Sie sind auf Seite 1von 6

On Zadeh’s Compositional Rule of Inference ∗

Robert Fullér† Hans-Jürgen Zimmermann


rfuller@ra.abo.fi zi@buggi.or.rwth-aachen.de

Abstract
This paper deals with Zadeh’s compositional rule of inference [12] under triangular
norms: IF X is P AND X and Y have relation R, THEN Y is Q, where P and Q are
fuzzy sets of the real line IR, R is a fuzzy relation on IR and the conclusion Q is defined
via sup-T composition of P and R:

µQ (y) = sup T (µP (x), µR (x, y)), y ∈ IR.


x∈IR

It is shown that (i) when the triangular norm T and the membership function of the
observation P are continuous, then the conclusion Q depends continuously on the
observation; (ii) when T and the membership function of the relation R are continuous,
then the observation Q has continuous membership function. Furthermore, we present
similar results for the discrete case and for multiple fuzzy reasoning schemes.

Keywords: Compositional rule of inference, fuzzy relation, fuzzy interval, triangular norm,
extension principle

1 Preliminaries
In this Section we set up the notations and present a lemma needed in order to prove stability
and continuity properties of the compositional rule of inference under continuous triangular
norms.

Definition 1.1 A fuzzy interval A is a fuzzy quantity with a continuous, finite-supported,


fuzzy-convex and normalized membership function µA : IR → [0, 1].
The family of all fuzzy intervals will be denoted by F. Fuzzy intervals are often used to
represent linguistic variables [11]. An α-level set of a fuzzy interval A is a non-fuzzy set
denoted by [A]α and is defined by
[A]α = {t ∈ IR | µA (t) ≥ α}
for α ∈ (0, 1] and [A]α = cl(suppµA ) for α = 0.

The final version of this paper appeared in: R.Lowen and M.Roubens eds., Fuzzy Logic: State of the
Art, Theory and Decision Library, Series D, Kluwer Academic Publisher, Dordrecht, [ISBN 0-7923-2324-6],
1993 193-200.

Supported by the Hungarian Research Foundation under the projects OTKA T 4281, OTKA I/3 2152 and
by the German Academic Exchange Service (DAAD).

1
Definition 1.2 A function T : [0, 1] × [0, 1] → [0, 1] is said to be a triangular norm [10]
(t-norm for short) if T is symmetric, associative, non-decreasing in each argument, and
T (x, 1) = x for all x ∈ [0, 1].

Definition 1.3 Let A be a fuzzy interval, then for any θ ≥ 0 we define ωA (θ), the modulus
of continuity of A by
ωA (θ) = max |µA (u) − µA (v)|.
|u−v|≤θ

Definition 1.4 We metricize F by the metric [6],

D(A, B) = sup d([A]α , [B]α ),


α∈[0,1]

where d denotes the classical Hausdorff metric in the family of compact subsets of IR2 , i.e.

d([A]α , [B]α ) = max{|a1 (α) − b1 (α)|, |a2 (α) − b2 (α)|},

and [A]α = [a1 (α), a2 (α)], [B]α = [b1 (α), b2 (α)].

When the fuzzy sets A and B have finite support {x1 , . . . , xn }, then their Hamming distance
is defined as
n
H(A, B) = |µA (xi ) − µB (xi )|.
i=1

The following lemma [2] shows that if the distance between the α-level sets of fuzzy inter-
vals A and B is small, then the distance between their membership grades can not be too
big.

Lemma 1.1 Let δ ≥ 0 be a real number and let A, B be fuzzy intervals. If

D(A, B) ≤ δ,

then
sup |µA (t) − µB (t)| ≤ max{ωA (δ), ωB (δ)}.
t∈IR

where ωA and ωB denote the modolus of continuity of A and B, respectively.

2 Stability and continuity properties of the compositional


rule of inference
In this section we show two very important features of the compositional rule of inference
under triangular norms. Namely, we prove that (i) if the t-norm defining the composition
and the membership function of the observation are continuous, then the conclusion de-
pends continuously on the observation; (ii) if the t-norm and the membership function of
the relation are continuous, then the observation has a continuous membership function.
We consider the compositional rule of inference with different observations P and P  :

2
Observation: X has property P X has property P 
Relation: X and Y have relation R X and Y have relation R
Conclusion: Y has property Q Y has property Q

According to Zadeh’s compositional rule of inference, Q and Q are computed as

Q = P ◦ R, Q = P  ◦ R

i.e.,
µQ (y) = sup T (µP (x), µR (x, y)), µQ (y) = sup T (µP  (x), µR (x, y)).
x∈IR x∈IR

The following theorem shows that when the observations are close to each other in the
metric D, then there can be only a small deviation in the membership functions of the
conclusions.

Theorem 2.1 Let δ ≥ 0 and T be a continuous triangular norm, and let P , P  be fuzzy
intervals. If
D(P, P  ) ≤ δ
then
sup |µQ (y) − µQ (y)| ≤ ωT (max{ωP (δ), ωP  (δ)}).
y∈IR

Proof. Let y ∈ IR be arbitrarily fixed. From Lemma 2.1. it follows that

|µQ (y) − µQ (y)| = | sup T (µP (x), µR (x, y)) − sup T (µP (x), µR (x, y))| ≤
x∈IR x∈IR

sup |T (µP (x), µR (x, y)) − T (µP  (x), µR (x, y))| ≤ sup ωT (|µP (x) − µP  (x)|) ≤
x∈IR x∈IR

sup ωT (max{ωP (δ), ωP  (δ)}) = ωT (max{ωP (δ), ωP  (δ)}).


x∈IR

Which proves the theorem.

Remark 2.1 It should be noted that the stability property of the conclusion Q with respect
to small changes in the membership function of the observation P in the compositional rule
of inference scheme is independent from the relation R (it’s membership function can be
discontinuous).

Remark 2.2 Since the membership function of the conclusion in the compositional rule of
inference can have unbounded support, it is possible that the maximal distance between the
α-level sets of Q and Q is infinite, but their membership grades are arbitrarily close to
each other.

The following theorem establishes the continuity property of the conclusion in the compo-
sitional rule of inference scheme.

Theorem 2.2 Let R be continuous fuzzy relation, and let T be a continuous t-norm. Then
Q is continuous and
ωQ (δ) ≤ ωT (ωR (δ)) for each δ ≥ 0.

3
Proof. Let δ ≥ 0 be a real number and let u, v ∈ IR such that |u − v| ≤ δ. Then

|µQ (u) − µQ (v)| = | sup T (µP (x), µR (x, u)) − sup T (µP (x), µR (x, v))| ≤
x∈IR x∈IR

sup |T (µP (x), µR (x, u)) − T (µP (x), µR (x, v))| ≤ sup ωT (|µR (x, u) − µR (x, v)|) ≤
x∈IR x∈IR

sup ωT (ωR (|u − v|)) = ωT (ωR (|u − v|)) ≤ ωT (ωR (δ)).


x∈IR

Which ends the proof.

Remark 2.3 From Theorem 3.2. it follows that the continuity property of the membership
function of the conclusion Q in the compositional rule of inference scheme is independent
from the observation P (it’s membership function can be discontinuous).

The next theorem shows that the stability property of the conclusion under small changes
in the membership function of the observation holds in the discrete case, too.

Theorem 2.3 Let T be a continuous t-norm. If the observation P and the relation matrix
R are finite, then
H(Q, Q ) ≤ ωT (H(P, P  )) (1)
where H denotes the Hamming distance and the conclusions Q and Q are computed as

µQ (yj ) = max T (µP (xi ), µR (xi , yj )),


i=1,...,m

µQ (yj ) = max T (µP  (xi ), µR (xi , yj )),


i=1,...,m

for j = 1, . . . , n, supp(µQ ) = supp(µQ ) = {y1 , . . . , yn } and supp(µP ) = supp(µP  ) =


{x1 , . . . , xm }.

The proof of this theorem is carried out analogously to the proof of Theorem 3.1.
It should be noted that in the case of T (u, v) = min{u, v} (1) yields

H(Q, Q ) ≤ H(P, P  ).

The above theorems can be easily extended to Multiple Fuzzy Reasoning (MFR) schemes.
Consider the following MFR schemes:

Observation: P P
Implication 1: P1 → Q1 P1 → Q1
... ...
Implication m: Pm → Qm Pm → Qm
Conclusion: Q Q

where → denotes a fuzzy implication operator [8].


According to Zadeh’s compositional rule of inference, Q and Q are computed by sup-T
composition as follows

m 
m
 
Q=P◦ Pi → Qi , Q =P ◦ Pi → Qi ,
i=1 i=1

4
i.e.,
µQ (y) = sup T (µP (x), min µPi (x) → µQi (y)),
x∈IR i=1,...,m

µQ (y) = sup T (µP  (x), min µPi (x) → µQi (y)),
x∈IR i=1,...,m

The following two theorems can be proved similarly to Theorems 3.1 and 3.2.

Theorem 2.4 Let δ ≥ 0, let T be a continuous triangular norm, let P , P  , Pi , Pi , Qi , Qi ,
i = 1, . . . , m, be fuzzy intervals and let → be a continuous fuzzy implication operator. If

max{D(P, P  ), max D(Pi , Pi ), max D(Qi , Qi )} ≤ δ,


i=1,...,m i=1,...,m

then
sup |µQ (y) − µQ (y)| ≤ ωT (max{ω(δ), ω→ (ω(δ))}),
y∈IR

where
ω(δ) = max{ωPi (δ), ωPi (δ), ωQi (δ), ωQi (δ)},
and ω→ denotes the modulus of continuity of the fuzzy implication operator.

Theorem 2.5 Let → be a continuous fuzzy implication operator, let P , P  ,Pi , Pi , Qi , Qi ,
i = 1, . . . , m, be fuzzy intervals and let T be a continuous t-norm. Then Q is continuous
and
ωQ (δ) ≤ ωT (ω→ (ω(δ)) for each δ ≥ 0,
where
ω(δ) = max{ωPi (δ), ωPi (δ), ωQi (δ), ωQi (δ)},
and ω→ denotes the modulus of continuity of the fuzzy implication operator.

Other results along this line have appeared in [1, 3, 4, 5, 7, 8, 14].

Remark 2.4 From limδ→0 ω(δ) = 0 and Theorem 3.4. it follows that

sup |µQ (y) − µQ (y)| → 0


y∈IR

if
D(P, P  ) → 0, D(Pi , Pi ) → 0 and D(Qi , Qi ) → 0,
for i = 1, . . . , m, which means the stability of the conclusion under small changes of the
observation and rules.

The stability property of the conclusion under small changes of the membership function
of the observation and rules guarantees that small rounding errors of digital computation
and small errors of measurement of the input data can cause only a small deviation in the
conclusion, i.e. every successive approximation method can be applied to the computation
of the linguistic approximation of the exact conclusion.

5
References
[1] D.Dubois, R.Martin-Clouarie and H.Prade, Practical computing in fuzzy logic, In:
M.M.Gupta and T.Yamakawa eds., Fuzzy Computing: Theory, Hardware, and Ap-
plications, North-Holland, Amsterdam, 1988 11-34.

[2] M.Fedrizzi and R.Fullér, Stability in Possibilistic Linear Programming Problems


with Continuous Fuzzy Number Parameters, Fuzzy Sets and Systems, 47(1992) 187-
191.

[3] R.Fullér and H.-J.Zimmermann, Computation of the Compositional Rule of Infer-


ence, Fuzzy Sets and Systems Fuzzy Sets and Systems, 51(1992) 267-275

[4] R.Fullér and H.-J.Zimmermann, On Zadeh’s compositional rule of inference, in:


R.Lowen and M.Roubens eds., Proceedings of the Fourth IFSA Congress, Vol. Ar-
tifical Intelligence, Brussels, 1991 41-44.

[5] H.Hellendoorn, Closure properties of the compositional rule of inference, Fuzzy Sets
and Systems, 35(1990) 163-183.

[6] O.Kaleva, Fuzzy differential equations, Fuzzy Sets and Systems, 24(1987) 301-317.

[7] J.B.Kiszka, M.M.Gupta, P.N.Nikiforuk, Some properties of expert control systems,


In: M.M.Gupta, A.Kandel and J.B.Kiszka eds., Approximate Reasoning in Expert
Systems, North-Holland, Amsterdam, 1985, 283-306.

[8] R.Martin-Clouarie, Semantics and computation of the generalized modus ponens:


The long paper, International Journal of Approximate Reasoning, 3(1989) 195-217.

[9] P.Margrez and P.Smets, Fuzzy modus ponens: A new model suitable for applications
in knowledge-based systems, International Journal of Intelligent systems, 4(1989)
181-200.

[10] B.Schweizer and A.Sklar, Associative functions and abstract semigroups, Publ.
Math. Debrecen, 10(1963) 69-81.

[11] B.Werners, Modellierung und Aggregation Linguistischer Terme, Arbeitsbericht,


No.90/03, RWTH Aachen, Institut für Wirtschaftswissenschaften 1990.

[12] L.A.Zadeh, Outline of a new approach to the analysis of complex systems and de-
cision processes, IEEE Transactions on Systems, Man and Cybernetics, Vol.SMC-3,
No.1, 1973 28-44.

[13] L.A.Zadeh, The role of fuzzy logic in the management of uncertainty in expert sys-
tems, Fuzzy Sets and Systems, 11(1983) 199-228.

[14] H.-J.Zimmermann, Fuzzy sets, Decision Making and Expert Systems, Boston, Dor-
drecht, Lancaster 1987.

Das könnte Ihnen auch gefallen