Information Technology Reference
In-Depth Information
7.5 Measures of Possibility and Necessity
Letitbe
F ↂ P (
X
)
a Boolean algebra of subsets of X . A mapping
ˀ : F ₒ[
0
,
1
]
is called a measure of possibility , provided:
ˀ( ) = ∅
ˀ(
X
) =
1
ˀ(
A
B
) =
max
(ˀ(
A
), ˀ(
B
))
, for all A
,
B in
F
.
Notice that the last property does not require A
B
= ∅
. Actually, any of these
mappings are fuzzy measures, since:
A
B
A
B
=
B :
ˀ(
B
) =
max
(ˀ(
A
), ˀ(
B
)) ˀ(
A
)
,or
ˀ(
A
) ˀ(
B
).
From max
(ˀ(
A
), ˀ(
B
)) ˀ(
A
) + ˀ(
B
)
,itfollows
ˀ(
A
B
) ˀ(
A
) + ˀ(
B
)
even if A
B
= ∅
. Hence, possibility measures are sub-additive.
A C
A C
A C
From A
=
X ,itis1
=
max
(ˀ(
A
), ˀ(
)) ˀ(
A
) + ˀ(
)
,or1
ˀ(
A
)
A C
ˀ(
.
Obviously,
)
ˀ(
A 1
A 2 ...
A n ) =
max
(ˀ(
A 1 ), ˀ(
A 2 ),...,ˀ(
A n ))
,
for all
A 1 ,...,
A n in
F .
Hence, if X
={
x 1 ,...,
x n }
is a finite set, to have a possibility measure
ˀ
, its
ˀ(
x i )
values
do verify:
1
= ˀ(
X
) =
max
(ˀ(
x 1 ), ˀ(
x 2 ), . . . , ˀ(
x n )),
forcing that some of the values
ˀ(
x i )
should equal 1. For example, if X
={
x 1 ,
x 2 ,
x 3 }
,
the three values
ˀ(
x 1 ) =
0,
ˀ(
x 2 ) =
0
.
5,
ˀ(
x 3 ) =
1, define a particular mea-
sure of possibility on
P (
X
)
. It is, for example,
ˀ( {
x 1 ,
x 2 } ) =
max
(
0
,
0
.
5
) =
0
.
5,
ˀ( {
x 1 ,
x 3 } ) =
max
(
0
,
1
) =
1, etc.
Remark 7.5.1 Instead of a family
of crisp sets no problem arises in considering
a family of fuzzy sets. Measures of possibility can be applied to fuzzy sets with the
only changes of A
F
μ ˃ . The only caution
is to use the connectives min , max to preserve distributivity.
μ ˃ , and A
B
= ∅
by
B
= ∅
by
X
Theorem 7.5.2
For each
μ ∈[
0
,
1
]
such that Sup
μ =
1 , the mapping
ˀ μ : F ₒ
[
0
,
1
]
defined by
ˀ μ (
A
) =
Sup
x
min
(μ(
x
), μ A (
x
))
,A
∈ F ,
X
is a possibility measure.
Proof
ˀ μ ( ) =
Sup
x
min
(μ(
x
),
0
) =
0.
ˀ μ (
X
) =
Sup
x
min
(μ(
x
),
1
) =
X
X
Sup
x
X μ(
x
) =
1. Finally, since
μ A B (
x
) =
max
A (
x
), μ B (
x
))
for all x
X :
 
Search WWH ::




Custom Search