systemic approach to fuzzy logic formalization for approximate reasoning
TRANSCRIPT
Information Sciences 181 (2011) 1045–1059
Contents lists available at ScienceDirect
Information Sciences
journal homepage: www.elsevier .com/locate / ins
Systemic approach to fuzzy logic formalization for approximate reasoning
Rafik Aliev a,⇑, Alex Tserkovny b
a Department of Control Systems, Intelligent Systems Research Laboratory, Azerbaijan State Oil Academy, Baku, Azerbaijanb Dassault Systems, Boston, USA
a r t i c l e i n f o a b s t r a c t
Article history:Received 7 August 2009Received in revised form 16 November 2010Accepted 19 November 2010
Keywords:Fuzzy logicImplicationAntecedentConsequentModus-ponensFuzzy conditional inference ruleStabilityContinuity
0020-0255/$ - see front matter � 2010 Elsevier Incdoi:10.1016/j.ins.2010.11.021
⇑ Corresponding author. Tel.: +994 125984509.E-mail addresses: [email protected] (R. Aliev), a
L.A. Zadeh, E.H. Mamdani, M. Mizumoto, et al., R.A. Aliev and A. Tserkovny have proposedmethods for fuzzy reasoning in which antecedents and consequents involve fuzzy condi-tional propositions of the form ‘‘If x is A then y is B’’, with A and B being fuzzy concepts(fuzzy sets). A formulation of fuzzy antecedent/consequent chains is one of the most impor-tant topics within a wide spectrum of problems in fuzzy sets in general and approximatereasoning, in particular. From the analysis of relevant research it becomes clear that for thispurpose, a so-called fuzzy conditional inference rules comes as a viable alternative. In thisstudy, we present a systemic approach toward fuzzy logic formalization for approximatereasoning. For this reason, we put together some comparative analysis of fuzzy reasoningmethods in which antecedents contain a conditional proposition with fuzzy concepts andwhich are based on implication operators present in various types of fuzzy logic. We alsoshow a process of a formation of the fuzzy logic regarded as an algebraic system closedunder all its operations. We examine statistical characteristics of the proposed fuzzy logic.As the matter of practical interest, we construct a set of fuzzy conditional inference rules onthe basis of the proposed fuzzy logic. Continuity and stability features of the formalizedrules are investigated.
� 2010 Elsevier Inc. All rights reserved.
1. Introduction
In our daily life we often make inferences where antecedents and consequents are represented by fuzzy sets. Such infer-ences cannot be realized adequately by the methods, which are based either on two-valued logic or many-valued logic. Inorder to facilitate such an inference, Zadeh [76–81] suggested an inference rule called a ‘‘compositional rule of inference’’.Using this inference rule, Zadeh, Mamdani [48], Mizumoto et al. [26,54,55], Aliev and Tserkovny [1–3] suggested severalmethods for fuzzy reasoning in which the antecedent contain a conditional proposition involving fuzzy concepts:
Ant 1 : If x is P then y is Q
Ant 2 : x is P0
--------------------------Cons : y is Q 0:
ð1Þ
Those methods are based on implication operators present in various fuzzy logics. This matter has been under a thoroughdiscussion for the last couple decades. Some comparative analysis of such methods was presented in [6–11,26,29,34,36–38,50,51,54,55,67,74,75,85]. A number of authors proposed to use a certain suite of fuzzy implications to form fuzzy conditional
. All rights reserved.
[email protected] (A. Tserkovny).
1046 R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059
inference rules [2–4,25–27,40,42,43,48,54,55]. The implication operators present in the theory of fuzzy sets were investigatedin [2–5,12–23,28,30–33,39,41,44,46,47,49,51–53,56–59,61–66,68,70–73,75,82–84]. On the other hand, statistical features offuzzy implication operators were studied in [60,69]. In turn, the properties of stability and continuity of fuzzy conditional infer-ence rules were investigated in [24,27,35,40].
However, a systemic approach toward fuzzy logic formalization for an approximate reasoning task has never been pre-sented. The intention of this study is to fill this gap. We will begin with a formation of a fuzzy logic regarded as an algebraicsystem closed under all its operations. In the sequel An investigation of statistical characteristics of the proposed fuzzy logicwill be presented. Special attention will be paid to building a set of fuzzy conditional inference rules on the basis of the fuzzylogic proposed in this study. Next, continuity and stability features of the formalized rules will be investigated.
In what follows, we introduce some required notation. x and y are the names of objects, while P, P0, Q and Q0 are fuzzyconcepts represented by fuzzy sets defined in the universes of discourse U, U, V and V, respectively. This form of inferencemay be viewed as a generalized modus ponens, which reduces to modus ponens when P0 = P and Q0 = Q. Let P and Q be fuzzysets in U and V, respectively, while the corresponding fuzzy sets are represented as P � UjlP: U ? [0,1], Q � VjlQ: V ? [0,1],where
P ¼Z
UlPðuÞ=u; Q ¼
ZVlQ ðvÞ=v : ð2Þ
Let �, [, \, : and � denoted a Cartesian product, union, intersection, complement and bounded-sum for fuzzy sets, respec-tively. Then the following fuzzy relations in U � V can be derived from the fuzzy conditional proposition ‘‘If x is P then y is Q’’in Ant 1 of (1). The fuzzy relations Rm and Ra were proposed by Zadeh [45,47], Rc by Mamdani [48], Rs, Rg are by Mizumoto[26,54,55], while, RL2, RL3 are by Aliev and Tserkovny [1–3].
Rm ¼ ðP � QÞ [ ð:P � VÞ ¼Z
U�VðlPðuÞ ^ lQ ðvÞÞ _ ð1� lPðuÞÞ=ðu;vÞ; ð3Þ
Ra ¼ ð:P � VÞ � ðU � QÞ ¼Z
U�V1 ^ ð1� lPðuÞ þ lQ ðvÞÞ=ðu;vÞ; ð4Þ
Rc ¼ ðP � QÞ ¼Z
U�VðlPðuÞ ^ lQ ðvÞÞ=ðu;vÞ; ð5Þ
Rs ¼ ðP � V)s
U � QÞ ¼Z
U�VðlPðuÞ!s lQ ðvÞÞ=ðu; vÞ; ð6Þ
where
lPðuÞ!s lQ ðvÞ ¼1; lPðuÞ 6 lQ ðvÞ;0; lPðuÞ > lQ ðvÞ:
(
Rg ¼ ðP � V)g
U � QÞ ¼Z
U�VðlPðuÞ!g lQ ðvÞÞ=ðu;vÞ; ð7Þ
and
lPðuÞ!g lQ ðvÞ ¼1; lPðuÞ 6 lQ ðvÞ;lQ ðvÞ; lPðuÞ > lQ ðvÞ:
(
Rb ¼ ð:P � V [ U � QÞ ¼Z
U�Vð1� lPðuÞÞ _ lQ ðvÞ=ðu;vÞ; ð8Þ
RD ¼ ðP � V)D
U � QÞ ¼Z
U�V½lPðuÞ!D lQ ðvÞ�=ðu; vÞ; ð9Þ
where
lPðuÞ!D lQ ðvÞ ¼1; lPðuÞ 6 lQ ðvÞ;lQ ðvÞlP ðuÞ
; lPðuÞ > lQ ðvÞ:
(
R� ¼ ðP � V)�
U � QÞ ¼Z
U�V½lPðuÞ!� lQ ðvÞ�=ðu; vÞ; ð10Þ
where
lPðuÞ!� lQ ðvÞ ¼ ðlPðuÞ ^ lQ ðvÞÞ _ ð1� lPðuÞ ^ 1� lQ ðvÞÞ _ ðlQ ðvÞ ^ 1� lPðuÞÞ:
R} ¼ ðP � V)}
U � QÞ ¼Z
U�V½lPðuÞ!} lQ ðvÞ�=ðu; vÞ; ð11Þ
R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059 1047
where
lPðuÞ!} lQ ðvÞ ¼1; lPðuÞ < 1 or lQ ðvÞ ¼ 1;0; lPðuÞ ¼ 1 or lQ ðvÞ < 1:
(
RL1 ¼ ðP � V)L1
U � QÞ ¼Z
U�V½lPðuÞ!L1
lQ ðvÞ�=ðu; vÞ; ð12Þ
where
lPðuÞ!L1lQ ðvÞ ¼
1� lPðuÞ; lPðuÞ < lQ ðvÞ;1; lPðuÞ ¼ lQ ðvÞ;lQ ðvÞ; lPðuÞ > lQ ðvÞ:
8><>:
RL2 ¼ ðP � V)L2
U � QÞ ¼Z
U�V½lPðuÞ!L2
lQ ðvÞ�=ðu; vÞ; ð13Þ
where
lPðuÞ!L2lQ ðvÞ ¼
1; lPðuÞ 6 lQ ðvÞ;ð1� lPðuÞÞ ^ lQ ðvÞ; lPðuÞ > lQ ðvÞ:
(
RL3 ¼ ðP � V)L3
U � QÞ ¼Z
U�V½lPðuÞ!L3
lQ ðvÞ�=ðu; vÞ; ð14Þ
where
lPðuÞ!L3lQ ðvÞ ¼
1; lPðuÞ 6 lQ ðvÞ;lQ ðvÞ
lP ðuÞþð1�lQ ðvÞÞ; lPðuÞ > lQ ðvÞ:
(
A necessary observation to be made in the context of this discussion is that with the only few exceptions for S-logic (6)and G-logic (7), and L1–L3 (12)–(14), all other known fuzzy logics (3)–(5) and (8)–(11) do not satisfy either the classical‘‘modus-ponens’’ principle, or other criteria which appeal to the human perception of mechanisms of a decision making pro-cess being formulated in [54]. The proposed fuzzy logic comes with an implication operator, which satisfies the classicalprinciple of ‘‘modus-ponens’’ and meets some additional criteria being in line with human intuition.
The material of the study is organized as follows. In Section 2, we offer some required prerequisites. Section 3 is dedicatedto a formalization of the proposed fuzzy logic. Next, Section 4 provides a certain introduction to a power sets based featuresof the fuzzy logic. The statistical analysis of proposed fuzzy logic is reported in Section 5. Section 6 covers the formalizationof fuzzy conditional inference rules where the rules exploit the proposed fuzzy logic. Section 7 is devoted to investigations ofstability and continuity of proposed fuzzy conditional inference rules.
2. Preliminary considerations
Consider a continuous function F(p,q) = p � q which defines a distance between p and q where p, q assume values in theunit interval. Notice that Fðp; qÞ 2 ½�1;1�, where F(p,q)min = �1 and F(p,q)max = 1. The normalized version of F(p,q) is definedas follows:
Fðp; qÞnorm ¼ Fðp; qÞ � Fðp; qÞmin
Fðp; qÞmax � Fðp; qÞmin ¼Fðp; qÞ þ 1
2¼ p� qþ 1
2: ð15Þ
It is clear thatF(p,q)norm 2 [0,1]. This function quantifies a concept of ‘‘closeness’’ between two values (potentially the onesfor the truth values of antecedent and consequent), defined within unit interval, which therefore could play significant role inthe formulation of the implication operator in a fuzzy logic.
Definition 1. An implication is a continuous function I from [0,1] � [0,1] into [0,1] such that for "p, p0, q, q0, r 2 [0,1] thefollowing properties are satisfied
(I1) If p 6 p0 then I(p,q) P I(p0,q) (antitone in first argument),(I2) If q 6 q0 then I(p,q) 6 I(p,q0) (monotone in second argument),(I3) I(0,q) = 1 (falsity),(I4) I(1,q) 6 q (neutrality),(I5) I(p, I(q,r)) = I(q, I(p,r)) (exchange),(I6) I(p,q) = I(n(q),n(p)) (contra positive symmetry), where n(�) – is a negation, which could be defined as n(q) = T(:Q) =
1 � T(Q).
1048 R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059
Definition 2 [7]. Given a fuzzy implication operator ? and a fuzzy subset Q of universe U, the fuzzy power set PQ of Q isgiven by the membership function lPQ, where
lPQ P ¼\x2U
ðlPx! lQ xÞ: ð16Þ
The degree to which P is subset of Q is defined as
pðP # QÞ ¼ lPQ P:
Definition 3 [7]. From Definition 2, the degree to which the fuzzy sets P and Q is the same, or their degree of similarity, is
pðP QÞ ¼ pðP # QÞ ^ pðP QÞ: ð17Þ
The following proposition is immediate.
Proposition 1 [7].
pðP QÞ ¼\x2U
ðlPx$ lQ xÞ: ð18Þ
Definition 4 [7]. The degree of disjointness of P and Q, or degree to which P and Q are disjointed, in the first and secondsense, is defined as follows
ð1Þ pðPdisj1QÞ ¼ pðP # Q CÞ ^ pðQ BCÞ; ð19Þð2Þ pðPdisj2QÞ ¼ pðP \ Q ;Þ: ð20Þ
Definition 5 [7]. A degree to which a set is a subset of its complement is defined in the form p(P # PC).
Definition 6 [7]. A degree to which a set is disjointed from its complement, in terms of the two interpretations providedabove is defined as p(Pdisj1PC) and p(Pdisj2PC)).
3. The fuzzy logic
Let us define the implication operation
Iðp; qÞ ¼ 1� Fðp; qÞnorm; p > q;
1; p 6 q;
(ð21Þ
where F(p,q)norm is expressed by (15). Before showing that operation I(p,q) satisfies axioms (I1)–(I6), let us show some basicoperations encountered in proposed fuzzy logic.
Let us designate the truth values of the antecedent P and consequent Q as T(P) = p and T(Q) = q, respectively. The relevantset of proposed fuzzy logic operators is shown in Table 1. To obtain the truth values of these expressions, we use well knownlogical properties such as p ? q = :p _ q; p ^ q = :(:p _ :q) and alike.
In other words, we propose a new many-valued system, characterized by the set of union ([) and intersection (\) opera-tions with relevant complement, defined as T(:P) = 1 � T(P). In addition, the operators ; and " are expressed as negations ofthe [ and \, respectively. It is well known that the implication operation in fuzzy logic supports the foundations of decision-making exploited in numerous schemes of approximate reasoning. Therefore let us prove that the proposed implication oper-ation in (21) satisfies axioms (I1)–(I6). For this matter, let us emphasize that we are working with a many-valued system,which values for our purposes are the elements of the real interval R ¼ ½0;1�. For our discussion the set of of truth valuesV11 = {0,0.1,0.2, . . . ,0.9,1} is sufficient. In further investigations, we use this particular set V11. Table 2 shows the implicationoperation developed in the proposed fuzzy logic.
Theorem 1. Let a continuous function I(p,q) be defined by (21) taking values present in Table 2, i.e.,
Iðp; qÞ ¼ 1� Fðp; qÞnorm; p > q
1; p 6 q
(¼
1�pþq2 ; p > q;
1; p 6 q:
(ð22Þ
where F(p,q)norm is a defined by (1).
Then axioms (I1)–(I6) are satisfied and, therefore (22) is an implication operation.
Table 1
Name Designation Value
Tautology PI 1Controversy PO 0Negation :P 1 � PDisjunction P _ Q pþq
2 ; pþ q – 1;1; pþ q ¼ 1
�Conjunction P ^ Q pþq
2 ; pþ q – 1;0; pþ q ¼ 1
�Implication P ? Q 1�pþq
2 ; p – q;1; p ¼ q
�Equivalence P M Q minððp� qÞ; ðq� pÞÞ; p – q;
1; p ¼ q
�Pierce arrow P ; Q 1� pþq
2 ; pþ q – 1;0; pþ q ¼ 1
�Shaffer stroke P " Q 1� pþq
2 ; pþ q – 1;1; pþ q ¼ 1
�
Table 2
p ? q 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1
0 1 .55 .6 .65 .7 .75 .8 .85 .9 .95 1.1 .45 1 .55 .6 .65 .7 .75 .8 .85 .9 .95.2 .4 .45 1 .55 .6 .65 .7 .75 .8 .85 .9.3 .35 .4 .45 1 .55 .6 .65 .7 .75 .8 .85.4 .3 .35 .4 .45 1 .55 .6 .65 .7 .75 .8.5 .25 .3 .35 .4 .45 1 .55 .6 .65 .7 .75.6 .2 .25 .3 .35 .4 .45 1 .55 .6 .65 .7.7 .15 .2 .25 .3 .35 .4 .45 1 .55 .6 .65.8 .1 .15 .2 .25 .3 .35 .4 .45 1 .55 .6.9 .05 .1 .15 .2 .25 .3 .35 .4 .45 1 .551 0 .05 .1 .15 .2 .25 .3 .35 .4 .45 1
R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059 1049
Proof
(I1) "p,p0 2 [0,1] jp0 P p > q) I(p,q) � I(p0,q) = 1 � p + q � 1 + p0 � q = p0 � p P 0) I(p,q) P I(p0,q),whereas q P p0 P p) Iðp; qÞ � Iðp0; qÞ 0) Iðp; qÞ Iðp0; qÞ.
(I2) "q,q0 2 [0,1]jq 6 q0 < p) I(p,q) � I(p,q0) = 1 � p + q � 1 + p � q0 = q � q0 6 0) I(p,q) 6 I(p,q0),whereas p 6 q 6 q0 ) Iðp; qÞ � Iðp; q0Þ 0) Iðp; qÞ Iðp; q0Þ.
(I3) I(0,q) 1, q P 0.
(I4) Ið1; qÞ ¼q2 ; q – 1;1; q ¼ 1:
�) Ið1; qÞ 6 q;
(I5) Notice that since p ? q = k(1 � p + q), where k = 0.5jp > q or 1jp 6 q, then "p,q,r 2 [0,1]jp > q > r) p ? (q ? r) =k(p ? (q ? r)) = k(p ? (1 � q + r)) = k(1 � p + 1 � q + r) = k(2 � p � q + r),whereas kðq! ðp! rÞÞ ¼ kðq! ð1� pþ rÞÞ ¼ kð1� qþ 1� pþ rÞ ¼ kð2� p� qþ rÞ.
(I6) IðnðqÞ;nðpÞÞ ¼ ð1� qÞ ! ð1� pÞ ¼qþ1�p
2 ; 1� q > 1� p1; 1� q 6 1� p
�¼
1�pþq2 ; p > q;
1; p 6 q:
�h
In addition, the proposed fuzzy logic is characterized by the following properties:Commutativity of both conjunction (\) and disjunction ([) operations, i.e.: p ^ q = q ^ p and
p _ q ¼ q _ p:
Assotiativity of these operations:
p ^ ðq ^ rÞ ¼ ðp ^ qÞ ^ r andp _ ðq _ rÞ ¼ ðp _ qÞ _ r:
Partial idempotency:
p ^ p ¼p; p > 0:5;0; p 6 0:5;
�
1050 R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059
p _ p ¼p; p < 0:5;1; p P 0:5:
�
Partial distributivity:
ðp _ qÞ ^ r ¼ ðp ^ rÞ _ ðq ^ rÞ; if pþ qþ 2r 6 1;ðp _ qÞ ^ r – ðp ^ rÞ _ ðq ^ rÞ; if pþ qþ 2r > 1;
ð23Þ
p _ ðq ^ rÞ ¼ ðp _ qÞ ^ ðp _ rÞ; if 2pþ qþ r 6 1;p _ ðq ^ rÞ– ðp _ qÞ ^ ðp _ rÞ; if 2pþ qþ r > 1:
ð24Þ
To prove (23), we note that
pþ qþ 2r 6 1jðp _ qÞ ^ r ¼ pþ q2
� �^ r ¼
pþq2 þ r
2¼ pþ qþ 2r
4: ð25Þ
On the other hand, we have
ðp ^ rÞ _ ðq ^ rÞ ¼ pþ r2
� �_ qþ r
2
� �¼ pþ r
2þ qþ r
2
� �.2 ¼ pþ qþ 2r
4: ð26Þ
Therefore expression (25) is the same as (26).To prove (24), we carry out the following calculations,
2pþ qþ r 6 1jp _ ðq ^ rÞ ¼ p _ qþ r2
� �¼ pþ qþ r
2
� �.2 ¼ ð2pþ qþ rÞ
4; ð27Þ
ðp _ qÞ ^ ðp _ rÞ ¼ pþ q2
� �^ pþ r
2
� �¼ ð2pþ qþ rÞ
4: ð28Þ
Therefore expression (27) is the same as (28).DeMorgan theorems,
:ðp ^ qÞ ¼ :p _ :q;
:ðp _ qÞ ¼ :p ^ :q;
hold as well. To prove them, let us notice that
:ðp ^ qÞ ¼ : pþq2
� �1
(¼ 1� pþq
2
� �1
(¼
2�p�q2
1
(: ð29Þ
On the other hand
:p _ :q ¼:pþ:q
2
1
�¼
2�p�q2
1
(: ð30Þ
Therefore expression (29) is the same as (30).By analogy
:ðp _ qÞ ¼ : pþq2
� �0
(¼ 1� pþq
2
� �0
(¼
2�p�q2
0
(: ð31Þ
On the other hand
:p ^ :q ¼:pþ:q
2
0
�¼
2�p�q2
0
(: ð32Þ
Therefore expression (31) is the same as (32).It should be mentioned that the proposed fuzzy logic could be characterized by yet some other three features.p ^ 0 0, p 6 1. Whereas p _ 1 1, p P 0 and ::p = p. As a conclusion, we should admit that all above features confirm
that resulting system can be applied to V11 for every finite and infinite n up to that (V11,:,^,_,?) is then closed under all itsoperations.
4. Fuzzy power sets and fuzzy logic
Considering (16)–(18) and taking into account (22), we can formulate the following propositions.
Proposition 2 (Degree of possibility of set-inclusion).
R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059 1051
pðP # QÞ ¼1�lP xþlQ x
2 ; lPx > lQ x;
1; lPx 6 lQ x:
(ð33Þ
Proposition 3 (Degree of possibility of set-equality).
pðP QÞ ¼
1�lPxþlQ x2 ; lPx > lQ x;
1; lPx ¼ lQ x;
1�lQ xþlPx2 ; lPx < lQ x:
8>>><>>>:
ð34Þ
From (34) it is clear that p(P Q) = 1 in case when lPx = lQQ and p(P Q) 6 0.5 otherwise. As it was mentioned in [9],there seem to be two plausible ways to define the degree to which sets P and Q may be said to be disjointed. One is thedegree to which each is a subset of the other’s complement. The second is the degree to which their intersection isempty.
Proposition 4 (Degree of disjointness of P and Q, or degree to which P and Q are disjointed). For the case (1) from (19)
pðP # Q CÞ ¼ p! ð1� qÞ ¼1�pþ1�q
2 ; p > 1� q
1; p 6 1� q
(¼
2�ðpþqÞ2 ; pþ q > 1;
1; pþ q 6 1;
(ð35Þ
pðQ BCÞ ¼ q! ð1� pÞ ¼1�qþ1�p
2 ; q > 1� p
1; q 6 1� p
(¼
2�ðpþqÞ2 ; pþ q > 1;
1; pþ q 6 1:
(ð36Þ
Therefore from (35) and (36), the definition (1) can arise in the form
pðPdisj1QÞ ¼2�ðpþqÞ
2 ; pþ q > 1;1; pþ q 6 1:
(ð37Þ
For case (2) from (20) and given that
p ^ q ¼pþq
2 ; pþ q > 1;0; pþ q 6 1:
�
We have
pðPdisj2QÞ ¼ pðp ^ q 0Þ ¼1; pþ q 6 1;0; pþ q > 1:
�ð38Þ
Proposition 5. Based on Definition 5 the degree to which a set is a subset of its complement is formulated as follows:
pðP # PCÞ ¼ p! ð1� pÞ ¼1�pþ1�p
2 ; p > 1� p1; p 6 1� p
(¼
1� p; p > 0:5;1; p 6 0:5:
�
Proposition 6. The degree to which a set is disjointed from its complement is based on the Definition 6. From expressions (37) and(38) we make an observation: p(Pdisj1PC) 1 whereas p(Pdisj2PC) 0.
5. Statistical properties of the fuzzy logic
In this section, we discuss some properties of the proposed fuzzy implication operator (21), assuming that the two prop-ositions (antecedent/consequent) in a given compound proposition are independent of each other and the truth values of thepropositions are uniformly distributed [20] in the unit interval In other words, we assume that the propositions P and Q areindependent from each other and the truth values v(P) and v(Q) are uniformly distributed across the interval [0,1]. Letp = v(P) and q = v(Q). Then the value of the implication I = v(p ? q) could be represented as the function I = I(p,q).
Because p and q are assumed to be uniformly and independently distributed across [0,1], the expected value of the impli-cation is
EðIÞ ¼ZZ
R
Iðp; qÞdpdq; ð39Þ
1052 R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059
Its variance is equal to
VarðIÞ ¼ E½ðI � EðIÞÞ2� ¼ZZ
R
ðIðp; qÞ � EðIÞÞ2dpdq ¼ E½I2� � E½I�2; ð40Þ
where R ¼ fðp; qÞ : 0 6 p 6 1;0 6 q 6 1g. From (39) and given (21) as well as the fact that Iðp; qÞ ¼ I1ðp; qÞ; p > q;I2ðp; qÞ; p 6 q
�. We
have the following
EðI1Þ ¼ZZ
R
I1ðp; qÞdpdq ¼Z 1
0
Z 1
0
1� pþ q2
dpdq ¼ 12
Z 1
0
Z 1
0ð1� pþ qÞdp
� �dq
¼ 12
Z 1
0p� p2
2þ q
� �jp¼1p¼0
� �dq
¼ 1
212þ q2
2
����q¼1
q¼0
" #¼ 1
2: ð41Þ
Whereas E(I2) = 1. Therefore E(I) = (E(I1) + E(I2))/2 = 0.75.From (40) we have
I21ðp; qÞ ¼
14ð1� pþ qÞ2 ¼ 1
4ð1� 2pþ 2qþ p2 � 2pqþ q2Þ;
EðI21Þ ¼
ZZR
I21ðp; qÞdpdq ¼ 1
4
Z 1
0
Z 1
0ð1� 2pþ 2qþ p2 � 2pqþ q2Þdp
� �
dq ¼ 14
Z 1
0p� 2
p2
2þ p3
3� 2
p2
2qþ 2qþ q2
����p¼1
p¼0dq ¼ 1
4
Z 1
0
13þ qþ q2
� �dq ¼ 1
4q3þ q2
2þ q3
3
����q¼1
q¼0¼ 7
24:
Here EðI22Þ ¼ 1. Therefore EðI2Þ ¼ ðEðI2
1Þ þ EðI22ÞÞ=2 ¼ 31
48.From (40) and (41) we have VarðIÞ ¼ 1
12 ¼ 0:0833.Both values of E(I) and Var(I) demonstrate that the proposed fuzzy implication operator could be considered as one of the
fuzziest from the list of the exiting implications [32]. In addition, it satisfies the set of important Criteria I–IV, which is notthe case for the most implication operators mentioned above.
6. The fuzzy logic and fuzzy conditional inference
As it was mentioned in [26] ‘‘. . . in the semantics of natural language there exist a vast array of concepts and humans veryoften make inferences antecedents and consequences of which contain fuzzy concepts . . .’’ A formalization of methods forsuch inferences is one of the most important issues in fuzzy sets theory, in particular Artificial Intelligence, in general.For this purpose, let U and V (from now on) be two universes of discourses and corresponding fuzzy sets P and Q are the sameas in (2).
Given (2), a binary relationship for the fuzzy conditional proposition of the type: ‘‘If x is P then y is Q’’ for proposed fuzzylogic is defined as
RðA1ðxÞ;A2ðyÞÞ ¼ P � V ! U � Q ¼Z
U�VlPðuÞ=ðu;vÞ !
ZU�V
lQ ðvÞ=ðu;vÞ ¼Z
U�VðlPðuÞ ! lQ ðvÞÞ=ðu;vÞ: ð43Þ
Given (21), expression (43) reads as
lPðuÞ ! lQ ðvÞ ¼1�lPðuÞþlQ ðvÞ
2 ; lPðuÞ > lQ ðvÞ;1; lPðuÞ 6 lQ ðvÞ:
(ð44Þ
It is well known that given a unary relationship R(A1(x)) = P one can obtain the consequence R(A2(y)) by applying a compo-sitional rule of inference (CRI) to R(A1(x)) and R(A1(x),A2(y)) of type (43):
RðA2ðyÞÞ ¼ P � RðA1ðxÞ;A2ðyÞÞ ¼Z
UlPðuÞ=u �
ZU�V
lpðuÞ ! lQ ðvÞ=ðu;vÞ ¼Z
V
[u2U
½lPðuÞ ^ ðlPðuÞ ! lQ ðvÞÞ�=v : ð45Þ
In order to have Criterion I satisfied, that is R(A2(y)) = Q from (45), the equality
[u2U½lPðuÞ ^ ðlPðuÞ ! lQ ðvÞÞ� ¼ lQ ðvÞ ð46Þ
has to be satisfied for any arbitrary v in V. To satisfy (46), it becomes necessary that the inequality
lPðuÞ ^ ðlPðuÞ ! lQ ðvÞÞ 6 lQ ðvÞ ð47Þ
R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059 1053
holds for arbitrary u 2 U andv 2 V. Let us define a new method of fuzzy conditional inference of the following type:
Ant 1 : If x is P then y is Q
Ant 2 : x is P0
-------------------------Cons : y is Q 0:
ð48Þ
where P, P # U and Q, Q # V. Fuzzy conditional inference in the form given by (48) should satisfy Criteria I–IV presented inthe Appendix. It is clear that the inference (48) is defined by the expression (45), whenR(A2(y)) = Q0.
Theorem 2. If fuzzy sets P # U and Q # V are defined by (43) and (44) , respectively and R(A1(x),A2(y)) is expressed as
RðA1ðxÞ;A2ðyÞÞ ¼ P � V!L4
U � Q ¼Z
U�VlPðuÞ=ðu;vÞ!L4
ZU�V
lQ ðvÞ=ðu;vÞ ¼Z
U�VðlPðuÞ!L4
lQ ðvÞÞ=ðu; vÞ;
where
lPðuÞ!L4lQ ðvÞ ¼
1�lP ðuÞþlQ ðvÞ2 ; lPðuÞ > lQ ðvÞ;
1; lPðuÞ 6 lQ ðvÞ;
(ð49Þ
then Criteria I, II, III and IV-1 [26] are satisfied.
Proof. For Criteria I–III let R(A1(x)) = Pa(a > 0) then
RðA2ðyÞÞ ¼ Pa � RðA1ðxÞ;A2ðyÞÞ ¼Z
Ula
PðuÞ=u �Z
U�VlpðuÞ!L4
lQ ðvÞ=ðu;vÞ ¼Z
V
[u2U
laPðuÞ ^ ðlPðuÞ!L4
lQ ðvÞÞ ,
v :
ð50Þ
9U1;U2 � UjU1 [ U2 ¼ U; jU1 \ U2 ¼ ; ) 8u 2 U1jlPðuÞ > lQ ðvÞ; 8u 2 U2jlPðuÞ 6 lQ ðvÞ: ð51Þ
From (50) and given subsets from (51), we have
RðA2ðyÞÞ ¼Z
V
[u2U1
laPðuÞ ^
1� lPðuÞ þ lQ ðvÞ2
� � �v
" #_Z
V
[u2U2
laPðuÞ ^ 1
�=v
" #: ð52Þ
Let us introduce the following function
f ðu; vÞ ¼1� lPðuÞ þ lQ ðvÞ
2: ð53Þ
Then the following relationship is satisfied:
8u 2 U1 laPðuÞ ^ f ðu; vÞ ¼
laPðuÞ; la
PðuÞ 6 f ðu; vÞ;f ðu; vÞ; la
PðuÞ > f ðu;vÞ;
����� ð54Þ
8u 2 U2 laPðuÞ ^ 1 ¼ la
PðuÞ;�� ð55Þ
From (54) and (55) we have
ð52Þ ¼Z
V
[u2U2
laPðuÞ=v
" #¼Z
Vla
Q ðvÞ=v ¼ Qa:
For Criteria IV-2 [26] let R(A1(x)) = :P = 1 � P then
RðA2ðyÞÞ ¼ :P �RðA1ðxÞ;A2ðyÞÞ ¼Z
Uð1�lPðuÞÞ=u �
ZU�V
lpðuÞ!L4lQ ðvÞ=ðu;vÞ ¼
ZV
[u2U
½ð1�lPðuÞÞ^ ðlPðuÞ!L4lQ ðvÞÞ�=v :
ð56Þ
From (56) and given subsets (51) we have
RðA2ðyÞÞ ¼Z
V
[u2U1
ð1� lPðuÞÞ ^1� lPðuÞ þ lQ ðvÞ
2
� � �v
" #_Z
V
[u2U2
ð1� lPðuÞÞ ^ 1 �
=v" #
¼Z
V
[u2U1
½ð1� lPðuÞÞ ^ f ðu;vÞÞ�=v" #
_Z
V
[u2U2
½ð1� lPðuÞÞ ^ 1�=v" #
: ð57Þ
1054 R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059
Since ð1� lPðuÞÞ ^ f ðu;vÞ �
6 ð1� lPðuÞÞ ^ 1� �
, therefore
ð5:17Þ ¼Z
V
[u2U2
½ð1� lPðuÞÞ ^ 1�=v" #
¼Z
V1=v
¼ unknown: �
Theorem 3. If fuzzy sets P # U and Q # V are defined by (43) and (44), respectively, and R(A1(x),A2(y)) is defined as
RðA1ðxÞ;A2ðyÞÞ ¼ ðP � V!L4
U � QÞ \ ð:P � V!L4
U � :QÞ ¼Z
U�VðlPðuÞ!L4
lQ ðvÞÞ ^ ðð1� lPðuÞÞ!L4ð1� lQ ðvÞÞÞ=ðu;vÞ:
ð58Þ
where
ðlPðuÞ!L4lQ ðvÞÞ ^ ðð1� lPðuÞÞ!L4
ð1� lQ ðvÞÞÞ ¼
1�lP ðuÞþlQ ðvÞ2 ; lPðuÞ > lQ ðvÞ;
1; lPðuÞ ¼ lQ ðvÞ;1�lQ ðvÞþlP ðuÞ
2 ; lPðuÞ < lQ ðvÞ:
8>><>>:
Then Criteria I, II, III and IV-2 [26] are satisfied.
Proof
9U1;U2;U3 � UjU1 [ U2 [ U3 ¼ U; jU1 \ U2 \ U3 ¼ ; ) 8u 2 U1jlPðuÞ > lQ ðvÞ; 8u 2 U2jlPðuÞ ¼ lQ ðvÞ;8u 2 U3jlPðuÞ < lQ ðvÞ: ð59Þ
Let us introduce the following functions
h1ðu;vÞ ¼1� lPðuÞ þ lQ ðvÞ
2; h2ðu;vÞ ¼
1� lQ ðvÞ þ lPðuÞ2
: ð60Þ
Therefore from (58)–(60) for Criteria I–III let R(A1(x)) = Pa(a > 0) then
RðA2ðyÞÞ ¼ Pa � RðA1ðxÞ;A2ðyÞÞ ¼Z
Ula
PðuÞ=u �Z
U�VðlPðuÞ!L4
lQ ðvÞÞ ^ ðð1� lPðuÞÞ!L4ð1� lQ ðvÞÞÞ=ðu; vÞ
¼Z
V
[u2U
½laPðuÞ ^ ðlPðuÞ!L4
lQ ðvÞÞ ^ ðð1� lPðuÞÞ!L4ð1� lQ ðvÞÞÞ�=v: ð61Þ
From (60), (61) and given subsets by (59) we have
RðA2ðyÞÞ ¼Z
V
[u2U1
½laPðuÞ ^ h1ðu;vÞ�=v
" #_Z
V
[u2U2
laPðuÞ ^ 1
�=v
" #_Z
V
[u2U3
½laPðuÞ ^ h2ðu;vÞ�=v
" #: ð62Þ
Then the following is satisfied
8u 2 U1 laPðuÞ ^ h1ðu;vÞ ¼
laPðuÞ; la
PðuÞ 6 h1ðu; vÞ;h1ðu;vÞ; la
PðuÞ > h1ðu;vÞ;
����� ð63Þ
8u 2 U2 laPðuÞ ^ 1 ¼ la
PðuÞ;�� ð64Þ
8u 2 U3 laPðuÞ ^ h2ðu;vÞ ¼
laPðuÞ; la
PðuÞ 6 h2ðu;vÞ;h2ðu;vÞ; la
PðuÞ > h2ðu; vÞ:
����� ð65Þ
From (63)–(65) we have
ð61Þ ¼Z
V
[u2U2
laPðuÞ=v
" #¼Z
Vla
Q ðvÞ=v ¼ Qa: �
For Criteria IV-2 let R(A1(x)) = :P = 1 � P then
RðA2ðyÞÞ ¼ :P � RðA1ðxÞ;A2ðyÞÞ ¼Z
Uð1� lPðuÞÞ=u �
ZU�VðlPðuÞ!L4
lQ ðvÞÞ ^ ðð1� lPðuÞÞ!L4ð1� lQ ðvÞÞÞ=ðu;vÞ
¼Z
V
[u2U
½ðlPðuÞ!L4lQ ðvÞÞ ^ ðð1� lPðuÞÞ!L4
ð1� lQ ðvÞÞÞ�=v: ð66Þ
R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059 1055
From (62), (66) and given subsets from (59) we have
RðA2ðyÞÞ ¼Z
V
[u2U1
½ð1� lPðuÞÞ ^ h1ðu; vÞÞ�=v" #
_Z
V
[u2U2
½ð1� lPðuÞÞ ^ 1�=v" #
_Z
V
[u2U3
½ð1� lPðuÞÞ ^ h2ðu;vÞÞ�=v" #
ð67Þ
As a conclusion we have
½ð1� lPðuÞÞ ^ h1ðu;vÞ� 6 ½ð1� lPðuÞÞ ^ 1�� and ½ð1� lPðuÞÞ ^ h2ðu;vÞ� 6 ½ð1� lPðuÞÞ ^ 1��;
therefore
ð67Þ ¼Z
V
[u2U2
½ð1� lPðuÞÞ ^ 1�=v" #
¼Z
Vð1� lQ ðuÞÞ=v
¼ :Q :
Theorems 2 and 3 show that fuzzy conditional inference rules, defined in (58) could adhere with human intuition to thehigher extent as the one defined by (49). The major difference between mentioned methods of inference might be explainedby the difference between Criteria IV-1 and IV-2. In particular, a satisfaction the Criterion IV-1 means that in case of logicalnegation of an original antecedent we achieve an ambiguous result of an inference, whereas for the case of the CriterionIV-2 there is a certainty in a logical inference.
7. Stability and continuity of fuzzy conditional inference
In this section, we revisit the fuzzy conditional inference rule (48). It will be shown that when the membership function ofthe observation P is continuous, then the conclusion Q depends continuously on the observation; and when the membershipfunction of the relation R is continuous then the observation Q has a continuous membership function. We start with somedefinitions. A fuzzy set A with membership function lA : R! ½0;1� ¼ I is called a fuzzy number if A is normal, continuous,and convex. The fuzzy numbers represent the continuous possibility distributions of fuzzy terms of the following type
A ¼Z
R
lAðxÞ=x:
Let A be a fuzzy number, then for any h P 0 we define xA(h), the modulus of continuity of A by
xAðhÞ ¼ maxjx1�x2 j6h
jlAðx1Þ � lAðx2Þj: ð68Þ
An a-level set of a fuzzy interval A is a non-fuzzy set denoted by [A]a and is defined by ½A�a ¼ ft 2 RjlAðtÞP ag for a 2 (0,1]and [A]a = cl(supplA) for a = 0. Here we use a metric of the following type
DðA;BÞ ¼ supa2½0;1�
dð½A�a; ½B�aÞ; ð69Þ
where d denotes the classical Hausdorff metric expressed in the family of compact subsets of R2, i.e. d([A]a, [B]a) =max{ja1(a) � b1(a)j, ja2(a) � b2(a)j}, whereas[A]a = [a1(a),a2(a)], [B]a = [b1(a),b2(a)j]. When the fuzzy sets A and B have finitesupport {x1, . . . ,xn}, then their Hamming distance is defined as
HðA;BÞ ¼Xn
i¼1
jlAðxiÞ � lBðxiÞj:
In the sequel we will use the following lemma.
Lemma 1 [15]. Let d P 0 be a real number and let A, B be fuzzy intervals. If D(A,B) 6 d, Then
supt2RjlAðtÞ � lBðtÞj 6 maxfxAðdÞ;xBðdÞg:
Consider the fuzzy conditional inference rule with different observations P and P0:
Ant 1: If x is P then y is Q
Ant 1: If x is P then y is Q Ant 2: x is P Ant 2: x is P0Cons: y is Q
Cons: y is Q0According to the fuzzy conditional inference rule, the membership functions of the conclusions are computed as
lQ ðvÞ ¼[u2R½lPðuÞ ^ ðlPðuÞ ! lQ ðvÞÞ�; lQ 0 ðvÞ ¼
[u2R½lP0 ðuÞ ^ ðlPðuÞ ! lQ ðvÞÞ�
1056 R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059
or
lQ ðvÞ ¼ supu2R½lPðuÞ ^ ðlPðuÞ ! lQ ðvÞÞ�; lQ 0 ðvÞ ¼ supu2R½lP0 ðuÞ ^ ðlPðuÞ ! lQ ðvÞÞ�: ð70Þ
The following theorem shows the fact that when the observations are closed to each other in the metric D(�) of (69) type, then therecan be only a small deviation in the membership functions of the conclusions.
Theorem 4 (Stability theorem). Let d P 0 and let P, P0 be fuzzy intervals and an implication operation in the fuzzy conditionalinference rule (70) is of type (44). If D(P,P0) 6 d, then
supv2RjlQ ðvÞ � lQ 0 ðvÞj 6maxfxPðdÞ;xP0 ðdÞg:
Proof. Given an implication operation in the fuzzy conditional inference rule (70) is of type (44), for the observation P wehave
Q ¼ P � RðA1ðxÞ;A2ðyÞÞ ¼Z
UlPðuÞ=u �
ZU�V
lpðuÞ!L4lQ ðvÞ=ðu;vÞ ¼
ZV
[u2U
½lPðuÞ ^ ðlPðuÞ!L4lQ ðvÞÞ�=v : ð71Þ
It was shown above that for U;V � R; 9U1;U2 � UjU1 [ U2 ¼ U; jU1 \ U2 ¼ ; ),
8u 2 U1jlPðuÞ > lQ ðvÞ;8u 2 U2jlPðuÞ 6 lQ ðvÞ;
8u 2 U1jlPðuÞ ^ f ðu;vÞ ¼lPðuÞ;lPðuÞ 6 f ðu;vÞ;f ðu;vÞ;lPðuÞ > f ðu; vÞ;
�
8u 2 U2jlPðuÞ ^ 1 ¼ lPðuÞ;
where f(u,v) is from (53). Applying the observation P0 to (71) we obtain the following
Q 0 ¼ P0 � RðA1ðxÞ;A2ðyÞÞ ¼Z
UlP0 ðuÞ=u �
ZU�V
lpðuÞ!L4lQ ðvÞ=ðu;vÞ ¼
ZV
[u2U
½lP0 ðuÞ ^ ðlPðuÞ!L4lQ ðvÞÞ�=v: ð72Þ
We also have
8u 2 U1 lP0 ðuÞ ^ f ðu; vÞ ¼lP0 ðuÞ;lP0 ðuÞ 6 f ðu;vÞ;f ðu; vÞ;lP0 ðuÞ > f ðu; vÞ;
�����8u 2 U2jlP0 ðuÞ ^ 1 ¼ lP0 ðuÞ:
From (71) and (72), we see that the difference of the values of conclusions for both P and P0 observations for arbitrary fixedv 2 RjlQ ðvÞ � lQ 0 ðvÞj. is defined as follows
8u 2 U1 jlQ ðvÞ � l0Q ðvÞj ¼jlPðuÞ � l0PðuÞj;0;
�����8u 2 U2jjlQ ðvÞ � l0Q ðvÞj ¼ jlPðuÞ � l0PðuÞj;
and therefore from Lemma 1 we have
supv2RjlQ ðvÞ � lQ 0 ðvÞj ¼ sup
u2RjlPðuÞ � lP0 ðuÞj 6 maxfxPðdÞ;xP0 ðdÞg �
Theorem 5 (Continuity theorem). Let binary relationship Rðu; vÞ ¼ lpðuÞ!L4lQ ðvÞ be continuous. Then Q is continuous and
xQ(d) 6xR(d) for each d P 0.
Proof. Let d P 0 be a real number and let v1;v2 2 R such that jv1 � v2j 6 d. From (68) we have xQ ðdÞ ¼maxjv1�v2 j6djlQ ðv1Þ � lQ ðv2Þj. Then
jlQ ðv1Þ � lQ ðv2Þj ¼ j supu2R½lPðuÞ ^ ðlPðuÞ!L4
lQ ðv1ÞÞ� � supu2R½lPðuÞ ^ ðlPðuÞ!L4
lQ ðv2ÞÞ�j
6 supu2R½lPðuÞ ^ jðlPðuÞ!L4
lQ ðv1ÞÞ � ðlPðuÞ!L4lQ ðv2ÞÞj� 6 sup
u2R½lPðuÞ ^xRðjv1 � v2jÞ�
6 supu2R½lPðuÞ ^xRðdÞ� ¼ xRðdÞ: �
8. Concluding remarks
R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059 1057
In this study, we proposed the fuzzy logic as an algebraic system closed under all its operations in which truth values of animplication operator need to be based on truth values of both the antecedent and the consequent. This implication operatorhas to be considered to be one of the fuzziest implication from a list of implication operators known so far. It was shown thatthe fuzzy logic presented here forms a basis for fuzzy conditional inference rules, which satisfy the set of known importantcriteria and seem to be suitable for capturing human intuition. The important features of stability and continuity of the pro-posed fuzzy conditional inference rules were also investigated.
Appendix A
Criterion I
Ant 1: If x is P then y is QAnt 2: x is P————————-Cons: y is Q.Criterion II-1
Ant 1: If x is P then y is QAnt 2: x is very P—————————Cons: y is very Q.Criterion II-2
Ant 1: If x is P then y is QAnt 2: x is very P————————–Cons: y is Q.Criterion III
Ant 1: If x is P then y is QAnt 2: x is more or less P—————————–Cons: y is more or less Q.Criterion IV-1
Ant 1: If x is P then y is QAnt 2: x is not P——————————Cons: y is unknownCriterion IV-2
Ant 1: If x is P then y is QAnt 2: x is not P————————-Cons: y is not Q.References
[1] R.A. Aliev, A. Tserkovny, The knowledge representation in intelligent robots based on fuzzy sets, Sov. Math. Doklady 37 (1988) 541–544.[2] R.A. Aliev, G.A. Mamedova, A.E. Tserkovny, Fuzzy Control Systems, Energoatomizdat, Moscow, 1991.[3] R.A. Aliev, B. Fazlollahi, R.R. Aliev, Soft Computing and its Application in Business and Economics, Physica-Verlag, A Spriger-Verlag Company, 2004.[4] I. Aguilo, J. Suner, J. Torrens, A characterization of residual implications derived from left-continuous uninorms, Information Sciences 180 (2010) 3992–
4005.[5] I. Azadeh, I.M. Fam, M. Khoshnoud, M. Nikafrouz, Design and implementation of a fuzzy expert system for performance assessment of an integrated
health, safety, environment (HSE) and ergonomics system: the case of a gas refinery, Information Sciences 178 (22) (2008) 4280–4300. 15.[6] W. Bandler, L.J. Kohout, Fuzzy relational products as a tool for analysis of complex artificial and natural systems, in: P.P. Wang, S.K. Chang (Eds.), Fuzzy
Sets; Theory and Applications to Policy Analysis and Information Systems, Plenum Press, New York, 1980, pp. 311–367.[7] W. Bandler, L.J. Kohout, The identification of hierarchies in symptoms and patients through computation of fuzzy relational products, in: R.D. Parslow
(Ed.), BCS’81: Information Technology for the Eighties, Heyden & Sons, 1980, pp. 191–194.[8] W. Bandler, L.J. Kohout, Semantics of fuzzy implication operators and relational products, International Journal of Man–Machine Studies 12 (1980) 89–
116.[9] W. Bandler, L.J. Kohout, Fuzzy power sets and fuzzy implication operators, Fuzzy Sets and Systems 4 (1980) 13–30.
1058 R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059
[10] W. Bandler, L.J. Kohout, The four modes of inference in fuzzy expert systems, in: R. Trappl (Ed.), Cybernetics and Systems Research, vol. 2, NorthHolland, Amsterdam, 1984, pp. 97–104.
[11] W. Bandler, L.J. Kohout, Probabilistic vs. fuzzy production rules in expert systems, International Journal of Man–Machine Studies 22 (1985) 347–353.[12] B.C. Bedregal, G.P. Dimuro, R.H.N. Santiago, R.H.S. Reiser, On interval fuzzy S-implications, Information Sciences 180 (2010) 1373–1389.[13] R. Belohlavek, E. Sigmund, J. Zacpal, Evaluation of IPAQ questionnaires supported by formal concept analysis, Information Sciences, in press,
doi:10.1016/j.ins.2010.04.011.[14] I. Bloch, Lattices of fuzzy sets and bipolar fuzzy sets, and mathematical morphology, Information Sciences, in press, doi:10.1016/j.ins.2010.03.019.[15] F. Bobillo, U. Straccia, Reasoning with the finitely many-valued Lukasiewicz fuzzy description logic, Information Sciences 181 (2011) 758–778.[16] H. Bustince, E. Barrenechea, J. Fernandez, M. Pagola, J. Montero, C. Guerra, Contrast of a fuzzy relation, Information Sciences 180 (2010) 1326–
1344.[17] T. Chen, Optimistic and pessimistic decision making with dissonance reduction using interval-valued fuzzy sets, Information Sciences 181 (2011) 479–
502.[18] I. Chajda, R. Halas, I.G. Rosenberg, On the role of logical connectives for primality and functional completeness of algebras of logics, Information
Sciences 180 (2010) 1345–1353.[19] B. Davvaz, J. Zhan, K.P. Shum, Generalized fuzzy Hv-submodules endowed with interval valued membership functions, Information Sciences 178 (1)
(2008) 3147–3159 (Nature Inspired Problem-Solving).[20] Deng-Feng Li, Guo-Hong Chen, Zhi-Gang Huang, Linear programming method for multiattribute group decision making using IF sets, Information
Sciences 180 (2010) 1591–1609.[21] J. Dian, A meaning based information theory – inform logical space: basic concepts and convergence of information sequences, Information Sciences
180 (15) (2010) 984–994 (Special Issue on Modelling Uncertainty).[22] M. Eduardo Valle, Permutation-based finite implicative fuzzy associative memories, Information Sciences 180 (2010) 4136–4152.[23] Z-P. Fan, B. Feng, A multiple attributes decision making method using individual and collaborative attribute data in a fuzzy environment, Information
Sciences 179 (2009) 3603–3618.[24] M. Fedrizzi, R. Fuller, Stability in possibilistic linear programming problems with continuous fuzzy number parameters, Fuzzy Sets and Systems 47
(1992) 187–191.[25] O.I. Franksen, Group representation of finite polyvalent logic, in: A. Niemi (Ed.), Proceedings of the 7th Triennial International Federation of Automatic
Control World Congress, Pergamon, IFAC, Helsinki, 1978.[26] S. Fukami, M. Mizumoto, K. Tanaka, Some considerations of fuzzy conditional inference, Fuzzy Sets and Systems 4 (1980) 243–273.[27] R. Fuller, H.-J. Zimmermann, On Zadeh’s compositional rule of inference, in: R. Lowen, M. Roubens (Eds.), Fuzzy Logic: State of the Art, Theory and
Decision Library, Series D, Kluwer Academic Publisher, Dordrecht, 1993, pp. 193–200.[28] Francisco J. Valverde-Albacete, C. Pelaez-Moreno, Extending conceptualization modes for generalized formal concept analysis, Information Sciences, in
press, doi:10.1016/j.ins.2010.04.014.[29] M. Gerhke, C.L. Walker, E.A. Walker, Normal forms and truth tables for fuzzy logics, Fuzzy Sets and Systems 138 (2003) 25–51.[30] P. Grzegorzewski, On possible and necessary inclusion of intuitionistic fuzzy sets, Information Sciences 181 (2011) 342–350.[31] M. Grabisch, J. Marichal, R. Mesiar, E. Pap, Aggregation functions: construction methods, conjunctive, disjunctive and mixed classes, Information
Sciences 181 (2011) 23–43.[32] Q. Hu, D. Yu, M. Guo, Fuzzy preference based rough sets, Information Sciences 180 (15) (2010) 2003–2022 (Special Issue on Intelligent Distributed
Information Systems).[33] B. Jayaram, R. Mesiar, I-Fuzzy equivalence relations and I-fuzzy partitions, Information Sciences 179 (2009) 1278–1297.[34] J. Jantzen, Array approach to fuzzy logic, Fuzzy Sets and Systems 70 (1995) 359–370.[35] S. Jenei, Continuity in Zadeh’s compositional rule of inference, Fuzzy Sets and Systems 104 (1999) 333–339.[36] A. Kandel, M. Last, Special issue on advances in Fuzzy logic, Information Sciences 177 (2007) 329–331.[37] M. Kallala, L.J. Kohout, The use of fuzzy implication operators in clinical evaluation of neurological movement disorders, in: International Symposium
on Fuzzy Information Processing in Artificial Intelligence and Operational Research, Christchurch College, Cambridge University, 1984.[38] M. Kallala, L.J. Kohout, A 2-stage method for automatic handwriting classification by means of norms and fuzzy relational inference, In: Proceedings of
the NAFIPS ’86 (NAFIPS Congress, New Orleans, 1986).[39] A. Kehagias, Some remarks on the lattice of fuzzy intervals, Information Sciences, in press, doi:10.1016/j.ins.2010.05.007.[40] J.B. Kiszka, M.E. Kochanska, D.S. Sliwinska, The infuence of some fuzzy implication operators on the accuracy of a fuzzy model, Fuzzy Sets and Systems
15 (Part 1) (1985) 111–128 (Part 2) 223–240.[41] A. Kolesarova, R. Mesiar, Lipschitzian De Morgan triplets of fuzzy connectives, Information Sciences 180 (2010) 3488–3496.[42] L.J. Kohout, W. Bandler, Relational-product architecture for information processing, Expert Systems Information Science 37 (1985) 25–37.[43] L.J. Kohout (Ed.), Perspectives on Intelligent Systems: A Framework for Analysis and Design, Abacus Press, Cambridge, Mass, USA and Tunbridge Wells,
Kent, UK, 1986.[44] J. Lai, Y. Xu, Linguistic truth-valued lattice-valued propositional logic system lP(X) based on linguistic truth-valued lattice implication algebra,
Information Sciences 180 (2010) 1990–2002 (Special Issue on Intelligent Distributed Information Systems).[45] C. Lee, Fuzzy logic in control systems: fuzzy logic controller, IEEE Transactions on Systems, Man & Cybernetics 20 (2) (1990) 404–435.[46] P. Levy, From social computing to reflexive collective intelligence: the IEML research program, Information Sciences 180 (2) (2010) 71–94 (Special
Issue on Collective Intelligence).[47] Z. Long, X. Liang, L. Yang, Some approximation properties of adaptive fuzzy systems with variable universe of discourse, Information Sciences 180
(2010) 2991–3005.[48] E.H. Mamdani, Application of fuzzy logic to approximate reasoning using linguistic syntheses, IEEE Transactions on Computers C-26 (12) (1977) 1182–
1191.[49] H. Ma, An analysis of the equilibrium of migration models for biogeography-based optimization, Information Sciences 180 (2010) 3444–3464.[50] M. Mas, M. Monserrat, J. Torrens, E. Trillas, A Survey on Fuzzy Implication Functions, IEEE Transactions on Fuzzy systems 15 (6) (2007) 1107–1121.[51] M. Mas, M. Monserrat, J. Torrens, The law of importation for discrete implications, Information Sciences 179 (2009) 4208–4218.[52] J. Medina, M. Ojeda-Aciego, Multi-adjoint t-concept lattices, Information Sciences 180 (2010) 712–725.[53] J.M. Mendel, On answering the question ’Where do I start in order to solve a new problem involving interval type-2 fuzzy sets?’, Information Sciences
179 (2009) 3418–3431.[54] M. Mizumoto, S. Fukami, K. Tanaka, Some methods of fuzzy reasoning, in: R. Gupta, R. Yager (Eds.), Advances in Fuzzy Set Theory Applications, North-
Holland, New York, 1979.[55] M. Mizumoto, H.-J. Zimmermann, Comparison of fuzzy reasoning methods, Fuzzy Sets and Systems 8 (1982) 253–283.[56] A.A. Molai, E. Khorram, An algorithm for solving fuzzy relation equations with max-T composition operator, Information Sciences 178 (2008) 1293–
1308.[57] S. Munoz-Hernandez, V. Pablos-Ceruelo, H. Strass, RFuzzy: An Expressive simple fuzzy compiler, in bio-inspired systems: computational and ambient
intelligence, Lecture Notes in Computer Science 5517 (2009) 270–277.[58] M. Nachtegael, P. Sussner, T. Melange, E.E. Kerre, On the role of complete lattices in mathematical morphology: from tool to uncertainty model,
Information Sciences, in press, doi:10.1016/j.ins.2010.03.009.[59] C. Noguera, F. Esteva, L. Godo, Generalized continuous and left-continuous t-norms arising from algebraic semantics for fuzzy logics, Information
Sciences 180 (2010) 1354–1372.
R. Aliev, A. Tserkovny / Information Sciences 181 (2011) 1045–1059 1059
[60] K-W. Oh, W. Bandler, Properties of fuzzy implication operators, Department of Computer Science, Florida State University, Tallahassee, FL 32304, USA,1988, pp. 24-33.
[61] Y. Ouyang, Z. Wang, H. Zhang, On fuzzy rough sets based on tolerance relations, Information Sciences 180 (2010) 532–542.[62] D. Pei, Unified full implication algorithms of fuzzy reasoning, Information Sciences 178 (2008) 520–530.[63] B.-S. Shieh, Infinite fuzzy relation equations with continuous t-norms, Information Sciences 178 (2008) 1961–1967.[64] N. Rescher, Many-Valued Logic, McGraw-Hill, New York, 1969.[65] L. Rutkowski, K. Cpalka, Flexible neuro-fuzzy systems, IEEE Transactions on Neural Networks 14 (3) (2003) 554–573.[66] B. Schweizer, Distribution functions: numbers of the future, in: A. Di Nola, A.G.S. Ventre (Eds.), La Matematica dei Sistemi Fuzzy Inst: di Matematica
del la Facoltà di Architettura, Universitàdegli Studi di Napoli, Italy, 1985, pp. 137–149.[67] M. Serruier, D. Dubois, H. Prade, T. Sudkamp, Learning fuzzy rules with their implication operator, Data & Knowledge Engineering 60 (2007) 71–89.[68] R.R. Stoll, Set Theory and Logic, Dover ed., Dover Publications, New York, 1979.[69] F. Wenstøp, Quantitative analysis with linguistic values, Fuzzy Sets and Systems 4 (2) (1980) 99–115.[70] R. Willmott, Two fuzzier implication operators in the theory of fuzzy power sets, In: Fuzzy Research Project, University of Essex, Colchester, UK, Dept.
of Mathematics, FRP-2 Report, 1978.[71] R. Willmott, Two fuzzier implication operators in the theory of fuzzy power sets, Fuzzy Sets and Systems 4 (1980) 31–36.[72] A. Xie, F. Qin, Solutions to the functional equation I(x,y) = I(x, I(x,y)) for a continuous D-operation, Information Sciences 180 (2010) 2487–2497.[73] Y. Shi, B. Van Gasse, D. Ruan, E. Kerre, On a new class of implications in fuzzy logic, IPMU 1 (2010) 525–534.[74] R.R. Yager, On global requirements for implication operators in fuzzy modus ponens, Fuzzy Sets and Systems 106 (1999) 3–10.[75] R.R. Yager, A framework for reasoning with soft information, Information Sciences 180 (8) (2010) 1390–1406.[76] L.A. Zadeh, Fuzzy sets, Information and Control 8 (1965) 338–353.[77] L.A. Zadeh, Outline of a new approach to the analysis of complex systems and decision processes, IEEE Transactions on Systems, Man & Cybernetics 1
(1973) 28–44.[78] L.A. Zadeh, The concept of a linguistic variable and its application to approximate reasoning, Information Sciences 8 (1975) 43–80.[79] L.A. Zadeh, Fuzzy logic, IEEE Computer 21 (4) (1988) 83–93.[80] L.A. Zadeh, Toward a generalized theory of uncertainty (GTU) – an outline, Information Sciences 172 (2005) 1–40.[81] L.A. Zadeh, Is there a need for fuzzy logic?, Information Sciences 178 (2008) 2751–2779[82] J. Zhang, X. Yang, Some properties of fuzzy reasoning in propositional fuzzy logic systems, Information Sciences 180 (2010) 4661–4671.[83] X. Zhang, Y. Yao, H. Yu, Rough implication operator based on strong topological rough algebras, Information Sciences 180 (2010) 3764–3780.[84] S. Zhao, E.C.C. Tsang, On fuzzy approximation operators in attribute reduction with fuzzy rough sets, Information Sciences 178 (15) (2008) 3163–3176
(Including Special Issue: Recent Advances in Granular Computing, Fifth International Conference on Machine Learning and Cybernetics).[85] H.-J. Zimmermann, Fuzzy Set Theory and Its Applications, second ed., Kluwer, Boston, 1993.