Chapter 5 : Fuzzy Relation

5.1 Crisp and Fuzzy Relation

Crisp relation:
+ degrees or strengths of relation
Fuzzy relation
。Cartesian product :
 X i  {( x1 , , xn ) | xi  X i , i  N n }
iN n

N n  {1, 2, , n}

。n-ary relation: a subset of iN X i n

R ( X 1 , X 2 , , X n )  X 1  X 2  X n
i.e., | |

a set the universal set

Characteristic function:
1 if ( x1 , x1 ,  xn )  R
 R ( x1 , x1 ,  xn )  
0 otherwise

。Binary, Ternary, Quaternary, Quinary, n-ary

relations
· Representation of a relation
 
R ( X 1 ,..., X n )
 

(ri1 ,i2 ,..,in ) : n-D membership array

ri1 ,i2 ,..,in = 1 iff ( x1 ,..., xn )  R

0 otherwise
○ Example 5.1 :
 
R ( X Y ) Z )Y1Y2Y3Y4 Z 2 Z1Z 3 Z 4 Z 5

  
R ({ X 1 , X 3} , X i J  N nY  X
  
 1 1 1 1 0.8 0.8 0.8 0.8
{ X i | j  J  N n }  R2        
 X , a,* X , a,$ Y , b,* Y , a,$ X , b,* X , b,$ Y , b,* Y , b,$
 
 y  { X 1 , X 3}
 

Y j  X j j  J
 
[ R  X  Y ]:[ R2  { X 1 , X 3}]
 
  
 y  { X 2 }, X  Y  { X 1 , X 3 }  {(*,*), ( x,$), (Y ,*), (Y , s)}
  


( x)  R ( y )  [ R  { X i }](Y )  max R( x) Rij
 x y


 Y j | j  J  X X j
jJ


1 0.7 0.4 0.8

R2,3     .
a,* a,$ b,* b,$
0.9 0.4 1 0.7 0.8
   
X , a,* X , b,* Y , a,* Y , a,$ Y , b,$
0.9 0.4 1 0.8 0.9 0 1 0.8
 R1,2     R1,3    
X , a X ,b Y , a Y ,b X ,* X ,$ Y ,* Y ,$
1 0 0.6 0.9 0.7
0  R ( x1 , x2 ,..., xn )  1     
( NY , Beijing ) ( NY , NY ) ( NY , London) ( Paris, Beijing ) ( Paris, NY ) ( Paris

X

={English , French} , Y ={ dollar , pound , franc , 
mark}
Z={US , France , Canada , Britain , Germary}
 
R( X Y Z )

={(English , dollar , US) , (French , franc ,
France)
(English , dollar , Canada) , (French ,
dollar , Canada)
(English , pound , Britain)}

Y1 Dollar 1 0 1 0 0 Dollar 0 0 1
0 0
Y2 Pound 0 0 0 1 0 Pound 0 0 0
0 0
Y3 Franc 0 0 0 0 0 Franc 0 1 0
0 0
Y4 Mark 0 0 0 0 0 Mark 0 0 0 0
0
US Fran Can Brit Ger US
Fran Can Brit Ger
Z1 Z2 Z3 Z4 Z5 Z1 Z2 Z3 Z4
Z5

English
Franch
X1 X2
 · Fuzzy Relations
  

Cartesian Product : X 1 X 2  X n

  

tuples : ( x1 , x2 ,..., xn )

membership grade :
0  R ( x1 , x2 ,..., xn )  1

Example 5.2: Birary relation R : represents the

concept “ very far”

X

= { New York , Paris}
Y

={Beijing , New York , London}
Relation in list notation
 
R( X Y )

=
1 0 0.6 0.9 0.7 0.3
    
( NY , Beijing ) ( NY , NY ) ( NY , London) ( Paris, Beijing ) ( Paris, NY ) ( Paris, Londom)

Relation in membership array

NY Paris
Beijing 1 0.9
NY 0 0.7
London 0.6 0.3
· Ordinary fuzzy relation
with valuation set [0,1]
L-fuzzy relation
With ordered valuation set L
5.2 Projection and Cyclindric Extensions


· set family X = { X 
i | i  Nn}

Let X =  xi | i  N n   X X i
jJ 

Let Y =  Yj | j  J   X X j
jJ


Where J  Nn ,|J|=r
Y a subsequence of X , YX

iff Y j  X j j  J

⊙ Projection : [ R  y] the projection of R on Y

  
R ( X 1 , X 2 , , X n )
  
: a relation


Y= { X 
i | j  J  Nn}

[ R  y] : a fuzzy relation (set)

[ R  y ](Y )  max R ( x )
x y

※ max can be generalized by other t-conorms

   

· Example 5.3 X1

={X,Y}, X ={a,b}, X ={*,$}

2

3

   0.9 0.4 1 0.7 0.8

R( X 1 , X 2 , X 3 )
  
= X , a,*  X , b,*  Y , a,*  Y , a,$  Y , b,$
  

Let R = [ R  { X , X }] ,
ij

i

j Ri  [ R  { X i }]


0.9 0.4 1 0.8

 R1,2    
X , a X ,b Y, a Y ,b
0.9 0 1 0.8
R1,3    
X ,* X ,$ Y ,* Y ,$
1 0.7 0.4 0.8
R2,3    
a,* a,$ b,* b,$
0.9 1
R1  
* y
1 0.8
R2  
a b
1 0.8
R3  
* $

⊙Cyclindric Extension [R  X  Y ] the CE of R into

X-Y


X-Y : sets Xi

that are in X but are not in Y
[ R  X  Y ]( x )  R ( y )

R: a relation defined on Y
·Example 5.4 ( Refer to example 5.3)
  
Let X = { X , X , X } 
1

2

3

 
And R= R 1,2  y  { X 1 , X 3}
 

∴ X-Y =

X3

= {*,$}
0.9 0.4 1 0.8
From example 5.3 R1,2    
X , a X ,b Y, a Y ,b

∴[R  X  Y ]  [R


1,2  { X 3}]

=
0.9 0.9 0.4 0.4 1 1 0.8 0.8
      
X , a,* X , a,$ X , b,* X , b,$ Y , a,$ Y , a,$ Y , b,* Y , b,$

    
[ R  X  Y ]:[ R12  { X 3}] [ R13  { X 2 }] [ R23  { X 1}] [ R1  { X 2 , X 2 }]
    

   
[ R2  { X 1 , X 3}] [ R3  { X 1 , X 2 }]
   

 
Consider [ R  X  Y ] = [ R 2  { X 1 , X 3}]
 

  
 y  { X 2 }, X  Y  { X 1 , X 3}  {(*,*), ( x,$), (Y ,*), (Y , s)}
  

{x,y} {x,$}
1 0.8
R= R 2  
a b

∴[R
 

2  { X 1 , X 3}]
 
=
1 1 1 1 0.8 0.8 0.8 0.8
      
X , a,* X , a,$ Y , b,* Y , a,$ X , b,* X , b,$ Y , b,* Y , b,$
5-7
Cyclindric closure
-A relation may be exactly reconstructed from
several of its projections by taking the set
intersection of their cyclindric extensions  Pi | i  I  :a

set of projections of a relation on X

 cyl{Pi }( X )  min[ Pi  X  Yi ]( X )  R
iI

Yi ：The family of sets on which Pi is defined.

‧ Example：
0.9 0.4 1 0.7 0.4 0.8
Cyl{R1, 2 , R1, 3 , R2 , 3 }      
x, a, x, b, y , a, y , a,$ y , b, y , b,
_ _ _
Refer to the original relation R( X , X , X )
 1  2  3
in example
5.3.
It is not fully reconstructable from its projections
become of ignoramus of R1 , R ,and
2
R3 .
5-8
 
5.3. Binary Relations R( X , Y )  

 
X  Y ：bipartite
 
graph
 
X  Y ：directed
 
graph
‧ Representations
i, matrices R  [rij ] , where rij  R ( xi , y j )

ii, sagittal diagrams

Examples：
i) y1 y2 y3 y4 y5

x1
.9 1 0 0 0
x2 0 .4 0 0 0
x3 0 0 1 .2 0
x4 0 0 0 0 .4
x5 0 0 0 0 .5
0 0 0 0 .2
x6

ii)
5-9
‧ Domain：dom R
Crisp – dom R = {x  X | ( x, y)  R, y  Y }
Fuzzy – dom R(x) = max R ( x, y )
yY

The domain of a fuzzy relation R(x,y) is a fuzzy set

on X; dom R(x) is its membership function.
e.g. dom R( X ) = max(0.9, 1) = 1
1

‧ Range：ran R
Crisp – ran R = { y  Y | ( x, y)  R, x  X }
Fuzzy – ran R(y) = max R( x, y )
xX

e.g. ran R( y ) = max(0.4, 0.5, 0.2) = 0.5

5

‧Height： h( R)  max max R( x, y)

yY xX

e.g., h( R )  1 normal fuzzy relation

5-10
‧Inverse： R 1
(Y , X )

R 1 ( y, x )  R ( x, y )

 R 1  R T , ( R 1 ) 1  R

0.3 0.2
e.g. R
 0 1 


0.6 0.4


 0.3 0 0.6
R 1  R T  
0.2 1 0.4

‧ Composition： R( X , Z )  P( X , Y )  Q(Y , Z )
R( x, z )  [ P  Q]( x, z )  max min[P( x, y ), Q( y, z )]
yY

Max-min composition
 P。Q  Q。P

Properties： 1 1
 ( P。Q )  Q 。P
( P。Q。
1

 ) R  P。(Q。R )

Matric form： [r ]  [ p ij ik ] [ q kj ]
。

Where [r ]  max min( p

ij
k
ik , qkj )
5-11
R ( x, z )  [ P。Q]( x, z )  max[ P( x, y‧
) Q( y , z )]
yY

max-product composition
matrix form [ rij ]  [ pik 。
] [ q kj ]

Where [rij ]  max( pik , qkj )

‧ Example
 0 .3 0 .5 0 .8   0 .9 0 .5 0 .7 0 .7 
 0 .0 0 .7 1 .0   0 .9 
 。0.3 0 .2 0 .0 

 0 .4 0 .6 0 .5 
 
1.0 0 .0 0 .5 0 .5 


 0 .8 0 .3 0.5 0 .5 
1.0 0.7 
Max-min =  0.2 0.5 
 0 .5
 0.4 0.5 0 .6 


0.8 0.15 0.4 0.45

1.0 0.63
Max-prod =  0.14 0.5 

0.5 0.2 0.28 0.54


‧ Relational join： R( X , Y , Z )  P( X , Y ) * Q(Y , Z )

R ( x, y, z )  [ P * Q ]( x, y , z )  min[P ( x, y ), Q( y, z )]

※The max-min composition can be obtained by

aggregating appropriate elements of the
corresponding join.
5-12
‧Example
※ [ P。Q]( x, z )  max[ P * Q]( x, y, z )
yY

5.4Binary Relation on a Simple Set

‧Representations
5-14
◎ characteristic Properties (Crisp case)
i, reflexive

irrflexive 

antiflexive 

ii, symmetric

asymmetric 

antisymmetric ( x, y )  R, ( y, x)  R  x  y

strictly antisymmetric x  y, ( x, y )  R or ( y, x)  R

iii, transitive

nontransitive 
 antitransitive 

◎ Fuzzy Relations
i, reflexive --- x , R ( x, x )  1

irreflexive --- x , R ( x, x )  1

antiflexive --- x , R ( x, x )  1

 -reflexive --- x , R ( x, x )  
 5-14
ii, symmetric -- x, y, R ( x, x)  R( y, x)

asymmetric -- x , y , R ( x, x )  R ( y , x )

 R ( x, x )  0
antisymmetric -- 
 R ( y , x)  0
x y

iii, max-min transitive -- x, z

R ( x, z )  max

min  R( x, y ), R( y, z ) 
yY

max-product transitive -- x , z

R ( x, z )  max

min  R( x, y )  R( y, z ) 
yY

nontransitive -- x , z

R ( x, z )  max

min  R( x, y ), R( y, z ) 
yY

antitransitive -- x, z

R ( x, z )  max

min  R( x, y ), R( y, z ) 
yY

◎ Example 5.7 R:very near

 reflexive, symmetric, nontransitive
5-15
◎Summary

Antireflexive

Symmetric

Antisymmetric

Transitive
Reflexive

Crisp:
equivalence;
Fuzzy:
similarity
Quasi-
equivalence
Compatibilit
y or
Tolerance
Partial
ordering
Preordering
or
Quasi-
ordering
Strict
ordering
Figure3.6 Some important types of binary relation
R(X,X)
◎ transitive closure: 
RT ( X )


Algorithm for computing RT

1. R /  R  ( R  R)

2. If R/  R , Let R/  R , go to step1
3. Stop, RT  R /

Where : component-wise max

5-16
◎Example 5.8
0.7 0.5 0.0 0.0 
 0.0 0.0 0.0 1.0 
R
 0.0 0.4 0.0 0.0 
 
 0.0 0.0 0.8 0.0 
0.7 0.5 0.0 0.5 
 0.0 0.0 0.8 0.0 
Step1: RR  
 0.0 0.0 0.0 0.4 
 
 0.0 0.4 0.0 0.0 
 0.7 0.5 0.0 0.5
 0.0 0.0 0.8 1.0 
R  ( R  R)    R/
 0.0 0.4 0.0 0.4
 
 0.0 0.4 0.8 0.0

Step2: R /
R , Let R  R/

repeat step1
0.7 0.5 0.5 0.5 
 0.0 0.4 0.8 0.4 
RR  
 0.0 0.4 0.4 0.4 
 
 0.0 0.4 0.4 0.4 
 0.7 0.5 0.5 0.5
 0.0 0.4 0.8 1.0 
R  ( R  R)    R/
 0.0 0.4 0.4 0.4 
 
 0.0 0.4 0.8 0.0 

Step3: R /
R , Let R  R/

repeat step1
0.7 0.5 0.5 0.5 
 0.0 0.4 0.8 1.0 
R 
'
R
 0.0 0.4 0.4 0.4 
 
 0.0 0.4 0.8 0.4 

Step4: Stop
RT  R /
 5-17
5.5 Fuzzy Equivalence Relation
◎Crisp binary relation
equivalence: reflexive, symmetric, and
transitive
equivalence classes
partition: X/R

◎ Example 5.9:

X   1, 2, ,10


 
R( X  X ) 
 
{ ( x, y) | x, y have the same remainder when
divided by 3}

R: reflexive, symmetric, transitive

 equivalence

partition X / R   (1, 4, 7,10), (2,5,8), (3, 6,9)

 5-18
◎ Fuzzy Binary Relation
。 Fuzzy
Similarity relation equivalence relation
Similarity classes equivalence classes
2 Interpretations of a similarity relation:
1.Group similar elements into crisp classes
whose members are similar to each other
to some specified degree.
2. 
x  X ,

associate a fuzzy set Ax

defined on X . 

。a fuzzy relation (Theorem 2.5, Eqs.


R  
 [0,1]
R

(2.1)(2.2))
If R: Similarity relation,
 R : equivalence relation


。Let  (  R) : the partition of 

X

w.r.t. 
R

  ( R)    (  R) |    0,1 

 (  R) : nested, i.e.,

 (  R) : a redefinement of  (  R) iff   
Prove that :A fuzzy relation R: X  X is a similarity
relation, then 
R is a equivalent relation
Pf : ∵ R : a similar relation
∴ R : reflexive, i.e., x  X , R ( x, x )  1

symmetric, i.e., x, y  X , R( x, y)  R( y, x)

transitive, i.e., x, z  X , R( x, z)  max[ R( x, y), R( y, z)]
2
yY

i, 
R : reflexive
x  X , R ( x, x)  1    [0,1], ( x, x)   R


R : reflexive

ii, 
R : symmetric
∵ R : symmetric
x, y  Z , R ( x, y )  R ( y , x)

Let R ( x, y )  R ( y , x )  

Then   or  

a, if   => ( x, y ), ( y, x)   R

b, if   => ( x, y ), ( y, x)   R

iii, 
R : transitive
∵ R : transitive
 x, z  X 2 , R ( x, z )  max[ R( x, y ), R( y, z )]
yY

Let R ( x, y )  1 , R ( y, z )   2

Assume 1   2

Then    1  2 , 1     2 , or 1   2  

a. if    1  2 => ( x, y )  R, ( y , z )   R --- (A)

R( x, z )  max[ R( x, y ), R ( y, z )]  min[ 1 ,  2 ]  1  
yY

 R ( x , z )   , ( x, z )   R --- (B)
(A) , (B) => 
R : transitive

b. if 1     2

 ( x, y )   R, ( y , z )   R , don’t care ( x, z )

c. if 1   2  

 ( x, y )   R, ( y , z )   R , don’t care ( x, z )

Example 5.10 : R( X , X ) : a fuzzy relation

R : reflexive , symmetric , transitive (
R '  R  ( R  R )  R )

∵ level set :  R  {0.0, 0.4, 0.5, 0.8, 0.9,1.0}

There are five nested partition 

 's

 The similarity class for each element is a fuzzy

set defined by the row of the membership matrix
corresponds to that element

Example : see Example 5.10

0 0 1 0 1 0.9 0.5
For c :     
a b c d e f

g
0 0 1 0 1 0.9 0.5
For e :     
a b c d e f

g

∴ c and e are similar at any level 

5.6 Compatibility Relations ---- reflexive ,

symmetric
compatibility
Alternatives : tolerance relation
proximity

 Crisp case :
 Maximal compatibility classes – not properly
contained within any other compatibility class
Complete cover – all the maximal compatibility
classes

 Fuzzy case :
α-compatibility class ---- a subset A of X ,
s.t.
x, y  A, if R ( x, y )  R( x, y )   , R : fuzzy compatibility
relation
maximalα ---- compatibility classes
completeα-cover

Example 5.11 : R( X , X ) : a fuzzy relation

∵ R : reflexive , symmetric
∴ a compatibility relation
∵ R  {0.0, 0.4, 0.5, 0.7, 0.8,1.0}

=> the completeα-covers

5-25

  0.5
0.7 0.9

b d
a  1 0.7 0 1 0.7 
b  0 1 0 0.9 0 
c 0.5 0.7 1 1 0.8 
 
d 0 0 0 1 0 
e  0 0.1 0 0.9 1 
1 1 0 0 0 0 0 0 0
1 1 0 0 0 0 0 0 0 

0 0 1 1 1 0 0 0 0
e.q.     0.4
0 0 1 1 1 1 1 0 0
0.5
R  0 0 1 1 1 1 1 1 0
 
0 0 0 1 1 1 0 0 0
0 0 0 1 1 0 1 0 0
 
0 0 0 0 1 0 0 1 0
0 0 0 0 0 0 0 0 1 


 x  y  xx  yy  xx  X ( x, y )  XS  {x1 , x2 }  Xy  Ax  y
(x,y)y  X ( x  y, or  y  xAx  XA  Xx  yR[ x ] ( y )  R ( y , x )
x  U ( R, A)( x )   R[ x ]y 
xA

1 1 0 0 0 0 0 0 0
1 1 0 0 0 0 0 0 0 

0 0 1 1 1 0 0 0 0
 
0 0 1 1 1 1 1 0 0
0.4
R  0 0 1 1 1 1 1 1 0
 
0 0 0 1 1 1 1 0 0
0 0 0 1 1 1 1 0 0
 
0 0 0 0 1 0 0 1 0
0 0 0 0 0 0 0 0 1 


Look for complete subgraphs

(1,2) , (3,4,5),(4,5,6,7),(5,8),(9)
(34,),(4,5,6),(4,5,7),(3,5),(5,6)
(4,5),(5,6,7),(4,6,7),(4,6),(6,7)
 maximal compatible classes (the complete 0.4-
cover):
(1,2),(3,4,5),(4,5,6,7),(5,8),(9)
These do not partition X.
5-26

e.q.   0.5

1 1 0 0 00 0 0 0
1 1 0 0 00 
0 0 0

0 0 1 1 10 0 0 0
 
0 0 1 1 10 1 1 0
0.5
R  0 0 1 1 10 1 1 1
 
0 0 0 1 10 1 0 0
0 0 0 1 10 0 1 0
 
0 0 0 0 10 0 0 1
0 0 0 0 0 0 0 0 1 


 maximal compatible classes

( The complete 0.5-cover)
(1,2),(3,4,5),(4,5,6),(4,5,7),(5,8),(9)
5-27

5.7. Ordering Relations

˙ partial ordering: reflexive , antisummetric ,
transitive

X Y : X : predecessor

precedes Y : successor

if exist First member : if x  y y  X (minimum)

unique Last member : if y  x y  X (maximam)

may not Minimal member : if yxx y

be unique Maximal member : if

x  y  x  y

˙properties :
1, if  , at most one first member
if  , at most one last member
2, There may be several maximal and minimal
member
3, if  a first member X ,  only one minimal
member Y exists and x=y
4, if  a last member x ,  only one maximal
member Y exists and x=y.
5, partial the first member  the last member
inverse
ordering the last member  the first member
partial ordering
5-28

※ In a partial ordering , it does not guarantee that

(x,y) , ( x  y, or y  x ).

If  ,  (x,y) : comparable (total ordering)

Otherwise (x,y) : non comparable

˙ A X

If x X , and y  A , x y ,
 x: lower bound of A on X
If ……… , x y

 x : upper bound of A on X

˙greatest lower bound ( or infimum ) GLB

- a lower bound which succeeds every other lower
bound

Least upper bound ( or supermum ) LUB

- a upper bound which preceeds every other upper
bound
˙Lattice – A partial ordering on X contains GLB
and LUB , S  {x1 , x2 }  X
5-29

˙ Connected – a partial ordering is said to be

connected
iff x, y  X , x y  x<y or y>x

˙ Linear ordering (total ordering , simple ordering ,

complete ordering )
- when a partial ordering is connected , then ( x, y ) :
comparable

˙ Hasse diagrams – representing partial orderings in

which  indicates 

˙ Example 5.12 : Crisp partial orderings

5-30

˙ Fuzzy partial ordering

- reflexive , antisymmetric , and transitive under
some form of transitivity.

※ any fuzzy partial ordering can be resolved into a

series of crisp partial ordering .

i.e. taking a series of  cut that produce increasing

levels of refinement

˙ In a fuzzy partial ordering , R

x  X , two fuzzy sets are associated with
R[ x ] : dominating class
R[ x ] ( y )  R( x, y )

R[ x ] : dominated class

R[ x ] ( y )  R( y, x)
5-31

˙ x undominated iff R(x,y) = 0 y  x

X undominating iff R(y,x) = 0 y  x

˙ Fuzzy upper bound for A X is a fuzzy set

U ( R, A)   R[ x ]
xA

※ If a least upper bound of A exists , it is the

unique element x  U ( R, A)

s.t. 1, 2,
U ( R, A)( x) >0 R(x,y) > 0 ,
y  support [ U(R,A) ]

˙ Example 5.13
a b c d e
a  1 0.7 0 1 0.7 
b  0 1 0 0.9 0 
Fuzzy partial ordering R: c  0.5 0.7 1 1 0.8 
 
d 0 0 0 1 0 
e  0 0.1 0 0.9 1 

1. row : dominating class for each element

column : dominated class for each element
2. d : undominated , C : undominating
3. For A = {a,b} , U(R,A) = the intersection of
0.7 0.9
The dominating classes of a and b = b

d

4, LUB(A) =b
5-32
5. Crisp ordering captured by the fuzzy ordering
e.g.   0.5
1 1 0 1 1
0 1 0 1 0


R  1 1 1 1 1
 
0 0 0 1 0
0 0 0 1 1

# is → 2 3 1 5 3
※The ordering become weaken with the increasing
α
5-33
Fuzzy preordering – reflexive and transitive
Fuzzy weak ordering –
i, an ordering satisfying the proportion of a fuzzy
total ordering except antisymmetry.
ii, a fuzzy preordering in which x  y , either
R(x,y)>0 or R(y,x)>0

Fuzzy strict ordering –

Antireflexive
Antisymmetric
Transitive

5.8. Morphisms
‧Crisp homomorphism h from (X,R) to (Y,Q)
Where R(X,X), Q(Y,Y)：binary relations
( x1 , x2 )  R  (h( x1 ), h( x2 ))  Q

‧ Fuzzy homomorphism h
If R(X,X), Q(Y,Y)：Fuzzy binary relations
And R ( x1 , x2 )  Q[ h( x1 ), h( x2 )]
※ It’s possible that a relation ( h( x1 ), h( x2 ))  Q which
( x1 , x2 )  R .
※ If this is never the case h is called a strong
homomorphism.
5-34
‧Crisp strong homomorphism h
If ( x1 , x2 )  R  (h( x1 ), h( x2 ))  Q

And ( y1 , y 2 )  Q  ( h 1 ( y1 ), h 1 ( y 2 ))  R

※ where h : many to one → h 1 ( y ) contains a set of

Xs

‧Fuzzy strong homomorphism h

H imposes a partition h on X
Let
A  {a1 , a2 ,  , an }

B  {b1 , b2 ,  , bn }   h

R,Q：fuzzy relations
h : strong homomorphism
max ( R (ai , b j ))  Q ( y1 , y 2 )
iff i, j

 y1  h(ai )ai  A
where  y 2  h(b j )b j  B
5-35
‧Example 5.14
R(X,X)
0 0 .5 0 0 
0 0 0.9 0 
R
1 0 0 0.5
 
0 0 .6 0 0 

Q(Y,Y)
0.5 0.9 0 
Q
 1 0 0.9

 1
 0.9 0 


→h：ordinary fuzzy homomorphism (one way)

R ( x1 , x2 )  Q(h( x1 ), h( x2 ))  strong

But Q ( ,  )  0.9 R ( d , c)  0

i,e, ( ,  )  Q while (d , c)  R where h( d )   , h( c )  

5-36
R(X,X)
0.8 0.4 0 0 0 0 
 0 0.5 0 0.7 0 0 
 
 0 0 0 .3 0 0 0 
 
 0 0.5 0 0 0.9 0.5
 0 0 0 1 0 0 
 

 0 0 0 0 1 0.8


Q(Y,Y)
 0 .7 0 0 .9 
 0 .4 0 .8 0 
 
 1
 0 1  

→h：strong fuzzy homomorphism (two way)

5-37
※Q represents a simplification of R
‧Isomorphism : (congruence)
h:1-1, onto  X Y

Endomorphism : (subgraph)
h:X→Y, Y  X

Automorphism :
Isomorphism and End Endomorphism
i.e.m X=Y nad R=Q
5-38
5.9 SUP-i Compositions of Fuzzy Relations
Generalize max-min Composition
i : t-norm
sup : t-conorm
‧ P(X,Y), Q(Y,Z)：fuzzy relations
:sup-i composition
i
P o Q( X  Z )
i
[ P o Q]( X , Z )  sup i[ P( x, y ), Q ( y, z )]
yY

‧ Properties
1.
i i i i
( P o Q ) o R  P o( Q o R )
i i
2. P o( Q j )  ( P o Q j )
j j

i i
3. P o( Q j )  ( P o Q j )
j j

i i
4. ( Pj ) o Q  ( Pj o Q)
j j

i i
5. ( Pj ) o Q  ( Pj o Q)
j j

6.
i i
( P o Q) 1  Q 1 o R 1
5-39
i i

3. Show Eq.(5.16), i.e., P ( Q j )  ( P  Q j ),

jJ jJ

 
Where and are fuzzy relations.

P( X , Y ) Q(Y , Z )


 i 
pf. From Eq.(5.13), i.e.,  P  Q  ( x, z )  sup i  P( x, y ), Q( y, z ) 
yY

 i   
  P ( Q j )  ( x, z )  sup i  P( x, y ), Q j ( y, z ) 
 jJ  yY 

jJ 

Q  Q j
Let jJ

 Q  Q1 , Q  Q2 , , Q  Q J

i.e., ( y, z ), Q( y, z )  Q1 ( y, z ), , Q( y, z )  Q J ( y, z )

 i is monotonically increasing
i[ P ( x, y ), Q j ( y , z )]  i[ P( x, y ), Q1 ( y, z )]
 jJ

 ........... ( x, y )
i[ P ( x, y ), Q ( y , z )]  i[ P( x, y ), Q ( y, z )]
  j J
 jJ

i[ P( x, y ), ( Q j )( y, z )]  i[ P( x, y ), Q j ( y, z )]
jJ jJ

 sup i[ P ( x, y ),( Q j )( y, z )]  sup i[ P( x, y), Q j ( y, z )]

 
yY jJ yY jJ

 sup i[ P( x, y ), Q j ( y, z )], ( x, y ), ( y, z )

jJ yY

 i   i 
  P ( Q j )  ( x, z )   ( P  Q j )  ( x, z ), ( x, z )
 jJ   jJ 
i i

i,e., P ( Q j )  ( P  Q j )
jJ jJ
5-40
。Sup-i composition monotonically increases
i i
P  Q1  P  Q2    (5.20)
i.e., i i if Q1  Q2
Q1  P  Q2  P    (5.21)

。Identity of i


1 0 
1, x  y 
E ( x, y )     
0, x  y 0
 1 

i.e., i
EP  PE  P
i

 2
。Relation R on X

: i-transitive

iff R ( x, z )  i  R ( x, y ), R( y, z )  , x, y, z  X


i
 RR  R

。i-transitive closure RT ( i )

--- The smallest i-transitive relation containing R

。Theorem 5.1: R: any fuzzy relation

 RT ( i )  R ( n )
n 1
, where
i
R ( n )  R  R ( n 1)
5-41
By (5.15) (5.17)

proof:
i   (n )  i   (m )   
 (n ) i (m ) 


i, RT ( i )  RT ( i )   R    R   
 n 1   m 1  n 1

m 1
 R R    R
  n ,m 1
(nm)

 
 R ( k )  R ( k )  RT (i )
k 2 k 1

i.e., RT ( i ) : i-transitive (R T (i )

i
 RT (i )  RT (i ) )
(5.20)(5.21) monotonically increasing
ii, Let S: i-transitive, RS

i i
 R (2)  R  R  S  S  S
mathematical
i-transitive induction
If R(n)  S ,

i i
 R ( n 1)  R  R ( n )  S  S  S

 R ( k )  S , k


 RT (i )  R ( k )  S
k 1

i.e., RT ( i ) : smallest

。Theorem 5.2: R: reflexive fuzzy relation on  2

X

,

X n


 R ( m )  R m 1 
 RT ( i )  R ( n 1)   n n 1
 m
 R  R 
5-42
proof : i, R : reflexive,
i i
 E  R, R  E  R  R  R  R (2)

 R ( n 1)  R ( n ) (By repetition)

ii, show R ( n 1)  R ( n )

proof: If x  y ,  R ( n 1) ( x, x)  1

reflexive
If x  y, i
Extension of definition


R ( n ) ( x, y )  sup i  R( x, z1 ), R( z1 , z2 ), , R( zn 1 , y ) 
Z1 ,, Z n1


X n


 X  Z 0 , Z1 , , Z n  y contains

at least 2 identical element.

Say Z r  Z s (r  s )

 i  R ( x, z1 ), , R ( zr 1 , z r ), , R( z s , z s 1 ), , R ( z n1, y )   R ( k ) ( x, y ), (k  n  1)


x, y  X , R ( n ) ( x, y )  R ( n 1) ( x, y ),


 R ( n )  R ( n 1)    ( B)  R ( n )  R( n 1)  ( A, B)

 RT ( i )  R ( n 1)
5-43
5.10 INF- w Compositions of Fuzzy Relations
i

a b b
。 w operation:
i a  b   1

wi (a, b)  sup  x  [0,1] | i( a, x)  b

where a, b  [0,1] , i : continuous t-norm

※ If i : logical conjunction (i.e.,  , and)
 w : logical implication (i.e.,  , if then)
i

。Theorem 5.3
1, i ( a, b)  d iff wi (a, b)  b

2, wi ( wi (a, b), b)  a

3, wi (i (a, b), d )  wi (a, wi (b, d ))

4, a  b,  wi (a, d )  wi (b, d ) ---- i

wi (d , a )  wi (d , b) --- ii

5, i ( wi (a, b), wi (b, d ))  wi (a, d )

6, wi (inf a j , b)  sup wi (a j , b)
j j

7, wi (sup a j , b)  inf wi (a j , b)
j j

8, wi (b,sup a j )  sup wi (b, a j )

j j

9, wi (b,inf a j )  inf wi (b, a j )

j j
10, i (a, wi (a, b))  b
5-44
proof: (1) i , If i ( a , b )  d ,  b   x | i ( a, x )  d 

( )  b  sup  x | i (a, x )  d   wi (a, d )

ii, If b  wi ( a, d ) i: continuous
monotone
d
i: monotone increasing
()

 ib(a, b)  ai (ad , wiBy , d ))  i (a,sup  x | i (a, x)  d  )  sup  i (a, x) | i (a, x)  d   d

(a(1)<=
b

(3)
i (a, x )a wi (bb, d d) Byi ((1)=>
b, i (a, x))  d

Associativity
communitation

 i (i (a, b), x)  d  x  wi (i (a, b)d )

By (A)
wi (i (a, b), d )

 wi (a, wi (b, d ))  sup  x | i (a, x )  wi (b, d )  sup  x | x  wi (i (a, b), d )  wi (i (a, b), d )

(7) Let By(4)

S  sup a j
j
---(B)  a j  s, j

 wi ( s, b)  wi (a j , b), j

 wi ( s, b)  inf wi (a j , b)
j
---- (C)
inf wi (a j , b)  wi (a j0 , b), j0  J
j
By(1)

 i ( a j0 ,inf wi (a j , b))  b, j0

j

 i ( s,inf wi (a j , b))  sup i (a j0 ,inf wi (a j , b))  b

j j0 j

By(1)

 wi ( s, b)  inf wi ( a j , b)
j
--- (D)
By(B)(C)(D)
 wi (sup a j , b)  wi ( s, b)  inf wi ( a j , b)
j
j
5-45
(2)Show wi ( wi (a, b), b)  a (Theorem 5.3 (2))
proof : w (a, b)  Sup  x [0,1] | i(a, x)  b and by Theorem
i

3.10
imin (a, b)  i (a, b)  min(a, b)

i, If a>b
wi (a, b)  Sup  x  [0,1] | i (a, x)  b  Sup  x  [0,1] | min(a, x)  b  b

i ( wi (a, b), a )
 i (b, a ) By Axiom i2 wi (a, b)  b

 i (b,1) By Axiom i2 ( a  1 )
b By Axiom i1
 i ( wi (a, b), a)  b

i, If ab

wi (a, b)  Sup  x  [0,1] | i (a, x)  b  Sup  x  [0,1] | min(a, x)  b  1

 i ( wi (a, b), a )  i ( wi (a, b), b)

 i (1, a ) By Axiom i2 wi (a, b)  1

 i (b,1) By Axiom i2
b By Axiom i1

 i ( wi (a, b), a)  b

By Theorem 5.3 property 1(i.e., i ( a, b)  d iff

wi (a, d )  b )
i ( wi (a, b), a )  b  wi ( wi (a, b), b)  a
(4) prove Theorem 5.3 (4) : ab => i, W (a, d )  W (b, d )
i i

ii, W (d , a)  W (d , b)
i i

proof : i, W (a, d )  W (b, d )

i i

Wi ( a, d )  sup{x | i (a, x)  d } ---- (A)

Wi (b, d )  sup{x | i(b, x)  d } ---- (B)

a, if d ab => (A)=d , (B)=d ,

∴ (A)=(B) ----- (1)
b, if ad b => (A)=1 , (B)=d
∴ (A)  (B) ----- (2)
c, if abd => (A)=1 , (B)=1
∴ (A)= (B) ----- (3)
(1),(2),(3) => (A)  (B)
i.e., W (a, d )  W (b, d )
i i

ii, see i

5. show i (Wi ( a, d ),Wi (b, d ))  Wi (a, d )

Proof : ∵ if a  b  Wi (a, b)  b

if a  b  Wi (a, b)  1

A, if a  b  i (Wi (a, b),Wi (b, d ))  i (b,Wi (b, d ))

 => i (b,Wi (b, d ))  i(b, d )  min(b, d )  d

Wi ( a, d )  d

=> i (b,Wi (b, d ))  i (b,1)  b

Wi ( a, d )  1

=> i (b,Wi (b, d ))  i (b,1)  b

Wi ( a, d )  1

B, if a  b  i (Wi (a, b),Wi (b, d ))  i (1, Wi (b, d ))  Wi (b, d )

=> Wi (b, d )  d

Wi ( a, d )  d

Wi (b, d )  d

Wi ( a, d )  1

Wi (b, d )  1

Wi ( a, d )  1
10. show i (a, Wi (a, b)  b

Proof : ∵ a  b  Wi (a, b)  b

a  b  Wi (a, b)  1

A, if ab

 i (a, Wi (a, b))  i (a, b)  min( a, b)  b

B, if ab

 i (a,Wi (a, b))  i( a,1)  a  b

 inf  Wi composition
Wi inf
( P  Q)( x, z )  y  Y Wi ( P ( x, y ), Q ( y , z ))

 Theorem 5.4 :
i Wi Wi
(1)( P  Q  R)  (Q  P 1  R)  ( P  (Q  R 1 ) 1 )

Wi Wi i Wi
(2)( P  (Q  S )  ( P  Q)  S

 Theorem 5.5 :
Wi Wi
( Pj )  Q  ( Pj  Q)
j j
Wi Wi
( Pj )  Q)  ( Pj  Q)
j j
Wi Wi
P  ( Q j )  ( Pj  Q j )
j j
Wi Wi
P  ( Q j )  ( Pj  Q j )
j j
 Theorem 5.6 : if Q1  Q2 => Wi
P  Q1  P  Q2
Wi

Wi Wi
Q1  R  Q2  R

Proof : Q1  Q2 => Q1  Q2  Q1 , Q1  Q2  Q2

∵ (P  Q )  (P  Q )  P  (Q  Q )  P  Q
Wi

1
Wi

2
Wi

1 2
Wi

=> Wi
P  Q1  P  Q2
Wi

∵ (Q  R)  (Q
1
Wi

2
Wi Wi
 R )  (Q1  Q2 )  R  Q2  R
Wi

=> Q  R  Q 1
Wi

2
Wi
R

 Theorem 5.7 :
1. i Wi
P 1 ( P  Q)  Q

2. Wi
R  P  ( P 1  R)
i

3. Wi
P  ( P  Q)  Q 1
Wi

4. Wi
R  ( R  Q 1 )  Q
Wi

Proof :
(1) P  Q  ( P
Wi
1 1
) Q
Wi
---- (A)
(5.26)  (5.25)

i.e.,   Wi  i  
(Q  P 1  R )  ( P  Q  R )

let 
Wi
P QQ
, 
P 1  P , 
QR
 i
 ( A)  P 1 ( P  Q)  Q

(2) ---- (B)

i i
P 1  R  P 1  R

 
Let P 1  P , QR , i 
P 1  R  R
Wi i
 ( B )  R  P  ( P 1  R )

(3) by (5.33) , [ P
i Wi
1
( P  Q)]1  Q 1

---- (C)
Wi i
 ( P  Q )1  P  Q 1

Let   
Wi
P  Q  P, P  Q, Q 1  R

Wi Wi
 (C )  P  ( P  Q) 1  Q 1

(4) follows (3)