INFORMATION AND cONTROI 8, 388-353 (1965) Fuzzy Sets* L. A. Zapew Department of Electrical Engineering and Electronies Research Laboratory, University of California, Berkeley, California A fuzzy set is a class of objects with a continuum of grades of membership. Such a set is characterized by a membership (charac- teristic) function which assigns to each object a grade of member- ship ranging between zero and one. The notions of inclusion, union, intersection, complement, relation, convexity, ete., are extended to such ‘sets, and various properties of these notions in the context of fuzzy sets are established. In particular, a separation theorem for convex fuzay sets is proved without requiring that the fuzzy sets be disjoint. I. INTRODUCTION More often than not, the classes of objects encountered in the real physical world do not have precisely defined criteria of membership. For example, the class of animals clearly includes dogs, horses, birds, etc. as its members, and clearly excludes such objects as rocks, fluids, plants, ete. However, such objects as starfish, bacteria, ete. have an ambiguous status with respect to the class of animals. The same kind of ambiguity arises in the case of a number such as 10 in relation to the “class” of all real numbers which are much greater than 1. Clearly, the “class of all real numbers which are much greater than 1,” or “the class of beautiful women,” or “the class of tall men,” do not constitute classes or sets in the usual mathematical sense of these terms. Yet, the fact remains that such imprecisely defined “classes” play an important role in human thinking, particularly in the domains of pattern recognition, communication of information, and abstraction. The purpose of this note is to explore in a preliminary way some of the basic properties and implications of a concept which may be of use in * This work was supported in part by the Joint Services Electronies Program (U.S. Army, U.S. Navy and U.S. Air Force) under Grant No. AF-AFOSR-139-64 and by the National Science Foundation under Grant GP-2413. 338FUZZY SETS 339 dealing with “classes” of the type cited above. The concept in question hat of a fuzzy set,’ that is, a “class” with a continuum of grades of -mbership. As will be seen in the sequel, the notion of a fuzzy set provides a convenient point of departure for the construction of a con- ceptual framework which parallels in many respects the framework used in the case of ordinary sets, but is more general than the latter and, potentially, may prove to have a much wider scope of applicability, particularly in the fields of pattern classification and information proc- essing. Essentially, such a framework provides a natural way of dealing with problems in which the source of imprecision is the absence of sharply defined criteria of class membership rather than the presence of random variables. We begin the discussion of fuzzy sets with several basic definitions. II. DEFINITIONS Let X be a space of points (objects), with a generic element of X de- noted by x. Thus, X = {z}. A fuzzy set (class) A in X is characterized by a membership (charac- teristic) function f,(x) which associates with each point® in X a real number in the interval 0, 1],° with the valuo of f,(#) at x representing the “grade of membership” of « in A. Thus, the nearer the value of fa(2) to unity, the higher the grade of membership of x in A. When A is a set in the ordinary sense of the term, its membership function can take on only two values 0 and 1, with f.() = 1 or 0 according as x does or does not belong to A. Thus, in this ease j4(z) reduces to the familiar characteristic function of a set A. (When there is a need to differentiate between such sets and fuzzy sets, the sets with two-valued characteristic functions will be referred to as ordinary sets or simply sets.) Example. Let X be the real line R' and let A be a fuzzy set of numbers 1 An application of this concept to the formulation of a elass of problems in pattern classification is described in RAND Memorandum RM-4307-PR, ‘‘Ab- straction and Pattern Classification,” by R. Bellman, R. Kalaba and L. A. Zadeh, October, 1964. ? More generally, the domain of definition of f, () may be restricted to a sub- set of X. * In a more general setting, the range of the membership function can be taken to be.a suitable vartially ordered set P. For our purposes, it is convenient and sufficient io revtxiet the range of f to the unit interval. If the values of f(x) are as truth values, the latter case corresponds to a multivalued logie with & continuum of truth values in the interval (0, 1)340 LADEH which are much greater than 1. Then, one can give a precise, albeit subjective, characterization of A by specifying f(z) as a function on R’. Representative values of such a function might be: f.(0) = 0;f4(1) = 0; fa(5) = 0.01; fa(10) = 0.2; f4(100) = 0.95; f4(500) = 1, It should be noted that, although the membership function of a fuzzy set has some resemblance to a probability function when X isa countable set (or a probability density function when X is a continuum), there are essential differences between these coneepts which will become clearer in the sequel once the rules of combination of membership functions and their basie properties have been established. In fact, the notion of a fuzzy set is completely nonstatistical in nature. We begin with several definitions involving fuzzy sets which are obvious extensions of the corresponding definitions for ordinary sets. A fuzzy set is emply if and only if its membership function is identically zero on X. ‘Two fuzzy sets A and B are equal, written as A = B, if and only if fa(2) = fo(2) for all x in X. (In the sequel, instead of writing fa(z) = fs(2) for all x in X, we shall write more simply fa = fs -) The complement of a fuzzy set A is denoted by A’ and is defined by fat = Li fas (1) As in the case of ordinary sets, the notion of containment plays a central role in the case of fuzzy sets. This notion and the related notions of union and intersection are defined as follows. Coniainment. A is contained in B (or, equivalently, A is a subset of B, or A is smaller than or equal to B) if and only if fs S fx . In symbols ACBeSfa Sha. (2) Union. The union of two fuzzy sets A and B with respective member- ship functions f,(x) and f(z) is a fuzzy set C, written as C = A U B, whose membership function is related to those of A and B by fo(x) = Max (f(x), fo(x)], «EX (3) or, in abbreviated form . leo = fav fo. (4) Note that U has the associative property, that is, A U (2 UC) = (AUB) UG. Comment. A more intuitively appealing way of defining the union isFUZZY SETS 341 the following: The union of A and B is the smallest fuzzy set containing both A and B. More precisely, if D is any fuzzy set which contains both A and B, then it also contains the union of A and B. ‘To show that this definition is equivalent to (3), we note, first, that C as defined by (3) contains both A and B, since Max [fa , fo] 2 Sa and Max [fu , fo] 2 fo. Furthermore, if D is any fuzzy sct containing both A and B, then fo 2 fa fo 2 fe and hence fo = Max (fs, fs] = fe which implies that C C D. Q.E.D. The notion of an intersection of fuzzy sets can be defined in an analo- gous manner. Specifically: Intersection. The intersection of two fuazy sets A and B with respective membership functions f4(x) and js(x) is a fuzzy set C, written as C = A 1 B, whose membership function is related to those of A and B by Fela) = Min [fa(x), fa(e)), 2 EX, (5) or, in abbreviated form : fo=ta A Sn. (6) ‘As in the case of the union, it is easy to show that the intersection of A and B is the largest fuzzy set which is contained in both A and B. As in the case of ordinary sets, A and B are disjoint if A 1 B is empty. Note that f, like U, has the associative property. The intersection and union of two fuzzy sets in R’ are illustrated in Fig. 1. The membership function of the union is comprised of curve seg- ments 1 and 2; that of the intersection is comprised of segments 3 and 4 (heavy lines). Comment. Note that the notion of “belonging,” which plays a funda- mental role in the case of ordinary sets, does not have the same role in342 ZADEH Tq &), fy (x) - xe ynvoeser Fra. 1, Illustration of the union and interseetion of fuzzy sets in Rt the case of fuzzy sets. Thus, it is not meaningful to speak of a point x “belonging” to a fuzzy set A except in the trivial sense of f(a) being positive. Less trivially, one can introduce two levels a and 8 (0 < @ < 1, 0 < 8 < 1, a > 6) and agree to say that (1) “x belongs to A” if fu(x) 2 a; (2) “x does not belong to A” if f.(x) < 6; and (3) “z has an indeterminate status relative to A” if 8 < f4(r) < a. This leads toa three-valued logic (Kleene, 1952) with three truth values: 7 (falar) 2 a), F (fa(w) S 8), and U (8 < faz) fa(x) and fu(t) < fa(x). Similarly, in the ease of (10), the corresponding relation in terms of fay fo, and fois: B ppPe Morgan’s lawsFUZZY SETS 343 Max [ic, Min [fs , fol'= Min [Max (fe, ful, Max [fo, fal} (12) which can be verified to be an identity by considering the six cases: Sale) > fale) > flee), ful) > felt) > Fale), fale) > fale) > fel), Jol) > fol) > fale), Folte) > faltt) > Sox), fe(t) > fale) > fale). Essentially, fuzzy sets in X constitute a distributive lattice with a 0 and 1 (Birkhoff, 1948). AN InrerPRETATION FOR Unions AND INTERSECTIONS In the case of ordinary sets, a set C which is expressed in terms of a family of sets Ai, +++, 4:, ++, An through the connectives U and N, can be represented as a network of switches a, --- , a, , with A, NA; and A; U A; corresponding, respectively, to series and parallel combina- tions of a, and a;. In the ease of fuzzy sets, one can give an analogous interpretation in terms of sieves. Specifically, let fi(z), i = 1, --- , n, denote the value of the membership function of A; at x. Associate with Ji(x) a sieve Si(az) whose meshes are of size f:(x). Then, f(x) v f,() and f(x) A f,(x) correspond, respectively, to parallel and series com- binations of S,(e) and S,(zr), as shown in Fig. 2. More generally, a well-formed expression involving A,, --- , An, U, and f corresponds to a network of sieves Si(x), ++ , Sn(v) which can be found by the conventional synthesis techniques for switching cir- cuits. As a very simple example, C = (A, U Az) NA U Ay (13) corresponds to the network shown in Fig. 3. Note that the mesh sizes of the sieves in the network depend on x and that the network as a whole is equivalent to a single sieve whose meshes are of size f(x). 3; 60 3\(0) 00) 5460 Fig. 2. Parallel and series connection of sieves simultating U and (844 ZADEH so) t Spl) Sq tx) Syl) Fic. 3. A network of sieves simultating {[fi(z) v fa(2)] A file)} Vv fale) IV. ALGEBRAIC OPERATIONS ON FUZZY SETS In addition to the operations of union and intersection, one can define a number of other ways of forming combinations of fuzzy sets and re- lating them to one another. Among the more important of these are the following. Algebraic product. The algebraic product of A and B is denoted by AB and is defined in terms of the membership functions of A and B by the relation Jan = fife. (14) Clearly, ABCANB. (15) Algebraic sum The algebraic sum of A and B is denoted by A + B and is defined by Fare =fat fo (16) provided the sum f, + Js is less than or equal to unity. Thus, unlike the algebraic product, the algebraic sum is meaningful only when the condition fx(«) + fs(x) S 1 is satisfied for all x, Absolute difference. The absolute difference of A and B is denoted by | A — B| and is defined by fia = Ja - fol. Note that in the case of ordinary sets | A — B | reduces to the relative complement of AN Bin A UB. ‘The dual of the algebraic product is the sum A @ B = (A’B!)’ = A+B — AB. (This was pointed out by T. Cover.) Note that for ordinary sets () and the alge- braic product are equivalent operations, as are J and ®.FUZZY SETS 345, Convex combination. By a convex combination of two vectors f and g is usually meant a linear combination of f and g of the form xf + (1 — X)g, in which 0 < 4 < 1. This mode of combining f and g can be generalized to fuzzy sets in the following manner. Let A, B, and A be arbitrary fuzzy sets. The convex combination of A, B, and A is denoted by (A, B; A) and is defined by the relation (A, B; A) = AA + A'B (17) where A’ is the complement of A. Written out in terms of membership functions, (17) reads Soassiar(e) = falx)fa(e) + — fale)Mfo(z), 2 € X. (18) A basic property of the convex combination of A, B, and Ais expressed by ANBc(A,B;A) CAUB for all A. (19) This property is an immediate consequence of the inequalities Min [fa(2), fo(x)] S Afa(@) + (1 — d)fo(2) = Max [fa(w), fa(w)], 2 € X (20) which hold for all \ in (0, 1). It is of interest to observe that, given any fuzzy set C satisfying AN BC CC A UB, one can always find a fuzzy set A such that C= (A, B; A). The membership function of this set is given by _ fol) = falc) Il) = Fay Fale)’ Puzzy relation. The concept of a relation (which is a generalization of that of a function) has a natural extension to fuzzy sets and plays an important role in the theory of such sets and their applications—just as it does in the case of ordinary sets. In the sequel, we shall merely de- fine the notion of a fuzzy relation and touch upon a few related concepts. Ordinarily, a relation is defined as a set of ordered pairs (Halmos, 1960); e.g., the set of all ordered pairs of real numbers 2 and y such that z = y. In the context of fuzzy sets, a fuzzy relation in X is a fuazy set in the product space X X X. For example, the relation denoted by # >> y, z,y € R', may be regarded as a fuzzy set A in R’, with the membership function of A, f.(x, y), having the following (subjective) representative values: f1(10, 5) = 0; f4(100, 10) = 0.7; f.(100, 1) = 1; ete. zeXx. (21)346 ZADEH More generally, one can define an n-ary fuzzy relation in X as a fuzzy set A in the product space X X X X --- X X. For such relations, the membership function is of the form fa(ai, --- , ta), where a: € X, t= Leryn In the case of binary fuzzy relations, the composition of two fuzzy re- lations A and B is denoted by Bo A and is defined as a fuzzy relation in X whose membership function is related to those of A and B by fuca(t, y) = Supy Min [fa(x, 2), fo, y)|- Note that the operation of composition has the associative property Ao(BeC) = (AB) eC. Puzzy sets induced by mappings. Let T be a mapping from X to a space Y. Let B be a fuzzy set in Y with membership function fo(y). The inverse mapping T™ induces a fuzzy set A in X whose membership function is defined by Sa(u) = faly), oY EY (22) for all x in X which are mapped by T into y. Consider now a converse problem in which A is a given fuzzy set in X, and T, as before, is a mapping from X to Y. ‘The question is: What is the membership function for the fuzzy set B in Y which is induced by this mapping? If T is not one-one, then an ambiguity arises when two or more dis- tinct points in X, say x; and 22, with different grades of membership in A, are mapped into the same point y in Y. In this case, the question is: What grade of membership in B should be assigned to y? To resolve this ambiguity, we agree to assign the larger of the two grades of membership to y. More generally, the membership function for B will be defined by foly) = Maxee-tayfa(2), yey (23) where 7”"(y) is the set of points in X which are mapped into y by 7’. V. CONVEXITY As will be seen in the sequel, the notion of convexity can readily be extended to fuzzy sets in such a way as to preserve many of the prop- erties which it has in the context of ordinary sets. This notion appears to be particularly useful in applications involving pattern classification, optimization and related problems.FUZZY SETS 347 convex fuzzy set non-convex _ X fuzzy set fy [Ax #1-A)x,] Dh fy be) tabs Xp x Fre. 4. Convex and noneonvex fuzzy sets in H? In what follows, we assume for concreteness that X is a real Euclidean space B”, DEFINITIONS Convexity. A fuzzy set A is convex if and only if the sets T. defined by Pa = {2|fa(z) 2 a} (24) are convex for-all « in the interval (0, 1). An alternative and moro direct definition of convexity is the follow- ing’: A is convex if and only if Fal + (1 — A)ao] 2 Min [fu(ar), Sa(ee)] (25) for all x, and 22 in X and all ) in (0, 1]. Note that this definition does not. imply that f,() must be a convex funetion of 2. This is illustrated in Fig. 4 forn = 1. To show the equivalence between the above definitions note that if A is convex in the sense of the first definition and @ = fu(e:) S Sales), then 2, € Pa and Ax, + (1 — A) € Ta by the convexity of I. . Hence falda + (1 — d)a2] 2 a = falar) = Min [fa(a1), fa(@e)]- Conversely, if A is convex in the sense of the second definition and a = f(a), then Tz may be regarded as the set of all points xz for which fala) = faa). In virtue of (25), every point of the form Aty + (1 — A)t2,O S A S 1, is also in [, and hence TY, is a convex set. Q.E.D. A basic property of convex fuzzy sets is expressed by the ‘Turoreo. Jf A and B are convex, so ts their intersection. ' This way of expressing convexity was suggested to the writer by his colleague, B, Berlekamp.348 ZADEH Proof: Let C = ANB. Then Foldty + (1 — A)22] = Min [faf\e + (1 — Adazl, folar + (1 — ddan]. (26) Now, since A and B are convex Salde, + (1 = Adee) B Min Yann), Fa()] . (27) Soldx, + CL — d)x] Z Min [fo(21), fo(ve)] and henee Felder + (1 — d)az) (28) = Min [Min [f4(x1), fa(are)|, Min [fx(a1), fn(a2) 1) or equivalently oldu, + (1 — A)2x] 29) 2 Min [Min [fa(z1), fo(m)], Min [fa(22), fo(2) |] and thus Sold + (1 = d)ea] Z Min [fela), fe(m)]. QE. D. (30) Boundedness. A fuzzy set A is bounded if and only if the sets Py = {x |fa(2) = e} are bounded for all a > 0; that is, for every @ > 0 there exists a finite R(a) such that || «|| <-R(q) for alleinT,. If A is a bounded set, then for each e > 0 then exists a hyperplane 1 such that fa() S e for all « on the side of H which does not contain the origin. For, consider the set T. = {x|fi(z) 2 ¢. By hypothesis, this set: is contained in a sphere S of radius R(e). Let H be any hyper- plane supporting S. Then, all points on the side of H which docs not contain the origin lie outside or on S, and henee for all such points falz) Se Lema. Let A be a bounded fuzzy sel and lel M = Supz fax). (M will be referred to as the maximal grade in A.) Then there is ai least one point x» at which M 7s essentially attained in the sense that, for each « > 0, every spherical neighborhood of x) contains points in the set Q(e) = te] f(z) 2 M — 4. Proof® Consider a nested sequence of bounded sets Ty, Te, «++ , where I, = {x | fala) 2 M — M/(n + 1)},n = 1, 2, --- . Note that § This proof was suggested by A. J. Thomasian.FUZZY SETS 349 T,, is nonempty for all finite » as a consequence of the definition of M as M = Sup. f(x). (We assume that M > 0.) Let 2, be an arbitrarily chosen point in Ty,» — 1, 2,---. Then, 2, %,, °-- , is a sequence of points in a closed bounded set T; . By the Bolzano-Weierstrass theorem, this sequence must have at least one limit point, say zo, in T, . Consequently, every spherical neighborhood of % will contain infinitely many points from the sequence 2; , 7, --- , and, more particularly, from the subsequence ry41, tw42, °*+ , Where N = M/c.Since the points of this subsequence fall within the set Q(¢) = {x | fa(z) = M — dj, the lemma is proved. Strict and strong convexity. A fuszy set A is strictly convex if the sets Vla,0 Min [f.(2), fa(ao)]- Note that strong convexity does not imply strict convexity or vice-versa, Note also that if A and B are bounded, sois their union and intersection. Similarly, if A and B are strictly (strongly) convex, their intersection is strictly (strongly) convex. Let A be a convex fuzzy set and let @ = Sup. f.(x). If A is bounded, then, as shown above, either M is attained for some 2, say 2», or there is at least one point x» at which M/ is essentially attained in the sense that, for each « > 0, every spherical neighborhood of xp contains points in the set Q(e) = [x | M — fa(x) S ¢}. In particular, if A is strongly convex and 2 is attained, then 2 is unique. For, if M = f4(2o) and M = fa(%), with a * 2, then f4(x) > M for x = 0.5% + 0.5%, which contradicts M = Max, f,(«). More generally, let C(A) be the set of all points in X at which M is essentially attained. This set will be referred to as the core of A. In the case of convex fuzzy sets, we can assert the following property of C'(A). Turorem. If A is a conver fuzzy set, then its core is a conver set. Proof: It will suffice to show that if 7 is essentially attained at x» and 2; , #; # xo, then it is also essentially attained at all x of the form r= t(1—-Ajm,05AS1. To the end, let P be a cylinder of radius with the line passing through xq and a} as its axis. Let ao’ be a point in a sphere of radius centering350 ZADEH on a and a1’ be a point in a sphere of radius ¢ centering on % such that Sax’) = M — cand falar’) = M,— «. Then, by the convexity of A, for any point u on the segment z'm', we have fa(u) = M ~ « Further. more, by the convexity of P, all points on x'2y' will lie in P. Now let x be any point in the segment «pv; . The distance of this point from the segment am’ must be less than or equal to ¢, since zon’ lies in P. Consequently, a sphere of radius « centering on z will contain at least one point of the segment x; and hence will contain at least one point, say w, at which f,(w) = M — «. This establishes that M is os- sentially attained at « and thus proves the theorem. Coronary. If X = E! and A is strongly convex, then the point at which M is essentially attained is unique. Shadow of a fuzzy set. Let A be a fuazy set in BE” with membership function f(z) = fa(t1, +++, tn). For notational simplicity, the notion of the shadow (projection) of A on a hyperplane H will be defined below for the special case where H is a coordinate hyperplane, e.g., H = {e[ a. = 0}. Specifically, the shadow of A on H = {x | 21 = 0} is defined to be a fuzay set Sx(A) in E”~ with fsqcay(x) given by Faycay(@) = Say (ay(42,°++ tn) = Supa, Sati, -+> , en). Note that this definition is consistent with (23). When A is a convex fuzay set, the following property of Sx(A) is an immediate consequence of the above definition: If A is a convex fuzzy set, then its shadow on any hyperplane is also a convex fuzzy set. An interesting property of the shadows of two convex fuzzy sets is expressed by the following implication Su(A) = Sx(B) forall H > A = To prove this assertion,’ it is sufficient to show that if there exists a point, say zo , such that f.(2) + fs(2o), then their exists a hyperplane H such that fsgcay(to") * fiayis)(40"), where ao” is the projection of x on A. Suppose that fa(x») = a > fe(ao) = B. Since B is a convex fuzzy set, the set Tz = [2 | f(a) > 6} is convex, and hence there exists a hyper- plane F supporting [ and passing through x . Let H be a hyperplane orthogonal to F, and let 29° be the projection of «> on H. Then, since ' This proof is based on an idea suggested by G. Dantzig for the case where A and B are ordinary convex sets.FUZZY SETS 351 fe(w) < for all x on F, we have fsy¢s(%0") < 8. On the other hand, fspiay(%0") = a. Consequently, fyc(te") * fsqca)(%o"), and similarly for the case where a < 8. A somewhat more general form of the above assertion is the following: Let A, but not necessarily B, be a convex fuzzy set, and let Sx(A) = Sa(B) for all H. Then A = conv B, where conv B is the convex hull of B, that is, the smallest convex set containing B. More generally, Sa(A) = Sz(B) for all H implies conv A = conv B: Separation of convex fuzzy sets. The classical separation theorem for ordinary convex sets states, in essence, that if A and B are disjoint con- vex sets, then there exists a separating hyperplane H such that A is on one side of H and B is on the other side. It is natural to inquire if this theorem can be extended to convex fuzzy sots, without requiring that A and B be disjoint, since the condition of disjointness is much too restrictive in the case of fuzzy sets. It turns out, as will be seen in the sequel, that the answer to this question is in the affirmative. Asa preliminary, we shall have to make a few definitions. Specifically, let A and B be two bounded fuzay sets and let H be a hypersurface in E” defined by an equation h(x) = 0, with all points for which h(z) = 0 being on one side of H and all points for which h(z) S 0 being on the other side.8 Let Kx be a number dependent on H such that f4(2) S Ku on one side of H and fs(x) < Ky on the other side. Let My be Inf Ky. ‘The number Dy = 1 — My will be called the degree of separation of A and B by H. In general, one is concerned not with a given hypersurface H, but with a family of hypersurfaces {Hy}, with \ ranging over, say, EZ. The problem, then, is to find a member of this family which realizes the highest possible degree of separation. ‘A special ease of this problem is one where the Hy are hyperplanes in E", with ) ranging over Z”. In this ease, we define the degree of separa- bility of A and B by the relation D=1-M (31) where _ M = InfyMy (32) with the subscript \ omitted for simplicity. * Note that the sets in question have H in common. \352 CADE fy (8) fy (x) My hyperplane H (point) Fic. 5, Ilustration of the separation theorem for fuzzy sets in 2 Among the various assertions that can be made concerning D, the following statement? is, in effect, an extension of the separation theorem to convex fuzzy sets. ‘Turorem. Let A and B be bounded conver fuzzy sets in B", with maximal grades M, and My, respectively (M4 = Sup. f(x), Ms = Sup. fo(x)]. Let M be the maximal grade for the intersection A NB (M = Sup, Min- [fa(w), fo(x)]). Then D = 1 — M. Comment. In plain words, the theorem states that the highest degree of separation of two convex fuzzy sets A and B that can be achieved with a hyperplane in &” is one minus the maximal grade in the inter- section A NM B. This is illustrated in Fig. 5 form = 1 Proof: It is convenient to consider separately the following two cases: (1) M = Min (M,, Mz) and (2) M < Min (M,, Mz). Note that the latter case rules out A C Bor BC A, Case 1. For concreteness, assume that My < Mz, so that M = My. Then, by the property of bounded sets already stated there exists a hyperplane H such that fs(x) S M for all z on one side of H. On the other side of H, fa(t) < M because f.(z) S My = M forall x. It remains to be shown that there do not exist an M’ < M and a hyperplane H’ such that f4(a) < M’ on one side of H’ and fa(z) < M’ on the other side. ‘This follows at once from the following observation. Suppose that such H’ and M' exist, and assume for concreteness that the core of A (that is, the set of points at which 17, = M is essentially attained) is on the plus side of H’. This rules out the possibility that f.(a) < M’ 9 This statement is based on a suggestion of E. BerlekampFUZZY SETS 353, for all x on the plus side of H', and hence necessitates that f:(a) S M' for all x on the minus side of H’, and f(z) S M’ for all x on the plus side of H’. Consequently, over all z on the plus side of H’ Sup. Min [fa(x), fa(w)] M’. Case 2, Consider the convex sets Ta = {a | fu(w) > M} and Ty = {2 | fo(c) > M}. These sets are nonempty and disjoint, for if they were not there would be a point, say u, such that f.(u) > M and f,(u) > M, and hence f4ng(w) > M, which contradicts the assumption that M7 = Sup: fana(#). Since I’, and Ty are disjoint, by the separation theorem for ordinary convex sets there exists a hyperplane H such that I’, is on one side of H (say, the plus side) and I is on the other side (the minus.side). Fur- thermore, by the definitions of T, and Ys, for all points on the minus side of H, f(x) S M, and for all points on the plus side of H,fo(z) S M. Thus, we have shown that there exists a hyperplane H which realizes 1 — M as the degree of separation of A and B. The conclusion that a higher degree of separation of A and B cannot be realized follows from the argument given in Case 1. This concludes the proof of the theorem. ‘The separation theorem for convex fuzzy sets appears to be of particeu- lar relevance to the problem of pattern discrimination. Its application to this class of problems as well as to problems of optimization will be explored in subsequent notes on fuzzy sets and their properties. Recervep: November 30, 1964 RErErences Birkuorr, G. (1948), ‘Lattice Theory,” Am. Math. Soc. Colloq. Publ., Vol. 25, ‘New York. Haumos, P. R. (1960), “Naive Set Theory.” Van Nostrand, New York Kuzeng, 8. C. (1952), “Introduction to Metamathematics,” p. 334. Van Nos- trand, New York.