26.08.2013 Views

On the cardinality of fuzzy sets - EUSFLAT

On the cardinality of fuzzy sets - EUSFLAT

On the cardinality of fuzzy sets - EUSFLAT

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>On</strong> <strong>the</strong> <strong>cardinality</strong> <strong>of</strong> <strong>fuzzy</strong> <strong>sets</strong><br />

Pavol Kráˇl<br />

Matej Bel University, Tajovského 40, 97401 Banská Bystrica<br />

kral@fpv.umb.sk<br />

Abstract<br />

It seems that a suitably constructed <strong>fuzzy</strong><br />

<strong>sets</strong> <strong>of</strong> natural numbers form <strong>the</strong> most complete<br />

and adequate description <strong>of</strong> <strong>cardinality</strong><br />

<strong>of</strong> finite <strong>fuzzy</strong> <strong>sets</strong>.(see [11]) Never<strong>the</strong>less,<br />

in many applications one needs a<br />

simple scalar evaluation <strong>of</strong> that <strong>cardinality</strong><br />

by nonnegative real number, e.g scalar<br />

<strong>cardinality</strong>. There are many approaches to<br />

this evaluation-sigma count <strong>of</strong> <strong>fuzzy</strong> set,psigma<br />

count <strong>of</strong> <strong>fuzzy</strong> <strong>sets</strong>, <strong>cardinality</strong> <strong>of</strong><br />

its core or support, <strong>cardinality</strong> <strong>of</strong> its αcut<br />

set,etc.(see [2]), [3], [4], [5], [7], [9],<br />

[10] for more details). Wygralak in [8]<br />

present an axiomatic approach to <strong>the</strong> scalar<br />

<strong>cardinality</strong> <strong>of</strong> <strong>fuzzy</strong> <strong>sets</strong> which contains as<br />

particular cases all standard concepts <strong>of</strong><br />

scalar <strong>cardinality</strong>. The best approximation<br />

<strong>of</strong> scalar <strong>cardinality</strong> <strong>of</strong> <strong>fuzzy</strong> set is presented<br />

in [12]. The aim <strong>of</strong> this paper is to select<br />

numbers from Wygralak’s best approximation<br />

which are coherent with human intuition.<br />

This selection is based on <strong>the</strong> dependency<br />

among objects which can be represented<br />

by <strong>fuzzy</strong> compatibility.<br />

Keywords: <strong>cardinality</strong> <strong>of</strong> <strong>fuzzy</strong> <strong>sets</strong>, dependency<br />

among objects, <strong>fuzzy</strong> compatibility.<br />

1 Introduction<br />

Let X be an universal set. Fuzzy <strong>sets</strong> in X will be identified<br />

with <strong>the</strong>ir membership functions A : X → [0,1].<br />

The <strong>cardinality</strong> <strong>of</strong> <strong>fuzzy</strong> set A will be denoted by<br />

card(A). The <strong>cardinality</strong> <strong>of</strong> crisp set B will be denoted<br />

by |B|. Fuzzy compatibility is a function C :<br />

X 2 → [0;1] fulfilling <strong>the</strong> following properties<br />

C(a,a) = 1,<br />

C(a,b) = C(b,a),<br />

for each a,b ∈ X. We use <strong>the</strong> following basic definitions<br />

<strong>of</strong> <strong>fuzzy</strong> set <strong>the</strong>ory:<br />

Aα = {x ∈ X : A(x) ≥ α}<br />

with α ∈]0,1] (α-cut set <strong>of</strong> A),<br />

A α = {x ∈ X : A(x) > α}<br />

with α ∈ [0,1[ (sharp α-cut set <strong>of</strong> A ),<br />

(core <strong>of</strong> A),<br />

core(A) = Aα<br />

supp(A) = A 0<br />

(support <strong>of</strong> A ). If supp(A) is finite, we shall say that<br />

A is a finite <strong>fuzzy</strong> set. Measuring <strong>the</strong> <strong>cardinality</strong> <strong>of</strong> a<br />

finite <strong>fuzzy</strong> set has many applications. (<strong>fuzzy</strong> querying<br />

in databases, expert systems, evaluation <strong>of</strong> natural<br />

language statements, aggregation, decision making in<br />

<strong>fuzzy</strong> environment, metrical analysis <strong>of</strong> grey images,<br />

etc. (see [3], [7], [15] for more details and fur<strong>the</strong>r reference).<br />

<strong>On</strong>e <strong>of</strong> <strong>the</strong> scalar cardinalities is <strong>the</strong> <strong>cardinality</strong><br />

described by Wygralak in [12]. Wygralak proposes<br />

one <strong>of</strong> <strong>the</strong> values <strong>of</strong> <strong>the</strong> interval<br />

[|A 0.5 |,|A0.5|],<br />

as <strong>the</strong> best scalar approximation <strong>of</strong> <strong>cardinality</strong> <strong>of</strong><br />

<strong>fuzzy</strong> set. Some authors think that only valid values<br />

for <strong>the</strong> <strong>cardinality</strong> <strong>of</strong> <strong>fuzzy</strong> set are <strong>the</strong> cardinalities<br />

<strong>of</strong> its α-cuts because only this representation preserve<br />

<strong>the</strong> dependency among objects from <strong>the</strong> point <strong>of</strong> view<br />

<strong>of</strong> <strong>cardinality</strong>.


Example 1.1 [1] Let X={John, Mike, Peter} be a set<br />

<strong>of</strong> friends, and let us suppose that John is blonde, and<br />

Mike and Peter are fairly blonde. We could say that<br />

John is in <strong>the</strong> set <strong>of</strong> blonde people with degree 1 and<br />

Mike and Peter are in <strong>the</strong> same set with degree 0.5.<br />

Suppose we want to buy a cap for every blonde person<br />

in X. Deciding how many caps we shall buy is<br />

equivalent to obtaining <strong>the</strong> <strong>cardinality</strong> <strong>of</strong> <strong>the</strong> <strong>fuzzy</strong> set<br />

XBLONDE = 1/John + 0.5/Mike + 0.5/Peter.<br />

Following Wygralak’s <strong>cardinality</strong>, <strong>the</strong> possible cardinalities<br />

for XBLONDE are [1,3]. If we are strict with<br />

<strong>the</strong> concept <strong>of</strong> blonde, we shall buy only one cap (for<br />

John). But if we relax our criterion we could think <strong>of</strong><br />

buying three caps. It is clear that buying two caps<br />

is not a solution for <strong>the</strong> problem, i.e <strong>cardinality</strong> <strong>of</strong><br />

XBLONDE can be one or three, but not two. (Mike is<br />

in XBLONDE iff Peter also is )<br />

In our opinion, only for α “near”1 is valid to take into<br />

account <strong>the</strong> dependency among objects with respect<br />

to <strong>the</strong> property “to be in A ”. In o<strong>the</strong>r cases <strong>the</strong> <strong>fuzzy</strong><br />

compatibility is more suitable to express dependency<br />

among objects. Let us consider <strong>the</strong> following example:<br />

Example 1.2 Let A = 0.5/x1 + 0.5/x2 + 0.5/x3 +<br />

0.5/x4 be a <strong>fuzzy</strong> set. The possible <strong>cardinality</strong> for A<br />

belong to [0,4]. If we assume <strong>the</strong> dependency among<br />

objects (from <strong>the</strong> point <strong>of</strong> view <strong>of</strong> <strong>cardinality</strong>) <strong>the</strong> possible<br />

cardinalities are 4 and 0, i.e. each xi is in A<br />

or none <strong>of</strong> xi is in A. Following properties <strong>of</strong> an entropy<br />

measure, proposed in [2], A is <strong>the</strong> fuzziest set<br />

over four objects, i.e we can assume each integer from<br />

[0,4] as a possible <strong>cardinality</strong> <strong>of</strong> A (integer from [0,4]<br />

are equivalent) and we need additional property <strong>of</strong> objects<br />

xi to select <strong>the</strong> best approximation <strong>of</strong> <strong>cardinality</strong>.<br />

The natural additional property <strong>of</strong> objects xi is a <strong>fuzzy</strong><br />

compatibility for example.<br />

We will define <strong>the</strong> selection from Wygralak’s best approximation<br />

which is more appropriate.<br />

Definition 1.1 Let A be a finite <strong>fuzzy</strong> subset <strong>of</strong> X.<br />

Let A(x1),A(x2),...A(xn) are <strong>the</strong> values <strong>of</strong> A. Let<br />

C : A 2 → [0,1] be a <strong>fuzzy</strong> compatibility. Let 0.5 A =<br />

{x ∈ X : A(x) = 0.5}. The possible scalar <strong>cardinality</strong><br />

<strong>of</strong> <strong>fuzzy</strong> set A with respect to <strong>the</strong> <strong>fuzzy</strong> compatibility<br />

is:<br />

<br />

D + A<br />

card(A) =<br />

0.5 i f D > 1<br />

A0.5 else<br />

where D = max<br />

x∈0.5 <br />

|{y ∈0.5 A,: C(x,y) ≥ β : |} and<br />

A<br />

β ∈ [0,1] is a treshold.<br />

Example 1.3 Let A from example 1.2 be a set <strong>of</strong> middle<br />

aged people. Let x1, x2 are 20 year old and x3,<br />

x4 are 60 year old. Let β = 1. Let C1(xi,xj) = 1 iff<br />

xi,xj are in <strong>the</strong> same age and C1(xi,xj) = 0 else. It<br />

is easy to see that card(A) = 2. Let C2(xi,xj) = 1 iff<br />

A(xi) = A(x j) and C2(xi,xj) = 0 else. It is easy to see<br />

that card(A) = 4.<br />

Proposition 1.1 Let A be a finite <strong>fuzzy</strong> set. Let C :<br />

A 2 → [0,1] be a fixed <strong>fuzzy</strong> compatibility.<br />

1. If β → 0 card(A) → |A0.5|<br />

2. If β → 1 card(A) → |A 0.5 |<br />

Proposition 1.2 Let A be a finite <strong>fuzzy</strong> set. Let β be a<br />

fixed treshold<br />

1. If for all i, j,i = j C(xi,xj) < β card(A) = |A 0.5 |<br />

2. If for all i, j,i = j C(xi,xj) ≥ β card(A) = |A0.5|<br />

Proposition 1.3 Let C be a fixed compatibility. Let β<br />

be a fixed treshold. We can construct a naive <strong>cardinality</strong><br />

<strong>the</strong>ory.<br />

The selection <strong>of</strong> a compatibility and a treshold depends<br />

on a concrete case.<br />

References<br />

[1] Delgado, M., Sánchez, D., Martin - Bautista, M.,<br />

Vila, M., A. (2002) A probabilistic definition <strong>of</strong><br />

a nonconvex <strong>fuzzy</strong> <strong>cardinality</strong>. In Fuzzy Sets and<br />

Systems , volume 126, pages 177-190.<br />

[2] De Luca, A., Termini, S.(1972) A definition <strong>of</strong><br />

non-probabilistic entropy in <strong>the</strong> setting <strong>of</strong> <strong>fuzzy</strong><br />

<strong>sets</strong> <strong>the</strong>ory. In Inform. Control, volume 20, pages<br />

301-312.<br />

[3] Dubois, D., Prade, D. (1985) Fuzzy <strong>cardinality</strong><br />

and <strong>the</strong> modeling <strong>of</strong> imprecise quantification.<br />

Fuzzy Sets and Systems, volume 16, pages 199-<br />

230.<br />

,


[4] Gottwald, S. (1980) A note on <strong>fuzzy</strong> cardinals.<br />

In Kybernetika, volume 16, pages 156-158<br />

[5] Kaufmann, A. (1977) Introduction a la <strong>the</strong>orie<br />

des sous-ensembles flous. In Complement et<br />

Nouvelles Applications, volume 4<br />

[6] Liu, Y., Kerre, E., E.(1998) An overview <strong>of</strong><br />

<strong>fuzzy</strong> quantifiers.(I). Interpretations. In Fuzzy<br />

Sets and Systems, volume 95,pages 1-21.<br />

[7] Ralescu, D.(1995) Cardinality, quantifiers, and<br />

<strong>the</strong> aggregation <strong>of</strong> <strong>fuzzy</strong> criteria. In Fuzzy Sets<br />

and Systems, volume 69, pages 355-365.<br />

[8] Wygralak, M. (2000) An axiomatic approach to<br />

scalar cardinalities <strong>of</strong> a <strong>fuzzy</strong> set. In Fuzzy Sets<br />

and Systems, volume 110, pages 175-176.<br />

[9] Wygralak, M. (1997) Cardinalities <strong>of</strong> <strong>fuzzy</strong> <strong>sets</strong><br />

evaluated by single cardinals. In Proc. IFSA<br />

Congress Praha, pages 73 - 77<br />

[10] Wygralak, M. (2001) Fuzzy <strong>sets</strong> with triangular<br />

norms and <strong>the</strong>ir <strong>cardinality</strong> <strong>the</strong>ory. In Fuzzy Sets<br />

and Systems, volume 124, pages 1-24.<br />

[11] Wygralak, M. (1993) Generalized cardinal numbers<br />

and operation on <strong>the</strong>m. In Fuzzy Sets and<br />

Systems, volume 53, pages 49-85.<br />

[12] Wygralak, M. (1997) <strong>On</strong> <strong>the</strong> best scalar approximation<br />

<strong>of</strong> <strong>cardinality</strong> <strong>of</strong> a <strong>fuzzy</strong> set. In<br />

International journal <strong>of</strong> uncertainty, fuzziness<br />

and knowledge-based systems, volume 56, pages<br />

681-687.<br />

[13] Wygralak, M. (1999) Question <strong>of</strong> <strong>cardinality</strong> <strong>of</strong><br />

finite <strong>fuzzy</strong> <strong>sets</strong>. In Fuzzy Sets and Systems, volume<br />

102, pages 185-210.<br />

[14] Wygralak, M. (1999) Scalar cardinalities <strong>of</strong><br />

<strong>fuzzy</strong> <strong>sets</strong> with triangular norms and conorms.<br />

In Proc. Inter. Eur<strong>of</strong>use - SIC 99 Conf. Budapest,<br />

pages 322-327.<br />

[15] Zadeh,L. A. M (1983) A computational approach<br />

to <strong>fuzzy</strong> quantifiers in natural languages.<br />

In Comput. Math. Appl., volume 9, pages 149-<br />

184.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!