IEICE TRANS. INF. & SYST., VOL.E82D, NO.

5 MAY 1999
955
PAPER
Special Issue on Multiple-Valued Logic and Its Applications
Interval and Paired Probabilities for Treating Uncertain
Events
Yukari YAMAUCHI

, Nonmember and Masao MUKAIDONO

, Member
SUMMARY When the degree of intersections AB of events
A, B is unknown arises a problem: how to evaluate the proba-
bility P(A B) and P(A B) from P(A) and P(B). To treat
related problems two models of valuation: interval and paired
probabilities are proposed. It is shown that the valuation corre-
sponding to the set operations (intersection), (union) and
(complement) can be described by the truth functional (AND),
(OR) and (negation) operations in both models. The proba-
bilistic AND and OR operations are represented by combinations
of Kleene and Lukasiewicz operations, and satisfy the axioms of
MV (multiple-valued logic)-Algebra except the complementary
laws.
key words: interval probability
1. Introduction
Representing uncertain knowledge has been a very ac-
tive research area in Articial Intelligence, since, in life
knowledge is rarely exact and certain. Some attempts
to use probability theory in this area revealed certain
unpleasant eects. One of them is the loss of exibil-
ity and another is the loss of functionality in probabil-
ity operations. In probability theory, since an event is
given as a set, the probability of a compound event can
not be uniquely determined from the probabilities of its
component events.
Other attempts to nd structures weaker than
probability to represent uncertainty are Dempster-
Shafers evidence theory (DS theory) [1], possibility the-
ory [2]. The main dierence from the probability theory
is that the evidence and possibility theories do not sat-
isfy additivity. Instead of the standard point-valued
probability, Giang suggested the interval probability[3]
given by a pair of lower and upper bounds. He, based
on the semantic model of probability for sentences pro-
posed by Nilsson, shows that the functionality is re-
tained if one use his probability.
In this paper, the interval probability and the
paired probability are introduced in order to evaluate
probabilities P(A B) and P(A B) from P(A) and
P(B). Here, the semantic models (so as the intersec-
tion AB of events A, B) are assumed to be unknown.
In both models three probability operations:
p
,
p
and
p
are introduced corresponding to the set operations
Manuscript received September 10, 1998.
Manuscript revised November 17, 1998.

The authors are with the Department of Computer Sci-

ence, Meiji University, Kawasaki-shi, 2140033 Japan.
, and - (complement) over the event space. It is
shown that the probabilistic AND and the probabilis-
tic OR operations can be represented by combinations
of Kleene and Lukasiewicz [4] operations dened in the
closed interval [0, 1]. The set of probabilistic opera-
tions proposed in this paper satises the axioms of MV
(multiple-valued logic)-Algebra [5] except the comple-
mentary laws.
2. Preliminaries of Probability Theory
In a probability sample space an event is a subset of
. We denote the set of all events (the power set of )
by F. A mapping P : F from a closed interval [0, 1]
is called a probability measure on (, F) and satises
P(A) 0 for all A F, P() = 1 and P() = 0 and
if A B = then P(A B) = P(A) +P(B).
The third property is called the additivity of prob-
ability measure. Note that the degree of membership
of a sample e to an event A is given either 1 or 0 here.
And the probability for a compound event is given only
for the compound of disjoint events.
In the case that the events A and B are inde-
pendent to each other, P(A B) and P(A B) are
interpreted truth-functionally as P(A) P(B) and
P(A) +P(B) P(A B) respectively. However, a cru-
cial point here is that from P(A) and P(B), P(A B)
and P(A B) are not determined when the degree of
intersection of A and B is unknown. In order to retain
this loss of truth functionality, the operations for inter-
val and paired probabilities are proposed in the sequel.
2.1 A Motivation Numerical Probability
Given the probabilities P(A) = a, P(B) = b for events
A, B (a, b [0, 1]), a compound event such as A and B
happen, is the intersection of the sets A and B whose
size depends on the degree of intersection. P(A B)
takes the minimum value when there is the least inter-
section between A and B (either 0: Fig. 1 (a), or the
possible least intersections: Fig. 1 (b)) and the max-
imum value when there is the most (either B A:
Fig. 2 (a) or A B: Fig. 2 (b)). Thus the minimum and
the maximum values of P(A B) are max(0, a +b 1)
and min(a, b), respectively.
Similarly, a compound event such as A or B hap-
pens, is the union of the sets A and B whose size de-
956
IEICE TRANS. INF. & SYST., VOL.E82D, NO.5 MAY 1999
Fig. 1 Minimum P(A B) = max(0, a + b 1).
Fig. 2 Maximum P(A B) = min(a, b).
Fig. 3 Minimum P(A B) = max(a, b).
Fig. 4 Maximum P(A B) = min(1, a + b).
pends on the degree of intersection. P(A B) takes the
minimum value max(a, b) when there is the most inter-
section between A and B (when P(A B) = min(a, b))
and the maximum value min(1, a+b) when there is the
least (when P(A B) = max(0, a + b 1)) as shown
in the Figs. 3 and 4. Thus the minimum and the maxi-
mum values of P(A B) are max(a, b) and min(1, a+b),
respectively.
For our purpose we introduce notations borrowed
from multiple-valued logic algebras. For a, b [0, 1]
dene Lukasiewiczs and Kleenes AND operations, re-
spectively, by
a b

=
max(0, a +b 1)
and
a b

=
min(a, b).
Similarly, dene Lukasiewiczs and Kleenes OR opera-
tions, respectively, by
a b

=
min(1, a +b)
and
a b

=
max(a, b).
Then the above discussion is summarized as
a b P(A B) a b
a b P(A B) a b.
Inspired by the above observation, in order to re-
tain the truth functionality of the probability opera-
tions for compound events, the probability operations
AND and OR, denoted as
p
and
p
respectively, assign
the interval probabilities to the events AB and AB.
Denition 2.1.1: Given numerical probabilities
P(A) = a and P(B) = b, the probability AND opera-
tion
p
and the probability OR operation
p
assign the
closed intervals of probability for the compound events
of A and B as
P(A)
p
P(B)

=
[a b, a b] (1)
P(A)
p
P(B)

=
[a b, a b] (2)
when the degree of intersection is unknown.
The probability negation operation
p
assigns a nu-
merical probability 1P(A) for P(A) since the oper-
ation does not involve a compound event. (Hereafter,
we use notation a for 1 a.)
P(A) =
p
a

=
a (3)
Example: if P(A) = 0.7 and P(B) = 0.8 then
P(A)
p
P(B) = [max(0, 0.7 + 0.8 1),
min(0.7, 0.8)]
= [0.5, 0.7]
P(A)
p
P(B) = [max(0.7, 0.8), min(1, 0.7 + 0.8)]
= [0.8, 1].
3. Interval Probability
3.1 Interval Probability
Use of the interval probability dened in the previous
section retains the truth functionality of the probabil-
ity operations. However, the domain of the numerical
probabilities is not closed as an interval, so a general-
ization of the operations on the interval probabilities is
needed.
Given the set of all interval-value I, I = {[a, b] |
0 a b 1}, a mapping IP: F I is called an
interval probability. An interval probability IP(E) can
be interpreted as the scope of P(E), i.e. IP(E) = [a, b]
means a P(E) b.
Let the interval probabilities be given as IP(A) =
[a
1
, a
2
], IP(B) = [b
1
, b
2
] for events A, B. P(A B)
YAMAUCHI and MUKAIDONO: INTERVAL AND PAIRED PROBABILITIES FOR TREATING UNCERTAIN EVENTS
957
Fig. 5 P(A B).
Fig. 6 P(A B).
attains the minimum value a
1
b
1
, when P(A) = a
1
,
P(B) = b
1
(when the probability for each event takes
the minimum value) and the intersection of A, B is at
the least. It attains the maximum value a
2
b
2
, when
P(A) = a
2
, P(B) = b
2
(when the probability for each
event takes the maximum value) and the intersection of
A, B is at the most (Fig. 5). Similarly, P(A B) attains
the minimum value a
1
b
1
when P(A) = a
1
, P(B) = b
1
and the intersection of A, B is at the most. It attains
the maximum value a
2
b
2
, when P(A) = a
2
, P(B) =
b
2
and the intersection of A, B is at the least (Fig. 6).
Denition 3.1.1: Given interval probabilities IP(A)
and IP(B), the probability AND operation
p
, the prob-
ability OR operation
p
and the probability negation op-
eration
p
assign the interval probabilities for the com-
pound events of A and B and the negation of A as
follows.
IP(A)
p
IP(B)

=
[a
1
b
1
, a
2
b
2
], (4)
IP(A)
p
IP(B)

=
[a
1
b
1
, a
2
b
2
], (5)
p
IP(A)

=
[a
2
, a
1
]. (6)
The above denition leads to the following identities.
IP(A B) := IP(A)
p
IP(B),
IP(A B) := IP(A)
p
IP(B),
IP(A) :=
p
IP(A).
The set of interval probability operations {
p
,
p
,
p
}
has truth functionality and the domain of interval prob-
abilities is closed under it. Although [1, 1] and 1 are
formally dierent, they represent the same geometric
object (the point 1). In such cases we will put the
sign, i.e. [1, 1] 1.
Note, that the interval probability is inclusive of
numerical probability as a special case, since P(A) =
a [a, a] = IP(A).
3.2 Algebraic Properties of Interval Probability Oper-
ations
The set of interval probability operations {
p
,
p
,
p
} and
constants [0, 0], [1, 1] satisfy the axioms of MV Algebra
1 6.
1. commutative laws
A
p
B = B
p
A, A
p
B = B
p
A.
2. associative laws
A
p
(B
p
C)=(A
p
B)
p
C, A
p
(B
p
C) = (A
p
B)
p
C.
3. maximum element
A
p
1 = 1, A
p
1 = A.
4. minimum element
A
p
0 = A, A
p
0 = 0.
5. double negation law
p
(
p
A) = A.
6. de Morgan laws
p
(A
p
B) =
p
A
p

p
B,
p
(A
p
B) =
p
A
p

p
B.
Instead of the following
7. complementary laws
A +
p
A = 1, A
p
A = 0,
it satises the Kleene laws.
7. Kleene laws
A
p

p
A
p
B
p

p
B = A
p

p
A,
(A
p

p
A)
p
B
p

p
B = B
p

p
B.
As it holds
A B = (AB) B
A B = (AB) B,
the set of Lukasiewicz operations {, } together with
the operations is able to represent Kleene operations
{, }. So the Lukasiewicz(MV) Algebra {, , } is
inclusive of the Kleene Algebra {, , }. However, the
set of interval probability operations can not represent
Kleene operations.
From the algebraic standpoint, it is natural to
compare the two system of operations {
p
,
p
,
p
} and
{, , , , } which is dened component-wisely as
follows AB [a
1
b
1
, a
2
b
2
], AB [a
1
b
1
, a
2
b
2
],
A B [a
1
b
1
, a
2
b
2
], A B [a
1
b
1
, a
2
b
2
],
A [a
2
, a
1
].
Lemma 3.2.1: Each of the two systems of operations
{
p
,
p
,
p
} and {, , , , } dened over intervals can
not represent the other system.
Proof: It is enough for the proof if we nd out one
counter example.
{, , , , } can not represent {
p
,
p
,
p
}:
Let A and B be two numerical values a and b (A
= [a,a] and B = [b,b]). Then the result of the
958
IEICE TRANS. INF. & SYST., VOL.E82D, NO.5 MAY 1999
operations {, , , , } are numerical values, as
shown bellow. That is, the results of any itera-
tions of {, , , , } to A and B are numerical
values.
A B = [a, a] [b, b] = [a b, a b] = a b,
A B = [a, a] [b, b] = [a b, a b] = a b,
AB = [a, a] [b, b] =[a b, a b] = a b,
AB = [a, a] [b, b] =[a b, a b] = a b,
A = [a, a] = [a, a] = a.
On the other hand, interval values are generated
from the set of operations {
p
,
p
,
p
} on numerical
values a and b as shown bellow.
A
p
B = [a, a]
p
[b, b] = [a b, a b],
A
p
B = [a, a]
p
[b, b] = [a b, a b].
Thus, the above interval truth values can not be
generated when a b = a b or a b = a b
from the set of operations {, , , , } which is
closed under the numerical values.
{
p
,
p
,
p
} can not represent {, , , , }:
The set of interval truth values A =
{[0, 1], [1, 1/2], [1/2, 1/2], [1/2, 1]} is closed under
the operation {
p
,
p
,
p
} as shown in the following
tables.
p
[0,1/2] [1/2,1/2] [1/2,1] [0,1]
[0,1/2] [0,1/2] [0,1/2] [0,1/2] [0,1/2]
[1/2,1/2] [0,1/2] [0,1/2] [0,1/2] [0,1/2]
[1/2,1] [0,1/2] [0,1/2] [0,1] [0,1]
[0,1] [0,1/2] [0,1/2] [0,1] [0,1]
p
[0,1] [0,1/2] [1/2,1/2] [1/2,1]
[0,1] [0,1] [0,1] [1/2,1] [1/2,1]
[0,1/2] [0,1] [0,1] [1/2,1] [1/2,1]
[1/2,1/2] [1/2,1] [1/2,1] [1/2,1] [1/2,1]
[1/2,1] [1/2,1] [1/2,1] [1/2,1] [1/2,1]
[0,1] [0,1/2] [1/2,1/2] [1/2,1]
p
[0,1] [1/2,1] [1/2,1/2] [0,1/2]
On the other hand, the minimum closed set in-
cluding A under the operations {, , , , }
is {[0, 1], [1, 1/2], [1/2, 1/2], [1/2, 1], [0, 0], [1, 1]} as
shown easily. That is, {
p
,
p
,
p
} can not gen-
erate [0, 0] and [1, 1] from A but {, , , , }
can. This shows that {
p
,
p
,
p
} can not represent
{, , , , }.
4. Paired Probability
4.1 Paired Probability
Now, we would like to consider a dierent model called
Fig. 7 Paired event A = (A
1
, A
2
).
Fig. 8 Paired events and a triangle diagram.
the paired probability which has the form of a pair
of probabilities (optimistic case and pessimistic case)
for an event. In the interval probability, the interval
value is considered as the scope of numerical probabil-
ity for an event. In the paired probability, the min-
imum (lower) and maximum (upper) probabilities are
given for lower and upper events of a paired event which
has two boundaries. A paired event is interpreted as
A = (A
1
, A
2
) such that A
1
, A
2
F and A
1
A
2
(Fig. 7). Thus a paired probability PP(A) should be
interpreted as (P(A
1
), P(A
2
)).
Given the set of all paired-values P, P = {(x
1
, x
2
) |
0 x
1
x
2
1}, and the set of all paired-events F
P
,
F
P
= {(X
1
, X
2
) | X
1
X
2
, X
1
, X
2
F}, a mapping
PP:F
P
P is called the paired probability measure on
(, F
P
).
Let PP(A) and PP(B) be paired probabilities
(a
1
, a
2
) and (b
1
, b
2
). We adopt a triangle diagram
drawn with the representative three points (0,0), (1,1)
and (0,1) to represent the distribution of paired prob-
abilities. A point in the diagram represents a paired
probability as exemplied in Fig. 8. A partial order
between paired probabilities is represented in horizon-
tal direction. That is PP(A) PP(B) in Fig. 8.
We dene compound events of A = (A
1
, A
2
) and
B = (B
1
, B
2
) by
A B = (A
1
B
1
, A
2
B
2
),
A B = (A
1
B
1
, A
2
B
2
).
We call the rst and second components the lower and
upper limit events of AB (AB). As each component
is now indexed by two events we can apply the idea
of the interval of probabilities (Denition 2.1.1) to its
evaluation. Thus dene
PP(A B)

=
(P(A
1
)
p
P(B
1
), P(A
2
)
p
P(B
2
))
PP(A B)

=
(P(A
1
)
p
P(B
1
), P(A
2
)
p
P(B
2
)).
The rst and second components of PP(A B) are
called lower and upper limits of PP(A B). They are
given as interval probabilities
YAMAUCHI and MUKAIDONO: INTERVAL AND PAIRED PROBABILITIES FOR TREATING UNCERTAIN EVENTS
959
Fig. 9 Possible PP(A B).
Fig. 10 Example of contradiction.
P(A
1
)
p
P(B
1
) = [a
1
b
1
, a
1
b
1
],
P(A
2
)
p
P(B
2
) = [a
2
b
2
, a
2
b
2
].
The set of paired probabilities represented by the
two intervals of PP(A B) is shown as a rectangle (an
aggregate of paired probabilities) in the triangle dia-
gram in Fig. 9 whose minimum paired probability is
PP(A B) : (a
1
b
1
, a
2
b
2
), and maximum paired
probability is PP(A B) : (a
1
b
1
, a
2
b
2
).
The highest point in the triangle diagram repre-
sents a paired probability (the minimum P(A
1
B
1
),
the maximum P(A
2
B
2
)) However the combination
of the minimum A
1
B
1
and maximum A
2
B
2
some-
times contradicts A
1
A
2
or B
1
B
2
as exemplied
in Fig. 10.
Since A
1
A
2
and B
1
B
2
holds for every paired
event, the dierence between P(A
1
B
1
) (P(A
1
B
1
))
and P(A
2
B
2
) (P(A
2
B
2
)) is limited to less than or
equal to (a
2
a
1
) + (b
2
b
1
).
Lemma 4.1.1: The maximum dierence between the
lower and upper limits of PP(A B) is less than or
equal to (a
2
a
1
) + (b
2
b
1
).
Proof:
P(A B) = P(A) +P(B) P(A B), (7)
P(A B) = P(A) +P(B) P(A B), (8)
A
1
A
2
, B
1
B
2
(9)
holds for every events.
P(A
1
B
1
) = P(A
1
) +P(B
1
) P(A
1
B
1
) from (7)
P(A
1
B
1
) P(A
2
B
2
) since A
1
B
1
A
2
B
2
Fig. 11 P(A
1
B
1
) = a
1
+ b
1

1
.
Fig. 12 Possible arrangement of P(A
2
B
2
).
Fig. 13 Maximum P(A
2
B
2
).
from (9)
P(A
2
B
2
) = P(A
2
) +P(B
2
) P(A
2
B
2
) from (8)
P(A
2
) +P(B
2
) (P(A
1
) +P(B
1
)
P(A
1
B
1
))
Thus
P(A
2
B
2
) P(A
1
B
1
)
P(A
2
) +P(B
2
) (P(A
1
) +P(B
1
))
(a
2
a
1
) + (b
2
B
1
)
In the following we will explain the proof of
Lemma 4.1.1 in detail by using gures and numerical
values. Let the lower limit P(A
1
B
1
) of PP(A B) be

1
[a
1
b
1
, a
1
b
1
], then P(A
1
B
1
) is a
1
+b
1

1
(Fig. 11). Since P(A
2
B
2
) is bigger than or equal
to P(A
1
B
1
), its range is restricted to [a
1
+ b
1

1
, a
2
+b
2

1
] (Fig. 12). The maximum P(A
2
B
2
)
is a
2
+b
2
(a
1
+b
1

1
) (Fig. 13).
The maximum dierence between the lower and
upper limits of PP(A B) is (a
2
+ b
2
) (a
1
+ b
1

1
)
1
= (a
2
a
1
) +(b
2
b
1
) thus not depending on

1
(Fig. 14). Notice that is less than or equal to 1 since
P(A
1
B
1
) and P(A
2
B
2
) are in [0,1].
Since PP(A B) can not be uniquely dened
as a paired probability, we adopt the paired inter-
val probability to represent it. The paired inter-
val probability PP
I
(A) should be interpreted as the
possible distribution of paired probability and repre-
sented by the minimum and maximum paired prob-
ability (a
1
, a
2
), (a
3
, a
4
) with the maximum dierence
between lower and upper limit, maxA, denoted as
960
IEICE TRANS. INF. & SYST., VOL.E82D, NO.5 MAY 1999
Fig. 14 The set of PP(A B).
Fig. 15 Paired probability operations.
PP
I
(A)

=

a
1
a
2
a
3
a
4

[maxA]
(a
1
a
2
, a
3
a
4
, a
1

a
3
, a
2
a
4
and maxA [0, a
4
a
1
]). Note, the corre-
spondence to IP(A
1
) = [a
1
, a
3
], and IP(A
2
) = [a
2
, a
4
].
Given the set of all paired-interval-value P
I
, a
mapping PP
I
: F
P
P
I
is called a paired interval
probability measure on (, F
P
).
Denition 4.1.1: Given paired probabilities PP(A)
and PP(B), the paired probability operation
p
and
p
assign the paired interval probabilities for the com-
pound events of A and B and the probability negation
operation
p
assigns a paired probability as,
PP(A)
p
PP(B)

=

a
1
b
1
a
2
b
2
a
1
b
1
a
2
b
2

[1]
, (10)
PP(A)
p
PP(B)

=

a
1
b
1
a
2
b
2
a
1
b
1
a
2
b
2

[2]
. (11)
PP(A)

=
( a
2
, a
1
). (12)

1
=min((a
2
b
2
)(a
1
b
1
), (a
2
a
1
)(b
2
b
1
))

2
=min((a
2
b
2
)(a
1
b
1
), (a
2
a
1
)(b
2
b
1
))
Examples are shown in Fig. 15.
4.2 Paired Interval Probability
The domain of paired probabilities is not closed under
the set of probability operations {
p
,
p
,
p
}. In the sub-
sequent part, these operations are generalized so that
the domain of the paired interval probabilities is closed
under them.
Given the paired events A and B, the paired inter-
val probabilities of A and B are written as PP
I
(A) =

a
1
a
2
a
3
a
4

[maxA]
, PP
I
(B) =

b
1
b
2
b
3
b
4

[maxB].
PP(A B) attains the minimum value, (a
1
b
1
, a
2

b
2
), and the maximum value, (a
3
b
3
, a
4
b
4
), as shown
Fig. 16 Minimum and maximum PP(A B).
Fig. 17 PP
I
(A B).
Fig. 18 Minimum and maximum PP(A B).
Fig. 19 PP
I
(A B).
in Figs. 16 and 17.
Similarly, PP(A B) attains the minimum value,
(a
1
b
1
, a
2
b
2
), and the maximum, (a
3
b
3
, a
4
b
4
),
as shown in Figs. 18 and 19.
The following denitions are based on these obser-
vations.
Denition 4.2.1: Given two paired interval proba-
bilities PP
I
(A) =

a
1
a
2
a
3
a
4

[maxA]
, PP
I
(B) =

b
1
b
2
b
3
b
4

[maxB]
, the paired interval probability op-
erations
p
,
p
,
p
assign the paired interval probabilities
for the paired events A B, A B and A as follows
when the degree of intersection is unknown.
PP
I
(A)
p
PP
I
(B)

=

a
1
b
1
a
2
b
2
a
3
b
3
a
4
b
4

[1]
, (13)
PP
I
(A)
p
PP
I
(B)

=

a
1
b
1
a
2
b
2
a
3
b
3
a
4
b
4

[2]
, (14)
p
PP
I
(A)

=

a
4
a
3
a
2
a
1

[maxA]
. (15)
YAMAUCHI and MUKAIDONO: INTERVAL AND PAIRED PROBABILITIES FOR TREATING UNCERTAIN EVENTS
961

1
= min((a
4
b
4
) (a
1
b
1
), maxAmaxB)

2
= min((a
4
b
4
) (a
1
b
1
), maxAmaxB)
Numerical, interval and paired probabilities can
be regarded as the special cases of the paired inter-
val probabilities. Thus the paired interval probabilities
can be regarded as a general form of the probabilities.
The domain of the paired interval probabilities is closed
under the set of paired interval probability operations
{
p
,
p
,
p
}.
The set of paired interval probability operations
{
p
,
p
,
p
} satises exactly the same axioms of the oper-
ations on interval probabilities. Algebraic structures of
< I;
p
,
p
,
p
, , > and < P;
p
,
p
,
p
, , > must be
similar to each other, and should be interesting topics
for further discussions.
5. Conclusion
For representing uncertain knowledge using probabil-
ity, two models of probabilities, the interval and paired
probabilities, and their operations are introduced. The
operations for these models are dened by combinations
of both Kleene and Lukasiewicz operations that repre-
sent Kleene Algebra and MV Algebra, respectively. In
this paper we examine the algebraic properties of these
two models.
In application these models may have a drawback
such that the results become more wide range, however,
have a good point such that the domain of multiple
valued logic {0, 1/m, 2/m, . . . , (m1)/m, 1} is closed
in both algebras like the algebra of interval truth values
in[6].
Further discussion of the interval probability may
be related to approximated reasoning in Articial Intel-
ligence. The models and operations for paired probabil-
ity may be applied to the rough sets, fuzzy measure [7]
and data mining in Articial Intelligence.
References
[1] G. Shafer, A mathematical theory of evidence, Princeton
Univ. Press, 1976.
[2] D. Dubois and H. Prade, Belief revision and update in nu-
merical formalisms: An overview with new results for the
possibilistic framework, Procs. IJCAI93, 1993.
[3] P.H. Giang, Representation of uncertain belief using interval
probability, Proc. twenty-seventh International Symposium
on Multiple-Valued Logic, IEEE, 1997.
[4] M. Mukaidono, Fuzzy Logic (Kouza Fuzzy vol.4),
Nikkan Kougyo Press, 1993.
[5] C.C. Chang, Algebraic analysis of many-valued logics,
Trans. of the American Mathematical Society, vol.88, 1958.
[6] M. Mukaidono, Algebraic structures of interval truth val-
ues in fuzzy logic, Proc. Sixth International Conference
on Fuzzy Systems (FUZZ-IEEE97), vol.II, pp.699705, July
1997.
[7] M. Sugeno and T. Murofushi, Fuzzy Measure (Kouza Fuzzy
vol.3), Nikkan Kougyo Press, 1993.
Yukari Yamauchi was born in Japan
on 1973. She received the B.E. degree
in Liberal Arts and M.E. degree in Nat-
ural Science from International Christian
University, in 1996 and 1998, respectively.
She is currently a Ph.D. student in the De-
partment of Computer Science, Meiji Uni-
versity. Her main research interests are in
multiple-valued logic, fuzzy logic, and its
applications, rough sets and data mining,
modal logic and approximated reasoning
in Articial Intelligence.
Masao Mukaidono was born in
Japan on 1942. He received the B.E.,
M.E. and Ph. D. degree in electrical en-
gineering from Meiji University, in 1965,
1967 and 1970, respectively. He is cur-
rently a Professor in the Department of
Computer Science, School of Science and
Technology, Meiji University. His main
research interests are in multiple-valued
logic, fuzzy set theory and its applica-
tions, fault tolerant computing, and com-
puter aided logic design. Dr. Mukaidono was president of Japan
Society of Fuzzy Theory and Systems, and member of the IEEE
Computer Society, the Information Processing Society of Japan
and Japanese Society of Articial Intelligence.