PROCEEDINGS OF THE IEEE, VOL.

12, 66, NO. DECEMBER

1619 1978

Fuzzy Sets, Fuzzy Algebra, and Fuzzy Statistics

aspects of a situation easily overlooked by more “objective”

methods.
It should be remembered that one of the fundamental tenets
of modern science and engineering is to characterize scientific
phenomenainquantitativeterms byutilizinga reservoir of
conceptsand techniques which can be drawnupon to con-
struct mathematical models of various types of systems.
Such models for evaluation of fuzzy events and basic proba-
bility logic have been discussed by Lofti A. Zadeh [ 221, [ 251,
and B. R. Gaines [9]. Theirconclusions indicatethatthe
formal relationship of probability theory to fuzzy logic is not
I. INTRODUCTION spurious.

IL OGIC, PROBABILITY theory, and information theory It is therefore our aim to develop analytical methods which
(the latter ina broader sense of the word) have one func- wiU enable us to investigate problems via qualitative and quan-
tion in common: namely, t o provide rational effective titative fuzzy methods. These methods arerelated tothe
tools in unraveling structural relationships existing in a collec- theory of subjective probability, as developed byBruno de
tion of propositions. The aspects of the structure with which Finette [ 61 and Leonard J. Savage [ 171.
these three disciplines deal, and the methods of attack which Even thoughthe problems onhand are specific ones, we
they apply are different, yet they are related in such a way shall developcertain general principles of “fuzzystatistics”
that probability and information theories presuppose and are withoutturning specific attention to particularproblems.
natural extensionsof logic. It should be pointed out, however, that among other fields of
For this reason, in order to present a rigorous exposition of research in which the concepts andresults of the present study
fuzzy algebra and statistics, it is necessary first to scrutinize may be expected to findapplications are the following.
critically and define carefully all the basic notions involved in Patternrecognition. Imprecise andfuzzy logic applied ana-
fuzzy logic and in probability theory. lytically to classification of patterns andfeature vector
In the following sections we shall consider the question of selection.
subjective evaluation of probabilities from evidence which is Decision making. Collective decision making models. Multi-
not simply a record of conditional relative frequencies, but is criteria decision problems. Decision making in a fuzzy environ-
instead a sample of facts that pertain to causes of the event in ment involving a dynamical structure in which subjective eval-
question. uation is time function.
A good example of subjective evaluation from both “hard” Quantum mechanics. Evaluation of the outcome of measure-
and “soft” data is the weather forecast which is given as a per- ments of canonically conjugate variables adopting the pointof
centage chance of rain. Many other examples of estimates that view that some degree of fuzziness is an intrinsic feature of
are made by knowledgeable people on subjective grounds can any measurement on a microsystem.
be found. Planning often requires such judgments as the likeli- Meteorologicalevaluationsandweatherprediction. Subjec-
hood that the time for completion of a part of a project will tive evaluation in meteorology and in the field of meteorologi-
exceeda given limit, orthe probability that more than a cal forecasting. The evaluation of meteorologists (identify
certain quantity of a particular item will be sold. The use of “good” assessors) and a technique to help meteorologists to
quantitative decision procedures in business, government, and become “better” assessors.
industry has been growing as part of efforts to provide a more Section I1 provides a unifying point of view to the notions
rational basis for allocating resources. But often the measure- of subjectivity and inexactness, based on the theory of fuzzy
ments that would be required to obtain objective estimates of sets as introduced by Lotfi A. Zadeh [ 21 ] . Section I11 includes
probabilities for inputs to those procedures are either impossi- the description of fuzzystatisticsand analytic analysis of
ble to make or are too costly in time or money. Subjective fuzzy logic.
estimatesare then necessary. Indeed, they may incorporate
11. FUZZY SETS ANDFUZZY ALGEBRA
A . Fuzzy Sets
Manuscript received December 27, 1977; revised June 6, 1978. Parts
of this paper are based on the Doctoral dissertation of A. Kandel sub- The purpose of this chapter is to introduce a unifying point
mitted t o the Electrical Engineering andComputerScience Depart- of view to the notions of subjectivity and inexactness, based
ment, University of NewMexico. The submission of this paper was
encouraged after review of an advanced proposal.
on the theoryof fuzzy sets as introduced by Zadeh [ 2 1 1 . The
This paper is dedicated t o F‘rofessor Low1 k Zadeh. reason supporting the representation of inexact concepts by
A. Kandel is with the Department o f Mathematics, Computer Science fuzzy sets has beengiven by Goguen [ 101 , [ 11 ] .
Division, the Flonia State University, Tallahassee, FL 32306.
W. J. Byatt is with the Electrical Engineering and Computer Science Perhaps Goguen’s most convincing argument is a representa-
Department, University of New Mexico, Albuquerque,NM 87131. tion theorem whichsays thatany systemsatisfyingcertain

0018-9219/78/1200-161~9$00.75 0 1978 IEEE

1620 PROCEEDINGS OF THE IEEE, VOL. 66, NO. 12, DECEMBER 1978

axioms is equivalent to asystem of fuzzy sets. Since the

axioms are intuitively plausible for the system of all inexact
concepts,thetheorem allows us to conclude thatinexact
concepts can be represented by fuzzy sets. The representation
theorem is a precise mathematical result in the theoryof cate-
gories, so that a very precise sense is given to the concepts
“system,”“equivalent,” and “represented.” Unfortunately, a
good amount of category theory is required to state Goguen’s
result exactly; details can be found in his dissertation.
In abstract (or conventional, or nonfuzzy) set theory, the
setsconsideredaredefined as collections of objects having Fig. 1. Membership function (Example I ) .
some very general property P; nothing special is assumed or
considered aboutthenature of the individualobjects. For upon there being some notion of nearness or contiguity in the
example, we define a set A as the set of streets. Symbolically universe.
We now consider several examples of fuzzy sets.
A = { x Ix is a street}. Example 1: In this example we consider the class of all real
numbers which are much greater than 1. We candefine this
Now what about,the-‘‘class of long streets?” First of all, is it
set as A = { x l x is a real number and x >> 1). But it is not a
a set in the ordinary sense? Before we answer it we may F i t
welldefined set forthe reasons mentioned before. This set
ask: “HOW ‘long’ is a long street? Is a one-mile street a long
may be defined subjectively by a membership function such as
street? If so, then is there any difference between a half-mile
street and a one-mile long street, etc.?” Frankly, we do not PA ( x )= 0 for x Q 1
know how to answer adequately these questionsfromthe
information “long street” because the “class of long streets”
does not constitute a set in the usual sense. In fact, most of
the classes of objects encountered in the real physical world
which is depicted in Fig. 1.
are of this fuzzy, not sharply defined type. They do not have The assignment of the membership function of a fuzzy set is
precisely definedcriteria of membership. In such classes, an
subjective in natureand, in general, reflects the context in
object need not necessarily either belong or not belong to a
which the problem’ is viewed. Although the assignment of the
class; there may be intermediate grades of membership. This
membership function of a fuzzy set A is “subjective”, it cannot
is the conceptof fuzzy set,which is a “class” with a continuum
be assigned arbitrarily. For example, it would be totally wrong
o f grades o f membership. to assign the membership function of Example 1 as
Fuzzy set theory, introduced by Zadeh [ 211, is a generaliza-
tion of abstract set theory. In other words, the former always
includes thelatter as a special case; definitions,theorems,
proofs, etc. of fuzzy set theory always hold for nonfuzzy sets.
Because of this generalization,fuzzy set theory hasa wider lo for x > 1.
scope of applicability than abstract set theory in solving prob- A function P A such as
lems which involve, to some degree, subjective evaluation.
Intuitively, a fuzzy set is a class which admits the possibility
of partialmembershipin it. Let !2 = {x} denote aspace of
objects. Then a fuzzy set A in 52 is a set of ordered pairs
which monotonically decreases as x increases, for x > 1 or

where ~ A ( xis) termed “the grade of membership of x in A.”

We shall assume for simplicity that p ~ ( xis) a number in the
interval [0, 11, with the grades 1 and 0 representing, respec- which increases monotonically, but is approximately equal to
tively,fullmembership and nonmembershipinafuzzyset. 1 for x = 1. l , should not be considered as they describe other
We have assumed that an exact comparison is possible for the classes of objects rather than the one required by Example 1 .
truths of any two inexact statements ‘x E A’ and ‘y E A ’ , and Functions such as these will be called nonadmissible functions
that the exact relation so obtained satisfies the minimal consis- of the fuzzy set A . Thefunction ~ A ( x such ) as defined in
tency requirements of transitivity and reflexivity; the ordering Example 1 and other functions suchas
2 means “at least as true as” with Q denoting “not so true as.”
In his philosophical paper, Max Black [ 41 distinguishes three
kinds of inexactness:generality, that a conceptapply to a
variety of situations; ambiguity, that it describes more than
one distinguishable subconcept; and vagueness, that precise
boundaries are not defined. All three types of inexactness are
represented by afuzzy set: generality occurs when the uni-
verse is not just one point; ambiguityoccurs when there is
2) PA(x) =
I“
1 - [cosh ( x - 1)I-l
for x Q 1
for x >1
(2.6)

more than one local maximum of membership function; and which satisfy the condition 0 <C(A( x ) < 1 for all x E a,
and
vagueness occurswhen the function takes values other than are consistent with the specification of the set, will be called
just 0 and 1. Ambiguityand vagueness, therefore, depend the admissible functions of A .
 KANDEL AND BYATT: FUZZY SETS, ALGEBRA, AND STATISTICS 1621

n e previous discussion indicates that we should not expect elements 0, 1 satisfy

particular numerical values of one attribute suchas “shortness”
to be meaningful (except 0 and l ) , but rather that their order- x+o=x x*o=o
ing must have importance. Competent observers will agree on x+l=l x*l=x
whether oneman is shorter than another, butit is possible that
the meaning of the term “short” w l never stabilize. In the
i for every element x.
terminology of Suppes and Zinnes1251, degree of member- Leta= {x} be a space of objects as defined above and let A
ship may be measured by an ordinal scale, a term used also by and B be two fuzzy sets in a. The following terminology
Murphy [22] , to describe the measuring tool of the relation- parallels closely thepresentation byZadeh [21]. Equality
ships between the “accuracy” (validity measure) and “value” ( A = B ) is defined by
(expected-utility measure) of meteorologicalforecasts or be- A=B*~A(x)=~B(x), VX€E.
tween two (or more)forecasters.
FuzzysetA is contained in B ( A C B ) iff p ~ ( x ) Q p ~ ( x ) ,
B. Fuzzy Algebra Vx E a. A fuzzy set x is the complement of a fuzzy set A
Definition I: A fuzzy algebra is defined to be the system a.
iff p ~ ( x =) 1 - p ~ ( x ) ,Vx E The union of two fuzzy sets
A and B in 52 is defined as the membership function of A + B
z= ( Z , +, *,-> given by

where Z has at least two distinct elements, and Vx, y , z E Z , C(A+B(~)

= Max[pA(x), &(x)].
system Z satisfies the following set of axioms: The intersection of Aand B in a,denotedby A B , is
d e f i e d similarly by
1)Idempotency:
P A *B(X) = d [ p A (x), c(B(x)I.
x+x=x (2.7)
In the sequel, the term “fuzzy variable” will replace the term
2)Commutativity: “membership grade of a variable” in a set. Conventionally, we
x * y = y *x x+ y = y + x (2.8) shall drop the * symbol, i.e., x * y will be written as xy.
It is quite clear that among the infinite number of distinct
3) Associativity: assignments of grade membership to the variables, there is a
( x + y ) + z = x + ( y + z()x * y ) * z = x * ( y * z ) (2.9) finite number of binary assignments (assignments of 0 or 1 to
every variable) to a finite set.
4)Absorption: Zadeh [23] defines a fuzzy relation R as a fuzzy collection
of ordered pairs. Thus if X = {x} and Y = { y } are collections
x+(x *y)=x x *(x + y ) = x (2.10)
of objects denoted byx and y , then a fuzzy relation from X to
5) Distributivity: Y is a fuzzy subset of X x Y characterized by a membership
(characteristic) function p~ which associates witheach pair
x + ( y * z)= x *(y+z)=
(x, y ) its “grade of membership,” c ( R ( x , ~ )in
, R. We shall
(x + Y ) (x + z) x*y+x*z (2.1 1) assume for simplicity thatthe range of p~ is the interval
[0, 1 ] ,and we will refer to the numberp~ (x, y ) as the strength
6)Complement: If x E Z then there is a unique of the relationbetween x and y .
complement X of x such that The domain of a fuzzy relation R is denoted by dom R and
XEZandF=x. (2.12) is a fuzzy set defined by
7) Identities: (3!e+)(Vx) such that p d o m R ( X ) = suP[c(R(x,y)l,
XEX. (2.15)
x+e+=e++x=x YEY
(3!e,)(Vx) such that
x * e , = e , * x = x (2.13) Similarly, the range of R is denoted by ranR and is defied by

8 ) De-Morgan Laws: = = X * J m = Y + J . (2.14) k R ( Y ) = SUP [c(R(x,y)l, Y E y. (2.16)

XEX

This system forms adistributive latticewith existence of The height of R is denoted by h (R) andis defined by
unique identities under the operations + and *. It is noted
that a Boolean algebra is a complemented distributive lattice h(R)= Sup{Sup[C(R(x,Y)I). (2.17)
with existence of unique identities under the operations+ and X Y

*. However, for every element x in a Boolean algebra we have A fuzzy relation is subnormal if h(R) < 1 andnormal if
(Vx)(3E) xE = 0 and x +X = 1 which is not true in fuzzy alge- h(R) = 1.
bra. Hence, every Boolean algebra is a fuzzy algebra, but not The support of R is denoted by S(R) and is defined to be
vice versa. the exact subsetof X X Y over which ~ R ( xy, ) > 0.
Here we shall use a particular fuzzy algebra defined by the It should be noted that when X and Y are finte sets. p~
system may be represented by a relation matrix whose (x, y)th ele-
z = ([O, 11, +, *,-) ment is p~ (X,y ) .

where +, *, and - are interpreted as max, min, and complement, C. Fuzzy Hierarchical Relations
(X = 1 - x, Vx E [0, 11 ), respectively. The unique identities Thenotion of “similarity” as defined by Zadeh [23] is
e+ and e, are 0 and 1, respectively. In this fuzzy algebra the essentially a generalization of the notion of equivalence. More
1622

specifically,asimilarity relation, S, in is a fuzzyrelation

which is 1) reflexive, i.e.,
p&,y)= 1 iffx=y (2.18)
2) symmetric,i.e.,
P S b , Y ) = P$(Y, x), vx, Y E a (2.19)
and 3) transitive, i.e.,
p S k z) 2 Max{min[lr,(x,y), P J Y , ~11). (2.20)
Y
Similarly, a dissimilarity relation D can be d e f i e d as the com-
plement of S with
&(X, v) = 1 - P$(X,v); x, Y E a. (2.21)
If p&, y ) is interpreted as a distance function d ( x , y ) , then
transitivity implies that
1 - d ( x , z)> Max{min[l - d ( x , y ) , 1 - d(y,z)l}
Y
(2.22)
and since
min[ 1 - d ( x , Y ) , 1 - d ( y , 211 = 1 - Max[d(x, Y ) , d ( y , z)l
(2.23)
we can conclude that

d(x, z) G Max[d(x, V I , d ( y , z)l , V X , Y z, E

(2.24)
which implies the triangle inequality.
D e f i g a proximity relation (describing subjective similari-
ties) as a reflexive, symmetric, but not necessarily transitive,
fuzzy relation, and then-purlieus relation over a frnite n , as
~ ; ( x , Y2
) SUP {min[ps(x,xl), * * ' ,~ls(~n-l,~)I}
;En
(2.25)
Before proceeding we would like to notethatthe above
where ~ = ( X ~ , X ~ ; ~ ~ , X (52 ~ - ~ € ~ ( n - 1)-fold
is ) the relations are similar to the idea of a-level-sets of a fuzzy rela-
Cartesian product of X with itself where x, y E X) implies that tion R , denoted by R , (and is a nonfuzzy set in X X Y ) as
forallx,yEXandallnZl
d e f i e d by Zadeh [ 231.
ogp~(x,Y)(Cr;+'(x,Y)gl. (2.26) Theorem 4: Let T 1 2 T2. Thenthepartition via R T ,
refines that of R T ~ .
Consequently, limn+- p;(x,y) i i J x , y ) exists
by the Proof:
monotone convergence principle;namely, for every E > 0,
there is an integer N such that I p ; ( x , y ) - &(x, y ) l < E for xR~,y-j&(x,y)> T1 Z T 2 + x R T , Y . Q.E.D.
n > N . Since the sequence is nondecreasing and is bounded It is clear that if R T is athresholdrelation inducedby
from above and from below, we can conclude that the exis- pJx, y ) and, R k is a threshold relation inducedby &,x, y ) and
tence of the limit is due to thefollowing theorem. <
pJx, y ) pJx, y ) for all x , y € a, then R T refines R T .
Theorem I : A bounded nondecreasing sequence {ai} has a The function fi,(x, y ) = 1 - & ( x , y ) acts as a distance func-
limit; the limit is the smallest number that is not less than any tion, since
ai.
Definition 2: Let x and y be two elements of and let a,
p : ( x , y ) be the n-purlieus relation d e f i e d above. Then we
define the propinquity &(x, y ) in [0, 11 as
i i s ( x , u ) = lim P ~ X V ,I . (2.27)
n+-
as
Definition 3: Let x, y E a.
Then x and y are said to have a
threshold relation (xR T Y ) iff &(x, y ) 2 T .
BYATT:
KANDEL AND ANDALGEBRA,
FUZZY SETS, STATISTICS 1623

D. Max and Min as Natural Tools thatthe intervals of variation of the membership grade of
After the introduction of the fuzzy set concept by Zadeh, fuzzy sets p E [0, 11 coincide. However, between the concepts
Bellman and Giertz [ 21 supplied the philosophical background of fuzziness and probability there are also essential differences.
to Zadeh’s definition, showing that not only are they natural, Probability is an objective characteristic; the conclusions of
but under quite reasonable assumptions the only ones possible. probability theory can,in general, be tested b y experience.
In their excellent paper, they prove that the membership func- The membership grade is subjective, although itis natural to
tion, which is a quantitativeform of asubjective degree of assign o lower membership grade to an event that, considered
belief,imposescertain natural restrictions on the subjective from the aspect of probability, would have a lower probability
value of compound statements. Thus the subjective value of of occurrence. The fact that the assignment of a membership
‘‘X and y ” , or “x or y ” , should not decrease if our subjective function of a fuzzy set is “nonstatistical” does not mean that
evaluation of x (or y ) increases. Clearly, evaluating “x and y” we cannot use probability distribution functions in assigning
requires more, and evaluating “x or y” less, than evaluating membership functions. As a matter of fact, a careful examina-
x or y alone. Given this basic idea, it is shown that the Max tion of the variables of fuzzy sets reveals that they may be
and Min definitions as introduced are not only natural when classified intotwotypes: statisticalandnonstatistical.The
we are implementing subjectiveevaluations, but under quite variable “magnitude of x” in Example 1 is an example of the
reasonable assumptions, the only ones possible. formertype. However, if one considers the “class of tall
It should be noted that when the rigid “either-or” condition men,” the “height of a man” can be considered to be a statis-
of set theory is relaxed, the concept of negation necessarily tical variable. In this case, for instance,if a man is under 5 feet
becomes a fuzzy one. The borderline between “x” and “not tall we would not call him a tall man by “everybody’s’’ stan-
x” is no longer sharp. dard, and we should assign to him a low grade of membership
What conditions are, then, naturalto impose on a function N in the class of tall men. Similarly, if a man is over 7 feet, he
connecting the truth values of “x” and“not x?” In other certainly deserves a high grade of membership in the class of
words, what choice should be made for the statement tall men.
The operations applied in the theory of fuzzy sets (finding
p(not
= N [ px () x ) ] ? (2.28) maxima and minima) have a special form and differ from oper-
ations applied in probability.
In order for ordinary set theory to remaina special case of
fuzzyset theory, we require that N ( 0 ) = 1 and N ( l ) = 0, It is only possible to judge the effectiveness of the methods
of fuzzy set theory when a sufficient number of functioning
N[N(p)] = p , and N be continuousandstrictlymonotonic
decreasing since the subjective evaluation of “not x” should devices shall have been constructed on the basis of this theory.
decrease when our evaluation of x increases. It is well known that probability theory can be constructed
The definition as an axiomatic theory,based on three axioms, coinciding with
the corresponding axioms of measure theory. Here an attempt
p(not x ) = 1 - p ( x ) (2.29)
is made to generalize to fuzzysetscertainpositions of the
appears reasonable in practical applications and will be used in axiomatic probability theory, utilizing the concept of fuzzy
this work. event, in a different way from Zadeh’s extension [22].
It should be noted that technical measures in fuzzy sets may The motivation for the development of fuzzy statistics is its
also be described by the so-called “linguistic variables” pro- philosophical and conceptual relation to subjective probability.
posed by Zadeh [24]. For detailedbibliographies on these In the subjectivistic view, probability represents the degree of
and related topics the reader is referred to the work of Zadeh belief that a given person has in a given event on the basis of
e t al. [26], Zimmermann [ 2 7 ] , Kaufmann [ 131, and Kandel given evidence. This view (called also personalistic or judgmen-
and Davis [ 121 . tal probability) can be best described by an earlyarticleby
James Bernoulli [3] who defines probability as degree of con-
111. FUZZYSTATISTICS
fidence in a proposition of whose truth we cannot be certain.
A . Motivation His “degree of confidence” is identified with the probability
Ordinarily, imprecision and indeterminacy are considered to of an event and depends on the knowledge that the individual
be statistical randomcharacteristics, and are taken into account has at his disposal. Thus it varies from individual to individual
bythemethods of probability theory. In real situations, a and can be best described as the art of guessing ( a n conjec-
frequent source of imprecision is not only thepresence of ran- fundi) [ 5 ] , [ 6 ] ,[ 141.
dom variables, but the impossibility, in principle, of operating The difficulty in applying subjective probability stems from
with exact data as a result of the complexity of the system, the vagueness associated with judgments made via subjective
imprecision of theconstraintsand objectives. At the same analysis. Thepostulates of subjectiveprobability cannot be
time, classes of objects appear in the problems that do not applied, as is widely recognized, to many interesting and useful
have clear boundaries; the imprecision of such classes is ex- theories of modem science which are inexact. Subjective
pressed in the possibility that an elementdoes not only belong probability can be regarded asapersonal way of treating
ornot belong to acertain class butintermediate grades of objective views inthatthey areconcerned with individual
membership are also possible. judgment. These judgments,however,are not additive since
Intuitively, a similarity is felt between the conceptsof fuzzi- human behavior is often contradictory to the assumption of
ness and probability. The problems in which they are used are subjective probability, that an individual is using additive mea-
similar or coincide. There are problems in which indeterminacy sures in his criteria for evaluation, as discussed in [ l ] and [ 7 ] .
is encountered, due to random factors, inexactknowledge, or We shall remove the restrictive device known as additivity
thetheoretical impossibility or lack of necessity to obtain and formulate the basic structure of afuzzy statistic to be
exact solutions. The similarity is also underscored by the fact applied as our analyticmodel of investigation of imprecise data.
1624 PROCEEDINGS O F THE IEEE, VOL. 66, NO. 12, DECEMBER 1978

B. Compatibility and Fuzzy Expectation measure of the degree to which one statement is confirmed
In the classical approacha probabilitysystem is atriple by other statements. This is called logical probability and is
(a’, a’
S,P), where is an arbitrary set (the sample space which defined even if the statements themselvesare false. It refers
includes all possible outcomes), S is a set of events, and P is a to relationsamongthestatementsrather than to relations
real valued function d e f i e d for each A E S such that: amongevents.The logical view is plausiblewhensituations
1) 0 < P ( A ) < 1 VA E S can be reduced t o a set of cases to which the evidence points
2) P(S2’) = 1 equally.
3) if A 1 , A ? , -
is any sequence 6f pairwise disjoint sets in The emerging consensus among decision theoristsis a view of
probabilitythatfranklyadmitsa subjectivecomponent. It
S, then
takes into account that thereis an element of human judgment
P(GI-4,) = Pun). even in the seemingly most objective procedures for determin-
n ingquantitativeprobabilities,anditdoesnotrequirethat
there be only one correct value unless the evidence logically
Afunction P satisfying the abovethreerules is called a
entails it. The essence of this subjective or personal view is
probability measure and the elements of S are referred t o as
that probability is intimately related t o human decision mak-
events. Property (3) is known as countableadditivity. First,
ing, reflectingaperson’sdegree ofbelief that the event in
aneventcan lje inexacta fuzzy event inthe sense that it
question will actually occur. Degree of belief, in this context,
belongs to a fuzzy set. Secondly, even if A is a nonfuzzy event
is interpreted as the extent t o which the belief would contrib-
(well-defined), P ( A ) may be illdefined (e.g., vague evaluations
ute toa disposition t o act rather than as an intensity of feeling.
of exact events, “soft” predictions).
A subjective element can be seen in the frequency view in
At this point we would like t o discuss the exhaustive power
the necessity t o hypothesize a limiting relative frequency and
of the probability system. Clearly, for finite spaces, P ( A ) of
to reassess the probability if the evidenceindicates it. This
any event A is the sum of the probabilitiesof all sample points
amounts t o apersonaldecision. When should weassess the
in it 181. We have no argument about the probability measure
probability? It is assumed that when evidence is available it
itself or about the calculus by which mathematical probabili-
will be used t o update probabilities, and, in essence, quantify
tiesmaybemanipulated.However,ouralternativemeasure
the subjective evaluation involved. In order to do so we use
and calculus will eventually eliminate the exhaustivity involved
the term “gradeofmembership”as the implicationof the
in the classical structure, which restricts every sample point
quantity which depends on a parameter involving human sub-
into a well-defined set.
jective evaluation.
Interestingly enough, we are not the only ones who do not
It is interesting to note that Sugeno and Terano [ 181-[201,
agree with classical probability. There have been, and continue
have used fuzzy integrals over the range (0, 1] to express their
to be, divergent opinions among theorists about the nature of
subjective evaluation of fuzzy objects. They use this concept
the relations between the mathematics of probability and the
in order t o achieve identification of human characteristics as
events t o which it is applied. There are, in fact, so many diver-
well as t o apply it to macroscopic optimization using condi-
gent views, special cases, andsubtlephilosophicalquestions
tional models andto theevaluation of fuzzy objects.
that a brief summary must necessarily ignore most of the fine
Definition 4: Let B be a Bore1 field (a-algebra) of subsets of
points.
The frequency view most often used is based on an experi-
the real line a. A set function p ( - ) defined on B is called a
fuzzy measure if it has the following properties:
ment in which trials are made and a record is kept of propor-
tion of trials whose outcome is the desired one t o the total 1) p ( @ ) = 0 (@is
empty
the
set) (3.2)
number of trials. The probability of the event is taken t o be
the limiting value of the proportion as the number of trials 2) P ( W = 1 (3.3)
increases without bound. If z is one of a set of exclusive and 3) If a, p E Bwith a C 6, then p ( a )< p @ ) (3.4)
exhaustive outcomes of trials and n, is the number of times it
has occurred inn trials, the probability of z is 4 ) If {ail 1 Q j < m} is a monotone sequence, then
n
p ( z ) G lim (3.1)
n+ca n
Clearly a, Q E B ; also if a j E B and ( 9 1 1 < j < m } is a
if the limit exists. If it does not, p ( z ) is undefined. It is never monotonic sequence then limi+ca(ai)E B. In the above defi-
possible to assign an exact probability on this basis unless the nition 1) and 2) mean that the fuzzy measure is bounded and
set of possible trials is exhausted, n being finte, since we can nonnegative, 3 ) means monotonicity (in a similar way t o f i t e
take only a finite amount of data. But we can hypothesize additive measures used in probability), and 4 ) means continu-
until further evidence suggests otherwise. The principal diffi- ity. It should be noted that if 51 is a f i t e set, then the con-
culty with the frequency view of probability is that it makes tinuity requirement can be deleted.
no sense for situations that happen only once, +rice it defines Definition 5: (a,B, p ) is a f u z z y measure space; its analog
probabilityintermsofcollectionsorsequencesofevents. in probability is (a,S, P). p ( * )is the fuzzy measure of (a, B).
From this viewpoint, the probability of a unique event, such The fuzzy measure p is defined on real intervals. Clearly we
as heads on a particular toss of a coin, must either be unde- must use a function ’ f (T)t o describe p[x > T ] when x is the
fined or else be zero or one, depending onwhich way the coin membership function of set A . Throughout our investigation
falls. Similarly,a purefrequency view must reject a priori we shall use p t o represent f~ (T)assuming that our set A is
probabilities which are not based on some number of trials but well specified.
on collateralevidence, such as the numberof sides of a die. Let x: G! + [O, 1 I and [T = { x Ix(x) 2 T } . The function x
Still another view has been advanced that probability is a is called a B-measurable function if E B V T E [0, 11 . Defi-
KANDEL AND BYATT: FUZZY SETS, ALGEBRA, AND STATISTICS 1625

nition 6 defines the fuzzy expected value (FEV) of when x since

X € [ 0 , 1 ] . Extension of this definition when X € [a,b],
0 Q a < b < 00 will be presented in (3.8).
Definition 6: Let x
be a B-measurable function such that
x € [0, 11 . The fuzzy expected value (FEV) of over a set A ,
x clearly
with respect to themeasure p(') is defined as

and thus we define

Note: The B-measurable function x is called the compatibility
function. In the case of linguistic variables [ 241 the numerical
variables whose values constitute what may be called the base
variable, describea fuzzy restriction on the meaning of the Since the turning point between ) Tis at H where
p ( [ ~and
linguistic variable. This fuzzy restriction (which is clearly sub- C ( ( [ H ) > H and ~ ( 5 ' f i ) Q H
jective) on the values of the base variable is characterized by
it is quite obvious that R and H must be identical. Namely
the compatibility function x which associates with each value
of the base variable a number in theinterval [ 0, 11, represent-
ing its compatibility with the fuzzy restriction.
If thecomposition established in Definition 6 underthe Q.E.D.
operation "Max-min" seems like a pessimistic evaluation of
the FEV,doesa similar composition using theoperation Theorem 5 can also be derived via a set of interesting lem-
"min-Max" act as an optimistic model? mas. We feel that the following is not just a repetition of a
The answer tothe above question lies inthe following proof, but that the lemmas give more insight to the properties
theorem. of the FEV [ 151.
Theorem 5: Let x:
a + [0, 1 ] and Lemma 1 : For every T I , T2 E [0, 1 ]
= {X Jx(x)> T} E B. (3.6) min[T1,1-((5~,)1 QM~X[T~,C((~T,)I.
Then Proof: Case 1:
T1 < T z .
In this case
(3.7)
~~~[TI,.c((ET,)I QT1 QT2 QMax[T2,p(f'~,)I.
where p is a fuzzy measure.
Case 2:
Proof: Let S U P T ~ O{, m ~ i]n [ T , p ( I ~ ) I )= H. Then
T2 < T i .
In this case
C((tT,)<P(fT,).

Since p ( f ~increases
) as T decreases we have Hence,
m i n [ T ~ ~ ~ c ( tQ~c,() (l ~ T , ) < c ( ( ~ T , ) < M ~ ~ [ T ~ , c ( ( ~ T
from which the lemma follows. Q.E.D.
and Lemma 2:

Thus Proof: By Lemma 1, every member of {min[T, p ( 5 ~ ) ] ,

T E [0, 11) is less than every member of {Max[T, p ( f ~ ) 1 ,
T E [ 0 , 1 I}. The lemma follows.
Lemma 3: There is no real numbers such that
and

Assume Proof: Suppose such an s exists. We distinguish two cases.

Inf {Max[T,p ( f ~ ) ] }= R such that R # H . Case 1 : p(&) Q s;
TEIO,11
Case 2: ~(6,) > s.
Then
Each case leads to a contradiction.
Case 1 :
s< Inf{Max[T, ~(FT)I) Max[#, ~ ( ( f ~ =
) ls.
TEIO,Il
1626 PROCEEDINGS OF THE IEEE, VOL. 6 6 , NO. 12, DECEMBER 1978

Case 2: Example 2: Using the base variable age we examineaset

having only x people of age a and y people of age p where a
s = h [ s , p(t,9)1 < SUP { h [ T , p ( t T ) ]} < 8.
TE[O,11 and 0 correspond t o the compatible values CYTand &-,respec-
tively. Assume
Since both cases lead t o a contradiction, thelemma follows.
Q.E.D.
Lemma 4: It is not true that and

Proof: If this inequality held, there would be a real number In this example p ( * ) acts as a standard probabilistic measure.
Clearly
betweentheleftandthe right-handmembers,contradicting
Lemma 3. Q.E.D
Hence, Theorem 5 follows from Lemmas 2 and 4.
Since the FEV is defined with no use of addition, subtrac-
tion, multiplication, or division, it is not to be expected that
it will satisfy identities involving these operations nor operators
derived from them such as d / ( d t ) . Thus the FEV is not addi-
tive or linear.
Clearly Definition 6 can be extended to any real interval
[ a , bl by extending T and p (or T and f A ( T ) )to this interval,
under the same transformation that x undergoes.Hence, we
can define the FEV of an extended grade of membership func-
tion 7) E [ a , b l , as

where [ T * = { x 1 ~ (2 .
x )T * } and p* E [ a , b ]
Inparticular, if C is aconstantand we want to evaluate In what follows we shall construct a combinatorial scheme
FEV(x + C) fromsomecompatibilityfunction x, then the to generalize the above example.
constant C acts as a factorto scale the compatibility curve and Assume a finite set of data points where there are n + 1 dis-
thus t o scale the FEV and t o evaluate C + FEV(x). Thus the tinct levels of compatibility suchthat
value of FEV(x + C) willbe in the interval [C, C + 11 if
x E [0, 1 1 . Similarly, if C isa positive constant, the FEV(C * x )
is basically a normalization of the compatibility curve x by a which imply n distinct levels of fuzzy measure p ( t ~ )exclud-
,
factor C whichacts as a normalization factor on the entire ing 0 and 1.
FEV(x). Thus we have the following results. Clearly the two sets representing
Proposition 1: Let C be a constant and x + [O, 1 I . Then
FEV(x + C) = FEV(x) + C.
are in increasing order, and we trivially exclude any permuta-
Proposition 2: Let C > 0 be a constant and x +. [0, 1I . Then tion in eachof the sets. Thus the set representing the unionof
- -
FEV(C X) = C FEV(x). these two sets has (2n + 1) elements; in order t o find all possi-
ble arrangements of these elements we can view the problem
These last results provide some insight into the natureof the of finding the number of arrangements as follows.
FEV as a tool for partitioning of a population into “above Find the number of n-arrangements of (2n + 1) objects, of
FEV” and “below FEV.” Thus the FEV is more of a typical which exactly (n + 1) are alike of one kind and exactly n are
value of the population. alike of the second kind.
In the followingsection we shall illustrate and obtain the Solution: Let y be the requirednumber of arrangements.
relationsbetweenthe FEV andthestandardprobabilistic -
Let the (2n + 1) objects be labeled X I , * * , x Z n + , with the
expectation on one hand, and between theFEV and measures n + 1 alike of one kind labeled x l , -
, x , + , , and the n alike
of central tendency. of the secondkindlabeled * -,X Z , + , . Thenthere will
be (2n + l)! n-arrangementsofthese(now distinct) labeled
C. FE V and Measures of Central Tendency objects, such that we can set
An average is a value which is typical of a set of data. Since (2n + l)! = m ! ( n + l)! (3.9)
such typical values tend to lie centrally within a set of data
arranged according to magnitude, averages are also known as since for each of the 7 required arrangements, the fmt (n + 1)
measures of central tendency. Several types of averages can be alike, being labeled, can be interchanged in (n + l ) ! ways and
defined, the mostcommonbeingthearithmeticmean,the the n alike of the second kind, being labeled, can be inter-
median, the mode,thegeometricmean,andtheharmonic changed in n! ways.Thus ? = [(2n + l)!]/[n!(n+ l ) ! ] . We
mean. present now our main claim.
Before discussing the relations of the FEV to the mean and Theorem 6: The median (middle value) of the set of 2n + 1
to the median, we shall illustrate the use of the concept by a numbers,representing {ai}?:: and { p j ( t ~ ) } T =as~ described
simple example. above, where 0 < al < a 2 < -* * a,+, < 1 for a finite n, ar-
KANDEL AND W A T T : FUZZY SETS, ALGEBRA, AND STATISTICS 1627

ranged in order of magnitude (i.e., in an array left t o right), is

the FEV.
Proof: In order t o find the FEV, we have t o compare ai’s,
>
i 1 with the proper p i ( g ~ )take
, the minimum of these two
and thenthe Maximumover all minimumvalues.Clearly
&(al, 1) = al . It should be noted that a1 cannot appear t o
1

0.9
the right of the middle of the array since it has to be followed
by at least all the membersof the sequence {ak};::. It is 0.8

important to note that the valuesof the two sequences are 0.7

independent but not their place in the array. This is due t o 0.6
the fact that if ai > ai then p j ( [ ~<) &(.$!T). Because of that 0.5
and because the array is arranged in an increasing order, the 0.4
fmt part of the array, including n numbers (n + 1 left t o right 0.3
and excluding a l ) is compared with the second part of the
0.2
array, including the last n numbers (right to left). Since we
0.1
take the minimum of every comparison, the results must lie in
the first n + 1 numbers of the array. Out of these n + 1 num- 0
bers we are interested in the maximum, which is obviously the
number in the (n + 1)th place, since the array is increasing in Fig. 2. Compatibility curve for hourly wages.
order. This number is the median of the array.
Example 3: Let A be ourtestedpopulationsuchthat
Q.E.D.

people are of age a (compatible value a ~ y) ,people are of age

0 (compatible value &), and z people are of agey (compatible
x
curve such that
1 person is making $3.00 -- al = 0.40
valueyT)wherex+y+z=A,1>yT>PT>aT>Oand 3 persons are making
4 persons are making
$4.00
$4.20 -- ai = 0.50
a3 = 0.55

Since R = 2 thereare10differentarrangements as given

below; the FEV’s can easily be verified graphically and thus
2 persons are making
2 persons are making $10.00
$4.50
- cq = 0.60
as = 1.oo.
Arranging the sequences {ai};=l and { pj(.$!T)}!=l is an ordered
will be omitted here. array, where p is represented by ~ A ( Tand) ~ A ( T=)l f T 1 / 1 2 ,
we get the values
{ 1 1 2 L 1 1 1 2 1 1 1 }
63 3 3 5’ 2 7 20’ 5 , 3 ’ 12’ *

Both the FEV and the medianare 8.

Theprobabilisticex-
pectedvalue(mean) is 0.61(using the samecompatibility
curve on [0, 1 1 interval).
In thisexamplethe FEV gives abetterindication of the
average hourly wage than the mean, by not being affected by
the extreme value of $10.00.
It should be noted that a different population distribution
might give a FEV which is different from the median.
Example 5: Thehourly wagesoffive peopleare $2.20,
>FEV = Dr. $2.50, $2.70, $3.50,and $10.00; using the same compatibility
curve of Example 4, these wages give the following sequence
of ai’s, 1 < i < 5:
0.25, 0.3, 0.35,0.45, 1.
Namely, the median is 0.35 ($2.70), the FEV is 0.4 ($3.00),
and the mean is 0.47 ($3.80).
Since the median of a data set is a quantile that, like the
expected value, acts as a “center” for a given distribution, we
claim that the first p j ( g ~ in
) the array (from left to right),
a
which is greater or equal to indicates the respective ai as the
median of the data set. If no such p j ( . $ ! ~exists
) then a, is the
median. We shall denote the above specified pj(&-) and ai as
pme& and a m e b , respectively. Clearly, if p m e h > FEV
then &,,.,e& < FEV and if p m e b < FEV then ct,,,e> h
4,
FEV; since p m e b > the mean as obtained is closer to the
FEV than to the median. Namely

Example 4: Using the base variable “Hourly Wages” let us

I Mean-FEV I < I Mean-Median I (3.10)
assumea given population and a given subjectivecompatibilityorusingtheemirical result forunimodalfrequencycurves
 1628 PROCEEDINGS OF THE IEEE, VOL. 66, NO. 12, DECEMBER 1978

which are moderately skewed (asymmetrical) [ 241 ,we have

I Mean-FEV I Q 4 I Mean-Mode I. (3.1 1)

Similar results can be derived for grouped data.

Itshouldbe pointed out that different compatibility curves

with the same fuzzy measure and the same frequency distribu- and
tion might yield a different FEV. This is not so when standard
median is used, since the FEV is also a function of the subjec-
tive evaluation. Hence,
We will now formulate as easier way to obtain the value of
the FEV.
Definition 7: A value T E [0, 11 is a subtypical value of a
function if T < p ( t ~ )supertypical
, if T 2 p&), and typical by use of Theorem 5 the result follows.
and Q.E.D.
if T = p ( t ~ ) .Note that a typical value is both subtypical and Now let the domain of a function x be the union of a finite
supertypical. number of subspacesK = {sl,sz, * ,s n } such that x: K +
Theorem 7: If a number T1 is a subtypical value of two [0, 11. Ifwe consider a fuzzy measure space (K, 2K, p ) we
functions t and q , and if for every T 2 T I , f~ = QT, then can write the FEVof x as
FEV([) = FEV(p).
Proof: For T < TI we have

T < TI Q IT, 1 <P ( b )

We assume that x(si) < X ( S ~ +for
~ ) 1 < i < n - 1. If this is
and not the case then rearrangement of x(si) is necessary. Then
the following holds.
Theorem 9: The FEV of x in (K, 2K, p ) can be written as
Thus
I
Max{h[X(si), c~(Ki)l
i

where
and
Ki = { s i , s i + l , * ,s,,}, 1 Q i Q n.

Proof: Let si) = h s E ~ X($). Then

g

and since {KiI 1 Q i Q n < 2K), the reverse inequality holds,

too. The equality follows. Q.E.D.
Even though the power set 2, has 2" members, Theorem 9
calls for a monotone sequence of subsets of K such that K1 >
-
K2 > * * > K, so that we can find the FEV by calculating
min[x(si), p(Ki)] at n points at most.
Theorem 10: The FEV of x in (K, 2 K , p ) can be written as

d [ X ( s j ) , ~ ( ( K j ) l iff x(sj-1) </J(K~)

Q~(sj)
or

e and result follows. Q.E.D. >

c ( ( K ~ ) X b j ) 2 c ( ( K ~1).
+
Theorem 7 implies that values of a function that are below Proof:
a subtypical value may beignored in calculating the FEV. a) Necessity: Clearly since p(Ki) is a monotonic decreasing
Similarly, we show in Theorem 8 , which is the dual of Theo- function and x(si) monotonicincreasing, the function
rem 7, that values of a function that are above a supertypical
value may be ignored in calculating the FEV. d[X(si), p(Ki)l
Theorem 8: If a number T2 is a supertypical value of two has a single peak for some i. Thus it is necessary that
functions 5 and q , and if for every T Q Tz, [T = VT, then
FEV(t7) = FEV(t). min[X(sj-l), ~ ( K j - 1 ) l Q min[x(s/), ~ ( ( K j l l
Proof: For T > Tz we have
2 min[x(sj+l),~ ( W j + l ) l -
T > T2 2 P ( ~ T , 2
) P(~T) b) Sufficiency: Assume x(sj) >p(Kj). Then we have
and x(sj-1) <Cc(Ki) Q X(S~)
from theabove inequality.
KANDEL AND BYATT:
ANDALGEBRA,
FUZZY SETS, STATISTICS 1629

If we assume since p ( $ ~increases

) as T decreases, and, therefore, it is trivial
that

then
P ( K j ) > X(sj) > P W j + l ) . and thus
Hence the proof is complete. Q.E.D.
Thus there is no need to evaluate min[x(sj), p ( K j ) ] for all i,
but only at point j fulfilling the requirements of Theorem 10. and
Moreover, since x is a known function there is a need to eval-
uate p ( K i ) for three different points only. P ( t H ) >H.
Clearly this technique, as far as complexity of computations Similarly p (&!H+)< H since t ~= +{ x Ix(x)> H } , and
is concerned, has advantages over classical evaluations of
probabilistic expected values.
Another relation between the FEV, when p ( * ) is taken to be
additive (probability measure), and the probabilistic expected Hence,
value can be established via the following theorem. lr-
Theorem 11: Let (52, B, p ) beaprobabilityspaceandlet
x: L?+ [0,11 be a B-measurable function. Then

where

4
implies H = and lAl< 4. Q.E.D.
This means that by a subjective evaluation of grades of mem-
bership via the FEV, using a probabilistic measure, we obtain
4
an average value different by no more thana from the classi-
cal probabilistic expected value.
The problem of selectinga proper measure is related to
selecting a criterion for optimality condition. Clearly a func-
tion of theform p ( - ) I p ( A ) = S U P , E A [ X ( X ) ] satisfies the
conditions of a fuzzy measure.
To summarize this point, we are confronted with anidentifi-
cation problem; namely, we have t o find an optimal set func-
Define the quantityH by
tion p ( * ) , amon! the set of all set functions satisfying the
fuzzy measures p ( * ) such that a certain optimality criterion is
that eOp=
satisfied, e.g., find p ( * ) such e. It should
Then be pointed out, however, that even in the cases where the
fuzzy measure used was a probability measure, the concept of
FEVhad advantages over the classical methods of"average
evaluation."
As a last remark in this chapteron fuzzystatistics it is
interesting to note that regarding the logic of uncertainty, a
formal propositional calculus has beendeveloped by Gaines
clearly [ 9 ] ,in which statements consisting of truth-values in the con-
tinuous interval [ 0, 1 ] have been assigned t o propositions in a
lattice. These truth-values form a continuous, order-preserving
1630 PROCEEDINGS OF THE IEEE, VOL. 66, NO. 12, DECEMBER 1978

valuation of the lattice giving a Basic Probability Logic (BPL) andtheir applications will concentrateontheseandmany
which satisfies the usual definitions of a probability logic or other areas, and particularly on theselection of an appropriate
probability over a language except for theLaw of the Excluded measure function under a suitable optimality criterion. This
Middle (LEM). will enable better applications and clear understanding of the
A common semantics for the BPL and all its derivatives can role of fuzzy statistics in manyengineering and scientific struc-
be given in terms of the binary responses of a population. tures, especially in man-machine systems.
Hence, fuzzy logic and probability logic have common formal In conclusion, the wider framework for analysis of uncer-
andsemanticfoundations. Defining thepopulationfurther tainty described in this work establishes a close link between
brings out the distinctions normally made between“degree of fuzziness and probability theory, to the mutual advantage of
membership” and “probability,” not now as absolute divisions both fields. It also makes clear theuniquecomputational
but rather, for example, as the distinction between “physical- advantage of fuzzy statistics, an important attribute when one
frequentist”and “subjective” interpretations of probability. is dealing with practical problems in the real world.
For example, the “population” may be one of physical events,
or of people giving opinions, or of decision-making elements ACKNOWLEDGMENT
(neurons). Portions of this research have been carried out at the EECS
To quote Gaines [91: Department,The University of New Mexico; the Computer
Science Department, New Mexico Institute of Mining and
. . .it is suggested that regardless of ita correctness in particular Technology, the EE Department,BenGurion University of
circumstances the advantages of fuzzy logic may be seen in its
strong truth-functionality(TF). It is unique amongst the various the Negev, Israel, and the Mathematics Department, Florida
logics of uncertainty in not requiring memory of the structures State University.
of two propositions (in terms of the lattice generators) when The authors are deeply indebted to all the people who have
computing connectives involving them. This is not only of prac- been involved in the project. Our thanks to Professors Donald
tical importance but may also explain the role of fuzzy logic in
modelling human reasoning where short-term memory capabili- D. R. Morrison and S. Karni, P. Dorato, and t o H. A. Davis for
ties are notoriously weak. theirsupportandcooperation. Special thanks are due to
Professor Lotfi A. Zadeh for his encouragement as well as for
In conclusion, Gaines [9] shows that a nonfunctional BPL his valuable suggestions.
provides a formal foundation fora general logic of uncertainty
encompassing both fuzzy and probability logics. LEM is con-
REFERENCES
sistent withweaker forms of TF leading to partially fuzzy
logics with Max-min connectivesbetweenprimitiveproposi- [ 11K. J . Arrow, Social Choiceand Individual Values, Cowles
Commission Monograph No.12.NewYork: Wiley and Sons,
tions, logics of statistical independence,mutual exclusion, First edition, 195 1.
etc. Gaines model also clarifies the distinction between fuzzy [ 2 ] R. Bellman, and M . Giertz (1973), “On the analytic formalism
o f the theory of fuzzy sets,” Inform. Sa’., vol. 5 , pp. 149-156.
“degree of membership” and conventional “probability,” [ 3 1 J. Bernoulli, Ars Conjectandi, Basel.
showing it to be one of detailed semantic interpretation rather 141 M. Black, “Vagueness: An exerciseinlogicalanalysis,”Phil.
than one of logic or basic semantics. Science, vol. 4 , pp. 427-455.
[ S I D. Van Dantzig,“Statisticalpriesthood (Savage on personal
Thus the wider frameworkfor logics of uncertainty de- probabilities),” Statistica Nederlandica, vol. 2, pp. 1-16.
scribed by Gaines [9] establishes a close link between fuzzy [ 6 ] B. de Fmetti, “La Revision:Ses lob logiques, ses sources
logic and probability theory, to the mutual advantage of both subjectifes,” Ann. Inst. H.Poineare, vol. 7 , pp. 1-66. Trans-
lated as “Foresight:Itslogicallaws, its subjectivesources,”
fields. It also makes clear theuniquecomputational advan- in Studies in SubjectiveProbability. NewYork: Wiley and
tage of fuzzy logic derived from its strong.truth-functionality, Sons, 1964.
[ 7 ] -, “Methodsfordiscriminatinglevels of partial knowledge
which is so much emphasized in this work. concerning a test item,” Brit. J . Math. Stat. Psychol., vol. 18,
pp. 87-123.
IV. CONCLUSIONS [SI W. Feller, AnIntroduction to ProbabilityTheory and Its
Applications. New York: Wiley and Sons.
In this paper, we reported on the development of analytical 191 B. R. Gaines, “General fuzzy logics,” EES-MMS-FUZZ-75,
methods in fuzzy set theory, fuzzyalgebra, and fuzzy statistics. University o f Essex, U.K.
Techniques based on these subject matters can be applied t o (101 J. A.Goguen, Jr., “L-fuzzysets,” J. MathAnal.and App.,
VOl. 18, pp. 145-174.
many processes in science and engineering. We shall not detail [ 1 1 ] -, “Categories o f fuzzy sets: Applications o f non-Cantorian
the wide range of applications using fuzzy systems and basic set Theory,” Ph.D. dissertation, Dept. o f Math., Univ. o f Cali-
fornia, Berkeley, CA.
models-the papers are listed in the bibliography. In particu- 1121 A. Kandel and H. A. Davis, “The first fuzzydecade,” (bibli-
lar, fuzzystudiesappear to have strong links with classical ography on fuzzy sets and their applications), CSR 140, Com-
multivalued logics, especially to Zukasiewicz’s i n f ~ t e - v a l u e d puter ScienceDepartment,NewMexicoInstitute of Mining
and Technology, Socorro, NM.
logic. It is clear that many mathematical structures may be [ 131Kaufmann, TheorieDesSws-EnsemblesFlous, vol. 1-3,
analyzed using fuzzy sets rather thanclassical sets. Masson, Paris, France, 1973.
We feel that newcomers, as well as established researchers in [ 1 4 ] H. E.Kyburg, Jr., and H. E. Smokler, Studies in Subjective
Probability. New York: Wiley and Sons, 1964.
this field, will find ourpresentation, as well as the bibliog- [.15] D. Morrison, private communication, 1976.
raphy,animportanttool inexploring themathematical [16] A. H. Murphy, “Ordinal relationships between measures o f the
propertiesandtheenormousamount of applications into “accuracy” and “value” of probabilityforecast: Preliminary
results,” Tellus, vol. 24, no. 6 , pp. 531-541, 1972.
several fields, including the evaluation of fuzzy phenomena in 171 L. J. Savage, The Foundations OfStatirriCs. New York: Dover,
pattern recognition and classification, weather prediction and 1972.
181 P. Suppes and Zinnes, “Basic measurementtheory,” in Hand-
forecast evaluation, hydrological forecasting, quality control, book of Math.Psychology, Luce,Buch, and Galinter,Eds.
the evaluation of eigenvalue spectrum for a particle in a fuzzy New York: Wdey and Sons, 1963.
medium, and the topic of the measurement problem in quan- i 19) M. Sugeno and T.Terano,“Anapproach t o theidentification
o f human characteristics by applying fuzzy integrals,” in Proc.
tum mechanicsare very promising, and it is our hope that 3rd IFAC Symp.onIdentificationandSystemParameter
future studies on the analytical properties of fuzzy set theory Esrimation, the Hague, theNetherlands, 1974.
KANDEL AND BYATT: FUZZY SETS, ALGEBRA, AND STATISTICS 1631

1201 -, “Analytical representation of fuzzy systems,” in Special Dept. of Civil Eng., University of Calif., Berkeley, CA.
Interest Discussion on Fuzzy Automataand Decision Processes, I241 Bellmann, R. E., Kalaba, R., andZadeh,L. A. (1966), “Ab-
6th IFAC World Congress, Boston,MA, 1975. straction and pattern classification,” J. Math. AM^. and Appl.,
121) L. A. Zadeh,“Fuzzysets,” Inform.Control, vol. 8, pp. 338- Pp. 1-7.
353, 1965. I251 Bellmann, R. andZadeh, L. A. (1970), “Decision-makingin
(221 -, “Probabilitymeasuresoffuzzyevents,” J. Math. AMI. a fuzzy environment,”Man. Sci., pp. B141-B164.
and App., vol. 10, pp. 421-427, 1968. 1261 Bellmann, R. (1971), “Law and
mathematics,”Technical
1231 -, “Similarityrelationsandfuzzyordering,” Znform. Sci., report 71-34, University of Southern California, Los Angeles,
VOl. 3. DD. 177-200. 1971. CA.
-
I . .

I , “A fuzzy-algorithmic approach t o t h e definition of com- I Bellmann,R.E.andMarchi, E. (1973), “Games of protocol:

plex o r impreciseconcepts,”Memo. No., ERL-M 4 7 4 , Elec- Thecity as a dynamiccompetitiveprocess,”Tech.report
tronics Res. Lab., Univ. of Calif., Berkeley, 1974. RB73-36, University of
Southern California,
Los Angeles,
1 - , “Fuzzy sets as a basis for a theory of possibility,” ERL CA.
Report, University of California, Berkeley, 1977. Bellmann, R. andGiertz, M. (1973), “On theanalyticfor-
, I L. A. Zadeh e t al., “FuzzySetsandTheirApplicationsto malism ofthetheoryoffuzzysets,” InformationSciences,
Cognitiveand Decision Process,’’ in Contributionsfromthe pp. 149-156.
U.S.-JapanSeminaronFuzzySets and TheirApplications, Bellmann, P. E. and Marchi, E. (1973), “Games ofprotocol:
UC Berkeley,July 1974, published by Academic Press, N.Y., Thecity as a dynamiccompetitiveprocess,”Tech.report
1975. RB73-36, University of
Southern California,
Los Angeles,
H. J. Zimmemann, “Bibliography on theory and applications CA.
of fuzzy sets,” RWTH Aachen, West Germany, 1975. 301 Be-hann,R. (1974), “Local logics,” TechnicalReport 74-9,
University of Southern California,Los Angeles, CA.
BIBLIOGRAPHY 311 Bezdek,J. C. (1973), “FUZZYmathematics in pattern classi-
fication,” Thesis,CenterforAppl.Math.,CornellUniversity,
[ 11 A d a m , E. W., (1968), “Elements of a theory of inexact mea- Ithaca, NY.
surement,” Phil. of Sci., pp. 205-228. I321 Bezdek, J. C. andDunn, J. C. (1974), “Optimalfuzzyparti-
[ 2 ] Adaric, P. N., Borisov, A. N., andGolender, V. E. (1968), tions: A heuristic for estimating the parameters in a mixture
“An adapJive algorithmforrecognitionoffuzzypatterns,” of normal distributions,” Cornell University, Ithaca,NY.
Kibernetika Z Diagnostika, pp. 13-1 8, (in Russian). I331 Bezdek, J. C. (1974), “Numerical taxonomy with fuzzy sets,”
[ 3 ] Adey, W. E. (1972), “Organizationofbraintissue: Is the J. Math. Biology, vol. 1, pp. 57-71.
brain a noisy processor?,” Znt. J. Neuroscience, pp. 271-284. 1,341 Bezdek,J. C. (1974), “Cluster validitywithfuzzysets,” J.
141 Allen, A. D. (1973), “A method of evaluating technical journals Cybernetics, pp. 58-73.
on the basis of published comments through fuzzy implications: 1351 Bezdek, J. C. (1975), “Mathematicalmodels for systematics
A survey of the major IEEE transactions,”ZEEE Trans. Systems, andtaxonomy,” in R o c .8 t h Annual Znt. Conf. Numerical
Man, Cybernetics, vol. SMC-3, pp. 422-425. Taxonomy, San Francisco, CA.
[SI Allen, A. D. (1974), “Measuring the empiricalpropertiesof 1361 Black, M. (1970), “ “Vagueness”: Anexercisein logical anal-
sets,” ZEEE Trans. Systems, Man, Cybernetics, vol. SMC-4, ysis,” Phil. ofscience, pp. 427-455.
PP. 68-73. 1371 Black, M. (1970), Margins ofPrecision,Ithaca:Cornell Univ.
161 Albin, M. (1975), “Fuzzy sets and their application t o medical Press.
diagnosis,” Ph.D. dissertation,Dept.of Math.,Univ. of Cali- (381 Black, M. (1973), Reasoning with Loose Concepts, Dialogue 2 ,
fornia, Berkeley, CA. pp. 1-12.
[ 7 ] Arbib, M. A. (1970), “Semi-Ring languages,” Elec. Eng.
~~ - Dept.,
- [ 3 9 ] B h , J. M. and Whinston, A. B. (1974), “Fuzzy sets and social
Stanford University, Stanford,CA. choice,” Jour. Cybernetics.
[S] Arbib, M. A. and Manes, E. G. (1974), “Fuzzymorphismsin 1401 B h , J. (1974), “Fuzzy relations in group decision,” Jour. of
automata theory,” in Proc.First Znt. Symp. Category Theory Cybernetics.
Applied to Computations and Control, pp. 98-105. (411 Boicescu, V. (1971), “Sur les algebres de lukasiewicz, logique,
191 Arbib, M. A. and Manes, E. G. (1975), “Fuzzymachinesin a automatique, informatique,” Ed. Acad. Rep. SOC.de Roumanie,
category,” Coins Tech. Report 75B-1, University Massachusetts, Pp. 71-97.
Amherst, MA. [ 4 2 ] Borghi, 0.(1972), “On a theory of functionalprobability,”
[ l o ] Arbib, M. A. and Manes, E. G. (1975), “Acategory-theoretic Revista Un. Mat. Argentina, pp. 90- 106.
approach t o systems in a fuzzyworld,” Synthese, pp. 381- 1431 Borisov, A. N. and Kokle, E. A. (1970), “Recognition of fuzzy
406. patternsbyfeature analysis, cybernetics and
diagnostics,”
I l l ] Asai, K. andKitajima, S. (1971), “A methodforoptimizing Riga, USSR.
control of multimodal systems using fuzzy automata,” Inform. I441 Borisov, A. N. and Kokle, E. A. (1970), “Recognition of fuzzy
Sciences, pp. 343-353. patterns,” Kibernetika IDiagnostika, pp. 135-147.
[ 121 Asai, K. andKitajima, S. (1971), “Learning control of multi- I451 Borisov, A. N. and Osis, J. J . (1970), “Methods for experimental
modal systems by automata,” Pattern Recognition and Model estimation of membership functions of fuzzy sets,”Kibernetika
Learning. New York: Plenum Press. ZDiagno~tika,PP. 125-134.
[ 131 Asai, K. andKitajima, S. (1971), “Optimizingcontrol using 1461 Borisov, A. N., Vulf, G.N., and Osis, J. J. (1972), “Prediction
fuzzy automata,” Autornatica, pp. 101-104. of the state of a complexsystem using the theory of fuzzy
[ 141 Assillian, S. (1974), “Artificialintelligence in thecontrolof sets,” Kibernetika Z Dfagnostika, pp. 79-84.
real dynamic systems,” Ph.D. dissertation, Queen Mary College, 1471 Bossel, H. H. and Hughes, B.B. (1973), “Simulation of value-
University of London, London, England. controlled decision-making,”Report SRC 73-1 1 , Systems
[ 151 Aubin, J. P. (1974), “Fuzzy games,” University of Wisconsin, Res. Center, Case Western Reserve University, Cleveland, OH.
Madison, WI. 1481 Bremermann, H. J. (1971), “Cybernetic functionals and fuzzy
[ 161 Aubin, J. P. (1974), “Theorie d e jeux,” C. R. Acad, Sc. Paris, sets,” ZEEE Symp.Record,Systems, Man and Cybernetics,
T. 279. Sene A-891. PP. 248-253.
[ 1 7 ] Aubin,’J. P. (1974), “Theorie d e jeux,” C. R. Acad. Sc. Paris, 1491 Brown, J. G. (19691, “Fuzzy sets on boolean lattices,” Ballastic
T. 279, Sene A-963. Research Laboratories, Memo Report 1957.
[ 181 Baas, S. M. and Kwakernaak, H. (1975), “Rating and ranking [SO] Brown, J. G. (1971), “A note on fuzzy sets,” Inform. Contr.,
of multiplaaspect alternatives using fuzzy sets,” Memo. NR.73, Pp. 32-39.
Dept. Appl. Math., Twente Universityof Technology, Enschede, 151) Bunge, M. C. (1966), “Categoriesofsetsvaluedfunctors,”
Netherlands. Ph.D. dissertation, University of California, Dept. of Math.
I191 Banaschewska, B. (1968), “Injectivehullsin thecategoryof 1521 Butariu, D. (1975), “L-fuzzy automata description of a neural
distributive lattices,” J. fur die Reine und Angewandte Mathe- model,” h o c . 3rd Znt. Cong. Cybernetics and Systems, Bu-
matik, pp. 102-109. charest, Rumania.
1201 Banaschewski, B. andBruns, G. (1967), “Categoricalchar- I531 Capocelli, R. M . andDeLuca, A. (1972), “Measures ofun-
acterization of t h e mac-neville completion,” Arch. der Math., certainty in the context of fuzzy sets theory,” Atti del ZL??
PP. 369- 377. C0ngres.v Nationale di CiberneticaDie C w i a ~ Tenne, Pisa,
121 ] Bang, S. Y. and Yeh, R. T. (1974), “Toward a theory of rela- Italy.
tional data structure, ” SESLTR-1, University of Texas, Austin, I541 Capocelli, R. M. and De Luca, A. (1972), “Fuzzysetsand
TX. decision theory (abstract),” Notices Amer. Math. SOC., A-709.
I221 Barnev, P., Dimitrov, V., andStanchev,V. (1974), “Fuzzy [ 5 5 ] Capocelli, R. M. and De Luca, A. (1973), “Fuzzy sets and de-
systemapproach t o decision-making based on publicopinion cision theory,” Inform. Control, pp. 446-473.
investigation
throughquestionnaires,” Inst. of Math. and 1561 Carlucci, D. andDonati, F. (1975). “A fuzzycluster of the
Mech., Bulgarian Acad. ofSci., Sofia, Bulgaria. demand within a regional service system,”specialinterest
[ 2 3 ] Becker,J. M. (1973), “Astructural design process,”Thesis, discussion on fuzzy automata anddecision processes, 6th
1632 PROCEEDINGS O F THE IEEE, VOL. 6 6 , NO. 12, DECEMBER 1978

IFAC World Cong., Boston, MA. Belgium.

I571 Carktrom, I. F. (1975), “Truthandentailmentfora vague [ 8 9 ] Dal Cin, N. (1975), “Fuzzy-state automata, their stability and
Quantifier,” Synthese, pp.461-495. fault-tolerance, Inst. ofSci., Univ. of Tubingen, Germany.
(581 Castonguay, C. (1972), “Meaning and
existance
in mathe- [ g o ] Dal Cin, N. (1975), “Modificationtolerances offuzzy-state
matics,” Library of Exact Philosophy, Springer-Verlag, vol. 9. automata,”Znt,.J. Comput.Inform. Sci.,pp.81-93.
1591 Chang, C. C. (1964), “Infinite valuedlogic as a basis for set (911 David, M. andThayse, A. (1973), “Representation of fuzzy
theory,”in Roc. 1964 Znt. Cong. for Logic,Methodology, functions,” Philips Research Reports, pp. 93-106.
and Philosophyof Science, Amsterdam, North-Holland. 1921 deKerf,J. (1974), “Vageverzamelingen,omega”(Vereniging
1601 Chang, C. L. (1967), “Fuzzysetsandpatternrecognition,” voor wis- en natuurkundigen Lovanienses), pp. 2-18.
Thesis, University of California, Berkeley, CA. [ 931 de Kerf, J. (1974), “Vage verzamelingen,” Ingenieurstifdingm,
1611 Chang, C.L. (1968), “Fuzzytopological spaces,” J. Math. Pp. 581-589.
AML Appl., pp. 182-190. (941 DeLuca, A. andTermini, S. (1971), “Algorithmicaspectsin
1621 Chang, C.L. (1970), “Fuzzy algebras, fuzzyfunctions,and complex systems analysis,” Scientia, vol. 106, pp. 659-671.
their applications to function approximation,” Dept. of H.E.W., I951 DeLuca, A. andTermini, S. (1972), “Algebraicproperties o f
Nat. Inst. of Health, Bethesda, MD. fuzzy sets,” J. Math. AMI. Appl., pp. 373-386.
I631 Chang,C. L. (1975), “Internationalandexecutionoffuzzy I961 De Luca, A. and Termini, S. (1972), “A definition of a non-
programs,” in Fuzzy Sets and Their Application to Cognitive probabilistic entropy in the setting of fuzzy set theory,” Zn-
and Decision Processes, L. A. Zadeh et al., Eds. New York: f o m . Contr., pp. 301-312.
Academic Press, pp. 191-218. (971 DeLuca, A. andTermini, S. (1974). “EntropyofI-fuzzy
[64] Chang,R.L.andPavlidis, T. (1976), “Fuzzydecisiontrees,” sets,”Znfown. Cont., pp. 55-13.
Technical Report 203, EECS Dept.,
Princeton
University, 981 Depalma, G. F. andYau, S. S. (1975), “Fractionallyfuzzy
Princeton, NJ. grammarswithapplication topatternrecognition,”inFuzzy
1651 Chang, R. L. (1976). “Application of fuzzy decision techniques Sets andTheirApplications to Cognitive and Decision Pro-
topattern recognitionandcurvefitting,” Ph.D. dissertation, cesses,L. A. Zadeh e t al., Eds. New York: Academic Press,
EECS Dept., Princeton University, Princeton, NJ. pp. 329-352.
(661 Chang, S. K. (1971), “Pictureprocessinggrammarand its 991 Diamond, P. (1975), “Fuzzy chaos,” Dept. Math.,Univ. Queens-
applications,”Znfom. Sci., pp. 121-148. land, Brisbane, Australia.
(671 Chang, S. K. (1971), “AutomatedInterpretationandediting 1’0 0 1 Diday, E. (1972), “Optimisationenclassificationautomatique
of fuzzy line drawings,” in Roc. Spain Joint Computer Conf., et reconnaissance des formes.” Notes Sci.. IRIA-6. Bull. 12.
pp. 393-399. 1011 Dimitrov, V., Wechler, W., h d Barnev, P. (1974), “Op&ized
I681 Chang, S. K. (1971), “Fuzzyprograms-Theoryandapplica- fuzzy control of humanistic systems,” Dept. of Math., Dresden,
tions,”
in Roc. Symp. Computers Automata,Polytechnic Technical Univ. Dresden, East Germany.
Inst. of Brooklyn, pp. 147-161. (1021 Dimitrov, V., Wechler, W., andBarnev, P. (1974), “Optimal
(691 Chang, S. K. (1972), “On theexecutionoffuzzyprograms fuzzycontrolofhumanisticsystems,”Inst. Math. Mech.,
usingfinite-statemachines,” ZEEE T M ~Comput.,
. pp. 241- Bulgarian Acad. Sci., Sofia, Bulgaria.
253. 11031 Dimitrov, V. (1975), “Efficient governing heuristic systems by
(701 Chang, S. S. L. (1969), “Fuzzy dynamic programming and the fuzzyinstructions,”Dept.ofSoftware,Inst.of Math., Bul-
decisionmakingprocess,”in Roc. 3rd. Princeton Con$ Znf. garian Academy of Sciences, Sofia, Bulgaria.
Sci. and Systems, pp.200-203. [ 1 0 4 1 Dravecky, J. and Riecan, B. (1975), “Measurability of functions
(711 Chang, S. S. L. (1972), “Fuzzymathematics, Man, and his with values in partially ordered spaces,” Capopis pro Pestovani
environment,” ZEEE Tmns. Syst.. Man, Cybernetics,SMC-2, Math., pp. 27-359.
pp. 92-93. [ 1 0 5 ) Drosselmeyer, E. and WoMeberger, “Studies on a fuzzy system
[ 7 2 ] Chang, S. S. L. (1972), “On fuzzy mathematics and control.” in the parochialfield,”specialinterestdisc. session on fuzzy
ZEEE Tmns Syst., Man, Cybernetics, SMC-2, pp. 30-34. automata and decision
processes, 6th
IFAC WorldCong.,
1731 Chang, S. S. L. and Zadeh, L. A. (1972), “On fuzzy mapping Boston, MA.
andcontrol,” BEE Tmm. Syst.,Man,Cybernetics,SMC-2, 11061 Dubois,T. (1974), “Unemethoded’waluationparles sous-
pp. 30-34. ensembles flow appliquee a la simulation,” Imago disc., Paper
(741 Chang, S. S. L. (1975), “Onfuzzyalgorithmandmapping,” 13. Univ. Catholique de Louvain,Belgium.
specialinterestdiscussion on fuzzyautomataanddecision I1071 Dunn,J. C. (1974), “Agraphtheoreticanalysisofpattern
processes, 6th IFAC World Cong., Boston, MA. classification via Tamura’s fuzzy relation,” ZEEE T M ~ sSyst., .
[ 7 5 ] Chang, S. S. L. (1975), “On risk anddecisionmakingin a Man, Cybernetics, SMC-4, pp. 310-313.
fuzzyenvironment,”in Fuzzy Sets and TheirApplications 11081 Dunn,J. C. (1973), “Afuzzyrelative of the isodata process
to Cognitive and Decision Processes, L. A. Zadeh et al., Eds. and its use indetectingcompactwell-separatedclusters,”
New York: Academic Press, pp. 219-226. J. Cybernetics, vol. 3, pp. 32-57.
[ 7 6 ] Chapin, E.W. (1971), “An axiomization of the set theory of I1091 Dunn,J. C. (1975), “Somerecentinvestigationsofa new
Zadeh,”NoticesAmer. Math. SOC.,Notice 6 8 7 - 0 2 - 4 , p. 753. fuzzy partioning algorithm and its application to pattern clazsi-
[ 7 7 ] Chapin, E. W. (1974), “Set-valued set theory: Part 1,” Notre fication problems,” J. cybernetics.
Dame Univ., Notre Dame J. Formal Logic, pp. 255-267. 1 lo] Dunn, J. C. (1977), “Well separated clusters and optimal fuzzy
[ 7 8 ] Chapin, E. W. (1975). “Set-valuedset theory: Part 2,” Notre partitions,” 1. Cybernetics, vol. 4, pp. 95-104.
Dame Univ., Notre Dame J. F o m l Logic, pp. 255-267. 1 1 1 1 Dunn, J. C. (1975), “Canonical forms of Tamura’s fuzzy rela-
[ 7 9 ] Chen, C. (1974), “Realizability ofcommunicationsnets: An tion matrix: A scheme forvisualizing cluster hierarchies, Cornell
applicationof the Zadehcriterion,” ZEEE l’kam. Circuits Univ., Ithaca, NY.
Syst., CAS-21,pp. 150-151. I 1121 Dunn, I . C. (1975), “Indices ofpartitionfuzzinesandthe
[SO] Cohen P. J. and Hersh, R. (1967), “Noncantorian set theory,” detection of clusters in large data sets,”special interest disc.
Sci. Amer., pp. 104-1 16. session on fuzzyautomataanddecision processes, 6th IFAC
[ S l ] Cleave, J. P. (1970), “The notion of validity in logical systems World Cong., Boston, MA.
with inexact predicates,” Brit. J. PhiL Sci., pp. 269-274. 1131 Dunst, A. J. (1971), “Application ofthefuzzysettheory,”
[SZ] Cohen, P. J. and Hirsch, R. (1967), “Non-Cantoriansettheory,” location unknown.
Sci. Amer., pp. 101-116. 1141 Endo, Y. andTsukamoto, Y. (1973), “Apportion models of
1831 Colavi, J. P. (1974), “The notion of logical consequence in the tourists by fuzzy integrala,” in Annual Con$ Records SZCE.
logic of inexact predicates.” Zenxhhift fur Moth Logik und 1151 Engel,A. B. and Buonomano, V. (1973), “Towards a general
Grvmddagen der Math., pp. 307-324. theory of fuzzy sets I,”Univ. de Campinas, SP. Brazil.
[ E 4 1 Conche, B. and Courant. P. (1973). “Application des concepts 1161 Engel, A. B. andBuonomano, V. (1973), “Towards a general
flous a la programmation en languagesquasi-natureh,” Semi- theory of fuzzy sets II, Univ. Estadual de Campinaa, SP, Brazil.
naire Bernard Roy, Univ. Paris-Dauphine. 1171 Esogbue. A. M. 0.(1970), “Dynamic programming and fuzzy
[ S S ] Conche, B., Jouault,J. P., andLuan, P. M. (1973), “Appli- allocationprocesses,”Tech. Memo 202, Oper. Res. Dept.,
cationdesconcepts flous alaprogrammationenlanguages Case Western Resene Univ., Cleveland, OH.
quasi-naturels,” Seminaire Bernard Roy, Univ. Puis-Dauphine. I1 181 Esogbue, A. 0.(1975), “On applicationoffuzzyallocation
[ 8 6 ] Conche, B. (1973). “Elements d’une methode de classification theorytothe modelling of cancer research appropriation
par utilisation d’un automate flou,” J.E.E.F.L.N., Univ. Paris- process," Georgia Inst. of Technology, Atlanta, GA.
Dauphine. 191 Fellinger, W. (1974), “Specifications for a fuzzy system model-
[ 8 7 ] Cools, M. (1973). “La semantiquedans les processus didac- ling language,” thesis, Oregon State Univ., Corvaflis, OR.
tiques,” Imago D i s c , pclp., Univ. CathoPquedeLowain, 201 Fine, K. (1975), “Vagueness, truth, and logic,” Synthese,
Bdgim. PP. 2 65- 300.
(881 Cools, M. andPeteau, M. (1973), “Stim 5 : Un programme d e 211 Floyd, R. W. (1967), ‘Won-Deterministicalgorithms,” J. A s r
stimulationinventiveutilisant la theoriedes sous-ensembles Comput. Mach., pp. 636-644.
flow,” Zmqgo Disc. Paper, Univ. Catholique de Loursin, [ 221 Fu, K. S. (1971), “Acritical review of learningcontrol re-
KANDEL AND BYATT: FUZZY SETS, ALGEBRA, AND STATISTICS 1633

search,”inPatternRecognitionandMachineLearning. New 541 Goguen, J . A. (1968), “The logic ofinexactconcepts,”Syn-

York: Plenum Press. these, pp. 325-373.
[123] Fu, K. S. and Li, T. J. (1969), “Formulation of learning auto- 5 5 1 Goguen, J . A. (1969), “Representing inexact concepts,” Z.C.R.
mata and games,”Znform. Sci., pp. 237-256. Quarterly Rep., 2 0 , Inst. for Comput. Res., Univ. of Chicago,
[ 1241 Fu, K. S. (1977), “Pattern recognition andsomesocio-economic Chicago, IL.
problems,” Purdue Univ., West Lafayette, IND. 561 Goguen, J . A. (1970), “Mathematicalrepresentationofhier-
I1251 Fujisake,H. (1971), “Fuzzinessinmedicalsciencesandits archicallyorganized
systems,”
GlobalSystems
Dynamics,
processing,”inProc.Symp. on Fuzziness in Systems and Its E. 0. Attinger, Ed.,pp. 1 1 1-129.
Processing, SICE. 571 Goguen, J . A. (1972), “Hierarchicalinexactdatastructures
[126] Fung,L. W. andFu, K. S. (1974), “TheKthoptimal Policy inartificialintelligenceprobelems,”inProc. 5 t h Hawaii Znt.
algorithm for decision making
in
fuzzy
environments,” in Conf Svstem Sciences, Honolulu, HI, pp. -?A<- .- -7 0.1. .-
Identification andSystemParameterEstimation. New York: [ l S 8 ] Goguen, J . A. (1973), “Somecommentsonapplyingmathe-
North Holland, pp. 1052-1059. matical system theory,” DOC. Computer Science Dept., Univ.
(1271 Fung, L.W. and Fu, K. S. (1974), “Characterisation of a class of Calif., Los Angeles, CA.
offuzzvo~timalcontrol
~ ~~ oroblems.”in Proc. 8thPrinceton [ 1 5 9 ] Goguen, J. A., Thatcher, J., Wagner, E., and Wright, J . (1973),
Conf. Zn> Sb. and Syst. “Ajunctionbetweencomputerscienceandcategorytheory:
[ 1281 Fung, L. W. and Fu, K. S. (1975), “An axiomatic approach t o Basic conceptsandexamples: 11,” Universal Constructions,
rational decision making,” in Fuzzy Sets andTheir Applications IBM Res. Remrt.
to Cognitive and Decision Processes, L. A.Zadeh e t al., US. I1601 Goguen, J. A. (1973). “Systemstheoryconceptsincomputer
New York: Academic Press, pp. 227-256. science,”in Proc. 6thHawaii Znt. Conf. SystemsSciences,
1291 Fung, L. W. and Fu, K. S. (1973), “Decision-making in a fuzzy Honolulu, HI, pp. 77-80.
environment,” Purdue Univ., West Lafayette, IN. [ 1 6 1 ] Goguen, J . A. (1973), “Axioms,extensionsandapplications
1301 Furukawa, M., Nakamura, K., andOda, M. (1973), “Fuzzy forfuzzysets: Languagesand representationonconcepts,”
variant process of memories,” in Ann. Conf. Records of SZCE. Doc. Computer SciencesDept.,Univ. of Calif.,LosAngeles,
1311 Furukawa, M., Nakamura, K., andOda, M. (1972), “Fuzzy CA.
modelofhumandecision-makingprocess,”inAnnual Conf. [ l a 2 1 Goguen, J. A. (1973), “ThefuzzyTychonofftheorem,” J.
Records of JAACE. Math. AnaL Appl., pp. 734-742.
1321 Gaines,R. E. (1975), “Stochastic and fuzzy logics,” Electron. [ 1 6 3 ] Goguen, J . A. (1974), “Concept representation in natural
Lett., pp. 188-189. andartificiallanguages:Axioms,extensionsandapplications
1331 Gaines, B. R. (1975), “Behaviour/structuretransformations for fuzzy sets,”Znt. J. Man-Machine Studies, pp. 5 13- 56 1.
underuncertainty,” EES-MMS-AUT-75, Dept.of Elec.Eng., [la41 Goguen, J. A. (1975), “On fuzzyrobotplanning,”inFuzzy
Univ. of Essex, Colchester, Essex, England. Sets andTheirApplicationsto Cognitive and Decision Pro-
1341 Gaines, R. B. (1975), “Fuzzyandstochasticmetric logics,” cesses, L.A.Zadeh e t al., Eds. New York:Academic Press,
EFS-MMS-FU71-75, Departmentof
Electrical Engineering p ~ 429-448.
.
Science, Univ. of Essex, Colchester, England. [ 1651 Goguen, J. A. (1975), “Complexity of hierarchically organized
1351 Gaines, B. R. (1976), “Fuzzyreasoningandthe logic of un- systemsand the structure of musicalexperiences,”in UCLA
certainty,” inProc. 6th Znt. S y m p . Multiple-ValuedLogic, Quarterly Report, Dept. ofC.S.
Loam. UT. [ 1661 Goodman, J. S. (1974), “From multiple balyage t o fuzzy sets,”
[ 1361 Gahes, B. R. (1976), “A bibliography of fuzzy systems theory Inst. Math., Univ. of Florence, Italy.
and closely related topics,” Dept. of Elec. Eng., Univ. of Essex, [ l a 7 1 Gottinger, H. W. (1975), “A fuzzy algorithmic approach to the
Colchester, Essex, England. definition of complexor imprecise concepts,”Univ. of Bielefeld,
[ 1371 Gaines, B. R. (1976), “Multivalued logics and fuzzy reasoning,” Conf. on Syst. Theory.
workshopondiscretesystemsandfuzzyreasoning,Dept. of [ l a 8 1 Gottinger, H. W. (1973), “Towardafuzzyreasoninginthe
Elec. Eng., Univ. of Essex, Colchester, Essex, England. behavioral science,” Cybernetic& pp. 1 1 3-1 35.
[138] Gaines, B. R. (1976), “Survey offuzzyreasoning,workshop [ 1691 Gottwald, S. (1971), “Zahlbereighskonstruktionen ineiner
on discrete systemsand
reasoning,”
QueenMary College, Mehnvertigen Mengenlehre,” Ztschr. Math. Logik Crdl. Math.,
Dept. of Elec. Eng. pp. 145-188.
[ 1391 Gearing, Ch. E. (1974), “Generalized Bayesian posterior analy- I1701 Gottwald, S. (1974), “Fuzzytopology:Productandquotient
sis with ambiguous information,” Dartmouth College, Hanover, theorems,”J. Math. Anal. Appl., pp. 512-521.
NH. [ 1711 Grattan-Guiness (1976), “Fuzzymembershipmapped onto
[ 1401 Gentilhomme, Y. (1968), “Les ensembles flous en linguistique,” interval and
many-valued quantities,”
EN3 4SF, Enfield,
Cahiers de linguistique theorie et applique, Bucharest, Rumania. England.
[141] Gerhardts, M. D. (1965), “ZurCharakterisierungDistributiver [ 1 7 2 ] Gusev, L. A. and Smimova, I. M. (1973), “Fuzzy sets: Theory
Scheirverdande,”Math. Annalen., pp. 231-246. and applications (a survey),’’ (in Russian), Avtomatika Z Tele-
[142] Gerhardts, M. D. (1969), “SchragverdandeundQuasiordnun- mekhanika, pp. 66-85.
gen,” Math. Annalen., pp. 65-73. I1731 G w v , L.A.and Smimova, J . M. (1973), “Simulationofbe-
I1431 Gluss, B. (1973), “Fuzzy multistagedecision
making and haviorandintelligence,”translatedfrom: A u t o m a t h Z Tele-
terminal regulators and their relationship t o n o fuzzy quadratic rnakhanika, pp. 66-85.
state and terminal regulators,”Znt. J. Contr., pp. 177-192. [ 1 7 4 ] Hamacher, H. (1975), “Uber LogischeVerknupfungenUn-
(1441 Georgescu,G. (1971), “Les algebres de lukasiewicz o - d e n t s , scharfer Ausagen und DerenZugehorigeBewertungsfunk-
logique,automatique,informatique,”Ed. Acad.
Rep.Soc. tionen,” 75/14, Lehrstuhl f i r Unternehmensforschung,Rwth
de Roumanie, pp. 99-176. Aachen.
[1451 Giles, F. (1973), “A logic for subjective belief,” in R o c . Conf. (1751 Harris, J . I. (1974), “Fuzzyimplicationcommentsonacon-
on Found. of Probability and Statistics and Statistical Theories ceptofZadeh,” DOAEResearch workingpaper cm., Min.
of Science, London, Ontario, Canada. of Defence, Byfleet, Surrey.
[ 146) Giles, R. (1974), “Lukasiewicz’slogicand fuzzysettheory,” [ 1761 Harris, J . I. (1974), “Fuzzy sets: How to be imprecise precisely,
Queen’s Univ., Math. Preprint 1974-29, Kingston, Canada. DOAE research working paper cm5., Min. of Defence, Byfleet,
I1471 Giles,R. (1974), “Anonclassicallogic for physics,”Sfudia Surrey.
Logica. (1771 Hendry, W. L. (1972), “Fuzzysetsand Russell’s paradox,”
I1481 Gitman, I. (1970), “Organization of data: A model and com- L o s Alamos Sci. Lab.,Univ. of Calif., Los Alamos, NM.
putationalalgorithmthat uses thenotationoffuzzysets,” I1781 Henry-Labordere, A. and deBacker, (1972), “Intelligence
Ph.D. dissertation, McGill Univ., Montreal, Canada. creatrice:Automatisationde proccuus d’apsociation,” Revue
(1491 Gitman, I. andLevine, M. D. (1970), “An algorithmforde- Metra Znt. Rapport, 69.
tectine unimodal fuzzy sets and an aoolication as a clustering [1791 Hirai, H., Asai, K., and Katajima, S. (1968), “Fuzzy automaton
and its application t o learning control systems,” memo fac. of
lSOj
techniiue,” ZEEE T r a k . Cornput., pp.583-593.
Gluss, B. (1973), “Fuzzymultistagedecision-making;Fuzzy ~-
ene.. Osaka Citv Univ.. bo. 67-73.
I ~ ~~

[ 180 Honda, N. ( 1 9 i l ) , “Fuzzy sets,” 1. ofZECE, pp. 1359-1363.

states and terminalregulatorsandtheirrelationship tonon- [ 181 Honda, N. andNasu, M. (1975), “Recognitionoffuzzylan-
fuzzyquadraticstageandterminalregulators,” Znt. J . Contr. guages,”in Fuzzy Sets and TheirApplications to Cognitive
(United Kingdom), pp. 177-192. andDecidon Processes, L. A. Zadeh et d.,Eds. New York:
151 ] Goguen, J. A. (1967), “L-fuzzy sets,” I. Math.Anal.Appl., Academic Press, pp. 279-300.
pp. 145-174. [ 182 Honda, N. andNasu, M. (1975), ‘IF-recognitionoffuzzy
1521 Goguen, J. A. (1968). “CategoriesofV-sets,”Univ.ofCali- languages,”special interest discussionsession o n fuzzyauto-
fornia, Berkeley, CA. mataanddecision processes, 6th IFACWorld Cong., Boston,
1531 Goguen, J . A. (1968), “Categories of fuzzy sets: Applications MA.
ofnon-cantorianset theory,.’ FkD.dissertation, Univ. of [ 183 Horejs, J. (1965), “Classificationsand theirrelationship t o a
California, Berkeley, CA. measure,” Publications of Faculty of S d m c e , University of
1634 PROCEEDINGS O F THE IEEE, VOL. 6 6 , NO. 12, DECEMBER 1978

Purkyne (Bmo), no. 168, pp. 475-493. and Technology, Socorro, NM.
[ 184) Hormann, A. M. (1971), “Machine-aided value judgments using I2151 Kandel, A. (1974), “Generation of the setrepresenting all
fuzzysettechniques,” SP-3590, System Dev. Corp.,Santa fuzzy prime implicants,” CSR 136, Dept. of Computer Science,
Monica, CA. New Mexico Institute Mining and Technology, Socorro, NM.
[ 1 8 5 ] Hughes, G. E.andCreswell, M. J. (1968), “An introduction [ 2 1 6 ] Kandel, A. (1974), “On the minimizationofincompletely
to modal logic,” London: Methuen. specified fuzzy functions,” inform. Cone., pp. 141-153.
[ 1861 Hughes, J. S. andKandel, A. (1975), “Applicationsoffuzzy [ 2 1 7 ] Kandel, A. andYelowitz,L. (1974), “Fuzzychains,” ZEEE
algebra to hazard detection in combinatorial switching circuits,” Trans. Syst., Man, Cybernetics, pp. 472-475.
Znt. J . Compuf. Inform. Sci.,to appear. [ 2 1 8 ] Kandel, A. (1974), “Codes over languages,” ZEEE Trans. Syst.,
(1871 Hung,N.T. (1975), “Informationfontionelleetensembles Man, Cybernetics, SMC-4, pp. 125-138.
flous,” seminar o n questionnaires, Univ. of Paris 6, Paris, [ 2 1 9 ] Kandel, A. (1974), “Fuzzy representation, CNF minimization,
France. andtheir
application tofuzzytransmitionstructures,” in
[I881 Hutton, B. andReilly, J. L. (1974), “Separationaxiomsin Proc. 1974 Znt. Symp.Multiple-valued Logic, Morgantown,
fuzzy topological spaces,” Univ. of Auckland, New Zealand. WV, pp. 361-379.
[ 1 8 9 ] Hutton, B. (1974), “Normalityinfuzzytopologicalspaces,” [ 2 2 0 ] Kandel, A. (1974), “Synthesisoffuzzy logic withanalog
Dept. of Math., Univ. of Auckland, New Zealand. modules:Preliminarydevelopments,” Comput.Educ. Trans.
[ 1 9 0 ] Inagaki, Y. and Fukumura, T. (1975), “On the descriptionof (ASEE), Pp. 71-79.
fuzzymeaning of context-free language,” in Fuzzy Sefs and I2211 Kandel, A. (1974), “Application of fuzzy logic to the detection
Their Applications t o Cognitive and Decision Processes, L. A. ofstatichazardsincombinationalswitchingsystems,” Znt.
Zadeh e t al., Eds. New York: Academic Press, pp. 301-328. J. CompuferandZnform. Sci.,pp. 129-139.
(1911 Ishikawa,A.andMirno,H. (1975). “Design of a video infor- [ 2 2 2 ] Kandel, A. (1974), “On theenumerationoffuzzyfunctions,
mation system and the fuzzy infohation theory,” Eurocomp 12th Holiday Symp.on
Developments in
Combinatorics,”
75, Brunel Univ., U.K. New Mexico State Univ., Las Cruces, NM.
1921 Jacobson, D. H. (1975), “Onfuzzygoalsandmaximizing [ 2 2 3 ] Kandel, A. (1974), “On the propertiesof‘fuzzyswitching
decisions in stochastic optimal control,” CSID Special Report, functions,” J. Cybernetics, pp. 119-126.
National Research Inst. for Math. Sci., Pretoria. [ 2 2 4 ] Kandel, A. (1975), “A noteonthe simplification of fuzzy
1931 Jain, R. (1975), “Outlineof anapproach for the analysis of switchingfunctions,” CSR 139, Dept.ofComputerScience,
fuzzysystems,”specialinterest discussionsession onfuzzy New Mexico Institute Mining and Technology, Socorro, NM.
automata and decision
processes,
6thIFAC World Cong., [ 2 2 5 ] Kandel, A. (1975), “Fuzzyhierarchicalclassificationofdy-
Boston, MA. namicpatterns,” NATO A.S.I. onpatternrecognitionand
1941 Jain,R. (1976), “Convolutionoffuzzyvariables,”JIETE, classification, France.
22. [ 2 2 6 ] Kandel, A. (1975), “Block decomposition of imprecise models,”
[ 1 9 5 ] Jain, R. (1976), “Decision making with fuzzy knowledge about in Con5 Record 9th ASILOMAR Con5 Circuits, Systems, and
the state of the system,” Nation System Conference,Elec. Eng. Computers, Pacific Grove, CA.
Dept., Univ. of Roorkee, India. [ 2 2 7 ] Kandel, A. (1975), “Propertiesoffuzzymatricesandtheir
[ 1 9 6 ] Jakubowski, R. andKasprzak, A. (1973), “Applicationof applications to hierarchicalstructures,” in Con5Record 9th
fuzzy programs to the design of machining technology,” Bull. ASZLOMARConf.Circuits,Systems,andComputers, Pacific
Polish Acad. Sci., pp. 17-22. Grove, CA.
[ 1 9 7 ] Jarvis, P. A. (1975), “Optimizationstrategiesinadaptivecon- [ 2 2 8 ] Kandel, A. (1976), “Fuzzysystemsandtheirapplications to
trol: A selective survey,” IEEE Trans. Syst., Man, Cybernetics, simulations,” in Proc. 9 f h Hawaii Znf. Con5 System Sciences,
SMC-5, pp. 83-94. Honolulu, HI.
(1981 Jouault, J. P. andPhamMinhLuan (1973), “Applicationdes [ 2 2 9 ] Kandel, A. (1976), “Inexactswitchinglogic,” ZEEE Trans.
concepts flous a la programmation en langages qusi-naturels,” Syst., Man, Cybernetics, SMC - 6, pp. 2 1 5- 2 19.
Inst. d’hformatique c’Enterprise, CNAM, Paris, France. [ 2 3 0 ] Kandel, A. (1976), “On the decomposition of fuzzy functions,”
[ 1 9 9 ] Kahne, S. (1975), “A procedureforoptimizingdevelopment ZEEE Trans. Computers, to appear.
decisions,” Aufomatica, pp. 261-269. [ 2 3 1 ] Kandel, A. (1976), “Fuzzy maps and their applications in the
[ 2 0 0 ] Kalmanson, D. (1973), “Recherche cardio-vasculaire et theorie simplification of fuzzy switching functions,” in Proc. 6th Int.
des ensembles flous,” La Nouvelle Presse Medicale, pp. 2757- Symp. Multiple-Valued Logic,Logan, UT.
2760. [ 2 3 2 ] Kandel, A. andDavis,H. A. (1976), “The fmt fuzzydecade
[ 2 0 1 ] Kandel, A. (1972), “Toward simplification of fuzzy functions,” (bibliography on fuzzysetsandtheirapplications),”Dept.
CSR 114, New Mexico Ins. Mining and Technology, Socorro, ofComputerScience,CSR 140, NewMexico Instituteof
NM. Mining and Technology, Socorro, NM.
(2021 Kandel, A. (1972), “On coded grammars and fuzzy structures,” (2331 Kaufmann, A. (1973), “Introductiona la theoriedes sous-
CSR 118, New Mexico Inst. Mining and Technology, Socorro, ensembles flous (a I’usage des igenieurs), Tome 1: Elements
NM . theoriques de base,” Masson et Cie,Paris.(Englishtranslation
12031 Kandel, A. (1973), “Fuzzy chains: A new concept in decision- published by Academic Press, New York, 1975.)
making under uncertainty,” CSR 123, New Mexico Inst. Mining [ 2 3 4 ] Kaufmann, A. (1975), “Introductiona la theoriedes sous-
and Technology, Socorro, NM. ensembles flow (a l’usage desingenieurs),Tome 2 : Appli-
[ 2 0 4 ] Kandel, A. (1973), “A new method for generating fuzzy prime cations a la linguistique, a la logique et a la semantique,” Masson
implicantsand an algorithmfortheautomaticminimization et Cie, Paris.
of inexact structures,” CSR 126, Dept. of Computer Science, I2351 Kaufmann, A. (1975), “Introductiona la theoriedes sous-
New Mexico Inst. Mining and Technology, Socorro, NM. ensembles flous (a I’usage desingenieurs),Tome 3: Appli-
I2051 Kandel, A. (1973), “A newalgorithmforminimizingincom- cations a la classification et la reconnaissance des formes, aux
pletelyspecified
fuzzy functions,” CSR 127, New Mexico automatesetauxsystemes,auxchoix des critares,”Masson
Inst. Mining and Technology, Socorro, NM. et Cie, Paris, (in preparation).
(2061 Kandel, A. andObenauf, T.A. (1973), “Onfuzzylattices,” [ 2 3 6 ] Kaufmann, A., Cools, M., and Dubois, T. (1973), “Stimulation
CSR 128, New Mexico Inst. Mining and Technology, Socorro, inventive dans un dialogue homme-machine utilisantla methode
NM. des morphologies et la theorie des sous-ensemble flous,” Imago
[ 2 0 7 ] Kandel, A. (1973), “On the analysis of fuzzy logic,”in Proc. Disc. Paper LDP-6, Centre Interfacultaire Imago, Univ. Catho-
6th Znt. Con5 System Sciences, Honolulu, HI. lique de Louvain,Heverlee, Belgium.
[ 2 0 8 ] Kandel, A. (1973), “Comment on an algorithm that generates [ 2 3 7 ] Kaufmann, A., Cools, M., andDubois, T. (1975), “Exercices
fuzzyprimeimplicantsby LeeandChang,” Inform. C o n e . , avec solutions sur la theorie des sous-ensembles flous,” Masson
pp. 279-282. et Cie, Paris, (in preparation).
[ 2 0 9 ] Kandel, A. (1973), “Fuzzyfunctionsandtheirapplication to [ 2 3 8 ] Kaufmann, A. (1975), “Introduction to a fuzzy theory of the
the analysisofswitchinghazards,”in R o c . 2ndTexasConf. human operator,” special interest discussion o n fuzzy automata
Computing Systems,Austin, TX, pp. 421t-4216. and decision processes, 6th IFAC World Cong., Boston, MA.
[ 2 1 0 ] Kandel, A. (1973), “Onminimizationoffuzzyfunctions,” I2391 Kaufmann, A. (1975), “A surveyoffuzzysetstheoryand
ZEEE Trans. Compuf., C-22, pp. 826-832. applications t o languages,” Automata and Algorithms, US-
[ Z l l ] Kandel, A. (1973), “Comments on ‘‘Minimization offuzzy Japan seminar on fuzzy sets and their applications, Berkeley,
functions,”ZEEE Trans. Cornput., (2-22, p. 217. CA. (originally scheduled but not presented)
[ 2 1 2 ] Kandel, A. (1974), “On fuzzymaps:Someinitialthoughts,” [ 2 4 0 ] Khatchadourian, H. (1965), “Vagueness, meaningandabsur-
CSR 131, Dept. ofComputerScience, NewMexico Institute dity,” Amer. Philosoph. Quar., vol. 2 , pp. 119-129.
Mining and Technology, Socorro, NM. I2411 Kickert, W. J. M. and van Nauta Lemke, H. R. (1975), “Appli-
[ 2 1 3 ] Kandel, A. (1974), “Simple disjunctive
decomposition of cationofafuzzycontroller in awarmwaterplant,” Delft
fuzzy functions,” CSR 132, Dept. Comput. Sci.,NewMexico Univ. of Technology, Dept. ofElec. Eng.,Control Lab., Holland.
Institute Mining and Technology, Socorro, NM. I2421 Killing, E. (1973), “Fuzzyplanner,”Tech.Report 168, Com-
[ 2 1 4 ] Kandel, A. (1974), “On thetheoryoffuzzymatrices,” CSR puter Science Dept.,Univ. of Wisconsin, Madison, WI.
135, Dept. of Computer Science, New Mexico Institute Mining 12431 Kim, H.H., Mizumoto, M., Toyoda, J., and Tanaka, K. (1974),
AND KANDEL 1635

“Lattice grammars,” Trans. IECE, 57-D. I2751 Kokawa, M., Nakamura, K., andOda, M. (1975), “Experi-
I2441 King, P. J.andMamdani, E. H. (1975), “Theappticationof mental approach to fuzzy simulation of memorizing, forgetting
fuzzy control systems t o industrial processes,” special interest and inference process,” in Fuzzy Sera and Their Applications
discussionsession on fuzzy automata and decisionprocesses, to Cognitive and Decision Processes, L.A.Zadeh et al., Eds.
6th IFAC World Cong., Boston, MA. New York: Academic Res,pp. 409-428.
(2451 f i e , V.A. and Osis, J. I. (1969), “Search methods for estab- (2761 Kokawa, M., Oda, M., and Nakamura, K. (1975), “Fuzzy theo-
lishina o f maximalSeDarabilitv of fuzzy sets.” Kibernetika I reticaldimensionalityreductionmethodofmulti-dimensional
Diag~om’ka,pp. 79-88. quantity,” 6rt, Triennial IFAC WorId Cong., Boston, MA.
[246 Kitagawa, T. (1973), “Three coordinate systems for informa- I277 Kotoh, K. andHiramatsu, K. (1973), “A representationof
tion science approaches,” Inform. Sci., pp. 159-169. pattern classesusing the fuzzy sets,” Tram.IECE, pp. 275-
[247 Kitagawa, T. (1973), “Biorobotsforsimulationstudiesof 282.
learning and intelligent controls,” US-Japan Seminar on Learn- [278 Lake,J. (1974), “Sets,fuzzysets,multi-setsandfunctions,”
ing Control and Intelligent Control, Gainesville, FL. Dept.of Math.,Polytechnicof theSouth Bank,Borough
[248 Kitagawa, T. (1975), “Fuzzinessininformative logics,” in Road, London, England.
Fuzzy Sets and Thew Applications to Cognitive and Decision [279 Lake, J. (1974), “Fuzzy sets and bald men,” Dept. of Math.,
Processes, L.A. Zadeh e t al., Ed. New York: Academic Press, Polytechnicof theSouth Bank,
Borough Road,London,
pp. 97-124. England.
( 2 4 9 I Kitajima, S. and Asai, K. (1972). “Learningmodeloffuzzy [ 280 Lakshmivarahan, S. andRajasethupathy, K. (0000), ‘‘Con-
automation with state-dependent output (3),” in Annual siderationsforfuzzifvinpformal languages andsvsthesisof
Joinr Conx Recordsof JAACE. fuzzy grammars,” Indian f&t.of Technology, Madras, India.
[ 2 5 0 ] Kitajima,-S. and A&, K. (1970), “Learning controls by fuzzy [ 2 8 l ] Lakoff, G. (1973), “Hedges: Astudyinmeaningcriteriaand
automata,” J. JAACE, pp. 551-559. the logic of fuzzy concepts,” J. of Philosophical Logic, pp. 458-
I2511 Kitajima, S. and Asai, K. (1973), “A method of learning control 508.
varying search domain by fuzzy automata,” Japan-US Seminar, [ 2 8 2 ] Lee,E.T.andZadeh, L.A. (1969), “Note on the fuzzy lan-
FL . guages,” Inform. Sa..,pp. 42 1-434.
[ 2 5 2 ] Klaua, D. (1965), “Uber Einen Ansatz zur Mehrwertigen Men- I2831 Lee, E. T. andZadeh,L. A. (1970), “Fuzzy languagesand
geniehre,” Monatsb.Deutsch.Akad. Wiss. Berlin, pp. 859- theiracceptancebyautomata,” in R o c . 4 t h PrincetonConf.
867. In$ Sci. Systems, p. 399.
[ 2 5 3 ] Klaua, D. (1966), “UberEinenZweitenAnsatzzurMehrwer- (2841 Lee, E. T. (1972), “Fuzzylanguagesand theirrelationto
tigen Mengenlehre,” MoMtsb. Deutsch. Akad. Wlrs. Berlin. automata,” thesis, Dept.-of Elec. Eng. and Computer Science,
[ 2 5 4 ] Klaua, D. (1966), “GrundbegriffeEinerMehrwertigenMen- Univ. of Calif., Berkeley, CA.
genlehre,”Mber. Dr. Akad. Wh.,pp. 782-802. [ 2 8 5 ] Lee,E.T. (1972), “Proximitymeasuresforthe classification
[2SS] Klaua, D. (1967), “Ein Ansatz m r Mehrwertigen Mengenlehre,” of geometric figures,” 1. of Cybernetics,pp. 43-59.
Math. Nachr., vol. 33, pp. 273-296. [ 2 8 6 ] Lee, E. T. (1974), “The shape-oriented dissimilarity of polygons
[ 2 5 6 ] Klaua, D. (1967), “Einbettung der KlassischenMengenlehrein and its application to the classification of chromosome images,”
die Mehrwertige,” MOMtSb. Deursch. Akad Wisp. Berlin, vol. 9 , Pattern Recognition, pp. 47-60.
pp. 258-272. [ 2 8 7 ] Lee, E. T. (1974), “An apptication of fuzzy sets to the classi-
I2571 Klaua, D. (1970), “Stetige Gleichmachtigkeiten Kontinuierlich- ficationofgeometricfiguresandchromosomeimages,” US-
Wertiger Mengen,” Nber. Dr. Akad.,vol. 12, pp. 749-758. Japan seminar on fuzzy sets and their applications, Berkeley,
[ 2 5 8 ] Klaua, D. (1970), “Stetige Gleichmachtigketten Continuierlich- CA.
Wertiger Mengen,”Mber. Dr. Akad. Wisp., pp. 749-758. [ 2 8 8 ] Lee, E. T. (1975), “Shape-oriented chromosome classification,”
(2591 Kling,R. (1973), “Fuzzyplanner,”Computer ScienceDept., IEEE Trans. Syst., Man, Cybernetics, SMC-5, pp. 629-632.
Tech. report 168, Univ. of Wisconsin. I2891 Lee,E.T. (1976), “Shapeorientedstorageand retieval of
I2601 Kling, R. (1973), “Fuzzyplanner:Reasoningwithinexact geometricfiguresandchromosome images,” Inform. Process.
conceptsinaproceduralproblem-solving language,” J.Cy- M a ~ g .(to
, appear).
bernetics, pp. 1-16. [ 2 9 0 ] Lee,R.T. C. andChang, C. L. (1971), “Somepropertiesof
I2611 Kling,R. (1973), “Fuzzyplanner:Computationalinexactness fuzzy logic,” Inform. Co?!?rol, pp. 417-431.
inaproceduralproblemsolvinglanguage,”Tech.report 168, [ 2 9 l ] Lee,R. T.C. (1972), Fuzzylogicand the resolutionprin-
Univ. Wisconsin, Dept. of Computer Science. ciple,” J. ACM, pp. 109-119.
[ 2 6 2 ] Klir, G. J. (1971), “On universal logic primitives,” IEEE Trans. I2921 Lee, S. C. andLee,E. T. (1970), “Fuzzyneuronsandauto-
Comput., C-20, pp. 467-469. mata,”in Proc. 4th Princeton Conf. Inf. Sci.Systems, pp.
I2631 Klir, G. J., “Processing of fuzzy activities of neutral systems,” 381-385.
State Univ. of NY. at Binghamton, Binghamton, NY. [ 2 9 3 ] Lee, S. C. (1974), “Fuzzysetsandneuralnetworks,” US-
[ 2 6 4 ] Knopfmacher, J . (1975), “On measures of fuzziness,” J. Math. Japan seminar onfuzzysetsandtheirapplications,Berkeley,
A n d A p p l . , pp. 529-534. CA.
I2651 Kochen, M. andDreyfuss-Raimi, G. (1974), “Onthepsycho- [ 2 9 4 ] Leenders, J . H. (1974), “VageVerzamelingen:Eenkritische
linguisticrealityoffuzzysets:Effectofcontextandset,” Benadering,kwartaalschriftwetenschappelijonderwijs,”Lim-
Univ. of Mich.,MentalHealthResearchInst.,Ann Arbor, burg, Belgium, pp. 441-455.
MI. [ 2 9 5 ] Lefaivre, R. (1974), “The representation of fuzzy knowledge,”
[ 2 6 6 ] Kochen, M. (1974), “On the precision of adjectives which de- US-Japan seminar on fuzzy sets and their application, Berkeley,
note fuzzy sets,” Mental Health Research Inst., Univ. of Mich., CA.
Ann Arbor, MI. I2961 Lefaivre, R. (1974), “Fuzzy:Aprogramming language for
I2671 Kochen, N. andDardre, A. N. (1974), “On the precision of fuzzy problem solving,” Tech. report 202, Dept. of Computer
adjectiveswhichdenotefuzzysets,” J.Cybernetics, pp. 49- Science, Univ. Wisconsin.
59. [ 2 9 7 ] Lefaivre,R. (1974), “Fuzzyproblem solving,”Tech. report
[ 2 6 8 ] Kochen, M. (1975), “Applications of fuzzy sets in psychology,” 37, Madison Acad. Computer Center, Univ. of Wisconsin.
in Fuzzy Sets and Their Applications to Cognitive andDecision I2981 Lewis, B. (1976), “Ordinarylanguagealgorithmsandcopying
Processes, L. A. Zadeh er al., Eds. New York: Academic Ress, with fuzziness in everyday life,” Workshop on discrete systems
pp. 395-408. and fuzzy reasoning, Queen Mary College, Dept. of Elec. Eng.
[ 2 6 9 ] Koczy, L. T. (1975), “R-fuzzy algebra as a generalized formu- [ 2 9 9 ] Lientz, B.P. (1972), “On time dependent fuzzy sets,” Inform.
lation of the intuitive logic,” Dept. of Process Control, Tech- Sci., pp. 367-376.
nical Univ., Budapest, Hungary. I3001 Loginov, V. I. (1966), “Probabilitytreatment ofZadehmem-
I2701 Koczy, L. T. and Hainal, M. (1975), “A new fuzzy calculus and bershipfunctions and their usein patternrecognition,” Eng.
itsapplication as a patternrecognitiontechnique,”in Proc. Cybernetics, pp. 68-69.
3rd Int. Cong. Cybernetics and Systems,Bucharest, Rumania. [ 3 0 1 ] Lombaerde, J. (1974). “Measuresd’entropieentheoriedes
I2711 Kokawa, M., Nakamura, K., and Oda, M. (1975), “Hint effect sous-ensembles flous,” Imagodisc.pap. IDP-12, Centre In-
and a jump of logic in a decision process,” Trans. IECE Japan, terfacultaireImago, Univ. CatholiquedeLouvain, Heverlee,
58-d. Belgium.
[ 2 7 2 ] Kokawa, M., Nakamura, K., andOda, M . (1973), “Fuzzy [ 3 0 2 ] Longo, G. (1975), “Fuzzysets,graphsandsourcecoding,”
expressionof
human experience-to-memory process,”
re- in NewDirections in Signal ProcessinginCommunicarions
search reports of the Automatic Control Lab., Nagoya Univ., and Control, J . K. Swirzynski,Noordhoff-Leyden, pp. 27-
Nagoya, Japan. 33.
[ 2 7 3 ] Kokawa, M., Nakamura, K., and Oda, M. (1974). “Fuzzy I3031 Lowen, M. R. (1974), “A theory of fuzzy topologies,” thesis,
theoreticalapproachestoforgetting processes on inference,” Dept. of Math., Free Univ. o f Brussels, Belgium.
Nagoya Univ., Nagoya, Japan. I3041 Lowen, P. (1974), “Convergence flous,” C.P. Acad.desSci.,
I2741 Kokawa, M., Nakamura, K., and Oda, M. (1974), “Fuzzy A-925.
theoreticaland
concept formational
approach t o memory [ 3 0 5 ] Lowen, M. R. (1975), “ropologiegenerale,” C.R.Acad.Sc.
and inference experiments,” IECE Jap., 57-D, pp. 487-493. Paris T.278 Serie A 1181.
KANDEL
ALGEBRA,
AND BYATT:
SETS, FUZZY STATISTICS
AND 1637

fuzzylinedrawingsforpicturedescription,” Trans. IECE, Their Applications to Cognitive and Decision Processes, L. A.

vOI. 57-A, pp. 216-223. Zadeh et al., Eds. New York: Academic Press, pp. 77-96.
[373] Okuda, T., Tanaka, H., and Asai, K. (1974). “Decision-making [405] Ruspini, E. H. (1969), “A new approach t o clustering,” Inform.
and information in fuzzy events,” Bul. Univ. Osaka Prefecture, and Control, pp. 22-32.
series A, Sakai, Osakai, Japan. [406] Ruspini,E. H. (1970), ‘Wumericalmethodsforfuzzycluster-
[374] Osis, J.J. (1968), “Faultdetection in complexsystems using ing,”Znform. Sci., pp. 319-350.
theory of fuzzy sets,”Kibernetika I Diagnostika, pp. 13-18. [407] Ruspini, E. H. (1972), “Optimization in sampledescriptions:
(3751 Otsuki, S. (1970), “Amodelforlearningandrecognizing Datareductionandpatternrecognition using fuzzyclustering
machine,” Inform. Process., pp. 664-671. (abstract),” IEEE Trans. Syst., Man, Cybernetics, “-2,
[376] Pask, G. (1976), “Comments on fuzzy relations in representing p. 541.
cognitive structures,” workshop on discrete systems and fuzzy [408] Ruspini, E. H. (1973), “Newexperimentalresults in fuzzy
reasoning, Queen Mary College, Dept. of Elec. Eng. clustering,” Inform. Sci., pp. 273-284.
13771 Paz,A. (1967), “Fuzzystarfunctions,probabilisticautomata [409] Ruspini, E. H. (1973), “A fastmethodforprobabilisticand
andtheirapproximationbynonprobabilisticautomata,” J. fuzzyclusteranalysis using associationmeasures,”Space Bio.
Comput. Syst. Sci.,pp. 371-390. Lab., Brain Res. Inst., Univ. of Calif., Los Angeles, CA.
13781 Peschel, M. (1975), “Some remarks t o “fuzzysystems” as a [410] Russell, 8. (1923). “Vagueness,” Austmlian J. Phil., pp. 84-
complement to the topicpaper from L. A. Zadeh,” Berlin. 92.
13791 Pinkava, G. (1976). “Aspects of multivalued logics,” workshop [411] Rutherford, D. A. andBloore,G.C. (1975), “Theimplemen-
on discrete systems and fuzzy reasoning,” Queen Mary College, tation
of
fuzzyalgorithmsfor
control,”
Control
Systems
Dept. of Elec. Eng. Centre, Univ. ManchesterInst. Sci. and Tech., Manchester,
[380] Post, E. L. (1921), “Introduction t o a general theoryof ele- U.K.
mentary propositions,” Amer. J. Math., pp. 163-185. [412] Rutherford, D. A. (1976), “Implementationandevaluation
[381] Poston, T. (1971),“Fuzzy geometry,” thesis, Univ. of Warwick, of a fuzzy control algorithm in a sinter plant,” workshop on
England. discretesystemsandfuzzyreasoning,Queen Mary College,
[382] Pratapa Reddy, V. C. V.andAmeling, W. (1974), “Electronic Dept. of Elec. Eng.
implementationoffuzzy logic,” submittedforpublication [413] Sanchez, E. (1974), “Equationsderelations flous,” thesede
1974. doctoratenbiolosehumaine,Facultede Medecine de Mar-
I3831 Preparata,F. P. andYeh, R. T. (1970), “Atheoryofcon- seille, Marseille, France.
tinuously valued logic,” Tech. report 89, Univ. of Texas, Austin, [414] Sanchez, E. (1976), “Resolutionofcompositefuzzyrelation
TX. equations,” Inform. Contr., vol. 30,pp. 38-48.
13841 Preparata, F. P. and Yeh, R. T. (1971), “On a theory of con- [415] Sanford, D. H. (1975), “Infinity and vagueness,” Phil. Review,
tinuouslyvaluedlogic,”in Con$ R e f . 1971 Symp.Theory pp. 520-535.
andApplicationsofMultiple-Valued Logic Design, pp. 124- [416] Sanford, D. H. (1975), “Borderline logic,” Amer. Phil. Quart.,
132. pp. 29-39.
[385] Preparata, F. P. and Yeh, R. T. (1972), “Continuouslyvalued (4171 Santos, E. S. (1968), “Maximin automata,” Inform.Contr.,
logic,” J. Comput. Syst. Sci.,pp. 397-418. pp. 363-377.
[386] Prugovecki, E. (1974), “Fuzzy sets in the theory of measure- I4181 Santos,E. S. (1969), “Maximin sequential-likemachinesand
mentofincompatible observables,” Found. Phys., vol. 4, chains,” Math. Syst. Theory, 3,pp. 300-309.
Pp. 9-18. [419] Santos, E. S. and Wee, N.G. (1970), “Generalformaulation
[387] Prugovecki, E. (1973), “A posulational framework for theories of sequential machines,” Inform. Contr., pp. 5-10.
of simultaneous measurement of several observables,” Found. I4201 Santos, E. S. (1970), “Fuzzyalgorithms,” Inform.Contr.,
PhyS., pp. 3-1 1. Pp. 326-339.
I3881 Prugovecki, E. (1974), “Measurementinquantummechanics [421] Santos, E. S. (1972), “On reductionofmaximinmachines,”
as a stochasticprocess on spacesoffuzzyevents,”Dept.of J. Math. Anal, Appl.
Math., Univ. of Toronto, Canada. [422] Santos, E. S. (1974), “Fuzzyalgorithms,” Inform.Contr.,
[389] Pun, L. (1975), “Experience in the use offuzzyformalismin pp. 1-1 1.
problems with various degrees of subjectivity,” special interest [423] Santos,E. S. (1974), “Fuzzyautomataand languages,” US-
discussion session on fuzzyautomataanddecision processes, Japan seminar o n fuzzy sets and their applications, Berkeley,
6th IFAC World Cong., Boston, MA. CA.
[390] Pun, L., Doumiengts, G., “Fuzzydecisionprocessfor solving [424] Santos, E. W. (1975), “Realizationoffuzzy languages by
production management problems in t h e industry of mechanical probabilistic,max-productandmaximautomata,” Inform.
construction,” 3rd Int. Conf. Prod. Res., Amherst, MA. Sci., pp. 39-53.
[391] Ragade, R. K. (1973), “A multiattribute perception and classi- [425] Santos, E. S. (1975), “Fuzzyprograms,” 6th IFAC World
ficationof visual similarities,” SystemsRes. and Planning Cong., Boston, MA.
Papers, S-001-73,Bell-Northem Research. [426] Santos, E. W. (1975), “Max-product grammars and languages,”
(392] Ragade, R. K. (1973), “On some aspects of fuzziness in com- Inform. Sci., pp. 1-23.
munication, 1. Fuzzy entropies,”
Bell-NorthernResearch, [427] Santos, E. W. (1976). “Fuzzysequentialfunctions,” J. Cy-
W-002-73,Ottawa, Canada. bernetics, t o appear.
[393] Ragade, R. K. (1973), “On some aspects of fuzziness in com- [428] Saridis, G. N. (1975), “Fuzzydecisionmaking in prosthetic
munication, 3. Fuzzy concept communication,” Bell-Northern devices and other
applications,” special interest discussion
Research, Ottawa, Canada. session onfuzzyautomataanddecision processes, 6thIFAC
I3941 Ragade,R. K. (1974), “A note on fuzzy information process- World Cong., Boston, MA.
ing,” Bell-Northern Research, Ottawa, Canada. [429] Saridis, G. N. (1975), “Fuzzynotions in nonlinearsystem
13951 Ragade,R. K. (1975), “Profiletransformation in groupsand classification,” Purdue Univ., West Lafayette, IN.
consensus formation by fuzzy sets, 3. Fuzzy concept communi- (4301 Sasama, H. (1975), “Fuzzysetmodelfortraincomposition
cation,” Univ. of Waterloo, Waterloo, Canada. in marshallingyard,”Report 1, workinggrouponfuzzy
~~ ~
sys-
[ 396‘I Rajasethupathy, K. S. andLakshmivarahan, S. (1974), “Con- tems, Tokyo,Japan, pp. 49-54.
nectness in fuzzytopology,”Dept.of Math., Vivekanamdha [431 Schimura, M. (1973), “Fuzzysetsconceptinrank-ordering
College, Madras, India. objects,”/. Math. Anal. Appl., pp. 717-733.
[ 397 Rajeck, R. K. (1975), “Benefitcostanalysisunderimprecise [432 Schutzenberger, M. P. (1962), “On a theoryofP.Jungen,”
conditions,” Univ. of Waterloo, Waterloo, Ontario. Proc. Amer. Math. SOC.,pp. 885-890.
[ 398 Reisinger, L. (1974), “On fuzzythesaurie,compstat 1974,” [433 Schwede, G. (1976), “N-Variable fuzzy maps with application
Proc. in Computational Statistics, Vienna, pp. 119-127. t o disjuctive decomposition of fuzzy switching functions,” in
[ 399 Rickman, S. M. and Kandel, A. (1975), “Column table approach Proc. 6th Znt. Symp. on Multiple-Valued Logic, Logan, UT.
for the minimization of fuzzy functions,” CSR 137, Dept. of [ 434 Serizawa, M. (1973), “A search technique of controlr o d pattern
Computer Sci., NewMexicoInstituteof Mining andTech- of smoothing core power distributions by fuzzy automation,”
nology, Socorro, NM. J. Nuclear Sci. Tech.
[400] Rieger, B.B. (1975),“On a tolerance topology model of natural I4351 Serfati, M. (1974), “Algebres de boole avec uneintroduction
language meaning,” Tech. Report Germanic Inst., Echn. Hoch- a la theorie des graphes orientes et aux sous-ensembles flous,”
schule, Aachen, Germany. ED. C.D. U.,Paris, France.
[401] Roberts,F. S. (1973), “Tolerancegeometry,” Notre Dame J. [436] Shimura, M. (1973), “Fuzzysetsconcept in rank-ordering
Formal Logic, pp. 68-76. objects,”J. Math. AnaL Appl., pp. 717-733.
[402] Rodder, W. (1975), “On “and” and “or” connectives in fuzzy [437] Shimura, M. (0000), “Applicationoffuzzyfunctions t o pat-
settheory,”Euro I, 75/07, LehrstuhlfurUnternehmensfor- tern classification,” Trans. IECE, vol. 55-D,pp. 218-225.
schung Rwth, Aachen, Germany. [438] Shimura, M . (1973), “Fuzzysets in rank-orderingobjects,”
[403] Rosenfeld,A. (1971), “Fuzzygroups,” J. Math.Anal. Appl., J. Math. Anal. Appl., pp. 717-733.
pp. 512-517. [439] Shimura, M . (1974), “An approach t o pattern recognition and
I4041 Rosenfeld, A. (197S), “Fuzzygraphs,” in FuzzySets and associative memories using fuzzy logic,” US-Japan seminar
1638 PROCEEDINGS
IEEE,
O F THE VOL. 66,NO. 12, DECEMBER 1978

of fuzzy sets and their applications,Berkeley, CA. (4721 Tanaka, H., Okuda, T., and Asai, K. (1973), “Fuzzymathe-
(4401 Shimura, M . (1975), “An approach to pattern recognition and matical programming,” Tmns. SZCE, 9,pp. 109-115.
associative memories using fuzzy logic,” in Fuzzy Setsand Their I4731 Tanaka, H., Okuda, T., and Asai, K. (1974), “Decision-making
Applications to Cognitive and Decision Processes, L. A. Zadeh andits goalina fuzzyenvironment,” US-Japanseminar on
e t al., Eds. New York: Academic Press, pp. 449-476. fuzzy sets and their applications, Berkeley, CA.
(441) Sinha, N. K. andWright, J. D. (1975), “Application of fuzzy I4741 Tanaka, H., Okuda, R., and Asai, K., “On fuzzy-mathematical
control to a heat exchanger,” special interest discussion session programming,”Univ.ofOsakaPrefecture, Sahai, Osaka 591,
o n .fuzzyautomataanddecision processes, 6th IFACWorld Japan.
Cong., Boston, MA. I4751 Tanaka, K. andMizumoto, M. (1975), “Fuzzyprogramsand
I4421 Siy, P. andChen, C. S. (1971), “Fuzzylogicapproach to theirexecution,”in FuzzySets andTheirApplications to
handwrittencharacterrecognitionproblem,”Dept.EE,Univ., CognitiveandDecisionProcesses, L. A.Zadeh etal., Eds.
of Akron, Akron,OH. New York: Academic Press, pp. 41-76.
I4431 Siy, P. and Chen, C. S. (1972), “Minimization of fuzzy func- I4761 Tanaka, H., Okuda, R., andAsai, K. (1975), “A formulation
tions,” E E E Trans. Comput., pp. 100-102. of fuzzy decision problems and its application to aninvestment
14441 Siy, P. (1973), “Fuzzylogicandhandwrittennumeralchar- problem,” Dept. of Industrial Eng., Univ. of Osaka, Prefecture,
acterrecognition,”thesis,Dept.ofElec., Univ. ofAkron, Sakai, Osaka 591,Japan.
Akron, OH. (4771 Tazaki, E. (1975), ‘Weuristic synthesis in a class of systems by
I445 1 Siy, P. and Chen, C. S. (1974), “Fuzzy logic for handwritten usingfuzzy automata,”Report 1, workinggrouponfuzzy
numeralcharacterrecognition,” ZEEE TransSyst., Man, Cy- systems, Tokyo, Japan, pp.61-66.
bernetics, pp. 570-575. I4781 Terano, T. (1972), “Fuzziness of systems,”Nikka-Giren Engi-
Skala, H. J. (1974), “On the problem of imprecision,”Dor- neers, pp. 21-25.
drecht, Netherlands. 14791
.~ Terano, T., “Fuzziness and its concept.,” in Pmc. S y m p . Fuzzi-
Smith, R.E. (1970), ‘Measure theory on fuzzy sets,” thesis, ness in Systems and Its Processing,SZCE.
Dept. of Math., Univ. of Saskatchewan, Saskatoon, Canada. I4801 Terano, T. and Sugeno, M . (1975), “Conditional fuzzy measures
Smith, R. W. (1970), “Measure theory on fuzzy sets,” thesis, and their applications,” in Fuzzy Sets and Their Applications
Dept. ofMath., Univ. of Saskatchewan, Saskatoon, Canada. to Cognitive and DecisionProcesses, L.A.Zadeh et al., Eds.
Sols, I. (1975), “Topology in complete lattices and continuous New York: Academic Press, pp. 151-170.
fuzzy relations,” Zaragoza Univ., Zaragoza, Spain. [481] Terano, T. and Sugeno, M . (1975), “Macroscopic optimization
Stoics, M. andScarlat,E. (1975), “Fuzzyconceptsinthe by using conditional fuzzy measures,” Report 1, working group
controlofproductionsystems,”in Proc.3rd Znt. Cong. Cy- on fuzzy systems, Tokyo, Japan, pp. 67-72.
bernetics and Syst.,Bucharest, Roumania. I4821 Terano, T. and Sugeno, M. (1975), “Macroscopic optimization
Stone, M. H. (1937), “Topological representation of distributive by using conditional measures,” 6th IFAC World Cong., Boston,
lattices and Brouwerian logics,” Cas. Math. Phys.,pp. 1-25. MA.
Sugeno, M. (1971), “On fuzzy nondeterministic problems,” in I4831 Thomason, M . G . and Marinos, P.N. (1,?72), “Fuzzy logic
Ann. Con$ Record of SICE. relationsandtheirutilityinroletheory, Proc. Con$ Syst.,
Sugeno, M. (1972), “Fuzzy measuresandfuzzyintegrals,” Man, Cybernetics, Washington, DC.
Trans. SICE, pp. 218-226. I4841 Thomason, M. G . (1974), “Fuzzy syntax-directed translations,”
Sugeno, M. and Terano, T. (1973), “An approach to the iden- J. Cybernetics, pp. 87-94.
tificationofhumancharacteristicsbyapplyingidentification 14851 Thomason, M. G . (1974), “Finitefuzzyautomata, regular
and system parameter estimation.” fuzzylanguages, and patternrecognition,”Dept.of Elec.
Sugeno, M. (1973), “Constructingfuzzymeasureandgrading Eng., Duke Univ., Durham, NC.
similarityofpatternsbyfuzzy integrals,” Trans SZCE, pp. (4861 Thomason, M. G . (1974), “Theeffect of logic operations
359-367. andfuzzy logic distributions,” IEEE Trans. Syst., Man, Cy-
Sugeno, M. and Terano, T. (1973), “An approach to the iden- bernetics,vol. “-5, pp. 309-310.
tification of human characteristics by applying fuzzy integr-als,” (4871 Thomason, M. G . andMarinos, P.N. (1974), “Deterministic
in Proc.3rd ZFAC Symp.IdentijkationSystem Parameter acceptorsof regular
fuzzy
languages,” IEEE Trans. Syst.,
Estimation. Man, Cybernetics, vol. SMC-4, pp. 228-230.
Sugeno, M. (eOOO), “Evaluationofsimilarityofpatternsby [488] Tsichritzis, D. (1968), “Fuzzypropertiesandalmost sohrable
fuzzy integrals,” in Ann. Con$ Records S E E . problems,”Tech. Report 70, Dept.of Elec. Eng., Princeton
Sugeno, M. (1974), “Theoryoffuzzyintegralsanditsappli- Univ., Princeton, NJ.
cations,” thesis, Tokyo Inst. ofTech., Tokyo, Japan. [489] Tsichritzis,D. (1969), “hieasures oncountablesets,” Tech.
Sugeno, M., Tsukamoto, Y., and Terano, T. (1974), “Subjective report 8 , Univ. ofToronto,Dept.Computer Sci., Toronto,
evaluation of fuzzy objects,” reprint of the IFAC Symp. Sto- Canada.
chastic Control. I4901 Tsuji, H., Mizumoto, M., Toyoda, J., andTanaka, K. (1972),
Sugeno, M. and Terano, T. (1975), “Analytical representation “Interaction between random environments and fuzzy automata
of fuzzy systems,” special interest discussion session o n fuzzy with variable structures,” Trans. ZECE, vol. 55-D, pp. 143-
automata and decision
processes,6th
IFAC World Cong., 144.
Boston, MA. (4911 Tsuji,H.,Mizumoto, M., Toyoda,J.,andTanaka, K. (1973),
Sugeno, S. (1976), “Fuzzysystemsandpatternrecognition,” “Linear fuzzy automation,” Trans. IECE, vol. 56-A, pp. 256-
workshopondiscretesystemsandfuzzyreasoning,Queen 257.
Mary College, Dept. Elec. Eng. I492 1 Tsukamoto, Y. (1972), “Identificationofpreferencemeasure
Tahani, V. (1971), “Fuzzy sets in informationretrieval,” by means of fuzzy integrals,” preprint JORS(in Japanese).
thesis,Dept.Elec. Eng. andComputer Sci., Univ. of Calif., (4931 Tsukamoto, Y. andLida, H. (1973), “Evaluationmodelsof
Berkeley, CA. fuzzy systems,” in Ann. Con$ Rec. SZCE.
Tamura, S., Higuchi, S., andTanaka, K. (1970), “Pattern [494] Tsukamoto, Y. (1975), “A subjective evaluation on attractivity
classificationbased on fuzzy relations,” Trans. ZECE, vol. of sightseeingzones,” Report 1, working group on fuzzy SYS-
5342,pp. 937-944. terns,” Tokyo, Japan.
Tamura, S. (1971), “Fuzzypatternclassification,”in hoc. [495] Uhr, L. (1975), “Towardintegratedcognitivesystems,which
Symp. Fuzzinessin Systems Its Processing, SZCE. mustmakefuzzydecisionsaboutfuzzyproblems,” in Fuzzy
Tamura, S., Higuchi, S., andTanaka, K. (1973), “Pattern Sets andTheirApplications to CognitiveandDecision Pro-
classificationbased on fuzzyrelations,” ZEEE Trans. Syst., cesses,” L. A. Zadeh et al., Eds. New York: Academic Press,
Man, Cybernetics, p. 98. pp. 353-394.
(4661 Tamura, S. and Tanaka, K. (1973),“Learning of fuzzy formal (4961 Van Frassen, B.C. (1975), “Comments: Lakoff‘s fuzzy propo-
language,”ZEEE Trans. Syst., Man, Cybernetics, pp. 98-102. sitionallogic,”
in Contemporary research in Philosophical
LogicandLinguisticSemantics, D. Hockney, W. Harper, B.
14671 Tanaka, K., Toyoda, J., Mizumoto, M., andTsuji, H. (1970), Freed,Eds.Holland:Reidel.
“Fuzzy automata theory and its application to automatic con- I4971 VanVelthoven, G. (1974), “Onderzoeknaartoepasbeerheid
trols,”J. JAACE, pp. 541-550. van detheorieder vage verzamelingen ophet parametrisch
(4681 Tanaka, K., Toyoda, J . , andMizumoto,H. (1970), “Fuzzy onderzoek inzake criminaliteit.
automatatheoryanditsapplicationtoautomaticcontrols,” I4981 VanVelthoven, G . (1975), “Applicationoffuzzysetstheory
J. JAACE, PP. 541-550. to criminal investigation,” Euro I, Brussels.
I4691 Tanaka, K. (1972), “Analogy and fuzzy logic,”Math. Sci. (4991 VanVelthoven, G . (1975), “Fuzzymodelsin personalman-
[470] Tanaka,H.,Okuda, T.,and Asai, K. (1972), “On the fuzzy agement,”in Proc.3rdInt.Cong.cyberneticsandSystems,
mathematical programming,” in AnnualCon5Records of Bucharest, Roumania, pp. 25-25.
SEE. [SOO] VanVelthoven, G., “Quelquesapplicationdelataxonomie
I4711 Tanaka, H. andKaneku, S. (1973), “A fuzzydecodingpro- flou,” Univ. des Sciences et Techniques de Lille.
cedure on error correcting codes,” Tech report on information [Sol] Verma, P. P. (1970), “Vaguenessand the principleofthe
of IECE. excluded middle,” Mind, pp.’66-77.
AND KANDEL ALGEBRA,
SETS, STATISTICS
AND 1639

[SO21 Vincke, P. (1973), “Une application d e la theorie des graphes [ 5 3 5 ] Zadeh, L. A. (1969), “Towardfuzziness in computersystems,
flous,” Univ. Libre de Bruxelles. fuzzy algorithms andlanguages,” Univ. of Calif., Berkeley, CA.
[SO31 Vincke, P. (1973), “La theorie des ensembles flous, memorie,” [ 5 3 6 ] Zadeh,L. A. (1970), “Biological application of the theory of
Faculte de Sci., Univ. Libre de Bruxelles. fuzzysetsandsystems,” in Biocybernetics of rhe Central
[SO41 Warren, R.H., “Closureoperatorandboundaryoperatorfor Nervous System, L. D. Proctor, Ed. Boston, MA: Little Brown
fuzzytopological
spaces,”
Appl.Math. Res.
Lab., Wright- and Co., pp. 199-212.
Patterson AFB, OH. [ 5 3 7 ] Zadeh, L. A. (1970), “Fuzzy languages andtheirrelation to
(5051 Warren, R. H. (0000). “Neighborhoods,basesandcontinuity human and machine intelligence.” in Proc. Cont on Man and
infuzzytopologicalspaces,”Appl.Math. Res. Lab., Wright- Computer, Inst. de la Vie, Birdeaux, France.
Patterson AFB, OH. [5381 Zadeh, L. A. (1971), “Toward fuzziness in computer systems-
1 506 Waken, R. H. (1974), “Optimalityinfuzzytopologicalpoly- Fuzzyalgorithmsand languages,” in Architecture and Design
systems,” Appl. Math. Res. Lab., Wright-Patterson AFB, OH. of DigitalComputers, G.Boulaye, Ed. Paris, France:Dunod,
[ 507 Watanabe, S. (1969). “Modified concepts of logic probability pp. 9-18.
and information based o n generalized continuous characteristic [539 1 Zadeh, L. A. (1971), “Quantitative fuzzy semantics,” Inform.
function,” Inform. Contr., pp. 1-2 1. Sci., pp. 159-176.
[ 508 Wechler, W. (1974), “Analyse und synthes zeitvariabler R-fuzzy [ 5401 1 Zadeh, L. A. (1971), “Toward a theory of fuzzy systems,” in
automaten,” Akad. D. Wiss. De DDR. AspectsofNetwork and SystemTheory. NewYork:Holt,
[ 509 Wechler, W. (1974), “R-fuzzy grammars,”
Tech. Univ. of Rinehart and Winston.
Dresden, DDR. 1541 Zadeh, L. A. (1971), “Similarity relations and fuzzy orderings,”
I510 Wechler, W. andDimitrov, V. (1974), “R-fuzzyautomata,” Inform. Sci., pp. 177-200.
Proc. IFIPCong., pp. 657-660. [ 542 Zadeh, L. A. (1972), “Asystem-theoretic view of behaviour
[511 Wechler, W. (1975), “R-fuzzyautomatawith a time-variant modification,” Memo. ERL-M320, Univ. of Calif.
structure,”in LectureNotesinComputerScience, vol. 28, [543 Zadeh, L. A. (1972). “Onfuzzyalgorithms,”Memo. ERL-
Springer-Verlag, pp. 73-76. M325, Univ. of Calif., Berkeley, CA.
[512 Wechler, W. (1975), “The concept of fuzziness in the theory I 544 Zadeh, L. A. (1972), “A fuzzy-set-theoreticinterpretationof
of automata,” Tech. Univ. Dresden, DDR. linguistic hedges,” J. Cybernetics, pp. 4-34.
I513 Wechler, H. (1975), “Applications of fuzzy logic t o technical [545 Zadeh,L.A. (1972), “A rationale for fuzzy control,” J. Dy-
diagnosis,” Univ. Calif., InformationandComputerSciences namic Systems, Measurementand Control, pp. 3-4.
Dept. [ 546 Zadeh, L. A. (1972), “Linguistic cybernetics,’’ in Proc.Int.
[514 Wee, W. G.andFu, K. S. (1969), “On a generalization of Symp. onSystemsSciencesand Cybernetics, Oxford Univ.,
adaptive algorithms and applications of the fuzzy set concept Oxford, England.
t o patternclassification,”Tech.report 67-7, Dept.of Elec. 1 547 1 Zadeh, L. A. (1972), “Mathematics-A call for reorientation,”
Eng., Purdue Univ., West Lafayette, IN. Math. Sci.
5 1 5 1 Wee,W. G: andFu, K. S. (1975), “Aformaulationoffuzzy [ 548‘ I Zadeh, L. A. (1973), “Outline of a new approach t o t h e anal-
automata and its application as a model of learning systems,” ysis of complex systems and decision processes,” IEEE Tmns.
IEEE Trans. Syst.,Sci.,Cybernetics, vol. SSC-5, pp. 215- Syst., Man, Cybernetics, SMC-3, pp. 28-44.
223. I549‘ 1 Zadeh, L. A. (1973), “On the analysis of large-scale systems,’’
5 1 61 1 Weiss,M.D. (1975), “Fixedpoints,separationandinduced in SystemsApproachesandEnvironmentalProblems, Int.
topologies for fuzzy sets,”J. Math. Anal. Appl., pp. 142-150. Symp., Schlob Reisenburg (Gunzburg/ULM).
517 1 Wensto, F. (1974), “Applications of linguistic variables in the [ 5 5 0 Zadeh, L. A. (1973), “The concept of a linguistic variable and
analysisoforganizations,”thesisproposal, Univ. of Calif., its application t o approximate reasoning,” Memo. ERL-M411,
Berkeley, CA. Univ. of Calif., Berkeley, CA. (published in 3 parts in Inform.
518‘ I Witten, I. H. (1976), “Learning to control sequential and non- Sci. 1975)
sequentialenvironments,”workshopondiscretesystemsand 1551 Zadeh, L. A. (1974), “On the analysis of large-scale systems,”
fuzzy reasoning, Dept. of Elec. Eng., Univ. of Essex, Colchester, Memo. ERL-M418,Univ. of Calif., Berkeley, CA.
Essex, England. I552 Zadeh, L. A. (1974), “Numerical versus linguistic variables,”
[519 Wong, C. K. (1973), “Covering properties of fuzzy topological Newsletter of the Circuits and Systems Society,pp. 3-4.
spaces,” J. Math. Anal. Appl., pp. 697-704. [ 5 5 3 Zadeh, L. A. (1974), “Fuzzy logic and its application t o approx-
[52a Wong, C. K. (1974), “Fuzzypointsandlocalpropertiesof imal reasoning,” in Proc. ZFIP Cong., Stockholm, Sweden.
fuzzy topology,” J. Math. AML Appl.,pp. 316-328. [554 Zadeh, L. A. (1974), “A fuzzy-algorithmicapproach to the
[521 Wong, C. K. (1974), “Fuzzytopology:Productandquotient defmition ofcomplex or unpreciseconcepts,” Memo. ERL-
theorems,” J. Math. Anal. Appl., pp. 512-521. M474, Univ. of Calif., Berkeley, CA.
I 522 Wong, C. K. (1974), ‘Categories of fuzzy sets and fuzzy topo- [ 5 5 5 1 Zadeh, L. A. (1974), “Fuzzy logic and approximate reasoning,”
logical spaces,” RC 5138, IBM Thomas 9. Watson Res. Center, Memo. ERL-M479, Univ. of Calif., Berkeley, CA.
Yorktown-Heights, NY. [ 5 5 61 1 Zadeh, L. A. (1975), “Calculus of fuzzy restrictions,” in Fuzzy
523 Wong, C. K. (1975), “Fuzzytopology,”in FuzzySets and SersandTheirApplications to CognitiveandDecision Pro-
Their Applications to Cognitive and Decision Processes, L. A. cesses, L. A. Zadeh etal., Eds.NewYork:Academic Press,
Zadeh e t al., Eds. New York: Academic Press, pp. 171-190. pp. 1-40.
5 24 Wong, G. A. and Shen, D. C. (1972), “On the learning behavior I557 1 Zadeh, L.A., Fu, K. S., Tanaka, K., and Shimura, M. (1975),
offuzzy automata,” in Proc. Inr.Cong. onCyberneticsand “Fuzzysetsandtheirapplications t o cognitive anddecision
Systems, Oxford Univ., Oxford, England. processes,” New York: Academic Press.
525 Woodhead,R. G., “Onthetheoryoffuzzysets t o resolve 1558‘ I Zadeh, L. A. (1975), “A relationalmodelforapproximate
ill-structured marine decision problems,” Dept. of Naval Arch. reasoning,” IEEE Int.Conf.onCybernetics and Society, San
and Shipbuilding, Univ. of Newcastle Upon Tyne, England. Francisco, CA.
5 2 6 Yeh, R. T. (1973), “Toward an algebraic theoryoffuzzy [ 5 5 9 ] Zadeh, L. A. (1976), “Semantic influence fromfuzzy premises,”
relational svstems.” Tech. rewrt TR-25. DeDt. of ComDuter presentedat6thInt.Symp.Multiple Valued Logic, Logan,
Science, Univ. of Texas at Ausiin, AustintTX. a
UT, pp. 217-218.
15273 Yeh, R. T.and Bang, S. Y. (1975). “Fuzzyrelations,fuzzy 15601 Zimmermann, H. J. (1974), “Optimization in fuzzyenviron-
graphs, and their applications t o clustering analysis,” in Fuzzy ments,” XXI Int. TIMS and 46’ ORSA Conf., San Juan, Puerto
Sets and TheirApplications to Cognitive and DecisionPro- Rico.
cesses, L. A. Zadeh e t al., Eds. NewYork:Academic Press, [ 5 6 1 ] Zimmermann, H. J. andGehring,H. (1974), “Fuzzyinforma-
pp. 125-170. tionprofileforinformationselection,”4thInternet Cong.,
[ 5 2 8 ] Yoeli, M. (1961), “Anoteon a generalizationof Boolean Paris, France.
matrix theory,” Amer. Math. Monrhly, pp. 552-557. [ 5 6 2 ] Zimmermann, H. J. (1975), “Optimale entscheidungen bei
15291 Zadeh, L.A. (1965), “Fuzzysets,” Inform.Contr., pp. 338- unscharfenproblembeschreibungen,” Lehrstuhl
fur
Unter-
353. nehmensforschung, Rwth Aachen, Germany.
15301 Zadeh, L. A. (1965), “Fuzzy sets and systems,”inProc. Symp. [ 5 6 3 ] Zimmermann, H. J.andRodder, W. (1975), “Analyse, Be-
on System Theory, Polytechnic Press of the Inst. of Brooklyn, schreibung und Optimierung von Unscharf Formulierten Prob-
Brooklyn, NY, pp. 29-37. lemen,”LehrstuhlfurUnternehmensforschun,RwthAachen,
[ 5 3 1 ] Zadeh, L. A. (1966), “Shadowsoffuzzysets,” Memo ERL- Germany.
M336, Univ. of Calif., Berkeley, CA. [ 5 6 4 ] Zimmermann, H. J. (1975), “Description andoptimization
[ 5 3 2 ] Zadeh, L. A. (1968), “Fuzzy algorithms,” Inform.Contr., offuzzy systems,’’ Inr. J . General Systems, ec 2 , pp. 209-
PP. 94-102. 215.
(5331 Zadeh,L. A. (1968), “Probabilitymeasuresoffuzzyevents,” [ 5 6 5 ] Zimmermann, H. J . (1975), “The potential of fuzzy decision-
J . Math. Anal. Appl.,pp. 421-427. makingintheprivateandpublicsector,documentation,”
[ 5 3 4 ] Zadeh, L. A. (1969), “Machineintelligenceversushumanin- S OAK-75, Lidingo, Sweden.
telligence,”in Proc.Conf.onScience and Society, Herceg- I5661 Zimmermann, H. J. (1975), “Bibliography:Theoryandappli-
Novi, Yugoslavia, pp. 127-134. cations of fuzzy sets,” Rwth, Aachen, Germany.