TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 8, August 201
4, pp. 6332 ~ 6337
DOI: 10.115
9
1
/telkomni
ka.
v
12i8.585
4
6332
Re
cei
v
ed Fe
brua
ry 24, 20
14; Re
vised
Ma
rch 26, 20
14; Accepted
April 15, 201
4
A Complete Lattice Lossless C
o
mpression Storage
Model
Zhi Huilai
Schoo
l of Com
puter Scie
nce
and T
e
chno
log
y
, He
nan Po
l
y
t
e
chn
i
c Univ
ersi
t
y
,
Jiaoz
uo He
na
n
P.R.China, +
8
6-03
91-3
9
8
7
7
1
1
email: zh
ihu
ila
i
@
12
6.com
A
b
st
r
a
ct
In this
pap
er, a
co
mpl
e
te
lattic
e
loss
less
co
mpressi
on stor
a
ge
mode
l is
pr
opos
ed t
o
i
m
prove th
e
storage
efficie
n
cy. In ord
e
r to bu
ild
the
pro
pose
d
mo
del, f
i
rst all th
e u
p
p
e
r an
d l
o
w
e
r ir
reduc
ibl
e
e
l
e
m
ents
of the co
mpl
e
te lattice ar
e id
entifie
d resp
ectively, t
hen
an i
s
omorp
h
ic
ma
ppi
ng for
m
the
compl
e
te lattic
e
to
a conc
ept lattic
e
is fou
nde
d, a
nd fin
a
lly
a
ma
trix is
used to
store the for
m
al co
nt
ext of the conc
ept lattic
e
.
Co
mp
ared w
i
th
using a
d
j
a
cent
matrix, exa
m
p
l
e an
d
an
alysis
show
that the pr
op
osed
meth
od can i
m
prov
e
the storage
efficiency of co
mp
lete lattice.
Ke
y
w
ords
: Lat
tice theory, co
mp
lete L
a
ttice, irreduc
ib
le el
e
m
e
n
t, lossless
compressi
on st
orag
e
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Lattice de
scri
bes the pa
rtial orde
r relat
i
ons bet
wee
n
objects, an
d
is widely used in
obje
c
t clu
s
tering and hi
era
r
chi
c
al
stru
ct
ure an
alysi
s
. In rece
nt years, lattice theory, especi
a
lly
the complete
lattice the
o
ry, is used i
n
many
fiel
d
s
, such a
s
grap
h q
ueryi
ng [1], situat
ion
hiera
r
chy ma
nipulate [2],
metaboli
c
pat
hway a
nalys
i
s
[3], set-val
u
ed varia
b
le re
pre
s
entatio
n [4],
and
so on. I
n
theoretical
resea
r
ch, adjacen
cy
ma
trix as the
storage
model
is acce
ptabl
e.
Ho
wever,
in
re
al a
ppli
c
ations the l
a
ttice
u
s
u
a
ll
y
contai
ns nume
r
ou
s n
ode
s
a
nd h
a
s a
compli
cate
d
stru
cture. At t
h
is
ci
rcumsta
n
ce,
adja
c
e
n
c
y mat
r
ix a
s
the sto
r
a
ge
model
will
co
st a
lot of stora
g
e
spa
c
e
and i
s
not co
ndu
cive to lattice re
trieval and l
a
ttice isomo
r
ph
ism jud
g
ment
.
Lattice sto
r
a
ge is no lo
n
ger a in
sig
n
ificant p
r
obl
e
m
, but a key theoreti
c
al i
s
sue
s
of pra
c
tica
l
appli
c
ation va
lue.
Formal
Concept Analysis
(FCA) after being prod
uced by professor R.
Wille [5], its core
stru
cture con
c
ept
l
a
ttice h
a
s attra
c
ted broa
d
a
ttenti
on a
nd
bein
g
used i
n
va
rious field
s
. A
s
its
uniqu
e adva
n
tage
s in da
ta analysi
s
a
nd kno
w
ledg
e system
de
velopment, it has
be
com
e
a
means for ex
ternal
recognition [6]. For
this reason, i
n
th
is
arti
cle we will
propose a compl
e
te
lattice sto
r
ag
e model ba
se
d on the theo
ry of FCA.
2. Preliminaries
Before
pro
c
e
eding,
we
bri
e
fly re
call the
lattice te
rmin
ology a
nd
propertie
s
[7], a
s
well
a
s
fundame
n
tal definition
s
in FCA [8, 9].
Definition
2.1
.
Let
P
be a
se
t. An order o
n
P
is a
bin
a
ry
rel
a
tion
s
u
ch that, for all
,,
x
yz
P
,
(1)
x
x
,
(2)
x
y
and
yx
imply
x
y
,
(3)
x
y
and
yz
imply
x
z
.
These condi
tions a
r
e
re
ferre
d to, re
spe
c
tive
ly, as reflexivity, anti-symm
e
t
ry and
transitivity.
Definition 2.2
.
Let
P
be an o
r
dere
d
set an
d let
,
x
yP
. We say
x
is cove
red
b
y
y
(or
y
cov
e
r
s
x
), and
write
x
y
or
yx
, if
x
y
and
x
zy
implies
zx
. More
over,
x
is
calle
d the lower neig
hbo
r o
f
y
and
y
is calle
d the upp
er nei
g
hbor of
x
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Com
p
lete Lattice Lo
ssl
ess Com
p
ressio
n Storage Mo
del (Zhi
Huilai
)
6333
Definition 2.3
.
Let
P
be an
o
r
de
red
set a
n
d
let
SP
. An ele
m
ent
x
P
is an u
pper
boun
d of
S
if
sx
for all
sS
. An lowe
r bo
und i
s
d
u
a
lly. The set of all upp
er
b
ound
s of
S
is
denote
d
by
u
S
and the set of all lowe
r bou
nds by
l
S
:
{|
(
)
:}
u
Sx
P
s
S
s
x
and
{|
(
}
: )
l
Sx
P
s
S
s
x
.
If
u
S
has a le
ast
element
x
, the
n
x
is called th
e least up
per
boun
d of
S
. Dually, if
l
S
has a g
r
eate
s
t element
x
, then
x
is call
ed the gre
a
test lo
wer b
oun
d of
S
.
The lea
s
t upp
er bo
und of
S
is also
called t
he su
pe
rmum
of
S
and is de
n
o
ted by su
p
S
; the greate
s
t lowe
r bou
nd
of
S
is also
call
ed the infimu
m of
S
and is d
enoted by inf
S
.
Notation 2.1. We write
x
y
in place of su
p
,
x
y
when it exists and
x
y
in place o
f
inf
,
x
y
when it exists. Similarly, we write
S
and
S
instea
d of sup
S
and inf
S
. I
t
is
sometim
e
s
n
e
ce
ssary to i
ndicate that the join
or m
e
et is bei
ng fo
und in
a pa
rticula
r
o
r
de
red
set
P
, in which
case we write
P
S
or
P
S
.
Definition 2.4.
Let
P
be an no
n-em
pty orde
red set.
(1) If
x
y
and
x
y
exis
ts
for all
,
x
yP
, th
en
P
is called a
lattice;
(2) If
S
and
S
exis
ts
for all
SP
, th
en
P
is
called a
c
o
mplete lattic
e
.
Definition 2.5.
Let
L
be a lattice. An eleme
n
ts
x
L
is join-irreduci
b
le if
(1)
0
x
(in ca
se
L
has a
zero)
(2)
x
ab
implies
x
a
or
x
b
for all
,
ab
L
.
A meet-irredu
cible el
ement
is defined d
u
a
lly.
Definition 2.6
.
Let
P
be an ordere
d
set an
d
QP
. Then
Q
is called join-dense in
P
if for every element
aP
there
is a sub
s
et
A
of
Q
such that
P
aA
. The dual
of join
-
den
se is me
e
t
-den
se.
Definition
2.7
.
A context i
s
a t
r
iple
,,
KG
M
I
where
G
and
M
are
sets and
I
GM
. The eleme
n
ts of
G
and
M
are
calle
d ob
jects
and attributes
re
spe
c
tively. As
usu
a
l, instea
d of
,
gm
I
we write
gIm
and say ‘the
object g ha
s
the attribute
m’.
Definition
2.8
.
Given a fo
rmal context
,,
KG
M
I
, the de
rivatio
n
functio
n
s
.
f
and
.
g
are defin
ed for
A
G
and
B
M
as
follows
:
{|
}
:
f
Am
M
g
A
g
I
m
;
{|
}
:
gB
g
G
m
B
g
I
m
.
Definition 2.9
.
A formal concept of formal co
ntext
,,
KG
M
I
is a pai
r
,
A
B
,
whe
r
e
A
G
,
B
M
,
f
AB
, and
gB
A
.
A con
c
ept
,
A
B
is sub
c
on
ce
pt of
,
CD
if
A
C
(eq
u
ivalently,
DB
). In thi
s
ca
se,
,
CD
is call
ed a su
pe
rco
n
ce
pt of
,
A
B
. We write
,,
A
BC
D
and
define the
relation
s
,
, and
as u
s
ual.
The
set of all
con
c
e
p
ts
ordere
d
by the
relation
forms a l
a
ttice, which i
s
denot
ed by
LK
and call
ed th
e con
c
e
p
t lattice of the co
n
t
ext
K
. The rel
a
tion define
s
the coveri
ng
graph
of
LK
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 633
2 –
6337
6334
3. Complete
Lattic
e
Stor
age Model Based on F
C
A
Definition 3.
1. Let
,,
KG
M
I
be a formal
con
t
ext, object
g
is call
ed a full
attributes
obj
ect, if and o
n
ly if
f
gM
. Dually, attribute
m
is called a la
rgest commo
n
attribute if and only if
gm
G
[10].
Definition 3.2
.
Let
,,
KG
M
I
be a formal contex
t, object
g
is
called a shaded
obje
c
t, if and only if the
r
e a
r
e a
se
ri
es of o
b
je
cts
{}
ii
T
g
and
T
is index set, that makes
()
(
)
i
iT
f
gf
g
. Dually, attri
bute
m
is call
e
d
a shad
ed
attribute, if a
nd only if the
r
e a
r
e a
seri
es of attri
butes
{}
ii
T
m
and
T
is
index set, tha
t
make
s
()
(
)
i
iT
gm
gm
.
Definition
3.3
.
Let
,,
KG
M
I
be a
f
o
rmal
contex
t,
K
is
calle
d a
pu
rified fo
rm
al
context, if and only if there is no full attributes
o
b
je
ct, no largest
comm
on attri
bute, no sh
a
ded
obje
c
t and no
shad
ed attrib
ute.
Definition 3.4
.
Let
,,
KG
M
I
be
a formal
co
ntext, a con
c
e
p
t is called
a o
b
ject
c
o
nc
ept if it
has
the form
,
gf
g
f
g
,
gG
, and
g
is called its obj
ect label;
a
con
c
e
p
t is
ca
lled a
p
r
op
ert
y
con
c
e
p
t if i
t
has that
ha
s the
form
,
gm
f
g
m
,
mM
,
and
m
is called
its attribute la
bel.
Propo
sition
3
.
1. In a co
m
p
lete lattice,
a join
-irredu
cible elem
ent
has
only on
e
lowe
r
neigh
bor, an
d
a meet-irred
ucibl
e
eleme
n
t has only o
ne upp
er nei
g
hbor [7].
Theo
rem 3.1.
Let
,,
KG
M
I
be a purified formal
context, a obj
ect co
ncept o
f
K
must b
e
a
joi
n
-irred
uci
b
le
element, a
nd
vice versa.
Dually, an attri
bute
con
c
ept
of
K
m
u
s
t
b
e
a
meet-irre
d
u
c
i
b
le eleme
n
t, and vice versa.
Proof: Pro
o
f
by co
ntra
ctio
n an
d a
s
sum
e
that
a o
b
je
ct con
c
ept
,
A
B
is not a join-
irre
du
cible el
ement, then
,
A
B
has at le
ast
two lower n
e
ighb
ors, an
d we d
enote
all these
lo
w
e
r
n
e
i
g
hbo
r
s
as
,
tt
tT
AB
, and
T
is the index set. Since
,
A
B
is a object co
nce
p
t, then it
must exist an
object
g
t
hat
make
s
f
gB
. By b
a
si
c theorem
of conce
p
t lattice, we ha
ve
()
tt
tT
t
T
B
Bf
A
, and also
f
AB
,
then we get
()
t
tT
fA
f
A
, w
h
ic
h
mean
s
,
A
B
is a sh
ade
d obje
c
t, and th
is is
contradi
ct
to the condit
i
on of the the
o
rem that
K
is not a
purified
cont
ext. So the
assumptio
n
f
a
ils, a
n
d
the
theo
rem
hol
ds. By
Dualit
y Princi
ple,
we
dire
ctly get that an attribute
con
c
ept of
K
must be a me
et-irredu
cibl
e element, and
vice versa.
Theo
rem 3.2.
[7] Let
V
be a
c
o
mplete lattic
e
, let
G
and
M
be sets a
nd
assume t
hat
there exi
s
t m
appin
g
s
:
GV
and
:
M
V
su
ch t
hat
()
G
is join-dense in
V
and
()
M
is me
et-d
ense in
V
. Define
I
by
gI
m
G
M
, for all
gG
mM
. T
h
en
V
is isom
orphi
c to con
c
ept lat
t
ice
,,
LG
M
I
.
We d
e
fine
:,
g
g
fg
fg
and
:,
mg
m
f
g
m
, then we
have
((
(
)
)
)
gI
m
g
g
m
g
f
g
g
f
g
m
g
m
g
m
and th
us we
have
gI
m
G
M
, which
mean
s
g
and
m
satisfy Theo
re
m 3.2. Moreo
v
er,
by Definition
3.4 we kno
w
that
g
is a obje
c
t con
c
e
p
t, and
m
is an attri
bute co
ncept.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Com
p
lete Lattice Lo
ssl
ess Com
p
ressio
n Storage Mo
del (Zhi
Huilai
)
6335
Acco
rdi
ng to
the above discu
ssi
on, g
i
ven a compl
e
te lattice
V
, by using ma
pping
g
and
m
, we ca
n get a con
c
ept lattice
,,
LG
M
I
whi
c
h is i
s
omorphi
c to the
c
o
mplete lattic
e
V
.
In the following Algorithm
1, by
labeling irre
du
cibl
e element o
n
V
, we can
get the
obje
c
ts a
n
d
attribute
s
containe
d in
,,
LG
M
I
, whi
c
h i
s
corre
s
p
ondin
g
the el
em
ents
contai
ned
in
G
and
M
. More
o
v
er, by u
s
in
g
the
co
nne
cti
on b
e
twe
e
n
join-i
rre
du
cibl
e ele
m
ent
and me
et-irre
duci
b
le el
em
ent whi
c
h i
s
embodi
ed in
V
, we can get
the relation
I
. So we
get
the context
,,
KG
M
I
.
Algorithm 1: formal
context
acqui
sition fo
rm co
mplete l
a
ttice
Input: c
o
mplete lattic
e
V
;
Output: formal c
ontext
K
Step 1: Traverse
com
p
let
e
lattice
V
upward fo
rm it
s
minimal
elem
ent, if the
cu
rre
ntly
visited eleme
n
t has only o
ne upp
er nei
g
hbor, then la
b
e
ling this el
e
m
ent with a u
n
ique letter;
Step 2: Traverse
com
p
l
e
te lattice
V
d
o
wn
wa
rd form its maximal element, if the
curre
n
tly visited eleme
n
t h
a
s only on
e lowe
r neig
h
b
o
r, then lab
e
l
i
ng this no
de
with a uniqu
e
digit;
Step 3: If m d
i
fferent digits
and n differe
nt le
tters are
use
d
, then establish a cont
ext with
m r
o
ws
an
d
n c
o
lu
mns
an
d s
t
or
e
it
b
y
usin
g
a ma
tr
ix
{}
ij
m
n
Aa
,
and ea
ch digi
t
co
rre
sp
ondi
ng
to a row
while
each letter
correspon
ding
to a column,
and initialize it to be a nil-matrix;
Step 4: T
r
ave
r
se
compl
e
te
lattice
V
, for ea
ch
nod
e la
bel
ed by
a di
git(assume
this
digit
is
i
), visit its upper
neig
hbo
rs u
n
til meet the maxima
l e
l
ement. And i
n
this process if there exi
s
t
an eleme
n
t labeled by a let
t
er(a
ssume th
is letter is
j
), then set the value of
ij
a
to 1;
Step 5: Return
A
.
Example 1. Let
V
be a complete and is shown in Figure
1. By using
Algorithm 1, firstly
we fin
d
its
irre
du
cible
el
ements,
whi
c
h i
s
sh
own
in Fi
gure
2. Seconda
ry, we g
e
t i
t
s
corre
s
p
ondin
g
formal co
n
t
ext
K
that is
sho
w
n in Ta
ble 1. And accordi
ng to
K
we get its
stora
ge matri
x
1
1
0
00
0
100
1
0
0
0011
0
1
1
0
0011
0
01
0
00111
10
10
10
0
0
1
1
10
10
0
0
0
1
1100
00
011
0
1
00
0
A
.
Figure 1. Co
mplete Lattice
V
Figure 2. Irre
duci
b
le Elem
ents of
V
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 633
2 –
6337
6336
Table 1. Fo
rmal Context
K
a
bc
de
f
g
h
1
*
*
2
*
*
*
3
* *
* *
4
*
*
*
*
5
*
*
*
6
* * *
*
7
* * *
8
*
*
*
If using
adj
ce
nt matrix to
s
t
ore complete
lattic
e
V
,
we mus
t
es
tablish
a
matrix
2
A
wi
th
19 ro
ws
an
d 19 colu
mn
s, and
th
e num
ber
of
1
el
em
ents i
s
twi
c
e
of the
numb
e
r
of
links bet
wee
n
the node
s, i.e. 62.
The sto
r
ag
e efficien
cy co
mpari
s
o
n
bet
wee
n
1
A
and
2
A
is given is Tabl
e 2.
Table 2. Storage Efficien
cy Compa
r
iso
n
size
non-zero eleme
n
t
s
Percentage of th
e
non-zero eleme
n
t
s
1
A
88
26 40.6%
2
A
19
19
62 17.1%
Other than
saving storage
spa
c
e, our
method is
al
so helpful to improve the e
fficiency
for judgin
g
complete latti
ce isomo
r
phi
sm. Com
p
let
e
lattice iso
m
orp
h
ism ju
dgment
can
be
conve
r
ted int
o
grap
h isom
orphi
sm jud
g
m
ent, and this is se
en a
s
a NP - com
p
l
e
te probl
em by a
majority of
schol
ars [11].
If we jud
g
e
gra
ph i
s
om
orphi
sm
by
perfo
rming
row a
nd
col
u
mn
excha
nge of adja
c
ent matrix, at
the worst ca
se, the total numbe
r of exchan
ge will rea
c
h
!!
rc
times
(
r
is the
numbe
r of
rows, a
s
c
is the of col
u
m
n
s), a
nd thi
s
is mu
ch g
r
eater tha
n
expone
ntial
time
com
p
lexity. Our method redu
ce
s th
e scale of the stora
ge ma
trix, thus it can
improve the e
fficiency of co
mplete lattice
isomo
r
phi
sm
judgment.
4. Conclusio
n
Based
on Fo
rmal Con
c
ept
Analysis, we pro
p
o
s
e a
compl
e
te lattice sto
r
a
ge m
e
thod.
The propo
se
d method o
n
l
y store
s
irre
duci
b
le elem
ents, and th
e relatio
n
shi
p
betwe
en th
em.
Comp
ared
wi
th the a
d
ja
ce
ncy mat
r
ix st
orag
e
m
e
tho
d
, the p
r
op
osed meth
od
can imp
r
ove
the
stora
ge effici
ency.
Ackn
o
w
l
e
dg
ements
The wo
rk prese
n
ted
in
this pape
r
i
s
su
ppo
rted b
y
Do
ctorial
Found
ation o
f
Hen
an
Polytechni
c University (B20
11-1
0
2
)
.
Referen
ces
[1]
Yuan
Da
yu,
Mi
tra Prase
n
jit. L
i
nd
e
x
: A l
a
ttice
-base
d
i
nde
x for gra
p
h
data
b
a
ses. VL
DB Jo
urna
l.
201
3;
22(2):2
29-2
52.
[2]
Ye J, Co
yle
L, Dobso
n
S, Ni
xo
n Pa
dd
y. R
epr
es
entin
g a
n
d
mani
pu
latin
g
situatio
n hi
era
r
chies us
in
g
situatio
n lattice
s. Revue d'
Intelli
ge
nc
e Artific
i
ell
e
. 200
8; 22(
5):647-
66
7.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Com
p
lete Lattice Lo
ssl
ess Com
p
ressio
n Storage Mo
del (Zhi
Huilai
)
6337
[3]
Yaron A, B. Goldste
i
n, Ale
x
a
nder Bockm
a
y
r.
A Lattice-T
heoretic F
r
ame
w
o
r
k for Metabolic P
a
th
w
a
y
Anal
ys
is. Com
putatio
na
l Met
hods
in S
y
ste
m
s Biol
og
y. L
e
cture N
o
tes i
n
Com
puter S
c
ienc
e. 20
13
;
130:1
78-
191.
[4
]
T
h
i
e
rry
De
n
oeu
x
,
Zou
l
fi
ca
r
Yo
u
n
e
s
, Fa
h
e
d
Abd
a
l
l
a
h
.
R
e
p
r
e
s
e
n
t
in
g
u
n
c
e
r
tai
n
ty
on se
t-va
l
ued
varia
b
les us
in
g
beli
e
f function
s. Artifici
al Intel
lige
n
ce. 2
010;
174(
7-8):47
9
-4
99.
[5]
Ganter, B., W
ille, R.. Formal Conce
p
t Anal
ys
i
s
: Mathematica
l
Found
atio
ns. Berlin: Spr
i
ng
e
r
. 1999.
[6]
Poelm
ans Jonas, Ignatov Dmitr
y
I, Kuznet
sov Sergei O. Formal conc
ept analy
sis in kno
w
ledge
process
i
ng: A surve
y
on a
p
p
l
i
c
ations. E
x
pert
S
y
stem
w
i
th A
pplic
atio
ns. 20
13; 40(1
6
): 653
8-65
60.
[7]
V. K. Garg, N.
Mittal,
A. Se
n. App
lic
ations
of l
a
ttice th
eo
r
y
t
o
d
i
strib
u
te
d com
puti
ng.
ACM SIGACT
Notes. 200
3; 3
4
(3):40-
61.
[8]
R. Wille. C
o
n
c
ept lattices
a
nd co
nce
p
tual
kn
o
w
l
e
d
ge s
y
stems. C
o
mp
uters & Math
ematics
w
i
t
h
Appl
icatio
ns. 1
992; 23(
6-9):4
93-5
15.
[9]
R. W
ille.
Restr
u
cturin
g L
a
ttic
e
T
heor
y
:
a
n
A
ppro
a
ch B
a
se
d
on
Hi
erarch
ies
of C
once
p
ts. I. Rival
(Ed.),
Ordered Sets, Reid
el, Dor
d
re
cht, Boston (19
82), pp. 44
5-47
0.
[10]
W
enxue H
o
n
g
, Jing
yi
ng Ma
o, Jianp
in
g Yu, Lins
ong
J
i
a. T
he comp
lete d
e
finiti
ons of attributes a
n
d
abstract descri
p
tion of attribut
e f
eatures of the formal co
ntext. ICIC
Expr
ess Letters. 20
13; 7(3):99
7
-
100
3.
[11]
Karp R M. Red
u
cibi
lit
y
amo
n
g
combin
atoria
l prob
lems. Ne
w
York: Plenum
Press, 1972
.
Evaluation Warning : The document was created with Spire.PDF for Python.