A New Penta-valued Logic Based Knowledge Representation
Vasile Patrascu
Department of Informatics Technology, Tarom Company
Bucurestilor Road, 224F, Bucharest-Otopeni, Romania
e-mail: patrascu.v@gmail.com
Abstract
In this paper a knowledge
representation model are proposed,
FP5, which combine the ideas from
fuzzy sets and penta-valued logic. FP5
represents imprecise properties whose
accomplished degree is undefined,
contradictory or indeterminate for some
objects. Basic operations of
conjunction, disjunction and negation
are introduced. Relations to other
representation models like fuzzy sets,
intuitionistic, paraconsistent and bipolar
fuzzy sets are discussed.
Keywords: Fuzzy set, multi-valued logics,
Frank t-norm, contradiction, uncertainty,
indeterminacy, intuitionistic fuzzy set.
1 Introduction
Fuzzy sets are a specially well-suited tool to
represent imprecise concepts with ill-defined
boundaries. When a property P is imprecise, its
negation ¬P is considered to be imprecise. The
fuzzy set theory assumes both P and ¬P are
related, namely: ¬P( x ) = 1 − P( x ) . However,
this is not always true in real life. Hence,
sometimes P and ¬P are represented
independently. On the other hand, fuzzy sets do
not allow to take into account the presence of
objects whose membership degree P is
undefined. Three valued logics can solve the
problem allowing three logical values: true, false
and undefined. However, this is not always
sufficient and we can find contradictions when a
certain value x verifies P and ¬P at the
same time. Four-valued logics can solve the
problem because it uses four logical values: true,
false, undefined and contradictory. There is a
special situation when the property P and its
negation ¬P are close to 0.5 for some values of
x . In this situation, we can detail the knowledge
representation using a penta-valued logic based
on five logical values: true, false, undefined,
contradictory and indeterminate. In conclusion,
the paper proposes a knowledge representation
model, where fuzzy sets and penta-valued logic
are combined to represent imprecise properties.
2 The Fuzzy Set and Its Extension
Let X be a crisp set. In the framework of Zadeh
theory [10], a fuzzy set A is defined by the
membership function µ A : X → [0,1] . The
non-membership function ν A : X → [0,1] is
obtained by negation and thus both functions
define a partition of unity, namely:
µA + νA =1 (2.1)
Atanassov has extended the fuzzy sets to the
intuitionistic fuzzy sets [1]. Atanassov has
relaxed the condition (2.1) to the following
inequality:
µA + νA ≤1 (2.2)
He has used the third function, the index of
uncertainty π A that verifies the equality:
πA = 1−µA − νA (2.3)
Similarly, we can consider instead of (2.1) the
following condition:
µA + νA ≥1 (2.4)
Thus, we obtain the paraconsistent fuzzy set and
one can define the index of contradiction :
κ A = µ A + ν A −1 (2.5)
There is a duality between intutitionistic fuzzy x = x • α + x oβ (3.4)
set and paraconsistent fuzzy sets. More
generally, in this paper, we will consider as Immediately, one results:
bipolar fuzzy set a set A , defined by two x• y = x− xo y (3.5)
functions totally independent µ A : X → [0,1]
xo y = x− x• y (3.6)
and ν A : X → [0,1] .
From (3.1) and (3.5) one obtains for the
A penta-valued fuzzy set was defined based on
conjugate:
Lukasiewicz penta-valued logic with the
following five descriptors: strong membership, ⎛ ⎛ 1 ⎞ ⎛ 1 ⎞⎞
weak membership, index of uncertainty, weak ⎜ ⎜ − 1⎟⎟ ⋅ ⎜⎜ − 1⎟⎟ ⎟
⎜ ⎜
non-membership and strong non-membership x • y = log 1 ⎜1 + ⎝
sx ⎠ ⎝sy ⎠⎟
1 ⎟
[9]. ⎜ − 1 ⎟
s
⎜ s ⎟
Belnap has defined a four-valued logic based on ⎝ ⎠
true, false, uncertainty and contradictory [2]. In
this paper we will add to these four values the Thus, two Frank t-norms are conjugated if one is
fifth: indeterminate. Thus, we will define a new computed with parameter s and the other is
penta-valued fuzzy set, constructing five-valued 1
computed with parameter . Thus, it results
fuzzy partitions based on true, false, undefined, s
contradictory and indeterminate. that the logics Godel and Lukasiewicz [6], [7],
[8] are conjugated and the Product logic is
identical with its conjugate.
3 Transformation from Bipolar
Knowledge Representation to a Penta- For s = 0 it results:
valued One
⎧ x ⊕ y = Max ( x, y )
⎪
Let there be the Frank t-norm [5] defined for ⎨ x o y = Min( x, y ) (3.7)
s ∈ (0, ∞ ) by: ⎪ x • y = Max (0, x + y − 1)
⎩
⎛ ( s x − 1) ⋅ ( s y − 1) ⎞
ts ( x, y ) = log s ⎜1 + ⎟ (3.1) From (3.5) one obtains
⎜ s −1 ⎟
⎝ ⎠ x = x• y + xo y (3.8)
Let there be a Frank t-norm denoted by „ o ”. and replacing x by x and y by y it results:
This t-norm verifies the Frank equation [5]:
x = x•y+xoy (3.9)
x o y − x o y = x + y −1 (3.2)
Replacing x by x o y and y by x o y in (3.8)
where x is the negation of x , namely:
it results:
x = 1− x
xo y = xo y•xo y+ xo yoxo y (3.10)
An equivalent form of Frank equation one can
obtain by replacing y with y , namely: or
xo y − yox = x− y (3.3) x o y = ( x o y) • ( x ⊕ y) + x o y o x o y (3.11)

Also, one defines its dual or its t-conorm „ ⊕ ” Replacing x by x o y and y by x o y in (3.8)
by: it results:
x ⊕ y = 1− x o y xo y = xo y•xo y + xo yoxo y (3.12)
and thus, the formula (3.2) has the equivalent or
form:
x o y = ( x o y) • ( x ⊕ y) + x o y o x o y (3.13)
x⊕ y+ xo y = x+ y
from (3.8) and (3.13) we obtain
Let there be two t-norms „ o ” and „ • ”. We say x = x • y + ( x o y ) • ( x ⊕ y ) + x o y o x o y (3.14)
that these two t-norms are conjugated if for
α + β = 1 there exists the equality: and replacing x by y in (3.14) it results:
y = y • x + ( x o y ) • ( x ⊕ y ) + x o y o x o y (3.15) (3.16) we can obtain S T , S F , S U , S C , S I
Now, we will denote: which characterize five logical values, namely:
true, false, undefined, contradictory and
⎧τ = x • y indeterminate. Thus, we came to have a penta-
⎪ϕ = x • y valued knowledge representation of bipolar
⎪⎪
imprecise information.
⎨π = ( x o y ) • ( x ⊕ y ) (3.16)
⎪κ = ( x o y ) • ( x ⊕ y ) In the bipolar fuzzy set theory, x could be the
⎪ membership function µ A and y could be the
⎩⎪ι = 2 ⋅ ( x o y o x o y )
non-membership function ν A . Using (3.16) we
From (3.8), (3.9), (3.11) and (3.13) it results a can obtain τ A , ϕ A , κ A , π A , ι A that define a
penta-valued partition of unity, namely:
penta-valued fuzzy set based on the five logical
τ+ϕ+ κ + π+ ι =1 (3.17) values that were mentioned above.
From (3.14),(3.15) and (3.16) it results the
inverse transform: 4 Penta-valued logic based on
contradiction, undefinedness and
⎧ ι
x = τ+κ+ indeterminacy
⎪⎪ 2
⎨ (3.18) In the framework of this logic we will consider
⎪y = ϕ + κ + ι the following five logical: true t , false f ,
⎪⎩ 2
undefined u , contradictory c and indeterminate
The functions defined by (3.16) have the i . Tables 1, 2 and 3 show tables for basic
following properties: operators in this logic.
⎧ τ ( x , y ) = ϕ( y , x )
⎨ (3.19)
⎩ τ ( x , y ) = ϕ( x , y ) Table 1: The OR operator
⎧ι( x, y ) = ι( y , x ) ∨ t i u c f
⎨ (3.20)
⎩ι( x, y ) = ι( x , y ) t t t t t t

⎧ π( x , y ) = π( y , x ) i t i i i i
⎪
⎨κ( x, y ) = κ( y , x ) (3.21) u t i u i i
⎪ π( x , y ) = κ ( x , y )
⎩ c t i i c i
In the case of the pair of logics Lukasiewicz- f t i i i f
Godel (3.7), one obtains the following particular
forms for the parameters considered in (3.16).
⎧τ = ( x − y ) +
⎪ϕ = ( y − x ) Table 2: The AND operator
⎪⎪ +
⎨ π = (1 − x − y)+ (3.22) ∧ t i u c f
⎪κ = ( x + y − 1) t t i i i f
⎪ +
⎪⎩ι = 1− | x − y | − | x + y − 1 | i i i i i f

where a + represents the positive part of a , u i i u i f

a+ | a | c i i i c f
namely: a+ =
2 f f f f f f
In bipolar preference theory x could be the
agreement function S + and y could be the One can see that the indeterminacy i is
non-agreement function S −
absorbent for the uncertainty u and contradiction
[3], [4]. Using
c.
 Table 3: The NOT operator ⎧µ AI B = µ A ∧ µB
⎪
¬ ⎪ ν AI B = ν A ∨ νB
⎨ (5.4)
t f ⎪ π AI B = π A ∧ πB
⎪κ = κ A ∧ κB
i i ⎩ AI B
u u In formulae (5.3) and (5.4), the symbols “ ∨ ”
and “ ∧ ” represent any couple of t-conorm, t-
c c
norm.
f t
The Complement
The logical values i, c, u are the same with The complement Ac for the set A ∈ FP5 is
their negations. defined by the formulae:
⎧µ Ac = νA
5 Penta-valued Fuzzy Set ⎪
⎪ ν Ac = µA
Let X be a crisp set. A penta-valued fuzzy set ⎨ (5.5)
A is defined by the following functions: the ⎪ κ Ac = κA
membership τ A : X → [0,1] , the non- ⎪π c = πA
⎩ A
membership ϕ A : X → [0,1] , the contradiction
κ A : X → [0,1] and the uncertainty In the set {0,1}4 there are four vectors having
π A : X → [0,1] . These functions verify the the form x = ( τ, ϕ, κ, π) , which verify the
following inequality: condition (5.1): T = (1,0,0,0) (True),
F = (0,1,0,0) (False), C = (0,0,1,0)
τA + ϕA + κA + πA ≤ 1 (5.1)
(Contradictory), U = (0,0,0,1) (Undefined) and
We will define the index of indeterminacy I = (0,0,0,0) (Indeterminate).
ι A : X → [0,1] by:
Using the operators defined by (5.3), (5.4) and
ιA = 1− τA − ϕA − κA − πA (5.2) (5.5) the same table results as seen in Tables 1, 2
and 3.
In this paper we will denote with FP5 the penta-
valued fuzzy sets defined in this section.
6 Fuzzy Set and Its Extension as FP5
For this kind of sets, one defines the union, the
intersection and the negation operators. Imprecise concepts represented by bipolar
fuzzy set can be translated into FP5. Particular
forms of bipolar fuzzy sets can also be translated
The Union to FP5, for example fuzzy sets, intuitionistic
The union A U B for two sets A, B ∈ FP5 is fuzzy sets and paraconsistent fuzzy sets. The
defined by formulae: operations defined by (5.3), (5.4) and (5.5)
supply new algebraic structures for the set types
⎧µ AU B = µ A ∨ µB mentioned above.
⎪
⎪ ν AU B = ν A ∧ νB
⎨ (5.3)
⎪ π AU B = π A ∧ πB 6.1 Bipolar Fuzzy Set as FP5
⎪κ = κ A ∧ κB
⎩ AU B One considers the bipolar fuzzy set A ∈ BFS
defined by the membership function µ A and the
The Intersection non-membership function ν A . Using the
The intersection A I B between two sets formulae (3.22) one obtains:
A, B ∈ FP5 is defined by the formulae:
 ⎧τ A = (µ A − ν A )+ We must emphasize that the index of
⎪ indeterminacy ι A is a symmetrical function. It
⎪ϕ A = (ν A − µ A )+
⎪ has the maximum value in the point (0.5,0.5) .
⎨κ A = (µ A + ν A − 1)+ (6.1.1)
This point is the middle of the line between the
⎪π = (1 − µ − ν )
⎪ A A A + points (1,0) and (0,1).
⎪ι = 1 − µ − ν − µ + ν − 1 From (6.1.2) it results:
⎩ A A A A A

The functions defined by (6.1.1) verify the ⎧µ A − ν A = τ A − ϕ A

⎨ (6.2.3)
condition (3.17) of a partition of unity. ⎩µ A + ν A = 1
We must underline that the index of From (6.1.3) one obtains the inverse transform:
indeterminacy ι A is a symmetrical function. It
⎧ ιA
⎪⎪µ A = τ A + 2
has the maximum value in the point (0.5,0.5) .
This point is the center of the square defined by ⎨ (6.2.4)
the points:(0,0), (1,0), (0,1) and (1,1). Also there ⎪ν = ϕ + ι A
⎩⎪
A A
are the following two equalities: 2

⎧µ A − ν A = τ A − ϕ A
⎨ (6.1.2) 6.3 Intuitionistic Fuzzy Set as FP5
⎩µ A + ν A = 1 + κ A − π A
We consider the intuitionistic fuzzy set A ∈ IFS
From (3.18) it results the inverse transform: defined by the membership function µ A and the
⎧ ιA non-membership function ν A . We will
⎪⎪µ A = τ A + κ A + 2 translate to a penta-valued fuzzy set using
⎨ (6.1.3)
⎪ν = ϕ + κ + ι A formulae (6.1.1).
⎩⎪
A A A
2 Thus, one defines the indexes of truth, falsity,
Thus, we have transformed a bipolar knowledge uncertainty and indeterminacy.
representation into a penta-valued one. ⎧τ A = (µ A − ν A )+
⎪
⎪ϕ A = (ν A − µ A )+
6.2 Fuzzy Set as FP5 ⎨ (6.3.1)
⎪π A = 1 − µ A − ν A
We consider the fuzzy set A ∈ FS defined by ⎪ι = µ + ν − µ − ν
⎩ A A A A A
the membership function µ A . One defines the
non-membership function ν A = 1 − µ A . Using Finally, due to the particularity (2.2) of
formulae (6.1.1) one define the indexes of truth, intuitionistic fuzzy sets, the penta-valued
falsity and indeterminacy. representation is reduced to a tetra-valued one:
τA + ϕA + πA + ιA = 1
⎧τ A = (µ A − ν A )+
⎪
⎨ϕ A = (ν A − µ A )+ (6.2.1) From (6.1.2) it results:
⎪ ⎧µ A − ν A = τ A − ϕ A
⎩ι A = 1 − µ A − ν A ⎨ (6.3.2)
⎩µ A + ν A = 1 − π A
having the following equivalent forms:
From (6.1.3) one obtains the inverse transform:
⎧τ A = (2 ⋅ µ A − 1)+
⎪ ⎧ ιA
⎨ϕ A = (1 − 2 ⋅ µ A )+ (6.2.2) ⎪⎪µ A = τ A + 2
⎪ ⎨ (6.3.3)
⎩ι A = 1 − 2 ⋅ µ A − 1 ⎪ν = ϕ + ι A
⎩⎪
A A
Finally, due to the particularity (2.1) of fuzzy 2
sets, the penta-valued representation is reduced
to a three-valued one:
τA + ϕA + ιA = 1
6.4 Paraconsistent Fuzzy Set as FP5
We consider the paraconsistent fuzzy set
A ∈ PFS defined by the membership function
µ A and the non-membership function ν A . We
will translate to a penta-valued fuzzy set using
formulae (6.1.1).
Thus, one defines the indexes of truth, falsity,
contradiction and indeterminacy.
⎧τ A = (µ A − ν A )+
⎪ modeling: Representation, fusion and optimal
⎪ϕ A = (ν A − µ A )+
⎨ (6.4.1)
⎪κ A = µ A + ν A − 1
⎪ι = 2 − µ − ν − µ − ν
⎩ A A A A A
Therefore, a bivalent knowledge representation
was transformed into a tetravalent one, due to
the particularity (2.4) of paraconsistent fuzzy
sets.
The four defined indexes verify the partition of
unity condition:
τA + ϕA + κ A + ιA = 1
From (6.1.2) one obtains:
⎧µ A − ν A = τ A − ϕ A
⎨ (6.4.2)
⎩µ A + ν A = 1 + κ A
From (6.1.3) it results the inverse transform:
⎧ ιA
⎪⎪µ A = τ A + κ A + 2
⎨ (6.4.3)
⎪ν = ϕ + κ + ι A
⎪⎩ A A A
2 Systems, (FUZZY-IEEE 2007), London, pp.
6 Conclusions
In this paper, a method was presented regarding
multi-valued knowledge representation. The
presented method is based on some properties of
the Frank t-norms. Also, a new penta-valued
logic was presented based on five logical values:
true, false, undefined, contradictory and
indeterminate. Using this new logic, new
representations were obtained for fuzzy set,
intuitionistic fuzzy set, paraconsistent fuzzy set
and bipolar fuzzy set. These new representations
and the operators defined on FP5 supply new
algebraic structures for these fuzzy sets types.
References
[1] K. Atanassov (1986). Intuitionistic fuzzy
sets, Fuzzy Sets and Systems 20, pp. 87-96,
1986.
[2] N. Belnap (1977). A Useful Four-valued
Logic, Modern Uses of Multiple-valued Logics
(D. Reidel, ed), Dordrecht-Boston, pp. 8-37,
1977.
[3] S. Benferhat, D. Dubois, S. Kaci, H. Prade
(2006). Bipolar possibility theory in preference
solutions, Information Fusion, Vol 7. pp. 135-
150, 2006.
[4] D. Dubois, S. Kaci, H. Prade (2004).
Bipolarity in reasoning and decision - An
introduction. The case of the possibility theory
framework. Proc. of the 10th Inter. Conf. on
Information Processing and Management of
Uncertainty in Knowledge-Based Systems
(IPMU'04), Perugia, Italy, July, 4-7, 2004.
[5] M. J. Frank (1979). On the simultaneous
associativity of f(x,y) and x+y-f(x,y). Aeq.
Math., 19:194-226, 1979.
[6] Gr. C. Moisil (1965). Old and New essays on
non-classical logics (Romanian), Ed. Stiintifica,
Bucharest, 1965.
[7] Gr. C. Moisil (1972). Essai sur les logiques
non-chrysippiennes, Ed. Academiei, Bucharest,
1972.
[8] Gr. C. Moisil (1975). Lectures on the logic
of fuzzy reasoning, Ed. Stiintifica, Bucharest,
1975.
[9] V. Patrascu (2007). Penta-Valued Fuzzy Set,
The IEEE International Conference on Fuzzy
137-140, U.K, July 24-26, 2007.
[10] L. A. Zadeh (1965). Fuzzy sets,
Information and Control, 8, pp. 338-353, 1965.