Lecture 8 - Operations On Fuzzy Sets: February 15, 2007

Fuzzy Sets Fuzzy Sets
Outline
and Fuzzy and Fuzzy
Techniques Techniques
Nataša Nataša
Sladoje Sladoje
Introduction Introduction
1 Introduction
Types of Set
Operations
Fuzzy Sets and Fuzzy Techniques Types of Set
Operations
Fuzzy Complements
Fuzzy Intersections
Lecture 8 – Operations on Fuzzy Sets Fuzzy Complements
Fuzzy Intersections 2 Types of Set Operations
Fuzzy Complements
Fuzzy Unions Fuzzy Unions
Combinations Combinations
of Operations of Operations Fuzzy Intersections
Aggregation
Nataša Sladoje Aggregation Fuzzy Unions
Operations Operations
An Centre for Image Analysis An

Application:
Fuzzy Uppsala University
Application:
Fuzzy 3 Combinations of Operations
Morphologies Morphologies
February 15, 2007 4 Aggregation Operations
5 An Application: Fuzzy Morphologies
Standard fuzzy operations Properties of the standard

and Fuzzy and Fuzzy
operations
Nataša Nataša
Sladoje Sladoje
Types of Set Types of Set

Operations Operations • They are generalizations of the corresponding
Recall:
(uniquely defined!) classical set operations.
Fuzzy Complements Fuzzy Complements
for fuzzy sets A, B on a reference set X , given by the

Fuzzy Intersections Fuzzy Intersections
Combinations corresponding membership functions A(x) and B(x): Combinations

• They satisfy the cutworthy and strong cutworthy
of Operations of Operations
properties. They are the only ones that do.
Aggregation Aggregation
Operations Ā(x) = 1 − A(x) − fuzzy complement Operations • The standard fuzzy intersection of two sets contains
An (A ∩ B)(x) = min[A(x), B(x)] − fuzzy intersection An (is bigger than) all other fuzzy intersections of those
Application: Application:
Fuzzy (A ∪ B)(x) = max[A(x), B(x)] − fuzzy union Fuzzy sets.
• The standard fuzzy union of two sets is contained in
for all x ∈ X .
(is smaller than) all other fuzzy unions of those sets.
• They inherently prevent the compound of errors of
the operands.
Other generalizations of the set Fuzzy complements

and Fuzzy and Fuzzy
operations Definition
Nataša Nataša
Sladoje Sladoje
Introduction
Aggregation operators Introduction
• A fuzzy complement cA of a fuzzy set A is given by a

Recall:
function c : [0, 1] → [0, 1]
Aggregation operators are used to combine several fuzzy
Combinations Combinations • Function c assigns a value to each membership value

of Operations sets in order to produce a single fuzzy set. of Operations
Aggregation Aggregation A(x) of x ∈ X to the fuzzy set A.

Operations Associative aggregation operations Operations
• A membership function of the set cA is defined as
An
Application: • (general) fuzzy intersections - t-norms An
Application: cA(x) = c(A(x)).
Fuzzy Fuzzy
Morphologies • (general) fuzzy unions - t-conorms Morphologies
Note:
Non-associative aggregation operations • the value cA(x ) is interpreted not only as the degree to which x ∈ X
• averaging operations - idempotent aggregation belongs to the fuzzy set cA, but also as the degree to which x does
not belong to the fuzzy set A;
operations
• the value cA(x ) does not depend on x, but only on A(x ).
Fuzzy complements Nested structure of the basic

and Fuzzy and Fuzzy
Axiomatic requirements classes of fuzzy complements

Nataša Nataša
Sladoje Sladoje
Types of Set Types of Set • All functions c : [0, 1] → [0, 1]

Operations
Ax c1. c (0) = 1 and c (1) = 0. boundary condition Operations
IsNotC(a) = a
Fuzzy Intersections
Ax c2. For all a, b ∈ [0, 1], if a ≤ b, then c (a) ≥ c (b). monotonicity Fuzzy Intersections
• All fuzzy complements (Ax c1 and Ax c2)

Combinations c1 and c2 are called axiomatic skeleton for fuzzy complements Combinations ½
of Operations of Operations 1 for a ≤ t
Ax c3. c is a continuous function. c(a) =
Aggregation Aggregation 0 for a > t
Operations Ax c4. c is involutive, i.e., c (c (a)) = a, for each a ∈ [0, 1]. Operations
An An • All continuous fuzzy complements (Ax c1- Ax c3)

Fuzzy Fuzzy c(a) = 12 (1 + cos πa) c( 13 ) =?
Morphologies Theorem Morphologies
Let a function c : [0, 1] → [0, 1] satisfy Ax c2 and Ax c4. Then • All involutive fuzzy complements (Ax c1- Ax c4)
c satisfies Axioms Ax c1 and Ax c3 too. Moreover, the 1−a
cλ (a) = , λ > −1 (Sugeno class)
function c is a bijection. 1 + λa
1
cω (a) = (1 − aω ) ω ω > 0 (Yager class)
cA(x) = 1 − A(x) Classical fuzzy complement
Generators Generators
and Fuzzy and Fuzzy
Nataša
Sladoje
Increasing generators Nataša
Sladoje
Decreasing generators
Introduction • Increasing generator Introduction • Decreasing generator

Types of Set
Operations
is a strictly increasing continuous function Types of Set
Operations
is a strictly decreasing continuous function
Fuzzy Complements
Fuzzy Intersections
g : [0, 1] → R, such that g(0) = 0. Fuzzy Complements
Fuzzy Intersections
f : [0, 1] → R, such that f (1) = 0.
Fuzzy Unions
• A pseudo-inverse of increasing generator g is defined Fuzzy Unions
• A pseudo-inverse of increasing generator f is defined
Combinations
of Operations as 
Combinations
of Operations as 
 0 for a ∈ (−∞, 0)  1 for a ∈ (−∞, 0)
Operations g (−1) = g −1 (a) for a ∈ [0, g(1)] Operations f (−1) = f −1 (a) for a ∈ [0, f (0)]
 
An 1 for a ∈ (g(1), ∞) An 0 for a ∈ (f (0), ∞)
Fuzzy
Morphologies
• An example: Fuzzy
Morphologies
• An example:
g(a) = ap , p > 0 f (a) = 1 − ap , p > 0

 
 0 for a ∈ (−∞, 0)  1 for a ∈ (−∞, 0)
1 1
g (−1) (a) = a p for a ∈ [0, 1] f (−1) (a) = (1 − a) p for a ∈ [0, 1]
 
1 for a ∈ (1, ∞) 0 for a ∈ (1, ∞)
Generators Generating fuzzy complements

and Fuzzy and Fuzzy
Nataša
Sladoje
Making one from another Nataša
Sladoje
Introduction Theorem Introduction

Theorem
Types of Set Let f be a decreasing generator f . The function g, defined by Types of Set
(First Characterization Theorem of Fuzzy Complements.)
Operations
Fuzzy Complements
Operations
Fuzzy Complements
Let c be a function from [0, 1] to [0, 1]. Then c is a
Fuzzy Intersections
Fuzzy Unions
g (a) = f (0) − f (a), for a ∈ [0, 1] Fuzzy Intersections
Fuzzy Unions
(involutive) fuzzy complement iff there exists an increasing
Combinations Combinations generator g such that, for all a ∈ [0, 1]
of Operations
is an increasing generator, with g (1) = f (0). Its pseudo-inverse is then of Operations
given by
Aggregation
Operations g (−1) (a) = f (−1) (f (0) − a), for a ∈ R .
Aggregation
Operations
c(a) = g −1 (g(1) − g(a)).
An An
Fuzzy Fuzzy
Morphologies Theorem Morphologies
Theorem
Let g be an increasing generator. The function f , defined by (Second Characterization Theorem of Fuzzy Complements.)
f (a) = g (1) − g (a), for a ∈ [0, 1] Let c be a function from [0, 1] to [0, 1]. Then c is a
(involutive) fuzzy complement iff there exists an decreasing
is a decreasing generator, with f (0) = g (1). Its pseudo-inverse is then generator f such that, for all a ∈ [0, 1]
given by
f (−1) (a) = g (−1) (g (1) − a), for a ∈ R . c(a) = f −1 (f (0) − f (a)).
Generating fuzzy complements Fuzzy intersections

and Fuzzy and Fuzzy
Nataša
Sladoje
Examples Nataša
Sladoje
Definition
Introduction Introduction An intersection of two fuzzy sets A and B is given by a

Types of Set Types of Set function of the form
i : [0, 1] × [0, 1] → [0, 1].

Fuzzy Intersections
Fuzzy Unions
Increasing generators Fuzzy Intersections
Fuzzy Unions
Standard fuzzy complement: g(a) = a.

Combinations
of Operations Sugeno class of fuzzy complements:
Combinations
of Operations A value is assigned to a pair of membership values A(x)
Aggregation 1 Aggregation and B(x) of an element x of the universal set X . It
Operations gλ (a) = ln(1 + λa), for λ > −1 Operations
λ represents membership of x to the intersection of A and B:
An Yager class of fuzzy complements: gω (a) = aω , for ω > 0. An
Fuzzy
Morphologies
Fuzzy
Morphologies
(A ∩ B)(x) = i(A(x), B(x)), for x ∈ X .
Decreasing generators Note:
Standard fuzzy complement: f (a) = −ka + k for k > 0. • Intuitive requirements to be fulfilled by a function i to qualify as an
Yager class of fuzzy complements: f (a) = 1 − aω . intersection of fuzzy sets are those of well known and extensively
studied t-norms (triangular norms); the names fuzzy intersection and
t-norm are therefore used interchangeably in the literature.
• The value (A ∩ B )(x ) does not depend on x, but only on A(x ) and
B (x ).
Fuzzy intersections Fuzzy intersections

and Fuzzy and Fuzzy
Nataša
Sladoje
Axiomatic requirements Nataša
Sladoje
Additional (optional) requirements

Operations Operations For all a, b, d ∈ [0, 1],
Fuzzy Complements
For all a, b, d ∈ [0, 1], Fuzzy Complements
Fuzzy Intersections
Ax i1. i (a, 1) = a. boundary condition Fuzzy Intersections Ax i5. i is a continuous function. continuity
Ax i6. i (a, a) ≤ a. subidempotency
Combinations Ax i2. b ≤ d implies i (a, b) ≤ i (a, d ). monotonicity Combinations
of Operations
Ax i3. i (a, b) = i (b, a). commutativity
of Operations Ax i7. a1 < a2 and b1 < b2 implies i (a1 , b1 ) < i (a2 , b2 ).
Aggregation Aggregation strict monotonicity
Operations Ax i4. i (a, i (b, d )) = i (i (a, b), d ). associativity Operations
An An
Application: Application: Note:
Fuzzy Axioms i1 - i4 are called axiomatic skeleton for fuzzy Fuzzy
Morphologies Morphologies Subidempotency is a weaker requirement than idempotency, i (a, a) = a.
intersections. A continuous subidempotent t-norm is called Archimedean t-norm
If the sets are crisp, i becomes the classical (crisp) The standard fuzzy intersection, i (a, b) = min[a, b], is the only
intersection. idempotent t-norm.
Fuzzy intersections Fuzzy intersections

and Fuzzy and Fuzzy
Nataša
Sladoje
Examples of t-norms frequently used Nataša
Sladoje
Properties

Fuzzy Intersections
Fuzzy Unions
• Drastic intersection
8
Fuzzy Intersections
Fuzzy Unions
< a if b = 1
of Operations i (a, b) = b if a = 1 of Operations
Aggregation
:
0 otherwise Aggregation
• imin (a, b) ≤ max(0, a + b − 1) ≤ ab ≤ min(a, b).
Operations • Bounded difference Operations
• For all a, b ∈ [0, 1], imin (a, b) ≤ i(a, b) ≤ min[a, b].
An i (a, b) = max[0, a + b − 1] An
Fuzzy • Algebraic product Fuzzy
i (a, b) = ab
• Standard intersection
i (a, b) = min[a, b]
Fuzzy intersections Fuzzy unions

and Fuzzy and Fuzzy
Nataša
Sladoje
How to generate t-norms Nataša
Sladoje
Definition
Introduction Introduction A union of two fuzzy sets A and B is given by a function of

Types of Set Types of Set the form
Operations Theorem Operations
Fuzzy Complements
Fuzzy Intersections (Characterization Theorem of t-norms) Let i be a binary
Fuzzy Complements
Fuzzy Intersections
u : [0, 1] × [0, 1] → [0, 1].
operation on the unit interval. Then, i is an Archimedean A value is assigned to a pair of membership values A(x)
of Operations t-norm iff there exists a decreasing generator f such that of Operations and B(x) of an element x of the universal set X . It
Aggregation
Operations
Aggregation
Operations
represents membership of x to the union of A and B:
i(a, b) = f (−1) (f (a) + f (b)), for a, b ∈ [0, 1].
An An
Application:
Fuzzy
Application:
Fuzzy
(A ∪ B)(x) = u(A(x), B(x)), for x ∈ X .
Morphologies Example: A class of decreasing generators fω (a) = (1 − a) , ω > 0 ω
Morphologies
generates a Yager class of t-norms Note:
• Intuitive requirements to be fulfilled by a function u to qualify as a
1
iω (a, b) = 1 − min[1, ((1 − a)ω + (1 − b)ω ) ω ], ω > 0. union of fuzzy sets are those of well known and extensively studied
t-conorms (triangular conorms); the names fuzzy union and t-conorm
It can be proved that imin (a, b) ≤ iω (a, b) ≤ min[a, b]. are therefore used interchangeably in the literature.
• The value (A ∪ B )(x ) does not depend on x, but only on A(x ) and
B (x ).
Fuzzy unions Fuzzy unions

and Fuzzy and Fuzzy
Nataša
Sladoje
Axiomatic requirements Nataša
Sladoje
Additional (optional) requirements
Types of Set Types of Set For all a, b, d ∈ [0, 1],

Fuzzy Complements
For all a, b, d ∈ [0, 1], Fuzzy Complements Ax u5. u is a continuous function. continuity
Ax u1. u (a, 0) = a. boundary condition
Fuzzy Unions Fuzzy Unions Ax u6. u (a, a) ≥ a. superidempotency
Combinations Ax u2. b ≤ d implies u (a, b) ≤ u (a, d ). monotonicity Combinations Ax u7. a1 < a2 and b1 < b2 implies u (a1 , b1 ) < u (a2 , b2 ).
Ax u3. u (a, b) = u (b, a). commutativity strict monotonicity
Operations Ax u4. u (a, u (b, d )) = u (u (a, b), d ). associativity Operations
An An Note:
Fuzzy Axioms u1 - u4 are called axiomatic skeleton for fuzzy unions. Fuzzy Requirements u5 - u7 are analogous to Axioms i5 - i7.
Morphologies Morphologies Superidempotency is a weaker requirement than idempotency.
They differ from the axiomatic skeleton of fuzzy intersections only
A continuous superidempotent t-conorm is called Archimedean
in boundary condition. t-conorm.
For crisp sets, u behaves like a classical (crisp) union. The standard fuzzy union, u (a, b) = max[a, b], is the only idempotent
t-conorm.
Fuzzy unions Fuzzy unions

and Fuzzy and Fuzzy
Nataša
Sladoje
Examples of t-conorms frequently used Nataša
Sladoje
Properties

Fuzzy Intersections
• Drastic union
8
Fuzzy Intersections
< a if b = 0
of Operations u (a, b) = b if a = 0 of Operations
• max[a, b] ≤ a + b − ab ≤ min(1, a + b) ≤ umax (a, b).
:
1 otherwise
Operations • Bounded sum Operations
• For all a, b ∈ [0, 1], max[a, b] ≤ u(a, b) ≤ umax (a, b).
An u (a, b) = min[1, a + b] An
Fuzzy • Algebraic sum Fuzzy
u (a, b) = a + b − ab
• Standard intersection
u (a, b) = max[a, b]
Fuzzy unions Combinations of set operations

and Fuzzy and Fuzzy
Nataša
Sladoje
How to generate t-conorms Nataša
Sladoje
De Morgan laws and duality of fuzzy operations
Types of Set
Operations
Theorem Types of Set
Operations De Morgan laws in classical set theory:
Fuzzy Complements (Characterization Theorem of t-conorms) Let u be a Fuzzy Complements
binary operation on the unit interval. Then, u is an

Fuzzy Unions Fuzzy Unions A ∩ B = Ā ∪ B̄ and A ∪ B = Ā ∩ B̄.
Combinations
of Operations
Archimedean t-conorm iff there exists an increasing Combinations
of Operations
Aggregation
generator g such that Aggregation
The union and intersection operation are dual with respect
Operations Operations to the complement.
An
Application:
u(a, b) = g (−1) (g(a) + g(b)), for a, b ∈ [0, 1]. An
Application:
Fuzzy
Morphologies
Fuzzy
Morphologies
De Morgan laws for fuzzy sets:
Example: A class of increasing generators fω (a) = a , ω > 0 generates
ω
a Yager class of t-conorms c(i(A, B)) = u(c(A), c(B)) and c(u(A, B)) = i(c(A), c(B))
1
uω (a, b) = min[1, (aω + bω ) ω ], ω > 0. for a t-norm i, a t-conorm u, and fuzzy complement c.
It can be proved that max[a, b] ≤ uω (a, b) ≤ umax (a, b). Notation: hi, u, ci denotes a dual triple.
Duality of fuzzy set operations Dual triples - Six theorems (1)

and Fuzzy and Fuzzy
Nataša
Sladoje
Examples of dual triples Nataša
Sladoje
Theorem
Introduction Introduction The triples hmin, max, c i and himin , umax , c i are dual with respect to any
Types of Set Types of Set fuzzy complement c.
Theorem
Fuzzy Unions
Dual triples with respect to the standard fuzzy complement Fuzzy Unions
Given a t-norm i and an involutive fuzzy complement c, the binary
of Operations of Operations operation u on [0, 1], defined for all a, b ∈ [0, 1] by
Aggregation hmin(a, b), max(a, b), cs i Aggregation
u (a, b) = c (i (c (a), c (b)))
An hab, a + b − ab, cs i An
Application: Application: is a t-conorm such that hi , u , c i is a dual triple.
Fuzzy
Morphologies hmax(0, a + b − 1), min(1, a + b), cs i Fuzzy
Morphologies
Theorem
himin (a, b), umax (a, b), cs i Given a t-conorm u and an involutive fuzzy complement c, the binary
operation i on [0, 1], defined for all a, b ∈ [0, 1] by
i (a, b) = c (u (c (a), c (b)))
is a t-norm such that hi , u , c i is a dual triple.

Dual triples - Six theorems (2) Aggregation operations

and Fuzzy and Fuzzy
Nataša
Sladoje
Nataša
Sladoje
Definition

Operations
Fuzzy Complements
Theorem Operations
Fuzzy Complements Aggregations on fuzzy sets are operations by which several
Given an involutive fuzzy complement c and an increasing generator g of
Fuzzy Intersections
Fuzzy Unions
c, the t-norm and the t-conorm generated by g are dual with respect to c.
Fuzzy Intersections
Fuzzy Unions fuzzy sets are combined in a desirable way to produce a
Combinations Combinations single fuzzy set.
Aggregation Theorem Aggregation

Operations Let hi , u , c i be a dual triple generated by an increasing generator g of the Operations Definition
An involutive fuzzy complement c. Then the fuzzy operations i , u , c satisfy the An Aggregation operation on n fuzzy sets (n ≥ 2) is a function
law of excluded middle, and the law of contradiction.
Fuzzy
Morphologies
Fuzzy
Morphologies
h : [0, 1]n → [0, 1].
Theorem Applied to fuzzy sets A1 , A2 , . . . , An , function h produces an aggregate
Let hi , u , c i be a dual triple that satisfies the law of excluded middle and the fuzzy set A, by operating on membership grades to these sets for each
law of contradiction. Then hi , u , c i does not satisfy the distributive laws. x ∈ X:
A(x ) = h(A1 (x ), A2 (x ), . . . , An (x )).
Axiomatic requirements Averaging operations

and Fuzzy and Fuzzy
Nataša Nataša
Sladoje Sladoje
Types of Set
Operations Ax h1 h(0, 0, . . . , 0) = 0 and h(1, 1, . . . , 1) = 1. boundary conditions
Types of Set
Operations
• If an aggregation operator h is monotonic and idempotent
Fuzzy Complements
Fuzzy Intersections Ax h2 For any (a1 , a2 , . . . , an ) and (b1 , b2 , . . . , bn ), such that ai , bi ∈ [0, 1]
Fuzzy Complements
Fuzzy Intersections
(Ax h2 and Ax h5), then for all (a1 , a2 , . . . , an ) ∈ [0, 1]n
Fuzzy Unions
and ai ≤ bi for i = 1, . . . , n, Fuzzy Unions
Combinations Combinations min(a1 , a2 , . . . , an ) ≤ h(a1 , a2 , . . . , an ) ≤ max(a1 , a2 , . . . , an ).

of Operations h(a1 , a2 , . . . , an ) ≤ h(b1 , b2 , . . . , bn ). of Operations
Aggregation h is monotonic increasing in all its arguments. Aggregation

• All aggregation operators between the standard fuzzy
Ax h3 h is continuous. intersection and the standard fuzzy union are idempotent.
An An
Application: Ax h4 h is a symmetric function in all its arguments; for any permutation p Application:
Fuzzy
on {1, 2, . . . , n} Fuzzy • The only idempotent aggregation operators are those
between standard fuzzy intersection and standard fuzzy
h(a1 , a2 , . . . , an ) = h(ap(1) , ap(2) , . . . , ap(n) ).
union.
Ax h5 h is an idempotent function; for all a ∈ [0, 1]
h(a, a, . . . , a) = a. Idempotent aggregation operators are called averaging
operations.
Averaging operations Ordered Weighted Averaging

and Fuzzy and Fuzzy
Operations - OWA
Nataša Nataša
Sladoje Sladoje
Generalized means: For

• a given weighting vector w = (ω1 , ω2 , . . . , ωn ),
Types of Set µ ¶1 Types of Set
Operations a1α + a2α + ··· + anα α Operations ωi ∈ [0, 1] for i = 1, . . . , n
Fuzzy Complements
hα (a1 , a2 , . . . , an ) = , Fuzzy Complements Xn
Fuzzy Intersections
Fuzzy Unions
n Fuzzy Intersections
Fuzzy Unions
ωi = 1
i =1
of Operations for α ∈ R, and α 6= 0, and for α < 0 ai 6= 0. of Operations • a permutation (b1 , b2 , . . . , bn ) of a given vector (a1 , a2 , . . . , an )
Aggregation • Geometric mean: For α → 0, Aggregation bi ≥ bj if i < j, for any pair i , j ∈ {1, 2, . . . , n}
lim hα (a1 , a2 , . . . , an ) = (a1 · a2 · · · · · an ) n ;
1
an OWA operation associated with w is defined as
An An
α→0
Fuzzy • Harmonic mean: For α = −1, Fuzzy hw (a1 , a2 , . . . , an ) = ω1 b1 + ω2 b2 + . . . ωn bn .
n
h−1 (a1 , a2 , . . . , an ) = 1 1 1
;
a1 + a2 + ··· + an Example: For w = (0.3, 0.1, 0.2, 0.4),
• Arithmetic mean: For α = 1,
hw (0.6, 0.9, 0.2, 0.7) = 0.3 · 0.9 + 0.1 · 0.7 + 0.2 · 0.6 + 0.4 · 0.2 = 0.54 .
1
h1 (a1 , a2 , . . . , an ) = (a1 + a2 + . . . an ).
n • OWA operations satisfy axioms Ax h1 - Ax h5
Functions hα satisfy axioms Ax h1 - Ax h5. • OWA operations cover the whole range between min and
max operations.
Norm operations Norm operations

and Fuzzy and Fuzzy
Nataša
Sladoje
Nataša
Sladoje
Example: λ-averages
Introduction Aggregation operations h on [0, 1]2 which are Introduction
Types of Set
Operations
monotonic Types of Set
Operations
Fuzzy Complements
commutative Fuzzy Complements
For a given t-norm i, and given t-conorm u
associative
Combinations Combinations for a, b ∈ [0, 1] and λ ∈ (0, 1)

of Operations
fulfil boundary conditions h(0, 0) = 0 and h(1, 1) = 1 of Operations

 min[λ, u(a, b)] when a, b ∈ [0, λ]
Operations are called norm operations. Operations
An An h(a,b) = max[λ, i(a, b)] when a, b ∈ [λ, 1]


Application:
Fuzzy A class of norm operations
Application:
Fuzzy λ otherwise
• contains t-norms and t-conorms, as special cases,
is a parametrized class of norm operations which are
which fulfill stronger boundary conditions; neither t-norms, nor t-conorms.
• covers the whole range of aggregation operators, from
imin to umax ; These operations are called λ-averages.
Averaging operations Decision as a fuzzy set

and Fuzzy and Fuzzy
Nataša
Sladoje
Do we need more than standard operations? Nataša
Sladoje
Introduction Making decisions in fuzzy environment: Introduction

Intersection: No positive compensation (trade-off)
Types of Set
Operations
Taking into account objectives and constraints expressed by Types of Set
Operations between the memberships of the fuzzy sets observed.
Fuzzy Complements fuzzy sets, make a new fuzzy set representing a decision. Fuzzy Complements
Union: Full compensation of lower degrees of
Fuzzy Unions
(H.-J. Zimmermann, P. Zysno, “Latent Connectivities in Human Decision Making”, FSS 4, 1980) Fuzzy Unions
membership by the maximal membership.

1. “The board of directors tries to find the ’optimal’ dividend to be paid to
Aggregation
Operations the shareholders. For financial reasons it should be attractive and for
Aggregation
Operations
In reality of decision making, rarely either happens.
An
reasons of wage negotiations it should be modest.” An
Application: The decision about optimal dividend is based on the fuzzy set Application: (non-verbal) “merging connectives” → (language) connectives
Fuzzy Fuzzy
Morphologies Morphologies {’and’, ’or’,...,}.
µOptimal (x ) = min[µAttractive (x ), µModest (x )].
2. “The teacher has to decide a student’s grade on a written test. The Aggregation operations called compensatory and are
given task was supposed to be solved by Method1 or by Method2. needed to model fuzzy sets representing to, e.g.,
The student provided two solutions, using both methods.”
The decision about his mark is based on a fuzzy set
managerial decisions.
µGoodSol (x ) = max[µGood1stSol (x ), µGood2ndSol (x )].
An example Theory vs. Experiments

and Fuzzy and Fuzzy
Nataša Nataša
Sladoje Sladoje
Introduction
• Two independent criteria, cr1 and cr2 , are used to Introduction

Operations determine the quality of a product. Operations
Fuzzy Intersections • For a number of products, the memberships to three Fuzzy Intersections
Combinations
fuzzy sets are experimentally determined: Combinations
of Operations
µcr1 , µcr2 , and µIdeal ; of Operations
Operations • Numbers: 60 people rating 24 products by assigning a Operations
An
Application:
value between 0 and 100, expressing memberships to : An
Application:
Fuzzy “FulfilledCr1”, “FulfilledCr2”, “IdealProduct”; Fuzzy
• Different theoretical aggregations (minimum,
maximum, arithmetic mean, geometric mean) are
computed and compared with the experimentally
obtained “Ideal”.
Conclusions Aggregation by γ-operator

and Fuzzy and Fuzzy
Nataša Nataša
Sladoje Sladoje
• Geometric mean gives the best prediction of the empirical data,
Types of Set among the operations tested. Types of Set
Fuzzy Complements • Humans use other connectives than “and” or “or”; they need other Fuzzy Complements
Fuzzy Unions aggregation operations than min and max. Fuzzy Unions γ -operation, for γ ∈ [0, 1], is defined as
Combinations • Even if the criteria are independent, and the corresponding fuzzy sets Combinations
of Operations
do not “interact”, the aggregation operation itself can “put them into
of Operations
µAθB = µ1A−γ γ
∩B · µA∪B .
Aggregation
Operations
interaction”. Aggregation
Operations
• Every concrete decision set may require a specific aggregation The value γ = 0.562 is determined from the
An An
Application: operation. Application: experimental data, for the chosen
Fuzzy Fuzzy interpretation of intersection and union.
Morphologies • A general connective is required to involve some parameter Morphologies
corresponding to “grade of compensation”.

• The suggested operation is a weighted combination of
non-compensatory ’and’ and fully compensatory ’or’ (in this work,
they are interpreted as the product and the algebraic sum).
An Application: Fuzzy Dilation and erosion

and Fuzzy and Fuzzy
morphologies Definitions
Nataša Nataša
Sladoje Sladoje
Introduction
Morphological operations Introduction
Types of Set
Operations
Types of Set
Operations Definition
Fuzzy Complements
Fuzzy Intersections
Fuzzy Complements
Fuzzy Intersections
Dilation of a set A by a structuring element B is
Fuzzy Unions
• Mathematical morphology is completely based on set Fuzzy Unions
Combinations
of Operations theory. Fuzzification started in 1980s. Combinations
of Operations
DB (A) = A ⊕ B = {x ∈ X |τx (B̂) ∩ A 6= ∅}.
Aggregation • Basic morphological operations are dilation and Aggregation
Operations Operations τa (S) = {y |y = a + s, s ∈ S}
An
erosion. Many others can be derived from them. An Ŝ = {y |y = −s, s ∈ S}.
Fuzzy • Dilation and erosion are, in crisp case, dual operations Fuzzy
Morphologies
with respect to the complementation: D(A) = c(E(cA)).
Morphologies
Definition
• In crisp case, dilation and erosion fulfil a certain
Erosion of a set A by a structuring element B is
number of properties. EB (A) = A ª B = {a ∈ X |τa (B) ⊆ A}.
Dilation and erosion How to construct fuzzy

and Fuzzy and Fuzzy
An example mathematical morphology

Nataša Nataša
Sladoje Sladoje
Introduction
Dilation and erosion of a shape by a circular structuring Introduction

Operations element. Operations
Fuzzy Intersections
Fuzzy Unions
Fuzzy Intersections
Fuzzy Unions
• Infinitely many fuzzy mathematical morphologies can
Combinations Combinations be constructed.
• It is desirable to understand the differences and to be
able to make choices (of operations, structuring
An
Application:
An
Application: elements,...) in the way that fits the task the best.
Fuzzy Fuzzy
Morphologies Morphologies • Main construction principles:
α-cut decomposition;
fuzzification of set operations.
How to construct fuzzy How to construct fuzzy

and Fuzzy and Fuzzy
mathematical morphology mathematical morphology

Nataša Nataša
Sladoje Sladoje
Introduction
α-cut decomposition Introduction
α-cut decomposition
Fuzzy Complements
Fuzzy Intersections For fuzzy sets µ(x) and ν(x) it holds that
Fuzzy Complements
Fuzzy Intersections
It is shown that this approach leads to definitions:
Z 1
of Operations [µ(x) ⊕ ν(x)]α (x) = µα (x) ⊕ να (x). of Operations
Dν (µ)(x) = sup µ(y )dα
Aggregation Aggregation 0 y ∈(να )x
Operations
Steps to take: Operations
Z 1
An An
Application: • α-cut decomposition of both the set and the structuring Application: Eν (µ)(x) = inf µ(y )dα
Fuzzy Fuzzy 0 y ∈(να )x
Morphologies element; Morphologies
• performance of (crisp) morphological operations; if a reconstruction of a fuzzy set is performed by using

Z 1
• reconstruction of a resulting fuzzy set from its obtained
µ(x) = µα (x)dα
α-cuts. 0
...

and Fuzzy and Fuzzy

Nataša Nataša
Sladoje Sladoje
Introduction
α-cut decomposition Introduction
Fuzzification of set operations
Fuzzy Unions
... or Fuzzy Unions
Perform "fuzzification" of all (set) operations and relations
of Operations of Operations used for definitions of morphological operations.
Aggregation
Operations
Dν (µ)(x) = sup[ν(y − x)µ(y )] Aggregation
Operations
• Replace unions and intersections by t-conorms and
y ∈X
An An t-norms,
Application: Eν (µ)(x) = inf [µ(y )ν(y − x) + 1 − ν(y − x)] Application:
Fuzzy y ∈X Fuzzy • Use fuzzy complementation
• Replace subsethood relation by inclusion indicators
if a reconstruction of a fuzzy set is performed by using
µ(x) = sup [αµα (x)]. • Use duality of morphological operations
α∈(0,1] • ...

and Fuzzy and Fuzzy

Nataša Nataša
Sladoje Sladoje
Introduction
Fuzzification of set operations Introduction
Fuzzification of set operations
Operations
Fuzzy Complements
Operations
Fuzzy Complements
• A fuzzy erosion can be defined as
Eν (µ)(x) = I(ν + x, µ)
Combinations Most general definitions of fuzzy dilation and fuzzy erosion Combinations
of Operations
obtained by fuzzification of set operations are: of Operations
Aggregation Aggregation where I(ν, µ) is an inclusion indicator, a function that

An Dν (µ)(x) = sup i(ν(y − x), µ(y )) An

fulfils a certain set of axioms. (Several exist in the
Application:
Fuzzy
y ∈X Application:
Fuzzy
literature.)
Morphologies Eν (µ)(x) = inf u(c(µ(y ), ν(y − x))), Morphologies Then, dilation is defined by duality principle.
y ∈X
• Alternatively, a fuzzy dilation is defined as
where i is any t-norm and u is its dual t-conorm with respect
to a fuzzy complement c. Dν (µ)(x) = O(ν + x, µ)
where O is an overlapping indicator.

Fuzzy Sets
A task on fuzzy morphology

and Fuzzy
Techniques
Nataša
Sladoje
Introduction
Use the paper:

Types of Set
Operations
Fuzzy Complements
I. Bloch and H. Maitre: “Fuzzy Mathematical Morphologies: A
Comparative Study”, Pattern Recognition, 28(9), 1995.
Fuzzy Intersections
Fuzzy Unions
Combinations • Design fuzzy morphological operations. Use α-cut decomposition principle,

of Operations and fuzzification of set operation by t-norms and t-conorms.
Aggregation • Implement at least three different morphological dilations and corresponding
Operations
erosions; use 2-3 different structuring elements; take simple (one-dimensional)
An examples; look at the examples in the paper.
Application:
Fuzzy • Comment the results. Observe visual differences and think of their effect in
Morphologies possible applications. Study the properties of the morphological operations,
using your obtained results and the paper.
• Make a short summary.

Lecture 8 - Operations On Fuzzy Sets: February 15, 2007

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Lecture 8 - Operations On Fuzzy Sets: February 15, 2007

Hochgeladen von

Copyright:

Verfügbare Formate

Fuzzy Sets Fuzzy Sets

An Centre for Image Analysis An

February 15, 2007 4 Aggregation Operations

5 An Application: Fuzzy Morphologies

Fuzzy Sets Fuzzy Sets

Standard fuzzy operations Properties of the standard

Types of Set Types of Set

for fuzzy sets A, B on a reference set X , given by the

Combinations corresponding membership functions A(x) and B(x): Combinations

Other generalizations of the set Fuzzy complements

Types of Set Types of Set

• A fuzzy complement cA of a fuzzy set A is given by a

Combinations Combinations • Function c assigns a value to each membership value

Aggregation Aggregation A(x) of x ∈ X to the fuzzy set A.

Fuzzy Sets Fuzzy Sets

Fuzzy complements Nested structure of the basic

Axiomatic requirements classes of fuzzy complements

Types of Set Types of Set • All functions c : [0, 1] → [0, 1]

• All fuzzy complements (Ax c1 and Ax c2)

An An • All continuous fuzzy complements (Ax c1- Ax c3)

Introduction • Increasing generator Introduction • Decreasing generator

g(a) = ap , p > 0 f (a) = 1 − ap , p > 0

Fuzzy Sets Fuzzy Sets

Generators Generating fuzzy complements

Introduction Theorem Introduction

Generating fuzzy complements Fuzzy intersections

Introduction Introduction An intersection of two fuzzy sets A and B is given by a

i : [0, 1] × [0, 1] → [0, 1].

Standard fuzzy complement: g(a) = a.

Fuzzy Sets Fuzzy Sets

Fuzzy intersections Fuzzy intersections

Types of Set Types of Set

Fuzzy intersections Fuzzy intersections

Types of Set Types of Set

Fuzzy Sets Fuzzy Sets

Fuzzy intersections Fuzzy unions

Introduction Introduction A union of two fuzzy sets A and B is given by a function of

Fuzzy unions Fuzzy unions

Types of Set Types of Set For all a, b, d ∈ [0, 1],

Fuzzy Sets Fuzzy Sets

Fuzzy unions Fuzzy unions

Types of Set Types of Set

Fuzzy unions Combinations of set operations

binary operation on the unit interval. Then, u is an

Fuzzy Sets Fuzzy Sets

Duality of fuzzy set operations Dual triples - Six theorems (1)

i (a, b) = c (u (c (a), c (b)))

is a t-norm such that hi , u , c i is a dual triple.

Dual triples - Six theorems (2) Aggregation operations

Types of Set Types of Set

Aggregation Theorem Aggregation

Fuzzy Sets Fuzzy Sets

Axiomatic requirements Averaging operations

Combinations Combinations min(a1 , a2 , . . . , an ) ≤ h(a1 , a2 , . . . , an ) ≤ max(a1 , a2 , . . . , an ).

Aggregation h is monotonic increasing in all its arguments. Aggregation

Averaging operations Ordered Weighted Averaging

Generalized means: For

Fuzzy Sets Fuzzy Sets

Norm operations Norm operations

Introduction Aggregation operations h on [0, 1]2 which are Introduction

Combinations Combinations for a, b ∈ [0, 1] and λ ∈ (0, 1)

An An h(a,b) = max[λ, i(a, b)] when a, b ∈ [λ, 1]

Averaging operations Decision as a fuzzy set