
of
t
inf
o
r
mation
p
r
e
s
e
nt
in
the
r
e
c
e
ive
d
s
ignal.
T
o
a
ddr
e
s
s
thes
e
li
mi
tations
,
r
e
s
e
a
r
c
he
r
s
ha
ve
inc
r
e
a
s
ingl
y
tur
ne
d
to
s
of
t
de
c
is
ion
de
c
oding
a
lg
or
it
hms
that
incor
por
a
te
the
c
onti
nuous
na
tur
e
of
the
r
e
c
e
ive
d
s
ignal's
a
mpl
it
ude
.
S
of
t
de
c
is
ion
de
c
oding
a
lgor
it
hms
leve
r
a
ge
a
dva
nc
e
d
tec
hniques
f
r
om
inf
or
mation
theor
y,
li
ne
a
r
a
lgebr
a
,
a
nd
s
ignal
pr
oc
e
s
s
ing
to
de
c
ode
r
e
c
e
ived
s
ignals
with
high
a
c
c
ur
a
c
y.
T
he
s
e
a
lgor
it
hms
a
r
e
pa
r
ti
c
ular
ly
e
f
f
e
c
ti
ve
in
nois
y
c
ha
nne
ls
,
w
he
r
e
the
r
e
li
a
bil
it
y
of
indi
vidual
b
it
s
is
unc
e
r
tain
a
nd
r
e
qui
r
e
s
pr
oba
bil
is
ti
c
tr
e
a
tm
e
nt.
T
he
pe
r
f
or
manc
e
opti
mi
z
a
ti
on
of
s
uc
h
a
lgor
it
hms
is
a
ls
o
a
c
onc
e
r
n
to
be
a
ddr
e
s
s
e
d.
P
e
r
f
or
manc
e
opti
m
iza
ti
on
of
s
of
t
de
c
is
ion
de
c
odi
ng
a
lgor
it
hms
is
c
r
uc
ial
f
or
a
c
hieving
im
p
r
ove
d
e
r
r
or
c
or
r
e
c
ti
on
c
a
pa
bil
it
ies
.
I
n
r
e
c
e
nt
ye
a
r
s
,
meta
he
ur
is
ti
c
a
nd
opti
mi
z
a
ti
on
tec
hniques
ha
ve
ga
ined
a
tt
e
nti
on
f
or
their
potential
to
e
nha
nc
e
the
pe
r
f
or
manc
e
of
thes
e
a
lgor
it
hm
s
.
Among
thes
e
tec
hniques
,
s
im
ulate
d
a
nne
a
li
ng
ha
s
e
mer
ge
d
a
s
a
pr
omi
ne
nt
method.
S
i
m
u
l
a
te
d
a
n
ne
a
l
i
n
g
[
3
]
,
a
t
e
c
hn
i
q
ue
m
o
d
e
le
d
a
f
te
r
t
h
e
m
e
t
a
l
lu
r
g
i
c
a
l
a
n
ne
a
li
n
g
p
r
o
c
e
s
s
,
i
s
we
l
l
s
u
i
t
e
d
f
o
r
f
i
n
d
i
n
g
n
e
a
r
-
o
p
t
i
ma
l
s
o
lu
t
i
o
ns
i
n
c
o
mp
l
e
x
s
e
a
r
c
h
s
p
a
c
e
s
.
F
o
r
e
x
a
m
pl
e
,
a
u
t
ho
r
s
i
n
[
4
]
,
[
5
]
h
a
v
e
d
e
mo
n
s
t
r
a
t
e
d
t
h
e
e
f
f
e
c
t
i
ve
n
e
s
s
o
f
s
i
m
u
la
t
e
d
a
n
ne
a
l
in
g
i
n
i
m
p
r
o
v
ing
e
r
r
o
r
c
o
r
r
e
c
t
io
n
p
e
r
f
o
r
ma
n
c
e
.
A
d
d
i
ti
o
n
a
l
l
y
,
C
he
n
e
t
a
l
.
[
6
]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Dual
s
imulat
e
d
anne
ali
ng
s
oft
de
c
ode
r
for
li
ne
ar
b
lock
c
ode
s
(
Hic
ham
T
ahir
i
A
laou
i
)
2777
h
a
v
e
c
o
mb
i
n
e
d
s
i
m
u
la
t
e
d
a
nn
e
a
l
i
n
g
w
i
t
h
g
e
n
e
t
ic
a
lg
o
r
i
t
h
ms
,
w
h
il
e
N
i
h
a
r
mi
n
e
e
t
al
.
[
7
]
i
n
t
r
od
u
c
e
d
a
s
i
m
u
l
a
te
d
a
n
n
e
a
l
i
ng
-
b
a
s
e
d
a
l
g
o
r
it
h
m
d
e
s
ig
n
e
d
f
o
r
s
o
f
t
de
c
is
i
o
n
d
e
c
o
di
n
g
o
f
l
i
ne
a
r
b
l
oc
k
c
o
d
e
s
.
F
ur
ther
mor
e
,
Az
oua
oui
e
t
al.
[
8
]
de
ve
loped
a
ne
w
e
f
f
e
c
ti
ve
tec
hnique
us
ing
the
dua
l
c
ode
to
r
e
duc
e
the
c
ompl
e
xit
y
of
de
c
oding
high
-
r
a
te
c
ode
s
.
T
his
a
ppr
oa
c
h
s
im
pli
f
ies
the
de
c
oding
pr
oc
e
s
s
while
maintaining
high
a
c
c
ur
a
c
y.
De
s
pit
e
thes
e
a
dva
nc
e
ments
,
the
int
e
gr
a
ti
on
o
f
s
im
ulate
d
a
nn
e
a
li
ng
with
dua
l
d
e
c
oding
tec
hniques
r
e
mains
unde
r
e
xplor
e
d
in
the
li
ter
a
tur
e
.
T
his
pa
pe
r
a
im
s
to
c
ombi
ne
the
s
im
ulate
d
a
nne
a
li
n
g
de
c
ode
r
de
ve
loped
by
Niha
r
mi
ne
e
t
al.
[
7
]
with
the
dua
l
de
c
oding
tec
hnique
us
e
d
by
A
z
oua
oui
e
t
al.
[
8]
to
c
r
e
a
te
a
n
e
f
f
icie
nt
a
nd
e
f
f
e
c
ti
ve
e
r
r
or
c
or
r
e
c
ti
on
meth
od
f
or
l
inea
r
block
c
ode
s
.
B
y
int
e
gr
a
ti
ng
the
s
e
tec
hniques
,
we
s
e
e
k
to
e
nha
nc
e
de
c
oding
pe
r
f
or
manc
e
a
nd
im
pr
ove
the
da
ta
tr
a
ns
mi
s
s
ion
r
e
li
a
bil
it
y
ov
e
r
nois
y
c
ha
nne
ls
.
S
ur
p
r
is
ingl
y,
thi
s
ne
w
a
lgor
it
hm
gives
ne
a
r
ly
the
s
a
me
pe
r
f
o
r
manc
e
a
s
in
[
7
]
.
T
he
f
ol
low
ing
s
e
c
t
ions
of
thi
s
pa
pe
r
a
r
e
o
r
ga
niz
e
d
a
s
f
ol
lows
:
s
e
c
ti
on
2
d
is
c
us
s
e
s
the
f
un
da
men
tals
o
f
the
s
i
mul
a
ted
a
nne
a
li
ng
a
l
gor
it
h
m
.
S
e
c
ti
on
3
p
r
e
s
e
nts
ou
r
pr
opos
e
d
de
c
o
de
r
ba
s
e
d
on
the
s
im
ulate
d
a
nne
a
li
ng
pr
oc
e
s
s
a
nd
d
ua
li
ty
pr
ope
r
ty
.
S
e
c
t
ion
4
e
xa
mi
ne
s
pa
r
a
me
ter
tu
nin
g
t
o
f
ind
t
he
op
ti
mal
va
lues
to
wo
r
k
w
it
h
,
s
im
ula
ti
on
r
e
s
ul
ts
,
a
nd
c
o
mpa
r
is
ons
wi
th
ma
in
c
om
pe
ti
t
or
s
'
de
c
ode
r
s
.
F
inal
ly
,
we
c
onc
l
ude
the
pa
pe
r
.
F
igur
e
1.
C
omm
unica
ti
on
s
ys
tem
model
2.
COM
P
RE
HE
NSI
VE
T
HE
OR
E
T
I
CA
L
B
ASI
S
2.
1.
B
as
ic
n
ot
at
ion
s
F
or
the
r
e
s
t
o
f
thi
s
a
r
ti
c
le
,
C
(
n,
k,
d
)
will
de
note
a
l
inea
r
c
ode
with
pa
r
a
mete
r
s
length
n
,
dim
e
ns
ion
k,
e
r
r
or
c
or
r
e
c
ti
on
c
a
pa
bil
it
y
t
a
nd
mi
nim
um
dis
tanc
e
d
ove
r
the
f
ield
F
2
.
T
h
is
c
ode
is
r
e
pr
e
s
e
ntable
by
a
k×
n
matr
ix
G
known
a
s
the
ge
ne
r
a
tor
mat
r
ix.
Fo
r
t
he
s
im
ulate
d
a
nne
a
li
ng
a
lgor
it
hms
,
N
i
is
the
nu
mber
of
it
e
r
a
ti
ons
,
T
s
the
s
tar
ti
ng
tempe
r
a
tur
e
,
T
f
the
f
inal
tempe
r
a
tur
e
,
α
the
c
ooli
ng
r
a
ti
o,
S
0
the
s
tar
t
s
olut
ion
a
nd
N
c
the
nu
mber
o
f
it
e
r
a
t
ions
r
e
quir
e
d
to
r
e
a
c
h
the
f
inal
tempe
r
a
tur
e
T
f
.
F
o
r
ge
ne
ti
c
a
lgor
it
hms
,
N
i
,
N
e
,
a
nd
N
g
r
e
pr
e
s
e
nt
the
s
ize
of
the
population,
the
tot
a
l
of
e
li
te
membe
r
s
,
a
nd
the
ge
ne
r
a
ti
ons
tot
a
l
,
r
e
s
pe
c
ti
ve
ly.
F
or
the
c
ompac
t
ge
ne
ti
c
a
lgor
it
hm
de
c
ode
r
(
C
GA
D)
a
lgor
i
thm
[
9]
,
T
c
r
e
pr
e
s
e
nts
the
a
ve
r
a
ge
numbe
r
of
ge
ne
r
a
ti
ons
.
2.
2.
S
i
m
u
lat
e
d
an
n
e
ali
n
g
S
im
ulate
d
a
nne
a
li
ng
is
a
ve
r
s
a
ti
le
meta
he
ur
is
ti
c
a
lgor
it
hm
wide
ly
us
e
d
f
or
s
olvi
ng
opti
mi
z
a
ti
on
pr
oblems
.
L
e
ve
r
a
ging
the
unde
r
s
tanding
of
a
nne
a
l
ing
f
r
om
the
f
ield
of
meta
ll
u
r
gy,
it
wa
s
f
ir
s
t
int
r
od
uc
e
d
by
Kir
kpa
tr
ick
e
t
al.
[
3]
.
T
he
a
lgo
r
it
hm
is
pa
r
ti
c
u
lar
ly
e
f
f
e
c
ti
ve
in
f
indi
ng
a
ppr
oxim
a
te
s
olut
ions
to
both
c
ombi
na
tor
ial
a
nd
c
onti
nuous
opti
mi
z
a
ti
on
pr
oble
ms
.
I
t
is
a
ls
o
known
f
o
r
it
s
a
bil
it
y
to
e
s
c
a
pe
loca
l
opti
ma,
a
c
omm
on
is
s
ue
in
opti
mi
z
a
ti
on
a
lgo
r
it
hms
,
e
s
pe
c
ially
in
gr
a
dient
-
ba
s
e
d
methods
.
L
oc
a
l
opti
ma
oc
c
ur
whe
n
a
n
a
lgor
it
hm
c
onve
r
ge
s
to
a
s
olut
ion
that
is
opt
im
a
l
withi
n
a
li
mi
ted
r
e
gion
but
not
ne
c
e
s
s
a
r
il
y
the
global
opti
mum
.
S
im
ulate
d
a
nne
a
li
ng
a
ddr
e
s
s
e
s
thi
s
by
pr
oba
bil
is
ti
c
a
ll
y
a
c
c
e
pti
ng
wor
s
e
s
olut
ions
,
a
ll
ow
ing
it
to
e
xplor
e
the
s
olut
ion
s
pa
c
e
mor
e
thor
oughly
.
Due
to
thes
e
pr
ope
r
ti
e
s
,
s
im
ulate
d
a
nne
a
li
ng
a
lgor
it
h
m
ha
s
a
s
igni
f
ica
nt
im
pa
c
t
in
va
r
ious
f
ields
,
wi
th
a
ppli
c
a
ti
ons
to
c
ombi
na
tor
ial
opti
mi
z
a
ti
on
p
r
oblems
,
inclu
ding
but
not
li
mi
ted
to
the
tr
a
ve
li
ng
s
a
les
man
pr
oblem
(
T
S
P
)
[
10]
–
[
12]
a
nd
the
qua
dr
a
ti
c
a
s
s
ignm
e
nt
pr
oble
m
(
QA
P
)
[
13]
,
[
14
]
,
ve
r
y
lar
ge
-
s
c
a
le
int
e
gr
a
ti
on
(
VL
S
I
)
c
ir
c
uit
de
s
ign
[
15
]
–
[
17]
,
to
na
me
a
f
e
w
.
He
r
e
is
a
de
s
c
r
ipt
ion
of
the
s
im
ulate
d
a
nne
a
li
ng
a
lgo
r
it
hm:
a)
I
nit
ializa
ti
on:
s
tar
t
with
a
n
ini
ti
a
l
s
olut
ion
a
nd
s
e
t
a
n
ini
ti
a
l
tempe
r
a
tur
e
(
T
)
a
long
with
a
c
ool
ing
s
c
he
dule
to
de
c
r
e
a
s
e
T
ove
r
ti
me.
b)
I
ter
a
ti
on:
r
e
pe
a
t
unti
l
a
s
toppi
ng
c
r
it
e
r
ion
is
met
(
e
.
g.
,
max
it
e
r
a
ti
ons
o
r
low
tempe
r
a
tur
e
)
:
i)
Ne
ighbor
hood
ge
ne
r
a
ti
on:
a
pply
a
pe
r
tu
r
ba
ti
on
to
t
he
c
ur
r
e
nt
s
olut
ion
,
yielding
a
ne
ighbor
ing
s
olut
io
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
4
,
Augus
t
2025
:
277
6
-
2787
2778
ii)
Obje
c
ti
ve
f
unc
ti
on
e
va
luation:
the
objec
ti
ve
f
unc
ti
on
va
lue
of
the
ne
w
s
olut
ion
is
c
a
lcula
ted
to
a
s
s
e
s
s
it
s
qua
li
ty.
Optim
iza
ti
on
pr
oblems
typi
c
a
ll
y
a
im
to
mi
nim
ize
thi
s
va
lue.
iii)
Ac
c
e
ptanc
e
or
r
e
jec
ti
on:
‒
A
be
tt
e
r
ne
w
s
olut
ion
is
a
c
c
e
pted
a
s
the
c
ur
r
e
nt
on
e
.
‒
A
wor
s
e
s
olut
ion
is
a
c
c
e
pted
with
a
pr
oba
bil
i
ty
e
xp(
-
Δ
E
/
T
)
,
whe
r
e
Δ
E
is
the
dif
f
e
r
e
nc
e
in
ob
jec
ti
ve
f
unc
ti
on
va
lues
.
c)
C
ooli
ng:
de
c
r
e
a
s
e
the
tempe
r
a
tur
e
a
c
c
or
ding
to
the
c
ooli
ng
s
c
he
dule,
whic
h
c
ont
r
ols
the
ba
la
nc
e
be
twe
e
n
e
xplor
a
ti
on
(
higher
tempe
r
a
tur
e
,
mo
r
e
a
c
c
e
ptanc
e
of
wor
s
e
s
olut
ions
)
a
nd
e
xploi
tation
(
lo
we
r
tempe
r
a
tur
e
,
les
s
a
c
c
e
ptanc
e
of
wor
s
e
s
olut
ions
)
.
d)
T
e
r
mi
na
ti
on:
s
top
ba
s
e
d
on
r
e
a
c
hing
a
maximu
m
number
of
it
e
r
a
ti
ons
,
a
s
pe
c
if
ic
tempe
r
a
tu
r
e
,
or
a
s
a
ti
s
f
a
c
tor
y
s
olut
ion
qua
li
ty.
e)
Output:
r
e
tur
n
the
be
s
t
s
olut
ion
f
ound.
T
he
s
ubs
e
que
nt
ps
e
udoc
ode
il
lus
tr
a
tes
a
ba
s
ic
im
p
leme
ntation
of
the
s
im
ulate
d
a
nne
a
li
ng
a
lgor
it
hm
,
with
a
ll
the
pr
e
vious
s
teps
:
Initialization of parameters (N
i
, T
s
, T
f
, S
0
)
Set T←T
s
and
S←S
0
While (T >T
f
)
{
While (iteration < N
i
)
{
Select an adjacent solution at random (s
n
);
Evaluate ΔE
nergy
= E
nergy
(s
n
)
–
E
nergy
(s);
If Δ
Energy
≤ 0 then s←s
n
;
else if random(0,1) ≤ Exp(
-
Δ
Energy
/T) ) then
s←s
n
; end if;
end if;
iteration←iteration+1;
}
T←cooling(T);
}
Ke
y
c
omponents
a
nd
pa
r
a
mete
r
s
of
the
s
im
ulate
d
a
nne
a
li
ng
a
lgor
it
hm
include
the
ini
ti
a
l
tempe
r
a
tur
e
,
c
ooli
ng
s
c
he
dule,
ne
ighbor
hood
ge
ne
r
a
ti
on
s
tr
a
tegy
,
a
nd
a
c
c
e
ptanc
e
pr
oba
bil
it
y
c
a
lcula
ti
on.
P
r
ope
r
ly
tuni
ng
thes
e
pa
r
a
mete
r
s
is
c
r
uc
ial
to
the
a
lgor
i
thm
's
e
f
f
e
c
ti
ve
ne
s
s
in
f
indi
ng
high
-
qua
li
ty
s
olut
ions
to
opti
mi
z
a
ti
on
pr
oblems
.
T
he
a
ppr
oa
c
h
is
e
xplaine
d
in
mor
e
d
e
tail
in
s
e
c
ti
on
3
.
2.
2.
3.
Dual
i
t
y
p
r
op
e
r
t
y
f
or
d
e
c
od
in
g
T
o
e
nc
ode
a
mes
s
a
ge
m
=
{m
i
}
1
k
,
we
c
a
n
us
e
(
1
)
:
=
(
1)
W
he
r
e
c
is
the
c
ode
wor
d.
Additi
ona
ll
y
,
in
o
r
de
r
to
de
ter
mi
ne
whe
ther
a
s
pe
c
if
ic
ve
c
tor
is
a
va
li
d
c
od
e
wor
d,
we
int
r
oduc
e
a
(n
-
k
)
×
n
ma
tr
ix
de
noted
by
H
(
pa
r
it
y
-
c
he
c
k
matr
ix
(
P
C
M
)
)
.
T
his
pa
r
ti
c
ular
mat
r
ix
ha
s
the
f
oll
owing
pr
ope
r
ty:
∀
2
,
v
is
a
c
ode
w
or
d
if
and
on
ly
if
:
ᵀ
=
0
(
2)
C
ons
ider
the
s
c
e
na
r
io
whe
r
e
we
tr
a
ns
mi
t
a
c
ode
wor
d
c
=
{c
i
}
1
n
us
ing
B
P
S
K
modul
a
ti
on
.
L
e
t's
de
note
z
=
{z
i
}
1
n
a
s
the
modul
a
ted
s
ignal
tr
a
ns
mi
tt
e
d
o
ve
r
a
Ga
us
s
ian
c
ha
nne
l,
s
ubjec
t
to
indepe
nde
nt
nois
e
c
omponents
n=
{n
i
}
1
n
.
He
r
e
,
both
z
a
nd
n
a
r
e
s
e
que
nc
e
s
that
a
r
e
s
tatis
ti
c
a
ll
y
indepe
nde
nt.
S
pe
c
if
ica
ll
y,
e
a
c
h
n
i
is
no
r
mally
d
is
tr
ibut
e
d
with
mea
n
0
a
nd
va
r
ianc
e
N
0
(
∼
(
Ɲ
(
0
,
0
2
)
)
)
whe
r
e
N
0
r
e
p
r
e
s
e
nts
the
nois
e
powe
r
de
ns
it
y.
T
he
r
e
c
e
ived
s
ignal,
de
noted
a
s
r
=
{r
i
}
1
n
,
is
given
by
the
e
qua
ti
on
r
=
z
+
n
.
W
e
in
tr
oduc
e
v
=
{v
i
}
1
n
a
s
the
binar
y
ha
r
d
de
c
is
ions
de
r
ived
f
r
om
r
(
qua
nti
z
a
ti
on
of
r
)
.
F
ur
ther
mo
r
e
,
we
e
xpr
e
s
s
the
e
r
r
or
s
yndr
ome
s
=
{s
i
}
1
n
-
k
a
s
in
(
3)
:
=
(
3)
I
n
the
latter
e
xpr
e
s
s
ion
the
s
yndr
ome
s
is
obtai
ne
d
ba
s
e
d
on
the
ha
r
d
de
c
is
ions
made
f
r
om
the
r
e
c
e
ived
s
ignal
r
.
“
s
=
0
”
mea
ns
that
the
r
e
c
e
ived
s
ignal
c
or
r
e
s
ponds
to
a
va
li
d
c
ode
wor
d
,
indi
c
a
ti
ng
a
n
e
r
r
or
-
f
r
e
e
tr
a
ns
mi
s
s
ion.
How
e
ve
r
,
in
the
pr
e
s
e
nc
e
of
tr
a
ns
mi
s
s
ion
e
r
r
or
s
,
our
de
c
ode
r
e
nde
a
vor
s
to
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Dual
s
imulat
e
d
anne
ali
ng
s
oft
de
c
ode
r
for
li
ne
ar
b
lock
c
ode
s
(
Hic
ham
T
ahir
i
A
laou
i
)
2779
de
ter
mi
ne
the
c
ode
wor
d
ĉ
that
maximi
z
e
s
the
pr
oba
bil
it
y
P
(
c
|r
)
with
in
the
c
ode
s
pa
c
e
C
.
Give
n
that
a
ll
c
ode
wor
ds
ha
ve
a
n
e
qua
l
li
ke
li
hood
o
f
be
ing
tr
a
ns
mi
tt
e
d,
we
c
a
n
wr
it
e
:
(
ĉ
|
)
=
(
|
)
=
(
|
)
(
)
/
(
)
(
4)
C
ons
ider
ing
a
dis
c
r
e
te
memo
r
yles
s
c
ha
nne
l
wit
h
a
ddit
ive
white
Ga
us
s
ian
nois
e
,
whe
r
e
binar
y
a
nti
poda
l
s
ignals
a
r
e
tr
a
ns
mi
tt
e
d
,
with
e
a
c
h
s
ymbo
l
be
ing
indepe
nde
ntl
y
a
f
f
e
c
ted
by
nois
e
,
the
maxi
mi
z
a
ti
on
of
P
(
r
|ĉ
)
oc
c
ur
s
whe
n
we
mi
nim
ize
the
s
qua
r
e
d
nor
m
o
f
the
di
f
f
e
r
e
nc
e
be
twe
e
n
r
a
nd
ĉ
(
∑
(
−
)
2
=
1
)
(
or
s
qua
r
e
d
E
uc
li
de
a
n
dis
tanc
e
be
twe
e
n
r
a
nd
ĉ
)
a
s
e
xplaine
d
in
[
18]
,
[
19]
.
C
ons
e
que
ntl
y,
the
c
ompl
e
x
tas
k
of
maximum
-
li
ke
li
hood
de
c
oding
s
im
pli
f
ies
to
the
mor
e
s
tr
a
ight
f
or
wa
r
d
ne
a
r
e
s
t
ne
ighbor
de
c
oding
,
us
ing
the
E
uc
li
de
a
n
metr
ic.
T
o
f
o
r
malize
thi
s
r
e
duc
ti
on
,
we
c
a
n
e
xpr
e
s
s
the
s
of
t
-
de
c
is
ion
de
c
oding
pr
oblem
a
s
(
5)
[
20]
:
=
{
}
1
;
ℎ
ℎ
∑
(
−
)
2
=
1
(
5)
T
his
opti
mi
z
a
ti
on
p
r
oblem
invol
ve
s
n
va
r
iable
s
,
w
it
h
only
k
va
r
iable
s
a
s
ge
ne
r
a
tor
ba
s
e
(
k
mos
t
indepe
nde
nt
a
nd
r
e
li
a
ble
bit
s
)
.
I
ns
tea
d
of
s
olvi
ng
thi
s
opti
mi
z
a
ti
on
pr
oblem
on
t
he
c
ode
s
pa
c
e
,
our
a
ppr
oa
c
h
is
to
s
e
a
r
c
h
f
o
r
the
e
r
r
or
ve
c
tor
,
de
noted
a
s
e
,
that
opti
mi
z
e
s
the
s
olut
ion.
T
o
th
is
e
nd,
we
will
c
ons
tr
uc
t
the
e
r
r
or
ve
c
tor
e
thr
ough
a
he
ur
is
ti
c
method
that
leve
r
a
ge
s
the
dua
l
pr
ope
r
ty
[
8]
.
T
he
e
r
r
or
e
ha
s
n
va
r
iable
s
,
wi
th
only
k
be
ing
indepe
nde
nt.
B
y
us
ing
thes
e
k
va
r
iable
s
a
nd
leve
r
a
ging
the
a
lgeb
r
a
ic
s
tr
u
c
tur
e
of
the
c
ode
,
we
c
a
n
de
duc
e
the
k
r
e
maining
va
r
iable
s
.
B
y
doing
s
o,
we
c
a
n
de
duc
e
,
to
a
c
e
r
tain
de
g
r
e
e
,
wha
t
wa
s
o
r
igi
na
ll
y
s
e
nt
us
in
g
(
6)
:
=
+
(
6)
S
ince
the
P
C
M
c
a
n
be
wr
i
tt
e
n
a
s
H
=
[
A
I
n
-
k
],
whe
r
e
A
is
a
bina
r
y
matr
ix
of
s
ize
(
n
–
k)
×
k
,
we
c
a
n
wr
it
e
:
(
+
)
=
0
⇔
=
(
7)
W
e
de
f
ine
the
r
e
li
a
ble
inf
or
mation
s
e
t
a
s
the
c
oll
e
c
ti
on
of
the
k
mos
t
r
e
li
a
ble
pos
it
ions
withi
n
the
r
e
c
e
ived
s
ignal
r
=
{r
i
}
1
n
.
Us
ing
thi
s
r
e
li
a
bil
it
y
inf
o
r
mation
s
e
t,
the
e
r
r
o
r
ve
c
tor
c
a
n
be
r
e
pr
e
s
e
nted
a
s
e
=
(e
X
, e
Y
)
,
whe
r
e
X
r
e
pr
e
s
e
nts
the
r
e
li
a
ble
inf
o
r
mation
s
e
t,
a
nd
Y
=
{
1
n}
\
X.
C
ons
e
que
ntl
y,
r
e
lation
(
7)
c
a
n
be
e
xpr
e
s
s
e
d
a
s
(
8)
:
(
,
)
(
,
−
)
=
⇔
=
+
(
8)
Ha
ving
the
f
ir
s
t
pa
r
t
o
f
the
e
r
r
or
ve
c
tor
(
e
X
)
,
a
n
d
us
ing
the
e
qua
ti
on
a
bove
,
we
c
a
n
de
duc
e
the
s
e
c
ond
pa
r
t
of
the
e
r
r
or
e
Y
a
nd
c
ompl
e
te
the
whole
e
r
r
or
ve
c
tor
e
=
(
e
X
,
e
Y
)
.
W
e
c
a
n
then
ve
r
i
f
y
that
(
v
+
e
)
is
a
va
li
d
c
ode
wor
d.
Our
e
nde
a
vor
is
to
s
e
a
r
c
h,
a
mongs
t
a
ll
the
e
r
r
or
s
e
t
E
rr
(
s
)
w
ith
s
yndr
ome
s
,
f
o
r
the
e
r
r
or
ve
c
tor
e
whic
h
mi
nim
ize
s
the
s
qua
r
e
d
nor
m
of
the
di
f
f
e
r
e
n
c
e
be
twe
e
n
the
r
e
c
e
ived
s
ignal
r
a
nd
the
r
e
late
d
c
ode
wor
d
c
=
(
v+
e
)
whic
h
is
given
by
(
9)
:
T
he
s
qua
r
e
d
nor
m
of
the
dif
f
e
r
e
nc
e
be
twe
e
n
r
a
nd
c
:
∑
(
−
)
2
=
1
(
9)
T
a
king
int
o
a
c
c
ount
the
a
na
lys
is
pr
ovided
e
a
r
li
e
r
,
the
a
lgo
r
it
hm's
de
s
c
r
ipt
ion
a
nd
s
teps
a
r
e
outl
ine
d
in
the
f
oll
owing
s
e
c
ti
on.
3.
T
HE
DU
AL
S
I
M
UL
A
T
E
D
AN
NE
A
L
I
NG
S
OF
T
DE
CODE
R
AL
GO
RI
T
HM
3.
1.
Cons
t
r
u
c
t
ion
of
t
h
e
DSAS
D
algori
t
h
m
B
uil
ding
on
the
pr
e
vious
s
e
c
ti
on,
whe
r
e
we
ou
tl
ined
the
two
f
unda
menta
l
c
omponents
of
our
pr
opos
e
d
a
lgor
it
hm,
the
ne
xt
s
e
c
ti
on
pr
e
s
e
nts
the
dua
l
s
im
ulate
d
a
nne
a
li
ng
s
of
t
de
c
ode
r
(
DSAS
D)
.
T
a
ble
1
outl
ines
the
c
ompl
e
te
mapping
be
twe
e
n
the
p
r
op
os
e
d
a
lgor
it
hm
a
nd
the
phys
ica
l
s
im
ulate
d
a
nne
a
l
ing
.
T
he
e
ne
r
gy
is
r
e
pr
e
s
e
nted
a
s
the
E
uc
li
de
a
n
dis
tanc
e
b
e
twe
e
n
the
r
e
c
e
ived
wor
d
a
nd
a
c
ode
wor
d,
a
nd
th
e
s
tate
is
r
e
pr
e
s
e
nted
a
s
a
k
-
bit
ve
c
tor
.
T
he
lowe
s
t
e
ne
r
gy
s
tate
c
or
r
e
s
ponds
to
the
ne
a
r
e
s
t
c
ode
wor
d.
T
a
ble
1.
M
a
p
ping
be
twe
e
n
the
pr
opos
e
d
a
lgor
it
hm
(
DSAS
D)
a
nd
phys
ica
l
s
im
ulate
d
a
nne
a
li
ng
a
lgor
i
thm
P
hys
ic
a
l
s
im
ul
a
te
d a
nne
a
li
ng
D
S
A
S
D
E
ne
r
gy
T
he
E
uc
li
de
a
n di
s
ta
n
c
e
be
twe
e
n a
c
ode
w
or
d a
nd t
he
r
e
c
e
iv
e
d w
or
d
S
ta
te
k
-
bi
t
ve
c
to
r
F
in
a
l
s
ta
te
T
he
de
c
od
e
d w
or
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
4
,
Augus
t
2025
:
277
6
-
2787
2780
T
he
DSAS
D
a
lgo
r
it
hm
is
de
s
c
r
ibed
a
s
f
oll
ows
:
i)
S
tep
1:
r
a
ndoml
y
ge
ne
r
a
te
a
nd
e
nc
ode
k
bina
r
y
in
f
or
mation
bit
s
us
ing
the
c
ode
's
matr
ix
G
,
r
e
s
ult
ing
i
n
a
n
n
-
bit
ve
c
tor
.
A
f
ter
tr
a
ns
f
or
m
ing
e
a
c
h
0
to
a
1
a
nd
e
a
c
h
1
to
a
-
1
,
int
r
oduc
e
s
im
ulate
d
Ga
us
s
ian
nois
e
to
pr
od
uc
e
a
r
e
c
e
ived
ve
c
tor
,
de
noted
a
s
r
,
whe
r
e
r
be
longs
to
ℝ
n
.
ii)
S
tep
2:
a
f
ter
r
e
c
e
ivi
ng
the
s
e
que
nc
e
r
=
{r
i
}
1
n
,
m
a
ke
a
binar
y
ha
r
d
de
c
is
ion
f
or
thi
s
r
e
c
e
ived
s
ignal
r
=
{r
i
}
1
n
to
obtain
v
=
{v
i
}
1
n
:
=
{
1
,
<
0
0
,
≥
0
iii)
S
tep
3:
c
omput
e
the
s
yndr
ome
a
s
s
=
v
Hᵀ
.
I
f
s
e
q
ua
ls
z
e
r
o,
output
v
a
nd
ter
mi
na
te;
other
wis
e
,
p
r
oc
e
e
d
f
ur
ther
.
iv)
S
tep
4:
a
pply
a
pe
r
mut
a
ti
on
to
the
c
oor
dinate
s
of
the
r
e
c
e
ived
ve
c
tor
r
,
e
ns
ur
ing
that
the
las
t
(
n
-
k)
pos
it
ions
hold
the
lea
s
t
r
e
li
a
ble
li
ne
a
r
ly
indepe
nde
nt
c
omponents
of
r
.
‒
r
i
is
c
ons
ider
e
d
mor
e
r
e
li
a
ble
than
r
j
if
|r
i
|
>
|
r
j
|
ba
s
e
d
on
the
a
s
s
umpt
ion
that
the
da
ta
is
c
o
r
r
upted
by
a
ddit
ive
white
Ga
us
s
ian
nois
e
dur
ing
tr
a
ns
mi
s
s
ion.
‒
S
or
t
the
s
e
que
nc
e
s
r
=
{r
i
}
1
n
in
de
s
c
e
nding
o
r
de
r
o
f
r
e
li
a
bil
it
y,
then
a
pply
a
s
e
c
ond
pe
r
mut
a
ti
on
that
lets
the
las
t
n
-
k
e
leme
nts
of
r
be
the
lea
s
t
li
ne
a
r
l
y
indepe
nde
nt
e
leme
nts
,
to
c
r
e
a
te
ne
w
s
e
que
nc
e
s
r
’
=
{r
’
i
}
1
n
.
L
e
t's
de
note
π
a
s
the
pe
r
mut
a
ti
on
mappi
ng
r
’
=
π(
r
)
.
‒
Apply
thi
s
pe
r
m
utation
π
to
H
to
obtain
H’
(
H
’
=
π(
H)
)
a
nd
v
’
(
v
’
=
π
(
v
)
)
.
‒
Us
e
Ga
us
s
i
a
n
e
li
mi
na
ti
on
on
H’
to
de
r
ive
a
s
ys
tem
a
ti
c
matr
ix.
v)
S
tep
5:
Ge
ne
r
a
te
a
n
e
r
r
or
ve
c
tor
of
k
-
bit
s
with
s
a
s
s
yndr
ome:
‒
T
he
f
i
r
s
t
ge
ne
r
a
ted
e
r
r
o
r
ve
c
tor
c
a
n
be
the
z
e
r
o
ve
c
tor
.
‒
R
a
ndoml
y
ge
ne
r
a
te
a
n
e
r
r
o
r
ve
c
tor
e
X
of
k
-
bit
s
.
‒
T
he
s
e
c
ond
pa
r
t
o
f
the
ve
c
to
r
is
e
Y
=e
X
A
T
+s
‒
T
he
e
r
r
o
r
ve
c
tor
e
is
f
or
med
a
s
(e
X
,
e
Y
)
vi)
S
tep
6:
a
pply
the
s
im
ulate
d
a
nne
a
li
ng
a
lgor
it
h
m
a
nd
us
e
(
8)
to
ge
t
the
be
s
t
e
r
r
o
r
c
a
ndidate
e
b
e
s
t
.
T
his
a
lgor
it
hm
is
de
picte
d
a
f
ter
wa
r
ds
.
vii
)
S
tep
7:
Obta
in
the
c
ode
wor
d:
‒
T
he
obtaine
d
c
ode
wor
d
c
’
=
v
'
+
e
b
e
s
t
is
a
s
s
oc
iate
d
with
the
mat
r
ix
H’
,
he
nc
e
we
e
s
ti
mate
the
c
ode
wor
d
ĉ
to
be
:
ĉ
=
−
1
(
’
)
(
10)
A
s
i
mul
a
ted
a
nne
a
l
ing
a
lgo
r
it
hm
,
us
i
ng
s
qua
r
e
d
E
u
c
li
de
a
n
dis
ta
nc
e
a
s
it
s
objec
ti
ve
f
u
nc
ti
on
,
f
inds
t
he
c
los
e
s
t
c
ode
w
or
d
.
Ou
r
s
im
ula
ted
a
nne
a
li
n
g
met
ho
d,
us
e
d
in
s
tep
6
,
d
if
f
e
r
s
f
r
om
c
las
s
ica
l
s
im
u
late
d
a
nne
a
li
ng
by
us
ing
r
e
li
a
bil
it
y
i
nf
o
r
ma
ti
o
n
to
gu
ide
s
olu
ti
o
n
ge
ne
r
a
t
ion
,
r
a
the
r
than
r
a
ndo
m
bi
t
f
l
ippi
ng
a
s
f
ol
lows
:
select_neighbor() {
For each bit i = 1 to k
If
(Random
(between
0
and
1)
>
1
1
+
e
x
p
(
−
2
r
i
′
N
0
)
)
th
en
(switch
the
bit r’
i
)
End for }
F
inally,
the
DSAS
D
a
lgo
r
it
hm
c
a
n
be
r
e
pr
e
s
e
nted
with
the
f
ol
lowing
ps
e
udoc
ode
:
Set the parameters {N
i
,T
s
,T
f
, α
, S
0
}; Set T = T
0
and S = S
0
While (T >T
f
) {
While (iteration < N
i
) {
S
n
= select_
neighbor();
ΔEnergy
= Energy(S
n
)
-
Energy(S);
error
= Evalu
a
teCorrectedError();
if (error < T) then break;
if
ΔEnergy ≤ 0
then S = S
n
;
else if random(0,1)
≤
Exp(
-
ΔEnergy/T) then
S = S
n
; end if
end if
iteration
= iteration + 1;
}
T = α
* T
}
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Dual
s
imulat
e
d
anne
ali
ng
s
oft
de
c
ode
r
for
li
ne
ar
b
lock
c
ode
s
(
Hic
ham
T
ahir
i
A
laou
i
)
2781
w
he
r
e
e
ne
r
gy(
s
)
=
∑
(
′
−
)
2
=
1
,
a
nd
c
=
{c
i
}
1
n
is
the
r
e
late
d
c
ode
wor
d.
3.
2.
DSAS
D
algorit
h
m
p
ar
am
e
t
e
r
t
u
n
i
n
g
Optim
izing
our
a
lgor
it
hm's
pa
r
a
mete
r
s
{α
i
, N
i
,
T
s
,
T
f
}
is
a
ke
y
c
ha
ll
e
nge
.
W
hil
e
pr
oba
bil
is
ti
c
models
li
ke
M
a
c
Ka
y's
[
21]
c
ould
be
us
e
d,
we
ins
tea
d
c
onduc
t
mul
ti
ple
s
im
ulations
due
to
our
a
s
s
umpt
ion
of
pa
r
a
mete
r
indepe
nde
nc
e
.
W
e
e
va
luate
bit
e
r
r
or
r
a
te
(
B
E
R
)
a
ga
ins
t
s
ignal
to
nois
e
r
a
t
io
(
S
NR
)
,
va
r
ying
one
pa
r
a
mete
r
a
t
a
ti
me
while
holdi
ng
the
other
s
a
t
their
de
f
a
ult
va
lues
,
a
s
outl
ined
in
T
a
ble
2.
T
he
va
r
i
a
ti
ons
of
pa
r
a
mete
r
s
α
a
nd
Ni
a
r
e
il
lus
tr
a
ted
in
F
igu
r
e
2,
whi
c
h
c
ons
is
ts
of
two
s
ub
-
f
igur
e
s
:
‒
P
a
r
a
mete
r
α
:
a
na
lyzing
F
igur
e
2(
a
)
s
ho
ws
that
c
hoos
ing
α
=
0.
95
is
a
n
e
f
f
e
c
ti
ve
opti
on
f
or
the
c
ooli
ng
r
a
ti
o,
a
pp
r
oa
c
hing
the
be
s
t
pe
r
f
or
manc
e
.
I
n
pr
a
c
ti
c
a
l
a
ppli
c
a
ti
ons
,
a
de
li
be
r
a
te
s
low
c
ooli
ng
mec
ha
ni
s
m
is
be
ne
f
icia
l,
a
s
it
he
lps
in
identi
f
ying
a
nd
uti
li
z
ing
c
ode
wor
ds
with
low
E
uc
li
de
a
n
dis
t
a
nc
e
.
‒
P
a
r
a
mete
r
N
i
:
we
typi
c
a
ll
y
de
ter
mi
ne
the
opti
mal
number
o
f
it
e
r
a
ti
ons
e
xpe
r
im
e
ntally.
Our
s
im
ulati
ons
,
s
hown
in
F
igur
e
2
(
b)
,
indi
c
a
te
that
s
e
tt
ing
N
i
to
25
0
yields
ne
a
r
-
opti
mal
r
e
s
ult
s
.
T
he
e
f
f
e
c
ts
of
pa
r
a
mete
r
s
T
s
a
nd
T
f
on
s
ys
tem
pe
r
f
or
manc
e
a
r
e
de
picte
d
in
F
igur
e
3
,
c
ompr
is
ing
two
s
ub
-
f
igur
e
s
:
‒
P
a
r
a
mete
r
T
s
:
e
xa
mi
ning
the
s
im
ulation
r
e
s
ult
s
pr
e
s
e
nted
in
F
igur
e
3(
a
)
,
it
is
e
vident
that
the
opti
mal
ini
ti
a
l
tempe
r
a
tu
r
e
T
s
is
0
.
2.
Give
n
the
s
igni
f
ic
a
nt
im
pa
c
t
of
thi
s
pa
r
a
mete
r
on
a
c
c
ur
a
c
y,
va
r
i
ous
meth
ods
f
or
e
s
ti
mating
T
s
e
f
f
e
c
ti
ve
ly
ha
ve
be
e
n
pr
opos
e
d,
s
uc
h
a
s
thos
e
int
r
oduc
e
d
in
[
22]
.
‒
P
a
r
a
me
te
r
T
f
:
i
n
t
he
c
o
n
tex
t
o
f
p
hys
ics
,
th
e
c
o
nc
e
p
t
o
f
a
f
r
e
e
z
in
g
t
e
m
pe
r
a
tu
r
e
is
i
nt
ui
t
ive
l
y
us
e
d
t
o
a
c
h
ie
ve
e
q
ui
l
ib
r
iu
m
.
Ou
r
s
i
m
ula
t
io
ns
,
a
s
s
ho
wn
in
F
i
gu
r
e
3
(
b
)
,
c
o
n
f
i
r
m
t
his
ide
a
,
de
mo
ns
t
r
a
t
in
g
th
a
t
o
p
ti
ma
l
pe
r
f
o
r
man
c
e
is
c
o
ns
is
te
nt
ly
a
tt
a
i
ne
d
a
c
r
os
s
v
a
r
i
ous
S
NR
va
lues
w
he
n
t
he
te
m
pe
r
a
tu
r
e
de
c
r
e
a
s
e
s
to
T
f
=
0
.
00
1
.
T
a
ble
2.
P
a
r
a
mete
r
va
lues
of
the
DSAS
D
a
lgor
i
th
m
P
a
r
a
me
te
r
V
a
lu
e
D
e
f
a
ul
t
c
ode
B
C
H
(
63,45,7)
C
ha
nne
l
A
W
G
N
M
odul
a
ti
on
B
P
S
K
M
in
im
um num
be
r
of
bi
t
e
r
r
or
s
200
M
in
im
um num
be
r
of
bl
oc
ks
1000
N
i
250
T
s
0.2
T
f
0.001
α
0.95
(
a
)
(
b)
F
igur
e
2.
Va
r
iation
of
(
a
)
pa
r
a
mete
r
α
a
nd
(
b)
num
be
r
of
it
e
r
a
ti
ons
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
4
,
Augus
t
2025
:
277
6
-
2787
2782
(
a
)
(
b)
F
igur
e
3.
E
vo
lut
ion
o
f
(
a
)
pa
r
a
mete
r
T
s
a
nd
(
b)
pa
r
a
mete
r
T
f
3.
3.
Com
p
lexit
y
an
a
lys
is
DSAS
D
a
lgor
it
hm's
s
tep
4
ha
s
O(
k²n)
ti
me
c
ompl
e
xit
y
[
23]
.
How
e
ve
r
,
pa
r
a
ll
e
li
z
a
ti
on
c
a
n
r
e
duc
e
the
ti
me
c
ompl
e
xit
y
to
O(
kn
)
,
a
c
os
t
ne
gli
gibl
e
c
omp
a
r
e
d
to
s
tep
6’
s
(
N
i
N
c
nk
)
ti
me
c
ompl
e
xit
y
.
As
s
hown
in
T
a
ble
3,
the
c
omput
a
ti
ona
l
c
os
t
o
f
the
C
ha
s
e
-
2
[
24]
a
nd
s
of
t
de
c
oding
ba
s
e
d
ge
ne
ti
c
a
lgor
it
hm
(
S
D
GA
)
[
25]
a
lgor
it
hms
s
c
a
les
e
xpone
nti
a
ll
y
with
t.
T
he
r
e
f
or
e
,
c
ode
s
with
high
e
r
r
o
r
c
or
r
e
c
ti
on
c
a
pa
bil
it
ies
pr
e
s
e
nt
the
mos
t
c
ha
ll
e
nging
c
omput
a
ti
ona
l
c
ompl
e
xit
y
a
nd
tend
to
e
xhibi
t
s
ubopti
mal
pe
r
f
or
manc
e
a
s
n
i
nc
r
e
a
s
e
s
.
C
onve
r
s
e
ly,
the
DSA
S
D,
s
im
ulate
d
a
nne
a
li
ng
s
of
t
de
c
ode
r
(
S
ASD)
[
7]
,
dua
l
domain
de
c
oding
ge
ne
ti
c
a
lgor
it
hm
(
DD
GA
)
[
8
]
,
C
GA
D
[
9]
,
M
a
ini
[
20]
,
a
nd
ge
ne
ti
c
a
lgor
it
hm
f
or
de
c
oding
s
ys
tema
ti
c
block
c
ode
s
(
AutDA
G)
[
26
]
a
lgor
i
thm
s
e
xhibi
t
l
inea
r
c
ompl
e
xi
ty,
with
r
e
s
pe
c
t
to
e
it
he
r
n
o
r
k
.
T
a
ble
3.
C
ompar
is
on
of
the
DSAS
D
a
lgor
it
hm
c
ompl
e
xit
y
with
othe
r
s
P
a
r
a
me
te
r
V
a
lu
e
C
ha
s
e
-
2
O
(
2
t
n²l
og
2
(
n)
)
D
D
G
A
O
(
N
i
N
g
[
k(
n −
k)
+
l
og(
N
i
)
]
)
M
a
in
i
O
(
N
i
N
g
[
kn +
l
og(
N
i
)
]
)
A
ut
D
A
G
O
(
N
i
N
g
kn)
S
D
G
A
O
(
2
t
(N
i
N
g
[
kn² +
kn +
l
og(
N
i
)
]
)
)
C
G
A
D
O
(
T
c
k(
n
-
k)
)
S
A
S
D
O
(
N
i
N
c
kn)
D
S
A
S
D
O
(
N
i
N
c
nk)
4.
RE
S
UL
T
S
AN
D
DI
S
CU
S
S
I
ON
T
his
s
tudy
e
xa
mi
ne
s
the
im
pa
c
t
o
f
c
ombi
n
ing
th
e
dua
l
pr
ope
r
ty
wi
th
s
im
ulate
d
a
nne
a
li
ng
in
the
DSAS
D
a
lgor
it
hm.
Although
pr
e
vious
r
e
s
e
a
r
c
h
ha
s
a
ppli
e
d
the
dua
l
pr
ope
r
ty
with
va
r
ious
de
c
ode
r
s
,
thi
s
s
pe
c
if
ic
c
ombi
na
ti
on
ha
s
not
be
e
n
pr
e
vious
ly
e
xplor
e
d.
T
he
DSAS
D
a
lgor
it
h
m
wa
s
im
pleme
nted
in
C
,
with
f
igur
e
s
ge
ne
r
a
ted
us
ing
o
c
tave
[
27]
.
Our
s
im
ul
a
ti
ons
we
r
e
pe
r
f
or
med
on
a
wo
r
ks
tation
with
a
n
I
ntel
C
or
e
(
T
M
)
i7
-
6920HQ
pr
oc
e
s
s
or
with
16
GB
of
m
e
mor
y
c
locke
d
a
t
2.
90
GH
z
,
r
unning
Ubuntu
18.
0
4.
6
L
T
S
x86_64
ope
r
a
ti
ng
s
ys
tem.
Optim
a
l
va
lues
f
or
a
lgor
it
hm
pa
r
a
mete
r
s
we
r
e
s
e
lec
ted
a
s
outl
ined
in
s
e
c
t
ion
3.
2.
P
e
r
f
or
manc
e
wa
s
a
s
s
e
s
s
e
d
ba
s
e
d
on
B
E
R
a
s
a
f
unc
ti
on
of
S
NR
(
E
b/N0)
.
W
e
c
onduc
ted
a
c
ompar
a
ti
ve
a
na
lys
is
of
the
pr
opo
s
e
d
DSAS
D
a
lgor
it
hm
a
ga
ins
t
c
las
s
ica
l
s
im
ulate
d
a
nne
a
li
ng
,
S
ASD
[
7]
,
a
nd
other
de
c
ode
r
s
in
the
s
a
me
c
a
tegor
y.
T
he
s
im
ulations
we
r
e
pe
r
f
or
med
u
s
ing
the
de
f
a
ult
pa
r
a
mete
r
s
s
pe
c
if
ied
in
T
a
ble
2.
S
tar
ti
ng
f
r
om
the
ne
xt
pa
r
a
gr
a
ph
,
we
de
tail
th
e
pe
r
f
o
r
m
a
nc
e
of
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Dual
s
imulat
e
d
anne
ali
ng
s
oft
de
c
ode
r
for
li
ne
ar
b
lock
c
ode
s
(
Hic
ham
T
ahir
i
A
laou
i
)
2783
DSAS
D
c
ompar
e
d
to
a
s
e
t
of
other
de
c
ode
r
s
,
inc
ludi
ng
S
ASD
[
7
]
,
c
las
s
ica
l
s
im
ulate
d
a
nne
a
li
ng
,
C
ha
s
e
-
2
[
24]
,
DD
GA
[
8]
,
C
GA
D
[9
]
,
M
a
ini
[
20]
,
c
ompac
t
ge
ne
ti
c
a
lgor
it
hm
with
high
s
e
lec
ti
on
pr
e
s
s
ur
e
(
c
GA
-
HSP
)
[
28]
,
AutDA
G
[
26
]
,
c
ompac
t
ge
ne
ti
c
a
lgor
it
h
m
wi
th
memor
y
(
c
GA
-
M
)
[
29]
,
ge
ne
ti
c
a
lgor
i
thm
meta
-
de
c
is
ion
de
c
ode
r
(
GA
M
D)
[
30]
,
S
DG
A
[
25]
,
a
nd
c
las
s
ica
l
binar
y
p
ha
s
e
s
hif
t
ke
ying
(
B
P
S
K)
de
c
oding
a
lgor
it
hms
.
T
he
c
ompar
is
on
of
DSAS
D
,
S
ASD,
a
nd
c
la
s
s
ica
l
s
im
ulate
d
a
nne
a
li
ng
f
o
r
B
C
H(
31,
21,
5
)
,
B
C
H(
63,
45,
7)
,
a
nd
L
DPC
(
60,
30)
c
ode
s
is
s
how
n
in
F
igur
e
4
.
T
his
c
ompar
is
on
unde
r
s
c
or
e
s
the
s
upe
r
ior
e
f
f
ica
c
y
of
both
the
D
S
ASD
a
nd
S
ASD
a
lgor
it
h
ms
ove
r
the
c
las
s
ica
l
method
f
or
thes
e
c
ode
s
.
S
p
e
c
if
ica
ll
y,
F
igur
e
4(
a
)
il
lus
tr
a
tes
the
r
e
s
ult
s
f
o
r
B
C
H(
31,
2
1,
5)
,
F
igur
e
4(
b)
f
or
B
C
H(
63,
45,
7
)
,
a
nd
F
igur
e
4(
c
)
f
or
L
DPC
(
60,
30)
.
(
a
)
(
b)
(
c
)
F
igur
e
4.
C
ompar
is
on
of
DSAS
D,
S
ASD
,
a
nd
c
las
s
ica
l
s
im
ulate
d
a
nne
a
li
ng
f
or
the
(
a
)
B
C
H(
31,
21,
5
)
,
(
b)
B
C
H(
63,
45
,
7)
,
a
nd
(
c
)
L
DPC
(
60,
30
)
T
he
e
va
luation
of
DSAS
D
de
c
oding
pe
r
f
or
manc
e
a
ga
ins
t
va
r
ious
c
ompetit
or
a
lgor
it
h
ms
is
p
r
e
s
e
nted
in
F
igu
r
e
5
.
F
igur
e
5(
a
)
de
mons
tr
a
tes
that
the
D
S
ASD
a
lgor
i
thm
ou
tper
f
or
ms
other
a
lgor
i
thm
s
b
y
1
dB
a
t
10⁻³
ove
r
the
GA
M
D
[
30]
a
lgor
it
hm
f
o
r
the
L
DPC
(
60,
30)
c
ode
,
a
nd
F
igur
e
5
(
b)
s
hows
that
DSAS
D
a
c
hieve
s
a
0.
25
dB
ga
in
ove
r
S
DG
A
a
t
10⁻⁴
a
nd
a
bout
a
1.
8
3
dB
ga
in
a
t
10⁻³
ove
r
B
P
S
K
de
c
oding
f
or
the
R
M
(
32,
16,
8
)
c
ode
.
T
he
pe
r
f
o
r
manc
e
c
ompar
is
on
of
DSAS
D
a
nd
other
c
ompeting
a
lgor
i
thm
s
on
B
C
H
c
ode
s
is
pr
e
s
e
nted
in
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14
,
No.
4
,
Augus
t
2025
:
277
6
-
2787
2784
F
igur
e
6
.
F
or
B
C
H(
31,
21
,
5)
,
a
s
s
hown
in
F
igur
e
6(
a
)
,
the
DSAS
D
a
lgor
it
hm
s
ur
pa
s
s
e
s
c
las
s
ic
a
l
s
im
ulate
d
a
nne
a
li
ng
(
by
1
dB
s
tar
ti
ng
f
r
om
10⁻³
)
,
C
ha
s
e
-
2
a
nd
C
GA
D
by
0.
5
dB
a
t
10⁻⁴,
a
nd
pr
oduc
e
s
r
e
s
ult
s
ne
a
r
ly
identica
l
to
thos
e
of
S
ASD
a
nd
M
a
ini
,
e
xc
e
pt
f
o
r
M
a
ini
with
a
0
.
31
dB
d
if
f
e
r
e
nc
e
a
t
10⁻⁵.
S
im
il
a
r
s
upe
r
ior
pe
r
f
or
manc
e
is
obs
e
r
ve
d
f
or
the
B
C
H(
63,
45
,
7)
c
ode
in
F
igur
e
6(
b
)
,
whe
r
e
DSAS
D
outper
f
or
ms
S
DG
A,
C
ha
s
e
-
2,
c
GA
-
M
,
c
GA
-
HSP
,
a
nd
AutDA
G
,
with
c
ompar
a
ble
pe
r
f
or
manc
e
to
DD
GA
[
8
]
a
nd
S
ASD
[
7]
.
(
a
)
(
b)
F
igur
e
5.
E
va
luation
of
DSAS
D
de
c
oding
pe
r
f
or
m
a
nc
e
a
ga
ins
t
c
ompetit
or
a
lgor
it
hms
a
ppli
e
d
to:
(
a
)
L
DPC
(
60,
30)
a
nd
(
b)
R
M
(
32,
16
,
8)
(
a
)
(
b)
F
igur
e
6.
P
e
r
f
or
manc
e
c
ompar
is
on
o
f
DSAS
D
a
nd
c
ompeting
a
lgor
it
hms
on
B
C
H
c
ode
s
:
(
a
)
B
C
H(
31,
21,
5)
a
nd
(
b)
B
C
H(
63,
45,
7)
.
T
he
s
im
ulation
r
e
s
ult
s
c
lea
r
ly
indi
c
a
te
that
th
e
pr
opos
e
d
DSAS
D
a
lgo
r
it
hm
of
f
e
r
s
s
upe
r
ior
pe
r
f
or
manc
e
c
ompar
e
d
to
c
las
s
ica
l
s
im
ulate
d
a
nne
a
li
ng
a
nd
many
other
de
c
ode
r
s
tes
ted
a
c
r
os
s
va
r
io
us
c
ode
s
.
Nota
bly,
the
DSAS
D
a
c
hieve
s
s
igni
f
ica
nt
ga
ins
,
s
uc
h
a
s
up
t
o
2
dB
f
o
r
B
C
H(
63,
45
,
7)
a
nd
L
DPC
(
60,
30)
,
de
mons
tr
a
ti
ng
c
ons
is
tent
im
pr
ove
ment
ove
r
other
methods
.
Additi
ona
ll
y,
the
c
los
e
a
li
gnment
in
o
utcome
s
be
twe
e
n
S
ASD
a
nd
DSAS
D
f
u
r
ther
va
li
da
tes
the
r
obus
tnes
s
a
nd
e
f
f
e
c
ti
ve
ne
s
s
of
the
pr
opos
e
d
a
lgor
it
hm.
T
he
s
upe
r
io
r
pe
r
f
or
manc
e
o
f
the
DSAS
D
a
lgor
it
hm
c
a
n
be
a
tt
r
ibut
e
d
to
the
e
f
f
e
c
ti
ve
int
e
gr
a
ti
on
of
the
dua
l
pr
ope
r
ty
with
s
im
ulate
d
a
nne
a
li
ng,
whic
h
c
a
n
be
noti
c
e
d
on
lar
ge
c
ode
s
.
T
he
c
ons
is
tenc
y
of
r
e
s
ult
s
a
c
r
os
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Dual
s
imulat
e
d
anne
ali
ng
s
oft
de
c
ode
r
for
li
ne
ar
b
lock
c
ode
s
(
Hic
ham
T
ahir
i
A
laou
i
)
2785
dif
f
e
r
e
nt
c
ode
s
indi
c
a
tes
that
the
pr
opos
e
d
meth
od
is
r
obus
t
a
nd
a
da
ptable
to
a
r
a
nge
of
e
r
r
or
c
or
r
e
c
ti
on
s
c
e
na
r
ios
.
T
his
s
tudy
de
mons
tr
a
ted
the
s
uc
c
e
s
s
f
ul
int
e
gr
a
ti
on
of
s
im
ulate
d
a
nne
a
li
ng
with
the
dua
l
p
r
ope
r
ty
o
f
li
ne
a
r
block
c
ode
s
.
W
hil
e
p
r
omi
s
ing,
f
u
r
ther
r
e
s
e
a
r
c
h
is
r
e
quir
e
d
to
c
onf
ir
m
it
s
a
ppli
c
a
bil
it
y
to
oth
e
r
c
ode
s
a
nd
to
im
pr
ove
c
onve
r
ge
nc
e
.
S
pe
c
if
ica
ll
y,
opti
mi
z
ing
the
c
ooli
ng
s
c
he
dule
c
ould
a
ddr
e
s
s
s
low
c
on
ve
r
ge
nc
e
is
s
ue
s
,
a
nd
a
da
pti
ng
the
de
c
ode
r
f
or
br
oa
de
r
c
ode
c
ompatibi
li
ty
c
ould
e
nha
nc
e
the
a
lgor
it
hm’
s
ove
r
a
ll
e
f
f
e
c
ti
ve
ne
s
s
a
nd
ve
r
s
a
ti
li
ty.
T
he
s
e
e
f
f
or
ts
will
be
ke
y
to
ove
r
c
omi
ng
the
c
ur
r
e
nt
li
mi
tations
a
nd
e
x
pa
nding
the
DSAS
D
a
lgor
it
h
m's
potential.
F
inally,
thi
s
s
tudy
highl
igh
ts
the
DSAS
D
a
lgor
it
hm's
s
upe
r
ior
pe
r
f
or
manc
e
,
a
c
hieving
up
to
2
dB
im
pr
ove
ment
ove
r
c
las
s
ica
l
s
im
ulate
d
a
nne
a
li
ng
a
nd
other
de
c
ode
r
s
,
pa
r
ti
c
ular
ly
f
or
lar
ge
c
ode
s
li
ke
B
C
H(
63,
45,
7)
a
nd
L
DPC
(
60,
30)
.
T
he
s
uc
c
e
s
s
f
ul
int
e
gr
a
ti
on
o
f
s
im
ulate
d
a
nne
a
li
ng
with
the
dua
l
pr
ope
r
ty
unde
r
s
c
or
e
s
it
s
r
obus
tnes
s
.
How
e
ve
r
,
f
ur
ther
r
e
s
e
a
r
c
h
is
ne
e
de
d
to
im
pr
ove
c
onve
r
ge
nc
e
a
nd
e
xtend
it
s
a
ppli
c
a
bil
it
y
to
a
wide
r
r
a
nge
of
c
ode
s
,
e
ns
ur
ing
th
e
DSAS
D
a
lgor
it
hm
r
e
a
c
he
s
it
s
f
ull
potential.
5.
CONC
L
USI
ON
I
n
thi
s
pa
pe
r
,
we
int
r
oduc
e
d
a
ne
w
s
of
t
de
c
ode
r
f
or
li
ne
a
r
block
c
ode
s
by
int
e
gr
a
ti
ng
the
s
im
ulate
d
a
nne
a
li
ng
pr
oc
e
s
s
with
the
dua
li
ty
p
r
ope
r
ty
of
l
in
e
a
r
block
c
ode
s
.
T
he
s
im
ulate
d
a
nne
a
li
ng
a
lgor
i
th
m
us
e
s
a
pr
oba
bil
is
ti
c
a
ppr
oa
c
h
to
e
xplor
e
the
s
olut
ion
s
pa
c
e
,
dr
a
wing
ins
pir
a
ti
on
f
r
om
the
a
nne
a
li
ng
pr
oc
e
s
s
.
S
ubopti
mal
s
olut
ions
a
r
e
a
c
c
e
pted
a
c
c
or
ding
to
a
tempe
r
a
tur
e
s
c
he
dule.
Our
pr
opos
e
d
DSAS
D
s
ur
pa
s
s
e
s
c
las
s
ica
l
s
im
ulate
d
a
nne
a
li
ng
a
nd
other
de
c
ode
r
s
s
uc
h
a
s
S
DG
A,
AutDA
G,
C
ha
s
e
-
2,
a
nd
C
GA
D
f
or
s
pe
c
if
ic
c
ode
s
,
a
c
hieving
ga
ins
of
up
to
2
.
5
dB
a
t
a
B
E
R
of
6×
10
-
6
f
or
the
L
DPC
(
60,
30)
c
ode
.
A
ke
y
a
dv
a
ntage
of
DSAS
D
is
it
s
a
bil
it
y
to
leve
r
a
ge
r
e
li
a
bil
i
ty
in
f
or
m
a
ti
on
f
r
o
m
r
e
c
e
ived
da
ta
to
ini
ti
a
te
the
s
e
a
r
c
h
a
nd
ge
ne
r
a
te
ne
ighbor
ing
s
olut
ions
.
Additi
ona
ll
y,
we
c
onduc
ted
a
c
ompar
a
ti
ve
a
na
lys
is
of
a
lgo
r
it
hmi
c
c
o
mpl
e
xit
y,
highl
ight
ing
DSAS
D's
e
f
f
icie
nc
y.
F
utu
r
e
wor
k
s
h
ould
f
oc
us
on
e
nha
nc
ing
the
c
onve
r
ge
nc
e
of
the
c
ooli
ng
mec
ha
nis
m
a
nd
e
xtending
the
a
lgor
it
hm
’
s
a
ppli
c
a
bil
it
y
to
a
br
oa
de
r
r
a
nge
of
c
ode
s
.
F
UN
DI
NG
I
NF
ORM
AT
I
ON
Author
s
s
tate
no
f
unding
invol
ve
d.
AU
T
HO
R
CONT
RI
B
U
T
I
ONS
S
T
AT
E
M
E
N
T
T
his
jour
na
l
us
e
s
the
C
ontr
ibut
o
r
R
oles
T
a
xo
nomy
(
C
R
e
diT
)
to
r
e
c
ognize
indi
vidual
a
uthor
c
ontr
ibut
ions
,
r
e
duc
e
a
utho
r
s
hip
dis
putes
,
a
nd
f
a
c
il
it
a
te
c
oll
a
bor
a
ti
on.
Nam
e
of
Au
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Hic
ha
m
T
a
hir
i
Ala
oui
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Ahme
d
Az
oua
oui
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
J
a
mal
E
l
Ka
f
i
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
li
z
a
ti
on
M
:
M
e
th
odol
ogy
So
:
So
f
twa
r
e
Va
:
Va
li
da
ti
on
Fo
:
Fo
r
ma
l
a
na
ly
s
is
I
:
I
nve
s
ti
ga
ti
on
R
:
R
e
s
our
c
e
s
D
:
D
a
ta
C
ur
a
ti
on
O
:
W
r
it
in
g
-
O
r
ig
in
a
l
D
r
a
f
t
E
:
W
r
it
in
g
-
R
e
vi
e
w
&
E
di
ti
ng
Vi
:
Vi
s
ua
li
z
a
ti
on
Su
:
Su
pe
r
vi
s
io
n
P
:
P
r
oj
e
c
t
a
dmi
ni
s
tr
a
ti
on
Fu
:
Fu
ndi
ng a
c
qui
s
it
io
n
CONF
L
I
CT
OF
I
NT
E
RE
S
T
S
T
AT
E
M
E
N
T
Author
s
s
tate
no
c
onf
li
c
t
of
int
e
r
e
s
t.
I
NF
ORM
E
D
CONSE
NT
Not
a
ppli
c
a
ble.
No
pe
r
s
ona
l
inf
or
mat
ion
wa
s
inclu
de
d
in
thi
s
s
tudy.
E
T
HI
CA
L
AP
P
ROVA
L
Not
a
ppli
c
a
ble.
No
r
e
s
e
a
r
c
h
r
e
late
d
to
human
us
e
h
a
s
be
e
n
done
in
thi
s
pa
pe
r
.
Evaluation Warning : The document was created with Spire.PDF for Python.