You are on page 1of 9

PROD.

TYPE: COM ED: Pabitra Borah


PP: 1-9 (col.fig.: nil)
FSS5257 MODC+(SC) PAGN: Padmashini -- SCAN: Nil

ARTICLE IN PRESS

Fuzzy Sets and Systems ( ) –


www.elsevier.com/locate/fss

Uniqueness in the generalized representation by fuzzy sets


3 Fathi B. Saidi∗ , Ali Jaballah
Department of Mathematics, University of Sharjah, P.O. Box 27272, Sharjah, United Arab Emirates

5 Received 21 July 2007; received in revised form 30 December 2007; accepted 10 February 2008

F
O
Abstract

O
7 The concept of representation of a given family of sets by a fuzzy set in such a way that the level sets of the fuzzy set are precisely
the given sets is widely used in the literature for the classification and study of fuzzy algebraic structures on the same underlying set.
9

PR
The problem of existence of such fuzzy sets was investigated by various authors. As far as we know, the problem of uniqueness of
such fuzzy sets was not studied before. We establish, among other things, necessary and sufficient conditions under which uniqueness
11 is guaranteed in the cases of complete as well as partial representations of the level sets of the fuzzy set.
© 2008 Elsevier B.V. All rights reserved.
D
13 Keywords: Representation; Fuzzy set; Level set; Uniqueness; Existence; Order; Chain
TE

1. Introduction
EC

15 Throughout this paper X is a nonempty set and F (X) is the family of fuzzy subsets of X. The level sets of a fuzzy
set f ∈ F (X) are denoted by ft , t ∈ [0, 1], and are given by

17 ft = {x ∈ X: f (x) t} = f −1 [t, 1].


R

The family of all level sets of f is denoted by Cf and is given by


R

19 Cf := (ft )t∈[0,1] .
O

The representation theorem for fuzzy sets which first appeared in Negoita and Ralescu [15] proved to be a useful tool
21 for the definition, analysis, and operation with fuzzy concepts (see [18] for an early paper on this subject and [7] for a
C

more recent historical perspective). The need for such representations was rediscovered by many authors: see Borillo
23 and Fuentes [1], Zhang and Lee [28], Dubois and Prade [5, pp. 136–137], Luo [10], Turksen [27, p. 9], and Swammy
N

and Raju [26, p. 191], among others.


U

25 The essence of the representation theorem has to do with the possibility of constructing a fuzzy subset of the set X
from a given family of subsets of X while insisting that the level sets of this fuzzy set be the members of the family
27

∗ Corresponding author. Tel.: +971 650 50372; fax: +971 650 50352.
E-mail addresses: fsaidi@sharjah.ac.ae (F.B. Saidi), ajaballah@sharjah.ac.ae (A. Jaballah).

0165-0114/$ - see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.fss.2008.02.001

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001
FSS5257
ARTICLE IN PRESS
2 F.B. Saidi, A. Jaballah / Fuzzy Sets and Systems ( ) –

1 we started with. Clearly some conditions on the given family of subsets are necessary in order for such a fuzzy set to
exit. Hence one needs to address the following problem:

Problem 1. Let A := (At )t∈[0,1] be a family of subsets of X.


3
(i) Does there exist a fuzzy set f ∈ F (X) such that ft = At for all t ∈ [0, 1]?
5 (ii) Moreover, if the answer to the question in Part (i) is affirmative, do we have uniqueness?

Or, more generally, one may consider the following problem:

Problem 2. Let A := (At )t∈[0,1] be a family of subsets of X and : [0, 1] → [0, 1] be a given function.
7
(i) Does there exist a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1]?
9 (ii) Moreover, if the answer to the question in Part (i) is affirmative, do we have uniqueness?

F
Note that Problem 1 is the special case of Problem 2 corresponding to the case where  = id. Problem 1 was

O
11 considered by Negoita and Ralescu, [15–17], and by Ralescu, [18], while Problem 2 was studied more recently by
Ralescu, [19]. For each of the two problems, necessary and sufficient conditions for the existence of the fuzzy set

O
13 were established. Characterizations of the structures of the collections of level sets that comply with the conditions of
Negoita and Ralescu were established in [2,3].

PR
15 It is well known that different fuzzy subsets of a set X can have equal families of level sets. Therefore, there were
many attempts to classify fuzzy sets and fuzzy algebraic structures on the same underlying set X, which have equal
17 collections of level sets, see e.g., Degang et al. [4], Gorjanac-Ranitovic and Tepavcevic [6], Šešelja and Tepavčević
[25], and the references cited there. Many global notions of fuzzy sets have been characterized by level sets, [9,11,12].
D
19 In [6], the authors gave necessary and sufficient conditions for the existence of a fuzzy set corresponding to a given
family of level sets in the general setting of lattice-valued fuzzy sets. In addition, they show that the general form of
TE

21 lattice-valued fuzzy sets are the type of fuzzy sets having the codomain {0, 1}c for a suitable chosen cardinal c.
Instead of using equality of the collections of level sets, some authors, see e.g., Murali and Makamba [14,13], use
23 equivalence relations on the collection of fuzzy sets defined in terms of conditions on the degrees of memberships
(range of values of the fuzzy sets), but these turn out to be, in general, equivalent to the equality of the collection
EC

25 of level sets. As mentioned in [24], the main reason behind such classification is the fact that, in general, there is an
uncountable number of distinct fuzzy subsets of the (finite or infinite) set R, and that the same holds true for fuzzy
27 algebraic structures. But not all of these fuzzy sets (or fuzzy algebraic structures) can be considered essentially different,
R

especially in the cases where they have equal collections of level sets. This led many authors to the classifications of
29 fuzzy algebraic structures by equality of level sets, see e.g. [16].
R

Šešelja and Tepavčević investigated the representation of partially ordered sets by fuzzy sets. They provided an
31 algorithm for the construction of a fuzzy set corresponding to a given partially ordered set [21]. They have also
O

investigated some algebraic properties of the given partially ordered set and the corresponding fuzzy set [22]. A survey
33 of results on the representation of a family of partially ordered sets by a fuzzy set has been recently published in [23].
C

In most previously established results only the question of existence of a fuzzy set corresponding to a given family
35 of subsets of X was addressed. Ralescu proved the following existence theorem for Problem 2 [19, Theorem 2]:
N

Theorem 1. Let : [0, 1] → [0, 1] be given, and let {At }t∈[0,1] be a family of subsets of X satisfying ∪t∈[0,1] At = X.
U

The necessary and sufficient conditions for the existence of a fuzzy set f ∈ F (X) such that f(t) = At , 0 t 1, are
37
(i) (s) (t) ⇒ As ⊇ At ; 
39 (ii) (t1 )(t2 ) · · · and (tn ) → (t) ⇒ ∞
n=1 Atn = At .

Before proceeding, we should mention here the fact that in the actual generalized theorem of Ralescu, there was no
41 mention of the assumption that
∪ At = X. (1)
t∈[0,1]

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001
FSS5257
ARTICLE IN PRESS
F.B. Saidi, A. Jaballah / Fuzzy Sets and Systems ( ) – 3

1 We have added this condition under the belief that Ralescu did indeed make that assumption while proving his theorem,
as without it his proof would be incomplete. Indeed, if Eq. (1) is not satisfied, then the fuzzy set

3 f (x) = sup{(): x ∈ A }

defined by Ralescu at the beginning of his proof needed to be defined also for x ∈ X\ ∪t∈[0,1] At , which was not done.
5 Alternatively, one may assume that sup ∅ = 0, as is customary in complete lattices, in which case we would get, for
each x ∈ X\ ∪t∈[0,1] At ,

7 f (x) = 0.

But in this case one has to impose, in addition, the condition that 0 ∈
/ range() to avoid any contradictions. For more
9 details, we refer the reader to Remark 2.
The problem of uniqueness under different considerations was investigated for the first time in [8,20]. There the

F
11 authors considered the following modified problem:

O
Problem 3. Let X be a nonempty set, S be a nonempty subset of the interval [0, 1], and A be a family of nonempty

O
subsets of X.
13
(i) Does there exist a fuzzy set f ∈ F (X) such that f (X) = S and Cf = A?

PR
15 (ii) If the answer to Part (i) is affirmative, do we have uniqueness?

They gave a complete characterization of the collection of all fuzzy sets whose level sets are precisely the members
D
17 of the given family A and whose collection of membership values is the given family S. Necessary and sufficient
conditions for the existence and uniqueness of such a fuzzy set were obtained.
TE

19 It is important to note that in Problem 3 we do not impose the condition that each member of the family A corresponds
to a specific level set as opposed to Problem 2 where it is required that each member A of the family A corresponds
21 specifically to the level set f() . It follows that the questions of uniqueness for the two problems are completely
different. In [8] the authors proved that, in the case of Problem 3, the fuzzy set is unique if and only if the collection S
EC

23 of membership values is a rigid set, in the sense that the only order isomorphism on S is the identity. If, for example,
S is the entire interval [0, 1], then the fuzzy set is not unique. The situation is different in the case of Problem 2 where
25 uniqueness is guaranteed under similar circumstances, namely when the range of  is the entire interval [0, 1], see
R

Corollary 2.
27 We aim in this paper to investigate the problem of uniqueness of the fuzzy set under the setting of Problem 2. In
R

general, such a fuzzy set, if it exists, does not have to be unique and one needs to impose some conditions on the given
29 family of subsets and/or the function  in order to obtain uniqueness. We show that in the cases where the function
O

 is onto, the fuzzy set, if it exists, is indeed unique, as expected. As for the cases where the function  is not onto,
31 we obtain necessary and sufficient conditions under which uniqueness is guaranteed. Note that, in general, Problem 2
requires only partial representation of the fuzzy set we seek. Indeed, if  is not onto, then the family A will represent
C

33 the level sets corresponding to the range of  only and not necessarily all the level sets. Therefore, in these cases, one
N

should impose the condition that the function  is onto in order to guarantee that all level sets are represented by the
35 family A. What if  is not one-to-one? In that case redundancy occurs in the sense that there may be many members
U

of A corresponding to the same level set. More precisely we have:

37 Remark 1. If f and  are as in Problem 2 and if t0 := () = () for some 


=  in [0, 1], then A = A = ft0 ,
hence both A and A correspond to the same level set ft0 .

39 Note that we may still have A = A even when ()


= (). This happens when f() = f() , which in turn
happens when

41 f −1 ([min{(), ()}, max{(), ()})) = ∅.

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001
FSS5257
ARTICLE IN PRESS
4 F.B. Saidi, A. Jaballah / Fuzzy Sets and Systems ( ) –

1 2. Characterization of uniqueness

Our main concern in this section is the investigation of Problem 2. We start with the following definition:

3 Definition 1. Let A = (At )t∈[0,1] be a family of subsets of the set X. We say that A is a decreasing chain if At ⊆ As
whenever t s in [0, 1], in other words if the mapping t → At from the totally ordered set ([0, 1], ) to the totally
5 ordered set (A, ⊆) is an order isomorphism.

The following observation addresses the question of uniqueness when Eq. (1) does not hold:

7 Remark 2. Given a family A = (At )t∈[0,1] of subsets of X and given a function : [0, 1] → [0, 1], it is clear that if
there exists a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1] and if ∪∈[0,1] A X then 0 ∈
/ range(), since
9 f0 = X, and
∪ A ) ⊆ [0, inf(range())]\range().

F
f (X\
∈[0,1]

O
11 In particular, if ∪∈[0,1] A X and if inf(range()) > 0, then the fuzzy set f is not unique and may be redefined
arbitrarily on X\ ∪∈[0,1] A with values in [0, inf(range())]\range(). Moreover, if ∪∈[0,1] A X and if f is a fuzzy

O
13 set defined on ∪∈[0,1] A and satisfying f(t) = At for all t ∈ [0, 1], then
the extension of f to X is unique if and only if inf(range()) = 0.

PR
15 It follows that, in order to obtain any kind of uniqueness result, we must restrict ourselves to the case where
∪∈[0,1] A = X.
D
17 In view of the previous remark, we assume for the remainder of this paper, and without loss of generality, that the
family A = (At )t∈[0,1] of subsets of X always satisfies Eq. (1), namely
TE

∪ At = X.
19 t∈[0,1]

The following remark shows, in particular, that, when  is not one-to-one, redundancy occurs in the sense that there
EC

21 may be more than one element of the family A corresponding to the same level set:

Remark 3. Let A = (At )t∈[0,1] be a family of subsets of the set X satisfying ∪t∈[0,1] At = X and let : [0, 1] → [0, 1]
23 be a given function. Suppose that there exists a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1]. Then, for
R

every  ∈ range(),
At = As , for every t, s ∈ −1 ().
R

25
O

We need to introduce some notations. Let A = (At )t∈[0,1] be a family of subsets of the set X and let : [0, 1] → [0, 1]
27 be a given function. Since the relation on the interval [0, 1] defined by
C

s ∼ t ⇔ As = At
N

29 is an equivalence relation, we obtain that the interval [0, 1] is the disjoint union of the distinct equivalence classes I ,
 ∈ J,
U


[0, 1] = I ,
31 ∈J

where J ⊆ [0, 1] may be (and is) chosen, by the axiom of choice, in such a way that
33  ∈ I for all  ∈ J.
Therefore, for every  ∈ J , we have
35 At = A for all t ∈ I . (2)

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001
FSS5257
ARTICLE IN PRESS
F.B. Saidi, A. Jaballah / Fuzzy Sets and Systems ( ) – 5


1 Also, for each x ∈ X, we let Tx := {(t): t ∈ [0, 1], x ∈ At } and we let h ∈ F (X) be the fuzzy set defined by

h(x) := sup(Tx ) := sup{(t): t ∈ [0, 1], x ∈ At } for each x ∈ X.
3 Finally, for each  ∈ J , we define the annulus Aann
 in X by

 := A \ ∪ {A :  ∈ J, A A }.
Aann
5  may be empty for some  ∈ J .
Note that Aann
The following lemma establishes some properties for the family A and the function :

7 Lemma 1. Let A = (At )t∈[0,1] be a family of subsets of the set X satisfying ∪t∈[0,1] At = X and let : [0, 1] → [0, 1]
be a given function. Suppose that there exists a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1]. Then:
9 (i) For every s, t ∈ [0, 1] we have
As ⊇ At if and only if (s)(t).

F
O
11 Therefore the family A is totally ordered.
(ii) For every ,  ∈ J we have

O
13 A A if and only if () < ().
(iii) For every ,  ∈ J satisfying A A , we have

PR
15 sup[(I )] <  for every  ∈ (I ).
Hence
D
17 sup[(I )] inf[(I )] whenever A A .
TE

Proof. Parts (i) and (ii) are trivial. As for Part (iii), let ,  ∈ J be such that A A and let x ∈ A \A . From the
19 assumptions and from Eq. (2) we obtain that, for each  ∈ J , f = A for all  ∈ (I ) and, consequently,
f = fsup[(I )] = A for every  ∈ (I ).
EC

21 It follows, since A A , that


fsup[(I )] f for every  ∈ (I ),
R

23 which implies that sup[(I )] <  for all  ∈ (I ) and, consequently,
R

sup((I )) inf((I )).


O

25 This ends the proof. 


C

We now establish some preliminary results on the fuzzy sets f satisfying f(t) = At for all t ∈ [0, 1] and their
27 connection with the fuzzy set h defined above:
N

Lemma 2. Let A := (At )t∈[0,1] be a family of subsets of X satisfying ∪t∈[0,1] At = X and let : [0, 1] → [0, 1] be a
U

given function. If there exists a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1] then:
29
(i) The fuzzy set f satisfies
31 f (x)h(x) for every x ∈ X.
(ii) The fuzzy set h ∈ F (X) does satisfy
33 h(t) = At for every t ∈ [0, 1].
Hence the fuzzy set f is unique if and only if f = h.

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001
FSS5257
ARTICLE IN PRESS
6 F.B. Saidi, A. Jaballah / Fuzzy Sets and Systems ( ) –

1 (iii) For every x ∈ X and every  ∈ J , x ∈ Aann


 if and only if

h(x) = sup[(I )] = sup[∪{(I ) :  ∈ J, A ⊇ A }].

3 (iv) For every x ∈ X, h(x) ∈ range() if and only if there exists  ∈ J such that

x ∈ Aann
 and sup[(I )] ∈ range().

5 Hence,
 
h−1 (range()) = h−1 ((I )) = Aann
 .
∈J ∈J


7 Proof. Part (i): Let x ∈ X and let (t) ∈ Tx for some t ∈ [0, 1]. Then

x ∈ At = f(t) ,

F
O
9 where the equality follows from the assumptions. It follows that f (x)(t) and consequently, since (t) was arbitrary

in Tx ,

O

11 f (x) sup(Tx ) = h(x).

PR
Part (ii): Let t ∈ [0, 1] be fixed and let x ∈ X be arbitrary. Then

13 x ∈ At ⇒ (t) ∈ Tx ⇒ (t)h(x) ⇒ x ∈ h(t) .

Conversely we have, since f (x) h(x) by Part (i),


D
15 x ∈ h(t) ⇒ h(x)(t) ⇒ f (x) (t) ⇒ x ∈ f(t) = At ,
TE

where the equality follows from the assumptions on f. This completes the proof of this part.
17 Part (iii): Let x ∈ Aann
 . By the definition of h, we have

EC

h(x) = sup(Tx ) = sup{(t): t ∈ [0, 1], x ∈ At }


= sup[∪{(I ):  ∈ J, A ⊇ A }] = sup[(I )],

where the last two equalities follow from Lemma 1, the fact that [0, 1] = ∈J I , and the fact that x ∈ Aann
 .
R

19 Conversely, let x ∈ X be such that


R

h(x) = sup[(I )].

Then x ∈ A and, by Lemma 1, x ∈ / A whenever A A , hence x ∈ Aann


O

21  . This ends the proof of the lemma.


Part (iv): Let x ∈ X be such that h(x) ∈ range(). Then h(x) = (t0 ) for some t0 ∈ [0, 1] and, consequently,
C

23 x ∈ h(t0 ) and x∈
/ h(t) whenever (t) > (t0 ).
N

It follows, by Part (ii), that


U

25 x ∈ At0 and x∈
/ At whenever (t) > (t0 ).

Since [0, 1] = ∪∈J I , there exists  ∈ J such that t0 ∈ I . Since At = At0 for all t ∈ I and since x ∈
/ At whenever
27 (t) > (t0 ), we obtain that (t)(t0 ) for all t ∈ I and, consequently,

h(x) = (t0 ) = sup[(I )].

29 It follows, by Part (iii), that

x ∈ Aann
 and sup[(I )] ∈ range().

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001
FSS5257
ARTICLE IN PRESS
F.B. Saidi, A. Jaballah / Fuzzy Sets and Systems ( ) – 7

1  and sup[(I )] ∈ range() for some  ∈ J . Then there exists s ∈ [0, 1]


Conversely, let x ∈ X be such that x ∈ Aann
such that
3 x ∈ As and sup[(I )] = (s).
It follows, by Lemma 1(i) and by the definition of I , that
5 x ∈ As and x∈
/ At for all t satisfying (t) > (s).
Hence, by Part (ii),
7 h(x) = sup{(t): x ∈ At } = sup{(t): x ∈ h(t) } = (s),
which implies that x ∈ range(). This completes the proof of the lemma. 

9 We are now ready to give our characterization of uniqueness:

F
Theorem 2. Let A = (At )t∈[0,1] be a family of subsets of the set X satisfying ∪t∈[0,1] At = X, : [0, 1] → [0, 1] be a

O
11 given function, and let h be the fuzzy set defined by
h(x) := sup{(t): x ∈ At } for each x ∈ X.

O
13 Suppose that there exists a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1]. Then the fuzzy set f is unique

PR
if and only if, for every x ∈ X, either h(x) = 1 or h(x) is a left-limit point of range(), in the sense that there exits
15 (t1 ) > (t2 ) > · · · such that
(tn )  h(x).
D
17 Proof. ⇒: Suppose that this is not true. Then there exists x0 ∈ X and ε > 0 such that
Kε := (h(x), h(x) + ε] ∩ range() = ∅ and Kε ⊂ [0, 1].
TE

(3)
19 Also, by Lemma 2(ii), we must have f = h. Now, define g on X by

h(x) if x
= x0 ,
EC

g(x) =
h(x0 ) + ε/2 if x = x0 .
21 Then
g
= h
R

23 and, for every t ∈ [0, 1], we have


R

g(t) = g −1 ([(t), 1]) = h−1 ([(t), 1]) = h(t) = At ,


O

25 where the second equality follows from Eq. (3) and Lemma 2(iii) and the last equality follows from Lemma 2(ii). This
C

contradicts the uniqueness assumption.


27 ⇐: Suppose that this not true. Then, by Lemma 2, there exists g ∈ F (X) such that g(t) = At for all t ∈ [0, 1] and
N

such that
29 g(x0 ) > h(x0 ) for some x0 ∈ X.
U

Clearly, we must have h(x0 ) < 1. It follows, by assumption, that h(x0 ) is a left-limit point of range(). Hence, there
31 exists  ∈ [0, 1] such that
h(x0 ) < () < g(x0 ).
33 It follows from the first and second inequalities, respectively, that
x0 ∈
/ h() = A and x0 ∈ g() = A ,
35 which is not possible. This completes the proof of the theorem. 

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001
FSS5257
ARTICLE IN PRESS
8 F.B. Saidi, A. Jaballah / Fuzzy Sets and Systems ( ) –

1 As a consequence of Theorem 2, we obtain:

Corollary 1. Let A = (At )t∈[0,1] be a family of subsets of the set X satisfying ∪t∈[0,1] At = X and let : [0, 1] → [0, 1]
3 be a given function. Suppose that there exists a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1]. Then the
fuzzy set f is unique if and only if, for every x ∈ X,
 
1 if x ∈ At ,
h(x) = t∈[0,1]
5 / At } if x ∈
inf{(t): t ∈ [0, 1], x ∈ / A for some  ∈ [0, 1].

In the special case where range() = [0, 1] we obtain immediately:

7 Corollary 2. Let A = (At )t∈[0,1] be a family of subsets of the set X and let : [0, 1] → [0, 1] be a given function
satisfying range() = [0, 1]. If there exists a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1], then the family
9 A does satisfy ∪t∈[0,1] At = X and the fuzzy set f is unique and is given by, for every x ∈ X,

F

f (x) = h(x) := sup(Tx ) := sup{(t): t ∈ [0, 1], x ∈ At }.

O
11 Combining Corollary 1 with Remark 2 we obtain:

O
Corollary 3. Let A = (At )t∈[0,1] be a family of subsets of the set X and let : [0, 1] → [0, 1] be a given function.

PR
13 Suppose that there exists a fuzzy set f ∈ F (X) such that f(t) = At for all t ∈ [0, 1]. Then the fuzzy set f is unique if
and only if inf(range()) = 0 and, for every x ∈ X,
 
1 if x ∈ At ,
h(x) = t∈[0,1]
D
15 / At } if x ∈
inf{(t): t ∈ [0, 1], x ∈ / A for some  ∈ [0, 1].
TE

We finish with the following comments:


17 1. It follows immediately that the fuzzy set f, if it exists, is always unique under the setting of Problem 1. We
note that in this case uniqueness also follows from Corollary 2, since Problem 1 is a special case of Problem 2
EC

19 corresponding to the case when  is the identity function on [0, 1].


2. In the general setting of Problem 2, it follows from Theorem 2 that it is possible to have uniqueness even when
21 the given family A does not represent all the level sets of the fuzzy set f . To clarify the reasons behind what
makes this possible and the ideas behind the proof of the theorem: First note that, by Lemma 2, the I ’s,  ∈ J ,
R

23 are not only disjoint but also do not interlace in the sense that, if 
=  in J, then either I < I or I < I .
R

Then the fuzzy set f is unique when there is no “gap” between the I ’s and between the number 1 and the I ’s.
25 This guarantees that, for every x ∈ X, h(x) cannot be pushed slightly upward, and hence the fuzzy set f must be
O

equal to h.
C

27 References
N

[1] P.J. Borillo, R. Fuentes, A short note on representation of L-fuzzy sets by Moore’s families, Stochastica 8 (1984) 291–295.
29 [2] B. De Baets, Chains, kite-tail lattices and dense chains, Tatra Mt. Math. Publ. 21 (2001) 1–18.
U

[3] B. DeBaets, E. Kerre, Kite-tail lattices and their characterization, Far East J. Math. Sci. 1 (1993) 1–11.
31 [4] C. Degang, J. Jiashang, W. Congxin, A.C.C. Tsang, Some notes on equivalent fuzzy sets and fuzzy subgroups, Fuzzy Sets and Systems 152
(2005) 403–409.
33 [5] D. Dubois, H. Prade, Fuzzy sets, probability and measurement, European J. Oper. Res. 40 (1989) 135–154.
[6] M. Gorjanac-Ranitovic, A. Tepavcevic, General form of lattice-valued fuzzy sets under the cutworthy approach, Fuzzy Sets and Systems 158
35 (2007) 1213–1216.
[7] U. Höhle, Editorial, Fuzzy Sets and Systems 40 (1991) 253–256.
37 [8] A. Jaballah, F.B. Saidi, Uniqueness results in the representation of families of sets by fuzzy sets, Fuzzy Sets and Systems 157 (2006) 964–975.
[9] R. Kumar, Certain fuzzy ideals of rings redefined, Fuzzy Sets and Systems 46 (1992) 251–260.
39 [10] C.Z. Luo, Book announcement, Fuzzy Sets and Systems 39 (1991) 235–236.
[11] D.S. Malek, Fuzzy ideals of Artinian rings, Fuzzy Sets and Systems 37 (1990) 111–115.

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001
FSS5257
ARTICLE IN PRESS
F.B. Saidi, A. Jaballah / Fuzzy Sets and Systems ( ) – 9

1 [12] T.K. Mukherjee, P. Bhattachary, Fuzzy groups: some group theoretic analogs, Inform. Sci. 39 (1986) 247–268.
[13] V. Murali, B.B. Makamba, On the equivalence of fuzzy subgroups I, Fuzzy Sets and Systems 123 (2001) 259–264.
3 [14] V. Murali, B.B. Makamba, On the equivalence of fuzzy subgroups II, Fuzzy Sets and Systems 136 (2003) 93–104.
[15] C.V. Negoita, D.A. Ralescu, Applications of Fuzzy Sets to Systems Analysis, Wiley, New York, Toronto, Ont., 1975 191pp.
5 [16] C.V. Negoita, D.A. Ralescu, Representation theorems for fuzzy concepts, Kybernetes 4 (1975) 169–174.
[17] C.V. Negoita, D.A. Ralescu, Simulation, Knowledge-Based Computing, and Fuzzy Statistics, Van Nostrand Reinhold, New York, 1987 pp.
7 89–91.
[18] D.A. Ralescu, A survey of the representation of fuzzy concepts and its applications, in: M.M. Gupta, R.K. Ragade, R. Yager (Eds.), Advances
9 in Fuzzy Sets Theory and Applications, North-Holland, Amsterdam, 1979, pp. 77–91.
[19] D.A. Ralescu, A generalization of the representation theorem, Fuzzy Sets and Systems 51 (1992) 309–311.
11 [20] F.B. Saidi, A. Jaballah, From fuzzy sets to the decomposition of non-rigid sets, Fuzzy Sets and Systems 158 (2007) 1751–1766.
[21] B. Šešelja, A. Tepavčević, On a representation of posets by fuzzy sets, Fuzzy Sets and Systems 98 (1) (1998) 127–132.
13 [22] B. Šešelja, A. Tepavčević, Completion of ordered structures by cuts of fuzzy sets: an overview, Fuzzy Sets and Systems 136 (1) (2003) 1–19.
[23] B. Šešelja, A. Tepavčević, Representing ordered structures by fuzzy sets: an overview, Fuzzy Sets and Systems 136 (1) (2003) 21–39.
15 [24] B. Šešelja, A. Tepavčević, A note on a natural equivalence relation on fuzzy power set, Fuzzy Sets and Systems 148 (2004) 201–210.
[25] B. Šešelja, A. Tepavčević, Equivalent fuzzy sets, Kybernetika 41 (2) (2005) 115–128.
17 [26] U.M. Swammy, D.V. Raju, Algebraic fuzzy systems, Fuzzy Sets and Systems 41 (1991) 187–194.

F
[27] I.B. Turksen, Measurement of membership functions and their acquisition, Fuzzy Sets and Systems 40 (1991) 5–38.
19

O
[28] W. Zhang, H. Lee, The structure of the norm system on fuzzy sets, J. Math. Anal. Appl. 127 (1987) 559–568.

O
PR
D
TE
EC
R
R
O
C
N
U

Please cite this article as: F.B. Saidi, A. Jaballah, Uniqueness in the generalized representation by fuzzy sets, Fuzzy Sets and Systems (2008),
doi: 10.1016/j.fss.2008.02.001

You might also like