See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225320983

On the Relation between Fuzzy and Quantum Logic

Chapter in Studies in Fuzziness and Soft Computing · March 2009

DOI: 10.1007/978-3-540-93802-6_20

CITATIONS READS
14 1,711

3 authors:

Ingo Schmitt Andreas Nürnberger

Brandenburg University of Technology Cottbus - Senftenberg Otto-von-Guericke-Universität Magdeburg
125 PUBLICATIONS 673 CITATIONS 376 PUBLICATIONS 3,792 CITATIONS

SEE PROFILE SEE PROFILE

Sebastian Lehrack
Brandenburg University of Technology Cottbus - Senftenberg
19 PUBLICATIONS 170 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Cross-language information retrieval of less-resourced languages View project

Intention-based Anticipatory Interactive Systems (Intentionale Antizipatorische Interaktive Systeme) View project

All content following this page was uploaded by Ingo Schmitt on 18 January 2016.

The user has requested enhancement of the downloaded file.

1
On the Relation between Fuzzy and Quantum
Logic
Ingo Schmitt, Andreas Nürnberger, Sebastian Lehrack

Summary. Fuzzy logic provides us with a means to deal with concepts like vague-
ness and uncertainty. Quantum logic was developed in the context of quantum me-
chanics and in contrast to fuzzy logic, it is not based on membership values but on
vector subspaces identified by projectors. Interestingly, there are relations between
both theories. The interaction of a projector with a normalized vector produces a
value which can be directly interpreted as fuzzy membership value. In this contribu-
tion we show that under some circumstances the conjunction of projectors directly
corresponds to the algebraic product in fuzzy logic while taking in addition the pro-
ducing projectors into consideration. Thus, quantum logic gives us means to handle
the semantics behind the fuzzy algebraic product and algebraic sum.

1.1 Introduction

Fuzzy logic is a well-established formalism in computer science being strongly

influenced by the work of Zadeh [17, 16]. It provides us with a means to deal
with vagueness and uncertainty. Fuzzy logic is based on t-norms and t-conorms
for intersection and union, respectively, on membership values of fuzzy sets.
Quantum logic was developed in the context of quantum mechanics. In
contrast to fuzzy logic, the logic is not based on membership values but on
vector subspaces identified by projectors. The lattice of all projectors provides
us with a lattice operations interpreted as conjunction and disjunction.
Interestingly, there are relations between both theories. The interaction
of a projector with a normalized vector produces a value which can be in-
terpreted directly as fuzzy membership value. This paper shows, that under
some circumstances the conjunction of projectors directly corresponds to the
t-norm algebraic product in fuzzy logic. However, in contrast to fuzzy logic
which is defined on fuzzy sets, quantum logic takes the producing projectors
into consideration. As result, we are able to overcome the problem of idem-
potence for the algebraic product. Furthermore, if we restrict projectors to
be mutually commuting we obtain a logic obeying the rules of the Boolean
algebra. Thus, quantum logic gives us more insights into the semantics behind
the fuzzy norms algebraic product and algebraic sum.
In the following, we first give in Sect. (1.2) a brief introduction to Fuzzy
Logic and then introduce in more detail in Sect. (1.3) the concepts of Quantum
2

Logic. Finally, we discuss in Sect. (1.4) the relation between both theories and
conclude our work in Sect. (1.5).

1.2 Conjunction and Disjunction in Fuzzy Logic

If humans describe objects, they effectively use linguistic terms like, for in-
stance, small, old, long, fast. However, classical set theory is hardly suited to
define sets of objects that satisfy such linguistic terms. Let us, for examples,
assume a person being assigned to the set of tall persons. If a second person
is only insignificantly smaller, it should also be assigned to this set, and thus
it seems reasonable to formulate a rule like “a person who is less than 1mm
smaller than a tall person is also tall” to define our set. However, if we repeat-
edly apply this rule, obviously persons of any size will be assigned to the set of
tall persons. Any threshold for the concept tall will be hardly justifiable. On
the other hand, it is easy to find persons that are tall and small, respectively.
Modelling the typical cases is not the problem, but the penumbra between the
concepts can hardly be appropriately modelled with classical sets.
The main principle of fuzzy set theory is to generalize the concept of set
membership [17]. In classical set theory a characteristic function

Ω → {0, 1}
1IA : (
1, if ω ∈ A (1.1)
1IA (ω) =
0, otherwise,

defines the memberships of objects ω ∈ Ω to a set A ⊂ Ω. In fuzzy set theory

the characteristic function is replaced by a membership function

µM : Ω → [0, 1], (1.2)

that assigns numbers to objects ω ∈ Ω according to their membership de-

gree to a fuzzy set M . A membership degree of one means that an object
fully belongs to the fuzzy set, zero means that it does not belong to the set.
Membership degrees between zero and one correspond to partial member-
ships. Membership degrees can be used to represent different kinds of im-
perfect knowledge, including similarity, preference, and uncertainty. However,
no framework is provided to model the semantics of an element or how the
membership values had been derived.
Common fuzzy sets are so-called fuzzy numbers (or fuzzy intervals) that
assume a value of one for a single value a ∈ IR (or interval [a, b] ⊂ IR), and
have monotonously decreasing membership degrees with increasing distance
from this core. Fuzzy numbers can be associated with linguistic terms like,
for example, “approximately a”. In fuzzy rule based systems, typically param-
eterized membership functions are used, where these are in most cases either
triangular, trapezoidal, or Gaussian shaped (cf. Figure 1.1):
 3

Fig. 1.1. Examples of typical fuzzy sets

(x − x0 )2
 
µx0 ,σ (x) = exp − . (1.3)
2σ 2
If the complete input range is covered by overlapping fuzzy sets, this is
called fuzzy partition. If their number is sufficiently small, the fuzzy sets M are
usually associated with linguistic terms, e.g. AM ∈ {small, medium, large}.
Conjunctions and disjunctions of fuzzy membership degrees are evaluated
by so-called t-norms and t-conorms, respectively:
Definition 1. A t-norm > : [0, 1]2 → [0, 1] is a commutative and associative
function that satisfies >(a, 1) = a and a ≤ b ⇒ >(a, c) ≤ >(b, c).
Definition 2. A t-conorm ⊥ : [0, 1]2 → [0, 1] is a commutative and associa-
tive function that satisfies ⊥(a, 0) = a and a ≤ b ⇒ ⊥(a, c) ≤ ⊥(b, c).
For a, b ∈ {0, 1}, all t-norms (t-conorms) behave like the traditional conjunc-
tion (disjunction). For the values in between, however, different behaviors are
possible. [17] suggest the usage of max for union, min for intersection and
1 − µ(x) for the complement. While there are more functions available [7, 19],
every intersection operator has to be a t-norm.
First, we will consider min/max the standard because it is the only idem-
potent and first proposed set of functions [17]. [8] shows that the application
of min/max differs from the intuitional understanding of a combination of
values (see below). Furthermore, the binary min/max functions return only
one value. This leads to a value dominance of one of the two input values while
the other one is completely ignored [12, 6, 8]. Thus, min/max cannot express
influences or grades of importance of both values on a result, e.g. max(0.01, 1)
gives the same result as max(0.9, 1) although the values of the second pair do
not differ very much from a human point of view.
The form of the complement shows that the fuzzy set theory and its logic
does not form a Boolean algebra because the conjunction of x with its com-
plement is not equal 0:
x ∧ ¬x = min(x, 1 − x) 6= 0 e.g. for x = 0.5
To overcome the problem of value dominance, parameterized functions
have been presented such as Waller-Kraft [15] or Paice [8, 7]. Their parameter
4

basically regulates the behavior of the function between the extrema of a

t-norm or t-conorm resulting in a more comprehensible behavior for a human.
Alternatively, another pair of norms has been proposed: the algebraic prod-
uct a · b for intersection and the algebraic sum a + b − a · b for union [7]. They
provide means to express statements that involve both values and therefore
attenuate the dominance problem of min/max. In contrast to min/max the
algebraic product is not idempotent and thus no distributivity holds. This can
be easily shown:
x ∧ x = x2 6= x.
If it is not possible to define exact membership degrees it is sometimes
useful to consider only the qualitative order of items. Thus we can define the
concept of an L-Fuzzy-Set using the lattice concept:
Definition 3. Let (L, u, t) be a lattice with lmin being the smallest element
and lmax being the biggest element. Then a L-Fuzzy-Set η of X is a mapping
from the base set X to the set L, i.e.

η : X → L.

L(X) represents the set of all L-Fuzzy-Sets of X.

1.3 Conjunction and Disjunction in Quantum Logic

The development of quantum mechanics dates back to the beginning of the
last century. The early theoretical foundations were strongly influenced by
physicists such as Einstein, Planck, Bohr, Schrödinger and Heisenberg. Quan-
tum mechanics deals with specific phenomena of elementary particles such
as uncertainty of measurements in closed microscopic physical systems and
entangled states. In recent years, quantum mechanics became an interesting
topic for computer scientists who try to exploit its power to solve computa-
tionally hard problems. Introductions to quantum logic for non-physicists can
be found, e.g., in [5, 2, 11].

1.3.1 Mathematical and Physical Foundations

This subsection gives a short introduction to the formalism of quantum me-

chanics and shows its relation to probability theory. After introducing some
notational conventions, we briefly present the four postulates of quantum me-
chanics. Here, we assume the reader being familiar with linear algebra.
The formalism of quantum mechanics deals with vectors of a complex
separable Hilbert space H. For simplicity we present in the following the real-
value view of the formalism. However, the approach can be defined likewise
on complex and real vector space.
The Dirac notation [3] provides us an elegant means to formulate basic
concepts of quantum mechanics:
 5

• A so-called ket vector |xi represents a column   vector identified

 by x. Let
1 0
two special predefined ket vectors be |0i = and |1i = .
0 1
• The transpose of a ket |xi is a row vector hx| called bra whereas the trans-
pose of a bra is again a ket. Both form together a one-to-one relationship.
• The inner product between two kets |xi and |yi returning a scalar equals
the scalar product defined as the product of hx| and |yi. It is denoted by
a bra(c)ket ’hx|yi’. The norm of a ket vector |xi is defined by || |xi || ≡
p
hx|xi.
• The outer product between two kets |xi and |yi is the product of |xi and
hy| and is denoted by ’|xihy|’. It generates a linear operator expressed by
a matrix.
• A projector p = i |iihi| is a symmetric (pt = p) and idempotent (pp = p)
P
linear operator defined over a set of orthonormal vectors |ii. Multiplying a
projector with a state vector |ϕi means to project the vector onto the re-
spective vector subspace. Each projector p is bijectively related to a closed
subspace via p ↔ vsp (H) := {p|ϕi | |ϕi ∈ H}. Despite a projector p can
P an arbitrary orthonormal basis |ii for vsp (H), the de-
be constructed from
rived projector i |iihi| will be always the identity operator of vsp (H). We
can conclude this from the following completeness relation for orthonor-
mal vectors. Let |ii be a vector of an orthonormal basis for vspP (H). Then
an arbitrary vector |ψi ∈ vsp (H) can be expressed as |ψi = i vi |ii in
vsp (H) for some set of scalars vi . Note that hi|vi = vi and therefore
!
X X X
p|ψi = |iihi| |ψi = |iihi|ψi = vi |ii = |ψi
i i i

Since the last equation is true for all |ψi it follows that p is the identity
operator for vsp (H).
• The tensor product between two kets |xi and |yi is denoted by |xi ⊗ |yi
or short by |xyi. If |xi is m-dimensional and |yi n-dimensional then |xyi
is an m·n-dimensional ket vector. The tensor product of two-dimensional
kets |xi and |yi is defined by:
 
    x1 y1
x1 y1  x1 y2 
|xyi ≡ |xi ⊗ |yi ≡ ⊗ ≡ x2 y1  .

x2 y2
x2 y2
The tensor product between two matrices A and B is analogously defined:
 
    x1 y1 x1 y2 x2 y1 x2 y2
x1 x2 y1 y 2  x1 y3 x1 y4 x2 y3 x2 y4 
AB ≡ A ⊗ B ≡ ⊗ ≡
 x3 y1 x3 y2 x4 y1 x4 y2  .

x3 x4 y3 y 4
x3 y3 x3 y4 x4 y3 x4 y4
Next, we sketch the famous four postulates of quantum mechanics:
6

Postulate 1:

Every closed physical microscopic system corresponds to a separable complex

Hilbert space1 and every state of the system is completely described by a
normalized (the norm equals one) ket vector |ϕi of that space.

Postulate 2:

Every evolution of a state |ϕi can be represented by the product of |ϕi and
an orthonormal2 operator O. The new state |ϕ0 i is given by |ϕ0 i = O|ϕi. It
can be easily shown that an orthonormal operator cannot change the norm of
a state: || O|ϕi || = || |ϕi || = 1.

Postulate 3:

A central concept in quantum mechanics is the nondeterministic measurement

of a state which means to compute the probabilities of different outcomes. If
a certain outcome is measured then the system is automatically changed to
that state. Here, we focus on a simplified measurement given by projectors
(each one represents one possible outcome). The probability of an outcome
corresponding to a projector p and a given state |ϕi is defined by
!
X X
hϕ|p|ϕi = hϕ| |iihi| |ϕi = hϕ|iihi|ϕi
i i

Thus, the probability value equals the squared length of the state vector
|ϕi after its projection onto the subspace spanned by the vectors |ii. Due
to normalization, the probability value, furthermore, equals geometrically the
squared cosine of the minimal angle between |ϕi and the subspace represented
by p.
Figure 1.2 illustrates the connection between quantum mechanics and
probability theory for the two-dimensional case. Please notice that the base
vectors |0i and |1i are orthonormal. The measurement of the state |ϕi =
a|0i + b|1i with || |ϕi || = 1 by applying the projector |0ih0| provides the
squared portion of |ϕi on the base vector |0i which equals a2 . Analogously,
the projector |1ih1| provides b2 . Due to Pythagoras and the normalization of
|ϕi both values sum up to one. In quantum mechanics where |0ih0| and |1ih1|
represent two independent outcomes of a measurement the values a2 and b2
give the probabilities of the respective outcomes.
1
For simplicity, we restrict ourselves to the vector space Rn .
2
An operator O is orthonormal if and only if Ot O = OOt = I holds where the
symbol ’t ’ denotes the transpose of a matrix and ’I’ denotes the identity matrix.
 7
hϕ|0ih0|ϕi = a2
|1i
hϕ|1ih1|ϕi = b2
a2 + b2 = 1
1
a

b
|ϕi

1 |0i
Fig. 1.2. Pythagoras and probabilities

Postulate 4:

This postulate defines how to assemble various quantum systems to one sys-
tem. The base vectors of the composed system are constructed by applying
the tensor product ’⊗’ to the subsystem base vectors.

1.3.2 Lattice of Projectors

Following [18], we develop here the main concepts of quantum logic originally
developed by von Neumann [14]. Applying quantum logic on projectors will
give us the capability to measure state vectors on complex conditions. The
starting point is the set P of all projectors of a vector space H of dimensions
greater than two. We want to remind that each projector p ∈ P is bijectively
related to a closed subspace via p ↔ vsp (H) := {p|ϕi | |ϕi ∈ H}. The subset
relation on the corresponding subspaces forms a complete partially ordered
set (poset) of the projector set P whereby p1 ≤ p2 ⇔ vsp1 (H) ⊆ vsp2 (H).
Thus, we obtain a lattice3 with the binary operations meet (u) and join (t)
being defined as

p1 u p2 := p ↔ vsp (H) := vsp1 (H) ∩ vsp2 (H)

p1 t p2 := p ↔ vsp (H) := closure(vsp1 (H) ∪ vsp2 (H))

whereby the closure operation generates here the set of all possible vector
linear combinations. Furthermore, the orthocomplement (¬) is defined as

¬p1 := p ↔ vsp (H) := {|ϕi ∈ H | ∀|ψi ∈ vsp1 (H) : hψ|ϕi = 0}.

In quantum logic the orthocomplement can be interpreted as negation oper-

ator.
3
The laws of commutativity, associativity, and absorption are fulfilled.
8

1.3.3 Boolean Sublattice

Quantum logic in general does not constitute a Boolean algebra since the
distribution law is violated. To confirm this statement, we consider three
projectors p1 , p2 and p3 in a two-dimensional vector space H. The projec-
tors are specified
√ as p1 = |0ih0|, p2 = |1ih1|, and p3 = |vihv| whereby |vi =
(|0i + |1i)/ 2. We can observe that the closure of (vsp1 (H) ∪ vsp2 (H)) spans
the whole vector space H. Contrarily, the intersections (vsp3 (H) ∩ vsp1 (H))
and (vsp3 (H) ∩ vsp2 (H)) collapse to the null vector expressed here by the
projector p0 . Thus, we obtain

p3 u (p1 t p2 ) = p3 6= p0 = p0 t p0 = (p3 u p1 ) t (p3 u p2 )

violating the distribution law.

Fortunately, there exist sublattices of projectors which set up a Boolean
algebra. To identify these convenient sublattices we have to take the commu-
tativity of projectors into account.
Definition 4 (commuting projectors). Two projectors p1 and p2 of a vec-
tor space H are called commuting projectors if and only if p1 p2 = p2 p1 holds.
P
P linear algebra we know that two projectors p1 =
From i |iihi| and
p2 = j |jihj| commute if and only if their ket vectors |ii and |ji are vec-
tors of the same orthonormal basis B = {|k1 i, . . . , |kn i} for the underlying n-
dimensional vector space [2]. In that case, we can define Bp1 and Bp2 , whereby
Bp1 ⊆ B andPBp2 ⊆ B, as sets of orthonormal
P vectors which form the pro-
jectors p1 = i∈Bp |iihi| and p2 = j∈Bp |jihj|. If two projectors commute
1 2
then their join corresponds to the union of the respective sets of underlying
base vectors and their meet to the intersection. Thus, we can redefine the
meet, join and orthocomplement operation for commuting projectors.
Corollary 1 (sublattice operations for commuting projectors). Let p1
and p2 be two commuting projectors. The lattice operations can be adapted to:
X
p1 u p2 := |kihk| (1.4)
k∈Bp1 ∩Bp2
X
p1 t p2 := |kihk| (1.5)
k∈Bp1 ∪Bp2
X
¬p1 := |kihk| (1.6)
k∈B\Bp1

All projectors over one given orthonormal basis form a Boolean algebra.
This is affirmed by Stone’s representation theorem for Boolean algebras [13].
It states that every Boolean algebra is isomorphic to a field of sets and its
corresponding union and intersection operation. Here, the field of sets is the
common orthonormal basis B = {|k1 i, . . . , |kn i} and the respective algebra is
given by its power set 2B forming a subset lattice.
 9

A sublattice of projectors is shown in Figure 1.3. Each projector is con-

structed by a subset of the same orthonormal basis which contains three vec-
tors. The bit code refers to the selected basis vectors from the underlying
orthonormal basis. The code [110], for example, refers to the vector subspace
spanned by the first two basis vectors.

p4 t p5 t p 6
[111]

p 4 ≡ p1 t p2 p5 ≡ p1 t p 3 p 6 ≡ p2 t p3
[110] [101] [011]
join (t)
meet (u)
p 1 ≡ p4 u p5 p2 ≡ p4 u p 6 p 3 ≡ p5 u p6
[100] [010] [001]

p1 u p2 u p 3
[000]
Fig. 1.3. Sublattice of commuting projectors

Actually, quantum logic can be seen as a generalization of a Boolean alge-

bra: The sublattice over every equivalence class comprising commuting pro-
jectors constitutes a Boolean algebra.
A concise overview of further important results for quantum logic is given
in [1, 9, 10].

1.3.4 Mapping Objects to State Vectors

In this subsection we want to briefly explain the main ideas of mapping objects
into the vector space formalism of quantum mechanics.
Following, we distinguish between single-attribute and multi-attribute ob-
jects. We start our considerations with the encoding of a single-attribute ob-
ject with attribute A into a separated local vector space HA . Later we will
merge different single-attribute spaces HAi to a global multi-attribute one
represented by H. Here we only exemplarily describe the mapping of an ar-
bitrary non-negative, numerical value a ∈ [0, ∞) to its corresponding state
vector |ai. The state vector |ai is located in HA and represents the current
value of the attribute A.
Please recall that state vectors need to be normalized. Therefore, we cannot
directly map a value to a one-dimensional ket vector. Instead we need at least
10

two dimensions. A two-dimensional quantum system in the field of quantum

computation is called a qubit (quantum bit). Since every normalized linear
combination of two basis vectors |0i = (1, 0)t and |1i = (0, 1)t is a valid qubit
state vector we can encode infinitely many values. That is, we take advantage
of the superposition principle of quantum mechanics. Please notice that no
more than two vectors can be encoded as pairwise independent (orthogonal)
state vectors within a one-qubit system. So, for the one-qubit encoding the
state vector |ai is embedded in a two-dimensional vector space spanned by |0i
and |1i.
Definition 5 (mapping numerical values to qubit states). The normal-
ized qubit state vector |ai for a numerical value a ∈ [0, ∞) is defined by
 
1 1
a 7→ |ai = √ .
2
a +1 a
Thus, the numerical value is expressed by the normalized ratio between the
two basis vectors |0i and |1i.
A more complex object contains more than one attribute value. There-
fore, we have to adapt our mapping to a multi-attribute version. A multi-
attribute object can be regarded as a state vector in a composite quantum
system. Adopting Postulate 4, we use the tensor product for constructing
multi-attribute state vectors and vector spaces out of single-attribute ones.
Definition 6 (multi-attribute objects as tensor products of single-
attribute states). Assume, an object o = (a1 , . . . , an ) contains n attribute
values and |a1 i, . . . , |an i are their respective state vectors which are embedded
in separated Hilbert spaces HA1 ,. . . ,HAn , respectively. Then, the ket vector
|oi = |a1 i ⊗ . . . ⊗ |an i = |a1 ..an i
represents the object o in a global Hilbert space H = IA1 ⊗ . . . ⊗ IAn whereby
IAi is the identity matrix of HAi .

1.3.5 Measurement of Projectors

In this subsection we will investigate the measurement of projectors in more

detail. In quantum logic projectors are combined to new projectors before any
measurement w.r.t. an object takes place. Thus, a projector can be constructed
from an arbitrary logical condition formula by applying the meet (u), join
(t) and orthocomplement (¬) on projectors. A projector therefore embodies
the complete semantics of a well-formed condition.
In general, the measurement of a projector p on a given state vector |ai is
already introduced (Postulate 3) as
!
X X
ha|p|ai = ha| |iihi| |ai = ha|iihi|ai.
i i
 11

Later we will describe a restriction on the structure of complex conditions

which allows us to simplify the measurement of combined projectors signif-
icantly. Before we will turn our attention to the measurement of projectors
generated by complex conditions, we investigate the single-attribute case.

Constructing and Measurement of Single-Attribute Projectors

The generation of a certain single-attribute projector corresponds to the en-

coding of the respective attribute. For instance, we explore here an object o
with a numerical attribute A (Def. 5) and a projector pc determined by the
numerical condition ’A = c’. Thus, the projector pc is given by pc = |cihc|. It is
related to an one-dimensional subspace in the single-qubit system HA . Com-
puting the degree of matching between state vector |oi=|ai and the projector
pc = |cihc| yields

(1 + ac)2
ho|pc |oi = ha|pc |ai = ha|cihc|ai =
(a2 + 1)(c2 + 1)
   
1 1
whereby |ai = √a12 +1 and |ci = √c12 +1 . The resulting expression is
a c
equivalent to the squared cosine of the enclosed angle between |ai and |ci.
There are different encoding techniques for further attribute domains
which influence the construction of projectors [12]. In every case we have
to preserve the Boolean character of our algebra which is based on commuting
projectors. In particular, it must be guaranteed that only orthogonal condi-
tions per attribute are used. Otherwise, the commutativity of the involved
projectors would be violated.
For example, it is not possible to support different conditions on the same
numerical attribute A. To exemplify that case we assume two conditions ’A =
c1 ’ and ’A = c2 ’ generating two one-dimensional projectors in HA . In general,
these projectors would not be orthogonal and therefore not commuting. That
is, their projectors cannot be based on one common set of orthonormal basis
vectors. In consequence of this fact, there is no proper way to express the
condition ’A = c1 ∨ A = c2 ’ in a single-qubit system HA .
But there also exists a special case for a measurement in which this effect
does not occur. Assume, we are only interested in a Boolean result (true ≡ 1
or f alse ≡ 0) for a measurement on a condition ’B = c’. The type of attribute
B is called Boolean condition attribute and the constant c is given by a value
of the attribute domain DB . Before we present the measurement of a state
vector |bi on condition ’B = c’ we have to briefly clarify the mapping of |bi
into its corresponding Hilbert space HB . The main idea is to bijectively assign
each possible attribute value dB ∈ DB to exactly one basis vector for HB .
Thus, a value of DB with |DB | = n is expressed by a vector of a predefined
basis of HB = Rn . So, the vector space HB is spanned by the predefined set
of n orthonormal basis vectors |dB i where each |dB i corresponds bijectively
12

to a value dB ∈ DB . Let now C ⊆ DB contain the required values of a

P over the attribute B. Such a condition is expressed by the projector
condition
pC = c∈C |cihc|.
Since all possible projectors pC on the domain DB are based on the same
basis they commute to each other. In consequence, the introduced adapted
meet, join and orthocomplement operation can be applied and those projec-
tors altogether constitute a Boolean algebra.
The following theorem shows that quantum measurement (Postulate 3) for
conditions on these special attributes yields either 1 or 0 as result.
Theorem 1 (measuring Boolean condition attributes). Let B be a
Boolean condition attribute and |bi an object state vector in HB . The mea-
surement result of a projector pC (C ⊆ DB ) is given by

1:b∈C
hb|pC |bi =
0 : otherwise.
Proof.
!
X X
hb|pC |bi = hb| |cihc| |bi = hb|cihc|bi
c∈C c∈C

Due to orthonormality of the basis vectors |ci we can write hb|ci = δ(b, c)
where δ is the Kronecker delta. That is, the measurement yields the value 1
only if b ∈ C holds. Otherwise, we obtain the value 0. u
t
Next we shift to a projector over a single-attribute Ai applying to a multi-
attribute object |oi = |a1 . . . an i. A condition ’Ai = c’ on a multi-attribute
object must be prepared accordingly to the definition of a multi-attribute
object (Def. 6). Thus, a single-attribute projector |cihc| needs to be combined
with all orthonormal basis vectors (expressed by the identity matrix IAj ) of
the non-restricted attributes.
Definition 7 (applying single-attribute projectors to multi-attribute
objects). Assume, ’Ai = c’ is a condition on attribute Ai . Its projector pc
expressing the condition against an n-attribute object is given by
pc = IA1 ⊗ . . . ⊗ IA(i−1) ⊗ |cihc| ⊗ IA(i+1) ⊗ . . . ⊗ IAn .
The following measurement formula yields the measurement value for a given
object |oi=|a1 ..an i.
ha1 ..an |IA1 ⊗ . . . ⊗ IA(i−1) ⊗ |cihc| ⊗ IA(i+1) ⊗ . . . ⊗ IAn |a1 ..an i =
ha1 |IA1 |a1 i . . . ha(i−1) |IA(i−1) |a(i−1) ihai |cihc|ai i ∗
ha(i+1) |IA(i+1) |a(i+1) i . . . han |IAn |an i = hai |cihc|ai i.
The result equals the measurement of the single-attribute object case. That
is, the computation of the measurement becomes very easy since we can com-
pletely ignore non-restricted attributes.
 13

Constructing and Measurement of Multi-Attribute Projectors

A projector over different attributes is based on a complex condition which is

constructed by recursively applying conjunction, disjunction and negation on
atomic conditions. Here, we want to regard a select-condition ’Ai = c’ with
an arbitrary constant c as an atomic condition. For combining two projectors
conjunctively (∧) we apply the meet operator returning a new projector. Anal-
ogously, disjunction (∨) corresponds to the join operator and the negation
(¬) of a condition is related to the orthocomplement of a projector. Despite
dealing with probability values, quantum logic behaves like Boolean algebra
if involved projectors do commute. We assume for the rest of this work a
sublattice of commuting projectors, respectively a Boolean algebra.
To support the measurement of a combined projector we can directly ex-
ploit the structure of the underlying condition. We require conditions to be
combined with disjoint sets of restricted attributes. That means, no attribute
is restricted by more than one operand of a conjunction or disjunction. We
will call this kind of conditions non-overlapping w.r.t. to a set of attributes.
Based on the requirement of disjoint conditions we develop simple evalua-
tion rules for logical operations (∧, ∨ and ¬) to measure a combined projector.
In particular, the measurement of atomic conditions and the application of
these evaluation rules are sufficient to compute the measurement of a projector
generated by a complex condition.

Negation:

The following theorem relates the orthocomplement of projectors to the mea-

surement of a negated condition.
Theorem 2 (measurement of negated projectors). Assume, a projec-
tor pc expressing an arbitrary condition c is given. The measurement of the
negated condition by applying p¬c on an object |oi equals the subtraction of
the non-negated measurement from 1:

ho|p¬c |oi = 1 − ho|pc |oi.

Proof. The orthocomplement for projectors can be also expressed as ¬p ≡ I − p

encompassing all projectors orthogonal to p. The expression I stands for the
identity matrix. Exploiting this formula and a state vector, we obtain

ho|p¬c |oi = ho|I − pc |oi = ho|I|oi − ho|pc |oi = 1 − ho|pc |oi.

t
u
The introduced negation for the measurement extends Boolean negation. How-
ever, if a measurement returns a probability value between 0 and 1 then the ef-
fect may be surprising. For example, assume an attribute A of the three-valued
domain {a, b, c} is given. Surprisingly, as shown in Table 1.1, the negated con-
dition ’¬A = b’ does not equal the condition ’A = a ∨ A = c’. Instead, that
14

condition yields the dissimilarity between the attribute value and the value
b. Thus, the measurement value of the value a is smaller than 1. This effect
is the direct consequence of dealing with values between 0 and 1.

query object value

condition a b c
A=b 0.75 1 0.75
¬(A = b) 0.25 0 0.25

Table 1.1. Negation values

Conjunction:

We will deduce from the following theorem that the measurement of a pro-
jector pa∧b generated by conjunctively combined conditions a and b can be
evaluated as algebraic product, if we require disjoint sets of restricted at-
tributes.

Theorem 3 (measurement of projectors generated by conjunctively

combined non-overlapping conditions).
Let pa = p1a ⊗ . . . ⊗ pna be a projector on n attributes and k restrictions on
the attributes {a1 , .., ak } ⊆ [1, .., n] with

i an ai -restriction : i ∈ {a1 , .., ak }
pa =
I : otherwise

and pb = p1b ⊗. . .⊗pnb be a further projector with l restrictions on the attributes

{b1 , .., bl } ⊆ [1, .., n]

i a bi -restriction : i ∈ {b1 , .., bl }
pb =
I : otherwise

and {a1 , .., ak } ∩ {b1 , .., bl } = ∅. Then, computing the measurement of the
projector pa∧b = p1a∧b ⊗ . . . ⊗ pna∧b on an object |oi yields

ho|pa∧b |oi = ho|pa |oiho|pb |oi.

Proof. The meet operation of projectors is defined over the intersection of the
corresponding subspaces. Thus, we obtain following derivation
 15

pa u pb = (p1a ⊗ . . . ⊗ pna ) u (p1b ⊗ . . . ⊗ pnb )

= (p1a u p1b ) ⊗ . . . ⊗ (pna u pnb )
= p1a∧b ⊗ . . . ⊗ pna∧b whereby

p1a∧b ↔ vsp1a∧b (H) = vsp1a (H) ∩ vsp1b (H),

...,
pna∧b ↔ vspna∧b (H) = vspna (H) ∩ vspnb (H)

Due to the disjointness {a1 , .., ak } ∩ {b1 , .., bl } = ∅ the vector space of every
attribute restriction is intersected with H producing identical restrictions.
Thus, all restriction are simply taken over and the projector pa∧b is obtained
as pa∧b = p1a∧b ⊗ . . . ⊗ pna∧b with

 an ai -restriction : i ∈ {a1 , .., ak }
pia∧b = a bi -restriction : i ∈ {b1 , .., bl }
I : otherwise


Due to these restrictions and the rule ha1 b1 |a2 b2 i = ha1 |a2 ihb1 |b2 i the mea-
surement of the projector pa∧b on an object |oi can be calculated by

ho|pa∧b |oi = ho|p1a∧b ⊗ . . . ⊗ pna∧b |oi

= ho|p1a |oi . . . ho|pka |oi ho|p1b |oi . . . ho|plb |oi ho|I 1 |oi . . . ho|I m |oi
| {z }| {z }| {z }
ho|pa |oi ho|pb |oi 1

= ho|pa |oiho|pb |oi

whereby m = n − (k + l) is the number of unrestricted attributes. Thus, the

measured results for conjunctively combined disjoint projectors are simply
multiplied. u
t

This important result can be exemplified by the following measurement of

multi-attribute object o. It is formed by two arbitrary numerical attributes
A1 and A2 . The state vector |oi = |a1 i ⊗ |a2 i = |a1 a2 i is located in the vector
space H = HA1 ⊗ HA2 whereby HA1 and HA2 stand for single-qubit systems.
The corresponding condition of interest is given by ’A1 = c1 ∧ A2 = c2 ’.
Initially, we can regard the conditions ’A1 = c1 ’ and ’A2 = c2 ’ as atomic
conditions integrated in HA1 and HA2 . Then, the conditions are expressed by
the two projectors pc1 = |c1 ihc1 | in HA1 and pc2 = |c2 ihc2 | in HA2 . Before
we can combine pc1 and pc2 in H, we have to map the both single-attribute
projectors to H. We label the extended projectors in H as p0c1 and p0c2 and
their respective sets of orthonormal vectors as Bp0c1 and Bp0c2 .
For the construction of p0c1 and p0c2 the original vectors |c1 i and |c2 i must
be combined with an orthonormal basis of the respective oppositional vector
space HAi (Def. 7). So, the vector |c1 i needs to be combined with all vectors
16

of an arbitrary orthonormal basis for HA1 , and an orthonormal basis for HA2
needs to be combined with the vector |c2 i. Here, we choose {|c1 i, |c1 i} for HA1
and {|c2 i, |c2 i} for HA2 , respectively. Please notice that the overline notation
denotes the negation of a vector: |ϕi = |¬ϕi. Thus, we obtain

A1 = c1 : Bp0c = {|c1 c2 i, |c1 c2 i}

1

⇒ p0c1 = |c1 c2 ihc1 c2 | + |c1 c2 ihc1 c2 |

A2 = c2 : Bp0c2 = {|c1 c2 i, |c1 c2 i}

⇒ p0c2 = |c1 c2 ihc1 c2 | + |c1 c2 ihc1 c2 |

The projectors p0c1 and p0c2 are commuting because they are based on the same
orthonormal basis {|c1 c2 i, |c1 c2 i, |c1 c2 i, |c1 c2 i} for H. Therefore, we are able
to combine the projectors p0c1 and p0c2 by applying the adapted meet Operation
(1.4) for commuting projectors:
X
pc1 ∧c2 = |kihk| = |c1 c2 ihc1 c2 |
k∈(Bp0 ∩Bp0 )
c1 c2

The expected result is obtained when we compute the measurement on the

state vector |oi = |a1 a2 i.

ho|pc1 ∧c2 |oi = ha1 a2 |c1 c2 ihc1 c2 |a1 a2 i

= ha1 |c1 iha2 |c2 ihc1 |a1 ihc2 |a2 i
= ha1 |c1 i2 ∗ ha2 |c2 i2
= ha1 |pc1 |a1 i ∗ ha2 |pc2 |a2 i

The last equation shows the simple multiplication of the single-attribute mea-
surement results for this example.

Disjunction:

We know that a Boolean algebra respects the de Morgan law [4]. Therefore,
we can compute the measurement for the disjunction of non-overlapping con-
ditions over conjunction and negation and obtain

ho|pa∨b |oi = 1 − (1 − ho|pa |oi)(1 − ho|pb |oi)

= ho|pa |oi + ho|pb |oi − ho|pa∧b |oi.

We are now able to define evaluation rules for the measurement of complex
non-overlapping conditions on multi-attribute objects.
 17

Definition 8 (negation, conjunction and disjunction of non-overlapping

conditions). Let c1 and c2 be two commuting conditions which do not con-
tain overlapping atomic conditions. For the evaluation w.r.t. a given object o
we define:

evalo (¬c1 ) = 1 − evalo (c1 ) (1.7)

evalo (c1 ∧ c2 ) = evalo (c1 ) ∗ evalo (c2 ) (1.8)
evalo (c1 ∨ c2 ) = evalo (c1 ) + evalo (c2 ) −
evalo (c1 ∧ c2 ) (1.9)

To evaluate overlapping conditions we have to apply an evaluation and

transformation algorithm which exploits the already introduced rules and the
following special case of mutually excluding conditions.

Theorem 4 (measurement of projectors generated by disjunctively

combined exclusive conditions). Assume, a projector pc1 ∨c2 is determined
by the condition c1 ∨ c2 whereby c1 ≡ (u ∧ e1 ) and c2 ≡ (¬u ∧ e2 ) are com-
muting exclusive subconditions. Moreover, the literals u and ¬u represent two
mutually excluding atomic conditions and the subformulas e1 and e2 can be
formed by arbitrary conditions. Computing the measurement of the projector
pc1 ∨c2 on an object |oi yields

ho|pc1 ∨c2 |oi = ho|pc1 |oi + ho|pc2 |oi.

Proof. Since the projectors pc1 and pc2 are commuting we can apply the
adapted join Operation (1.5) to measure pc1 ∨c2 . Let Bpc1 and Bpc2 the sets of
orthonormal basis vectors for pc1 and pc2 . We can state that the intersection
of Bpc1 and Bpc2 is always empty because the first component of each basis
vector |u . . .i for pc1 is different from the first component of each basis vector
|¬u . . .i for pc2 . Thus, we obtain

 
X
ho|pc1 ∨c2 |oi = ho|  |kihk| |oi
k∈Bpc1 ∪Bpc2
X X
= ho|kihk|oi + ho|kihk|oi
k∈Bpc1 k∈Bpc2
   
X X
= ho|  |kihk| |oi + ho|  |kihk| |oi
k∈Bpc1 k∈Bpc2

= ho|pc2 |oi + ho|pc2 |oi

t
u

Based on the last theorem we can formulate a further evaluation rule.

18

Definition 9 (disjunction of overlapping exclusive conditions). Let c1

and c2 be two commuting, exclusive and overlapping conditions. We can for-
mulate the following evaluation rule:

evalo (c1 ∨ c2 ) = evalo (c1 ) + evalo (c2 ). (1.10)

Our evaluation algorithm transforms expressions with overlapping condi-

tions into exclusive ones by applying Boolean rules. To compute the mea-
surement of the transformed conditions the rules of Definition 8 and 9 are
used.

Evaluation algorithm:

The algorithm evaluates a given condition w.r.t. a given object. We will show
that our evaluation is based on simple boolean transformations and basic
arithmetic operations. The algorithm for transforming an condition c is given
in Figure 1.4.

input: condition c
output: non-overlapping or mutually excluding condition c

(1) transform expression c into

disjunctive normal form x̂1 ∨ . . . ∨ x̂m
where x̂i are conjunctions of literals
(2) simplify expression c by applying
idempotence and invertibility4 rules
(3) if there is an overlap on a
attribute between some conjunctions x̂i then
(3a) let u be a literal of an attribute
common to at least two conjunctions
(3b) replace all conjunctions x̂i of c
with (u ∧ x̂i ) ∨ (¬u ∧ x̂i )
(3c) simplify c by applying idempotence,
invertibility, and absorption and obtain
c = (u ∧ x̂1 ) ∨ . . . ∨ (u ∧ x̂m1 )∨
(¬u ∧ x̂m1 +1 ) ∨ . . . ∨ (¬u ∧ x̂m2 )
(3d) replace c with (u ∧ e1 ) ∨ (¬u ∧ e2 ) where
e1 = x̂1 ∨ . . . ∨ x̂m1 , e2 = x̂m1 +1 ∨ . . . ∨ x̂m2
(3e) continue with step (3) for e1 and e2
(4) transform innermost disjunctions to
conjunctions and negations by applying
de-Morgan-law

Fig. 1.4. Transformation algorithm to resolve overlaps

 19

Analyzing the transformation result, we observe that the subformulas of

the innermost disjunctions (the leaves of the corresponding tree) are mutually
non-overlapping on attributes5 before we apply the fourth step. Thus, we can
directly apply Formula (1.9). All other disjunctions are based on exclusive
subformulas (generated by step (3d)). That is, we can apply Formula (1.10)
and simply add the scores. Since, furthermore, all conjunctions are based on
non-overlapping subformulas Formula (1.8) directly applies. The fourth step
is to simplify arithmetic calculations of multiple disjunctions.
Finally, we demonstrate the evaluation algorithm using an object o formed
by five attributes. Assume, the condition c is given by

c ≡ (A1 = d ∧ ((A2 = e ∧ (A3 = f )) ∨ (A2 = ¬e ∧ A4 = g))) ∨ A5 = h

whereby d, . . . , h are numerical constants. Note that A2 = e and A2 = ¬e are

orthogonal conditions. Hence, their corresponding projectors are commuting,
despite they restrict the same attribute. Consequently, we can still apply the
introduced evaluation rules for commuting projectors. In Figure 1.5 we ab-
breviate atomic conditions and attributes to the labels of the corresponding
constants d, . . . , h whereby do stands for the expression evaloA1 (d).

c ≡ (d ∧ ((e ∧ f ) ∨ (¬e ∧ g))) ∨ h

(1)(2)

(e ∧ d ∧ f ) ∨ (¬e ∧ d ∧ g) ∨ h
(3a)(3b)(3c) u=e

(e ∧ d ∧ f ) ∨ (e ∧ h) ∨ (¬e ∧ d ∧ g) ∨ (¬e ∧ h)
(3d)

(e ∧ ((d ∧ f ) ∨ h)) ∨ (¬e ∧ ((d ∧ g) ∨ h))

(4)

(e ∧ ¬ (¬ (d ∧ f ) ∧ ¬h)) ∨ (¬e ∧ ¬ (¬ (d ∧ g) ∧ ¬h))

arithmetic evaluation w.r.t. data object o:

evalo (c) = eo (1 − (1 − do f o ) (1 − ho )) +
(1 − eo ) (1 − (1 − do g o ) (1 − ho ))

Fig. 1.5. Example transformations and arithmetic evaluation

4
invertibility: a ∨ ¬a = 1, a ∧ ¬a = 0, ¬¬a = a
5
Otherwise the algorithm would not have stopped.
20

Summarising, we can emphasise again that we are now able to evaluate

an arbitrary commuting condition by means of the transformation algorithm
and simply arithmetic operations.

1.4 Fuzzy Logic versus Quantum Logic

After recapitulating fuzzy logic in Section 1.2 and introducing quantum logic
we will interrelate and compare concepts from both worlds. Both logics deal
with non-Boolean fulfillments of object conditions. Table 1.2 shows correspon-
dences between their underlying concepts.

quantum logic fuzzy logic

normalized vector object
projector measurement fuzzy set
projector complement complement of a fuzzy set
lattice operations fuzzy set operations
- meet on disjoint projectors - t-norm algebraic product
- join on disjoint projectors - t-conorm algebraic sum
Table 1.2. Correspondences between quantum and fuzzy logic concepts

The basic connection between a measurement by a projector p and a fuzzy

set s with respect to an object o is given by

µs (o) = ho|p|oi.

Both logics follow different ways of combining conditions being graphically

depicted in Figure 1.6:

µs1 ∩s2 (o) = >(µs1 , µs2 ) versus ho| u (p1 , p2 )|oi

µs1 ∪s2 (o) = ⊥(µs1 , µs2 ) versus ho| t (p1 , p2 )|oi

In fuzzy logic, conjunction and disjunction are directly based on a t-norm (>)
and a t-conorm (⊥), respectively, on membership values. In quantum logic,
however, these operation are performed on projectors before any evaluation
takes place. This fundamental difference gives quantum logic an advantage by
allowing it to consider query semantics during combining complex conditions.
Thus, we are able to see that the conjunctive combination only of disjoint
conditions in quantum logic yields the same result as the algebraic product
in fuzzy logic. The test on disjointness, however, is not feasible in fuzzy logic
since a t-norm is defined purely on membership values.
That property of the quantum approach allows us to differentiate seman-
tical cases during the evaluation. Thus, if we restrict our quantum conditions
to commuting projectors then all rules of a boolean algebra are obeyed. This
 21

P ho|p|oi
¬ p= i |iihi| [0, 1]

t ∩

µs (o)
fuzzy set s [0, 1] ¬

Fig. 1.6. Construction of complex conditions in quantum and in fuzzy logic and
their evaluations

is impossible in fuzzy logic because required semantics (conditions are com-

muting) is hidden behind the fuzzy sets. From this point of view we conclude,
that quantum logic can takes more condition semantics into account than
fuzzy logic can do.
A bridge between quantum logic and fuzzy logic can be established if we
use the generalized definition of a fuzzy set over conditions which is called
an L-fuzzy set. The lattice operations meet(∧), join(∨), and complement are
then used for conjunction, disjunction and negation on conditions. The lattice
is, of course, our projector lattice.
This bridge in combination with the algebraic product as t-norm and the
algebraic sum as t-conorm is depicted in Figure 1.7 where we assume disjoint
conditions. We use the by-pass over the projector lattice in order to prove that
the algebraic product provides correct answers. In practice, we can directly
apply the algebraic product on object evaluations but only if the underlying
conjunctively combined conditions are disjoint.

1.5 Conclusion
In our contribution we investigated the relation between concepts from fuzzy
logic and quantum logic. For commuting conditions we could show that quan-
tum logic follow the rules of a Boolean algebra. As main difference between
fuzzy and quantum logic we identified the way how conditions are combined
by conjunction and disjunction with respect to a given object: combination
in quantum logic is performed before and in fuzzy logic after object evalua-
tion takes place. Therefore, in quantum logic we are able to test conditions
to be combined on disjointness. In case of disjointness the effect of quantum
combination coincides with the the fuzzy combination using algebraic product
22
ho|p1 u p2 |oi
p1 u p2 [0, 1]

µp1 (o) = ho|p1 |oi

=
µp2 (o) = ho|p2 |oi

µp1 ∧p2 (o) = ho|p1 |oiho|p2 |oi

[0, 1] [0, 1]

ho|p1 u p2 |oi = µp1 ∧p2 (o)

= µp1 (o) ∗ µp2 (o)
= ho|p1 |oiho|p2 |oi

Fig. 1.7. CQQL evaluation of conjunctively combined and disjoint conditions on

object o

and norm. If disjointness is not fullfilled then an algorithm basing on rules

from Boolean algebra is presented which converts any complex condition into
a disjoint condition.
Besides theoretical insights into the relation between both worlds we learnt
we how to use the t-norms algebraic product and sum in order to obtain a
Boolean algebra.
In future work we will investigate how to deal with non-commuting condi-
tions. Furthermore, we plan to construct a complete database query language
in order to integrate concepts from information retrieval into classical database
systems.
References

1. E. Beltrametti and B.C. van Fraassen, editors. Current Issues in Quantum

Logic. Plenum Press, 1981.
2. I. Chuang, M. A. Nielsen, and I. L. Chuang. Quantum Computation and Quan-
tum Information. Cambridge University Press, 2000.
3. P. Dirac. The Principles of Quantum Mechanics. Oxford University Press, 4th
edition, 1958.
4. P. Dwinger. Introduction to Boolean algebras. Physica Verlag, Würzburg, 1971.
5. J. Gruska. Quantum Computing. McGraw-Hill, 1999.
6. Aljoscha Klose and Andreas Nürnberger. On the properties of prototype-based
fuzzy classifiers. IEEE Transactions on Systems, Man, and Cybernetics Part B,
37(4):817–835, 2007.
7. Rudolf Kruse, Joerg Gebhardt, and Frank Klawonn. Fuzzy-Systeme. Teubner,
Stuttgart, Germany, 1993.
8. Joon Ho Lee, Myoung Ho Kim, and Yoon Joon Lee. Ranking Documents in
Thesaurus-based Boolean Retrieval Systems. Inf. Process. Manage., 30(1):79–
91, 1994.
9. P.F. Lock. Connections among quantum logics, part 1: Quantum propositional
logics. International Journal of Theoretical Physics, 24:43–53, 1985.
10. P.F. Lock. Connections among quantum logics, part 2: Quantum event logics.
International Journal of Theoretical Physics, 24:55–61, 1985.
11. E. Rieffel and W. Polak. An introduction to quantum computing for non-
physicists. ACM Computing Surveys, 32(3):330–335, 2000.
12. Ingo Schmitt. Qql: A db&ir query language. VLDB J., 17(1):39–56, 2008.
13. M. H. Stone. The Theory of Representations of Boolean Algebras. Transactions
of the American Mathematical Society, 40:37–111, 1936.
14. J. von Neumann. Grundlagen der Quantenmechanik. Springer Verlag, Berlin,
Heidelberg, New York, 1932.
15. W. G. Waller and D. H. Kraft. A mathematical model for a weighted boolean
retrieval system. Information Processing and Management, 15(5):235–245, 1979.
16. Lofti A. Zadeh. Fuzzy Logic. IEEE Computer, 21(4):83–93, April 1988.
17. Lotfi A. Zadeh. Fuzzy Sets. Information and Control, (8):338–353, 1965.
18. Martin Ziegler. Quantum Logic: Order Structures in Quantum Mechanics. Tech-
nical report, University Paderborn, Germany, 2005.
19. H.-J. Zimmermann. Fuzzy Set Theory -and its applications (3rd ed.). Kluwer
Academic Publishers, Norwell, MA, USA, 1996.

View publication stats