Qian Lei, Zeshui Xu - Intuitionistic Fuzzy Calculus-Springer (2017)

Qian Lei Zeshui Xu
Intuitionistic Fuzzy Calculus
123
Qian Lei Zeshui Xu
College of Science Business School
PLA University of Science and Technology Sichuan University
Jiangsu, Nanjing Chengdu, Sichuan
China China
ISSN 1434-9922 ISSN 1860-0808 (electronic)

Studies in Fuzziness and Soft Computing
ISBN 978-3-319-54147-1 ISBN 978-3-319-54148-8 (eBook)
DOI 10.1007/978-3-319-54148-8
Library of Congress Control Number: 2017934465
© Springer International Publishing AG 2017
This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Since Zadeh proposed the concept of fuzzy set in 1965, the fuzzy set theory has
been rapidly developed and vastly applied in many fields. Over the last decades, a
variety of generalizations of classical fuzzy set have been derived from various
angles, one of which is the intuitionistic fuzzy set, which was given by Atanassov in
1983. It depicts the vagueness and uncertainty of things more comprehensively by
introducing a membership function and a non-membership function. Later on, Xu
and Yager defined the basic elements of an intuitionistic fuzzy set as intuitionistic
fuzzy numbers, which are essentially pairs of non-negative numbers belonging to
the closed unit interval [0,1]. Intuitionistic fuzzy calculus, which is analogous to the
calculus of the real numbers and the complex numbers in the classical mathematical
analysis, is established by regarding IFNs as the basic elements.
The main purpose of this book is to give a thorough and systematic introduction
to the latest research results on intuitionistic fuzzy calculus, which essentially focus
on two issues, one of which is to build the calculus theory under intuitionistic fuzzy
environment (denoted by Q1 in the structure diagram), the other is about how to
aggregate the continuous intuitionistic fuzzy data or information (denoted by Q2 in
the structure diagram). The book is constructed into six chapters that deal with the
related issues, which are listed as follows:
Chapter 1 mainly introduces the fundamental knowledge related to IFNs. We
first introduce the concepts of fuzzy sets and intuitionistic fuzzy sets. Later on, we
present the concept of the IFN, and its two representation methods, one of which
describes the IFNs as some points in two-dimensional plane, another expresses
them as the closed subintervals in the unit interval. Then, the operational laws of
IFNs, namely: addition, subtraction, multiplication, division, scalar-multiplication
and power operation, are provided in this chapter. Moreover, the geometrical and
algebraic properties of these operations are analyzed in detail. Also, we define the
change region and the non-change region of IFNs. Last but not least, we show three
kinds of order relations and utilize them to compare and rank IFNs, and then reveal
the relationships among the several orders.
Chapter 2 first gives the definition of intuitionistic fuzzy functions (IFFs), which
is just the object to be studied in the intuitionistic fuzzy calculus. Then, we
introduce the monotonically increasing IFFs and the continuous IFFs. By taking the
limit values of difference quotients of IFFs, the derivatives of IFFs are acquired.
Moreover, we make further efforts to give a criterion of differentiability of IFFs, and
research its important properties, including the chain rule of the derivatives of
compound IFFs. After getting the derivatives of IFFs, we define the differentials of
IFFs, and provide the relationship between the increment of one IFF and its dif-
ferential. In addition, the form invariance of differential in the intuitionistic fuzzy
calculus is revealed in this chapter.
Chapter 3 is devoted to the indefinite integrals of IFFs, which are essentially the
inverse operations of derivatives of IFFs. Then, some properties of the indefinite
integrals of IFFs are discussed, including the substitution rules. Afterwards, we
define the definite integrals of IFFs by utilizing two different methods, one is
developed based on a novel concept (intuitionistic fuzzy integral curves (IFICs)),
the other is introduced based on a closed interval of IFNs, and these two definitions
of the definite integrals of IFFs are completely equivalent to each other. There only
exists one difference between them, which is that the integrals of complex functions
along a curve are in the complex plane, while the other is more similar to the
integrals of real functions in a closed interval of real number axis. By building the
definite integral of IFF with the variable upper limit, we establish the fundamental
theorem (Newton–Leibniz formula) in the intuitionistic fuzzy calculus. Finally, the
definite integrals of IFFs are successfully utilized to aggregate information and data
in intuitionistic fuzzy environment.
Chapter 4 focuses on the methods aggregating continuous intuitionistic fuzzy
information. The study of this issue is essential and meaningful, which likes that the
probability theory and the mathematical statistics not only need to research the
discrete-type random variables, but also concern the continuous-type random
variables. We define the integral aggregating value of the region of IFNs, which
contains the aggregated IFNs. Moreover, many properties about it are provided.
The concept of integral aggregating value is utilized to generate a novel aggregation
technique (IFIA), which is able to deal with the continuous intuitionistic fuzzy
information. Lastly, we prove the idempotency, boundedness and monotonicity
of the IFIA, and utilize the given operator to handle some practical problems.
Chapter 5 mainly investigates the relationships among the definitions and con-
cepts proposed in the previous chapters. We show that there are closed connections
among the IFWA operator (a common aggregating technique utilized to aggregate
discrete IFNs), the IFIA operator and the definite integral of IFFs, and figure out
that the IFWA operator is only the integral of some specific IFFs. In addition, the
IFIA operator 4 is the integral of a special IFF, which is essentially the continuous
form of the IFWA. In other words, the IFWA is the discrete form of the IFIA.
Chapter 6 proposes the complement theory of the intuitionistic fuzzy calculus
built in the previous chapters, and studies the relationship between the calculus of
IFFs and its complement theory. Firstly, we study the complements of fundamental
knowledge of IFNs, and prove the closed connections among these operations of
IFNs and their complements. Secondly, we give the complements of derivatives,
differentials, indefinite integrals and definite integrals of IFFs. Then, the
Diagram of the organizational structure of this book
aggregation operators in the previous chapters are investigated based on the concept
of complement. In brief, this chapter manages to reveal the fact that any statement
or conclusion in the intuitionistic fuzzy calculus must have a counterpart in its
complement theory.
A diagram of the organizational structure of this book is provided to manifest the
structure of this book more clearly:
This book can be used as a reference for researchers and practitioners working in
the fields of fuzzy mathematics, operations research, information science, man-
agement science, engineering, etc. It can also be used as a textbook for postgraduate
and senior-year undergraduate students.
This work was supported by the National Natural Science Foundation of China
(No. 71571123).
Nanjing, China Qian Lei

Chengdu, China Zeshui Xu
July 2016
Contents
1 Basic Concepts Related to Intuitionistic Fuzzy Numbers . . . . . . . . . . 1

1.1 Introduction to Intuitionistic Fuzzy Numbers . . . . . . . . . . . . . . . . . 1
1.2 Basic Operations Between Intuitionistic Fuzzy Numbers . . . . . . . . 3
1.2.1 Geometrical Analysis of the Operations of IFNs . . . . . . . . . 5
1.2.2 Algebraic Analysis of the Operations of IFNs . . . . . . . . . . . 10
1.3 Order Relations of IFNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3.1 Three Kinds of Orders of IFNs and Relationships
Among Them . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 15
1.3.2 Properties of the Order Based
on the Operations of IFNs . . . . . . . . . . . . . . . . . . . . . . .... 17
1.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 20
2 Derivatives and Differentials of Intuitionistic Fuzzy Functions . . . . . 21
2.1 Intuitionistic Fuzzy Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2 Derivatives of Intuitionistic Fuzzy Functions . . . . . . . . . . . . . . . . . 25
2.3 Differentials of Intuitionistic Fuzzy Functions . . . . . . . . . . . . . . . . 33
2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3 Integrals of Intuitionistic Fuzzy Functions . . . . . . . . . . . . . . . . . . . . . 41
3.1 Indefinite Integrals of Intuitionistic Fuzzy Functions . . . . . . . . . . . 41
3.2 Definite Integrals of Intuitionistic Fuzzy Functions . . . . . . . . . . . . . 45
3.2.1 Generating Ways of Definite Integrals of IFFs . . . . . . . . . . 46
3.2.2 Properties of Definite Integrals of IFFs . . . . . . . . . . . . . . . . 53
3.3 Fundamental Theorem of Intuitionistic Fuzzy Calculus . . . . . . . . . 59
3.4 Application of the Definite Integrals of IFFs . . . . . . . . . . . . . . . . . 63
3.4.1 Aggregating Operator Based
on the Definite Integrals of IFFs . . . . . . . . . . . . ......... 64
3.4.2 Properties of Aggregation Operator Built
by the Definite Integrals of IFFs . . . . . . . . . . . . ......... 67
3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......... 72
4 Aggregation Operations of Continuous Intuitionistic
Fuzzy Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......... 73
4.1 Method Dealing with Continuous Intuitionistic Fuzzy
Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2 Properties of Integral Aggregating Value . . . . . . . . . . . . . . . . . . . . 80
4.3 Application of the Integral Aggregating Value . . . . . . . . . . . . . . . . 86
4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5 Relationships Among IFWA Operator, IFIA Operator
and Definite Integrals of IFFs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.1 IFWA Operator and Definite Integrals of IFFs . . . . . . . . . . . . . . . . 99
5.2 IFIA Operator and Definite Integrals of IFFs . . . . . . . . . . . . . . . . . 108
5.3 IFWA Operator and IFIA Operator. . . . . . . . . . . . . . . . . . . . . . . . . 120
5.3.1 Integral Aggregating Value Based
on Archimedean T-Norm and T-Conorm . . . . . ......... 122
5.3.2 IFIA Operator Based on Archimedean T-Norm
and T-Conorm . . . . . . . . . . . . . . . . . . . . . . . . . . ......... 128
5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......... 131
6 Complement Theory of Intuitionistic Fuzzy Calculus . . . . . . . . . . . . . 133
6.1 Complements of Fundamental Knowledge Related to IFNs . . . . . . 133
6.2 Complements of Derivatives and Differentials of IFFs . . . . . . . . . . 136
6.2.1 Complements of Derivatives of IFFs . . . . . . . . . . . . . . . . . . 137
6.2.2 Complements of Differentials of IFFs . . . . . . . . . . . . . . . . . 141
6.3 Complements of Integrals of IFFs . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.3.1 Complements of Indefinite Integrals of IFFs . . . . . . . . . . . . 143
6.3.2 Complements of Definite Integrals of IFFs . . . . . . . . . . . . . 147
6.4 Complements of Aggregation Operators . . . . . . . . . . . . . . . . . . . . . 151
6.4.1 Complements of IFWA Operator and IFIA Operator . . . . . 151
6.4.2 Integral Forms of Complements of IFWA Operator
and IFIA Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 156
6.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 164
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Abstract
Intuitionistic fuzzy calculus is investigated by utilizing intuitionistic fuzzy numbers

(IFNs) rather than real numbers in classical calculus, where the IFNs are the basic
elements of Atanassov’s intuitionistic fuzzy sets, which are very convenient and
comprehensive to depict the fuzzy characters of things in some actual applications.
In this book, we give a thorough and systematic introduction to the latest research
results on intuitionistic fuzzy calculus. Specifically, this book firstly introduces the
operational laws of IFNs and their geometrical and algebraic properties, which
provide a preparation for studying the calculus of IFNs. Next, the book defines the
concept of intuitionistic fuzzy functions (IFFs), which are the objects to be studied
in the intuitionistic fuzzy calculus, and then shows the research work on the
derivative, differential, indefinite integral, definite integral of IFFs, etc.
Significantly, this work gives methods to deal with continuous intuitionistic fuzzy
information or data successfully, which are different from the previous aggregating
operators focusing on discrete information or data. This book is suitable for the
engineers, technicians, and researchers in the fields of fuzzy mathematics, opera-
tions research, information science, management science and engineering, etc. It
can also be used as a textbook for postgraduate and senior-year undergraduate
students.
Chapter 1
Basic Concepts Related to Intuitionistic
Fuzzy Numbers
1.1 Introduction to Intuitionistic Fuzzy Numbers
The concept of fuzzy set, which was proposed by Zadeh (1965), has been paid more
and more attention. Zadeh tried to remind people that objective things are not
always black or white. For example, dogs, horses and birds are obviously animals,
and plants and rocks must not belong to the category of animals, however, for some
special objects (starfishes and bacteria), it is difficult to explain whether they are
animals or not. Zadeh also explained some concepts, namely: beautiful females and
tall males, which do not consist of a traditional set to describe them in math. Due to
various reasons, fuzziness or ambiguity is inevitable in practice. In such a situation,
Zadeh depicted the fuzziness by introducing a membership function as follows:
Definition 1.1 (Zadeh 1965) For any fixed non-empty set X, a fuzzy set A in X is
characterized by a membership function fA ðxÞ ðx 2 XÞ, which associates each ele-
ment x in X with a real number fA ðxÞ in the interval ½0; 1, with the value of fA ðxÞ
representing the “grade of membership” of x in A. And the nearer the value of fA ðxÞ
to unity, the higher “grade of membership” of x in A. When A is just a set in the
ordinary sense of the term, the membership function fA ðxÞ will only take on two
values (0 or 1), with fA ðxÞ ¼ 1 or 0 according as the element x does or does not
belong to A.
Later on, the concept of “fuzzy” has been rapidly combined with different dis-
ciplines to solve a multitude of application problems, which has sufficiently shown
the validity and the significance of the fuzzy theory. However, the membership
function fA ðxÞ of a fuzzy set does not fully reflect the ambiguity of things, because it
cannot express support, objection and hesitation information in a voting event.
After realizing the shortcomings of the fuzzy set, Atanassov (1986) extended the
fuzzy set to intuitionistic fuzzy set (IFS) through adding a non-membership
function.
2 1 Basic Concepts Related to Intuitionistic Fuzzy Numbers
Definition 1.2 (Atanassov 1986) Let X be a given non-empty set, then an IFS A has
the form: A ¼ fhx; lA ðxÞ; mA ðxÞijx 2 X g, each element of which is depicted by a
membership function lA : X ! ½0; 1 and a non-membership function mA : X !
½0; 1 with the conditions 0 lA ðxÞ þ mA ðxÞ 1 for all x 2 X. Moreover, lA ðxÞ and
mA ðxÞ respectively represent the membership degree and non-membership degree of
x in A. When 1 lA ðxÞ mA ðxÞ ¼ 0 for any x 2 X, an IFS reduces to the fuzzy set,
which shows that the concept of IFS is essentially a generalization of fuzzy set.
Because IFS can actually depict the vagueness and uncertainty of things more
exquisitely and more comprehensively, its theory has been rapidly developed and
vastly applied in various fields.
Before building the calculus theory in intuitionistic fuzzy environment, we first
introduce its basic numbers, called intuitionistic fuzzy numbers (IFNs), which are
just like the real numbers and complex numbers in the classical mathematical
analysis. Then we reveal how to understand the special “number” in several dif-
ferent ways.
Xu and Yager (2006, 2007) defined the basic elements of an IFS as intuitionistic
fuzzy numbers (IFNs) or intuitionistic fuzzy values (IFVs), which can be expressed
by an ordered pair of nonnegative real numbers ðl; mÞ for which l þ m 1. The real
number l, v and 1 l m are called the membership degree, the non-membership
degree and the indeterminacy degree of ðl; mÞ, respectively. Xu and Cai (2012)
provided a physical interpretation for each IFN. For example, (0.5, 0.1) can be
interpreted as “the vote for resolution is 5 in favor, 1 against, and 4 abstentions”.
Obviously, an IFN can be considered as a point in two-dimensional plane, which
indicates that any IFN a ¼ ðl; mÞ can be represented as one point in the lm plane.
In addition, we can get a conclusion that all IFNs would fall into the triangular area
in Fig. 1.1 for the conditions 0 l; m 1 and 0 l þ m 1 that all IFNs must meet.
Moreover, we denote the set, which consists of all IFNs, as N.
From another perspective, any IFN ðl; mÞ not only can be represented as a point
in the lm plane, but also can be regarded as an interval ½l; 1 m or ½m; 1 l,
which is the subinterval of ½0; 1 (as shown in Fig. 1.2). It is worth noting that the
lengths of ½l; 1 m and ½m; 1 l are both equal to 1 l m, which is exactly the
indeterminacy degree of ðl; mÞ.
Fig. 1.1 Representation of

a0 in the lm plane
1.1 Introduction to Intuitionistic Fuzzy Numbers 3
Fig. 1.2 Expressing a0 as the subintervals of ½0; 1
The above-mentioned two representations of an IFN play important roles in the

book. We will show some theorems and conclusions about IFNs from two different
angles.
1.2 Basic Operations Between Intuitionistic Fuzzy

Numbers
As we all know, the real numbers and the complex numbers have their own
operational laws, correspondingly, the IFNs also have some special operations,
which will be introduced in this section.
Based on the addition and multiplication of A-IFSs, Xu and Yager (2006, 2007)
defined the addition and multiplication operations between any two IFNs as
follows:

Definition 1.3 (Xu and Yager 2006, 2007). Let a ¼ðla ; ma Þ and b ¼ lb ; mb be
two IFNs. Then, the addition and multiplication operations between them are
defined as the following forms:

(Addition) a b ¼ la þ lb la lb ; ma mb ;

(Multiplication) a b ¼ la lb ; ma þ mb ma mb .
According to the addition
and multiplication
operations of IFNs, we can easily

2 2
get that a a ¼ 1 ð1 la Þ ; ma , a a a ¼ 1 ð1 la Þ3 ; m3a ,

a a ¼ l2a ; 1 ð1 ma Þ2 , a a a ¼ l3a ; 1 ð1 ma Þ3 and so on. Hence,
it is natural to give the following definition of scalar-multiplication and power
operation of IFNs:
Definition 1.4 (Xu and Yager 2006, 2007) Let a ¼ðla ; ma Þ be an IFN, and the
parameter k be a real number meeting k [ 0. Then we have

(Scalar-multiplication) ka ¼ 1 ð1 la Þk ; mka ;

(Power operation) ak ¼ lka ; 1 ð1 ma Þk .
In order
to understand
these operations better, we firstly transform a ¼ðla ; ma Þ
and b ¼ lb ; mb into ½ma ; 1 la and ½mb ; 1 lb , respectively. Then there are the
following processes:
Hence, the addition of IFNs actually multiplies the upper bound 1 la and the
lower bound ma of ½ma ; 1 la by the upper bound 1 lb and the lower bound mb
of ½mb ; 1 lb to an interval ½ma mb ; ð1 la Þð1 lb Þ. In addition, we can also get
the scalar-multiplication ka of IFNs by dealing with the upper and lower bounds of
½ma ; 1 la and ½mb ; 1 lb , respectively. On the other hand, if we transform a and
b into ½la ; 1 ma and ½lb ; 1 mb , we can analyze the multiplication and power
operations of IFNs in the same way. The processes can be shown in Fig. 1.3 (Lei
and Xu 2015b).
Based on the addition and multiplication between IFNs, we can define their
inverse operations (subtraction and division) as follows:

Definition 1.5 (Lei and Xu 2015b) Let a ¼ðla ; ma Þ and b ¼ lb ; mb be two IFNs.
Then we get (
lb la mb m 1l
; if 0 mba 1lb 1;
(Subtraction) b a ¼ 1la ma a
O; otherwise:
where O is the IFN ð0; 1Þ.
Fig. 1.3 Addition and multiplication between a and b

1.2 Basic Operations Between Intuitionistic Fuzzy Numbers 5
(
lb mb ma l 1mb
; ; if 0 lb 1;
(Division) b a ¼ la 1ma a 1ma
E; otherwise:
where E is actually ð1; 0Þ.
Obviously, the subtraction and the division defined in Definition 1.5 are the
inverse operations of addition and multiplication of IFNs, respectively. It means
that there are ða bÞ a ¼ b, ða bÞ b ¼ a, ða bÞ a ¼ b and
ða bÞ b ¼ a. In addition, we can calculate the difference between b and a by
using the following formula:

lb la mb
ba ¼ ;
1 la m a
m 1l
if only b and a satisfy that 0 mba 1lb 1. However, we notice that the result of
a
lb la mb
1l ; ma may not be an IFN, which means that at least one of three inequalities
a
lb la mb lb la mb
0 1la 1, 0 ma 1 and 0 1la þ ma 1 does not hold, if b and a do not
mb 1lb
meet 0 ma 1. Meanwhile, in order to let the subtraction operation of
1la
IFNs have the closure, Definition 1.5 defines the difference b a ¼ O when the
m 1l
condition 0 mba 1lb 1 does not hold, in order that the subtraction operation of
a
IFNs has the closure. However, in this case, the result O of b a ¼ O is almost
meaningless because the difference result completely loses the information of
minuend and subtrahend (b and a).
1.2.1 Geometrical Analysis of the Operations of IFNs
For any two given real numbers y and z, there must exist a real number x meeting
y ¼ x z, where the operation “ ” is one of the addition, subtraction, multiplication
and division operations between real numbers. Motivated by this, in this subsection,
we will investigate if the similar conclusion can be conducted for the complex
numbers, which is whether there exists an IFN b satisfying a ¼ a0 b, where “ ” is
one of the operations of IFNs about “⊕”, “⊝”, “⊗” and “⊘”, the conclusion with
real numbers is also applicable with complex numbers. In the following, we will
provide some analysis results for more detail. At first, some results are given as
follows (Lei and Xu 2015b, 2016a):

(1) In Fig. 1.4 (Lei and Xu 2016a), for any IFN b ¼ lb ; mb in the area S ðaÞ, it

m 1l lb la mb
must satisfy the condition 0 mba 1lb 1. Hence, bjja ¼ 1l ; ma must
a a
lb la mb
be an IFN. If we let x ¼ 1l ; ma , then there exists x meeting b ¼ a x.
a
In addition, for a set fa xjx 2 Ng, then b 2 fa xjx 2 Ng. Hence,

S ðaÞ
fa xjx 2 Ng. On the other hand, if only the IFN b belongs to the set
Fig. 1.4 Addition region

S ðaÞ of a
m 1l
fa xjx 2 Ng, then it must satisfy 0 mba 1lba 1 and fall into the area
S ðaÞ. So we also have the conclusion that fa xjx 2 Ng
S ðaÞ, and thus,
S ðaÞ ¼ fa xjx 2 Ng. We call S ðaÞ the addition region of a, which con-
tains the following two meanings:
(a) Any a x ðx 2 NÞ must fall into the area
S ðaÞ;
lb la mb
(b) For any b 2 S ðaÞ, b a ¼ 1l ; ma is still an IFN.
a
According to the definition of S ðaÞ, we have the corresponding notion of the

subtraction region S ðaÞ, which can be expressed as follows:
(2) If we let the set S ðaÞ be fa xjx 2 Ng, then there must exist an IFN x0 such
that b x0 ¼ a for any given b 2 fajjxjx 2 Ng. Hence, we have a 2 S ðbÞ
based on the definition of addition regions. Therefore, the equation S ðaÞ ¼
fajjxjx 2 Ng ¼ fbja 2 S ðbÞg holds, which successfully associates the
notion of subtraction regions with addition regions defined by (1) aforemen-
tioned. According to S ðaÞ ¼ fbja 2 S ðbÞg, we can get that the area of
S ðaÞ is just the shadow region of Fig. 1.5 (Lei and Xu 2016a), because there
Fig. 1.5 Subtraction region

S ðaÞ of a
Fig. 1.6 Addition region of

a0 in a subtraction region
Fig. 1.7 Multiplication

region S ðaÞ of a
must be a 2 S ða0 Þ for any a0 in the shadow region of Fig. 1.6 (Lei and Xu
2016a). Meanwhile, we can also have a conclusion that S ðaÞ
S ða0 Þ if only
a 2 S ða0 Þ.
Next, we will study the multiplication region S ðaÞ and the division region
S ðaÞ(Lei and Xu 2015b, 2016a) of IFNs in the same way.
(3) According to the multiplication and division operations of IFNs, the multipli-
cation region S ðaÞ is just the shadow region of Fig. 1.7.
(4) Similar to the method that defines the subtraction region based on the addition
region in (2), we can define the division region according to the multiplication
region in (3). Due to that any IFN a0 in the shadow region of Fig. 1.8 must
meet a 2 S ða0 Þ (as shown in Fig. 1.9), we can define the division region
S ðaÞ of a as the shadow region of Fig. 1.8. In addition, S ðaÞ
S ða0 Þ if only
a 2 S ða0 Þ.
In the above (1)–(4), we have analyzed some properties of the basic operations
between IFNs, including.“⊕”, “⊝”, “⊗” and “⊘”, As we know, the
Fig. 1.8 Division region

S ðaÞ of a
Fig. 1.9 Multiplication

region of a0 in a division
region
scalar-multiplication and the power operations of IFNs are essentially the addition
and the multiplication of IFNs, respectively, and some detailed analyses (Lei and
Xu 2016a) can be processed as follows:
Firstly, we introduce two symbols S ka and S a k , which represent the set
k
fbjb ¼ ka; k 2 ð0; 1Þg and bjb ¼ a ; k 2 ð0; 1Þ , respectively. For any given
IFN a0 ¼ ðl0 ; m0 Þ, we can get the following conclusions after analyzing the
mathematical expression of ka0 :
(1) ka0 can be considered as a function of the variable k, and the value of ka0 will
depend on the parameter k that varies from zero to the positive infinity.
(2) When ka0 ¼ ðl; mÞ, we can calculate k if only l0 6¼ 0, l0 6¼ 1, m0 6¼ 0 and
m0 6¼ 1.
(3) The image of ka0 can be represented as a function mðlÞ in the lm plane, whose
mathematical expression is
lnð1lÞ
lnð1l0 Þ
mðlÞ ¼ m0
Fig. 1.10 The image of the

scalar-multiplication S ka 0 of
a0
(4) ka0 can also be understated as a function lðmÞ in the lm plane, where
ln m
lðmÞ ¼ 1 ð1 l0 Þln m0
Next, we provide some analyses about the function mðlÞ, and lðmÞ can also be
analyzed in a similar way.
(1) mðlÞ satisfies mðl0 Þ ¼ m0 , which indicates 1ðl0 ; m0 Þ ¼ ðl0 ; m0 Þ when the
parameter k ¼ 1.
(2) mð1Þ ¼ 0 reveals that ka0 ! ð1; 0Þ when k ! þ 1.
(3) mð0Þ ¼ 1 represents that ka0 ! ð0; 1Þ if k ! 0.
(4) Because ka0 ¼ a0 ðk 1Þa0 (k [ 1), there must be ka0 2 S ða0 Þ.
(5) When 0\k\1, there exists ka0 2 S ða0 Þ due to ka0 ¼ a0 ð1 kÞa0 .
Based on the above (1)–(5), the images of the scalar-multiplication and the
power operation of IFNs can be shown in Fig. 1.10 (Lei and Xu 2016a) and
Fig. 1.11 (Lei and Xu 2016a), respectively.
Fig. 1.11 The image of the

power operation S a k of a0
0
Fig. 1.12 The change region

and the non-change region of
a0
Until now, we can answer the question mentioned at the beginning of this
section, i.e., whether there exists an IFN b satisfying a ¼ a0 b, where “ ” is one of
four basic operations (“⊕”, “⊝”, “⊗” and “⊘”,), for any two given IFNs a0 and a.
By the images of the operations of IFNs, it is easy to get that the answer is negative.
In order to improve this situation, we introduce a novel conception about the change
region of a0 (Lei and Xu 2015b), which is defined as the set fa0 xjx 2 Ng. It can
be actually represented by the area S 1 [ S 2 [ S 3 [ S 5 [ S 6 [ S 7 in Fig. 1.12 (Lei
and Xu 2015b). The region S 4 [ S 8 (Lei and Xu 2015b) is called the non-change
region of a0 . Then, we can get the following conclusions (Lei and Xu 2015b):
(1) For any IFN a, there must be one IFN b for which a ¼ a0 * b, if only a is in the
change region of a0 .
(2) If a is in the non-change region of a0 , then there must not be an IFN b, such that
a ¼ a0 * b.
1.2.2 Algebraic Analysis of the Operations of IFNs
In this subsection, we will give some algebraic properties of IFNs.

Theorem 1.1 (Lei and Xu 2015c) Let a be any IFN ðla ; ma Þ, O be ð0; 1Þ, and E be
ð1; 0Þ. Then we have
ð1Þ a O ¼ a; a O ¼ a; a a ¼ O; a E ¼ E
ð2Þ a E ¼ a; a E ¼ a; a a ¼ E; a O ¼ O
(3) When la 6¼ 1 and ma 6¼ 0, the expression 0a is meaningful and equal to O.
(4) If la 6¼ 0 and ma 6¼ 1, then a0 is meaningful and there is a0 ¼ E.
Proof The conclusions (1) and (2) can be easily proven according to the operational
laws of addition and multiplication between IFNs. Hence, their proofs are omitted
here. Next, we analyze (3) and (4). Because 00 and 1 ð1 1Þ0 are both mean-
ingless, we give a restriction on the parameter k of the scalar-multiplication ka and
the power operation ak of IFNs, that is k [ 0. However, in most cases, ka and ak
allow that k ¼ 0 if only k satisfies these conditions in (3) and (4). It is worth
pointing out that if there is no special instruction, a of 0a and a0 are respectively
assumed to satisfy the conditions in (3) and (4) in this book. ■
From Theorem 1.1, we can get a fact that O and E are respectively similar to
zero and unity in real number field to some extent.
Theorem 1.2 (Xu and Cai 2012; Lei and Xu 2015c) Let a, b and c be three IFNs,
k 1 and k 2 be two real number meeting k 1 0, k 2 0 and k 1 k 2 . Then
ð1Þ a b ¼ b a; a b ¼ b a
ð2Þ ða bÞ c ¼ a ðb cÞ; ða bÞ c ¼ a ðb cÞ
ð3Þ k 1 ða bÞ ¼ k 1 a k 1 b ; ða bÞk 1 ¼ ak 1 bk 1
ð4Þ k 1 ðb aÞ ¼ k 1 b k 1 a ; ðbaÞk 1 ¼ bk 1 ak 1
ð5Þ ðk 1 þ k 2 Þa ¼ k 1 a k 2 a ; ak 1 þ k 2 ¼ ak 1 ak 2
ð6Þ ðk 1 k 2 Þa ¼ k 1 a k 2 a ; ak 1 k 2 ¼ ak 1 ak 2
Proof According to the addition and the multiplication between IFNs, it is easy to
get (1), (2), (3) and (4), which shows actually the commutative law and the asso-
ciative law of IFNs. Next, we will prove (3) as follows:

k 1 a k 1 b ¼ 1 ð1 la Þk 1 ; lka 1 1 ð1 lb Þk 1 ; lkb 1

¼ 1 ð1 la Þk 1 ð1 lb Þk 1 ; lka 1 lkb 1

¼ 1 ðð1 la Þð1 lb ÞÞk 1 ; ðla lb Þk 1
¼ k 1 ða bÞ
Similarly, we can get ða bÞk 1 ¼ ak 1 bk 1 . Considering the conclusion (4), we

will prove it in two different cases (Case 1 and Case 2):
k 1
m 1l m 1l k 1
Case 1. Because 0 mba 1lb 1, there is 0 mba 1lb 1. Hence, we
a
a
k1

k1 m
ð1l Þ
also have a fact that the result of k 1 bk 1 a ¼ 1 ð1lb Þk 1 ; bk 1 is still an IFN if
a ma

lb la mb 1lb mb
only b a ¼ 1l ; ma ¼ 1 1l ; ma is an IFN. Then, we will prove
a a
lb la mb m 1l
(4) when b a ¼ 1l ; ma is still an IFN, which means 0 mba 1lb 1 holds,
a a
and we have the following process:

k 1 b k 1 a ¼ 1 ð1 lb Þk 1 ; lkb 1 1 ð1 la Þk 1 ; lka 1
k
! k 1 !
ð1 lb Þk 1 lb 1 1 lb k 1 mb
¼ 1 k1
; k1 ¼ 1 ;
ð1 la Þ la 1 l a ma
On the other hand, there is also

k 1 k 1 !
1 lb m b 1 lb mb
k 1 ðb aÞ ¼ k 1 1 ; ¼ 1 ;
1 l a ma 1 la ma
Hence, in this case, k 1 ðb aÞ ¼ k 1 b k 1 a holds.

m 1l
Case 2. When 0 mba 1lb 1 does not hold, which means that any result of

a k

lb la mb ð1l Þk 1 m 1
b a ¼ 1l ; ma and k 1 b k 1 a ¼ 1 ð1lb Þk 1 ; bk 1 is not an IFN, we get
a a ma
b a ¼ O and k 1 b k 1 a ¼ O according to the definition of subtraction in
Definition 1.5. Thus, k 1 b k 1 a ¼ O ¼k 1 O ¼ k 1 ðb aÞ.
According to Case 1 and Case 2, we have k 1 ðb aÞ ¼ k 1 b k 1 a holds. The
equation ðb aÞk 1 ¼ bk 1 ak 1 can be proven in the same manner.
In addition, based on the laws of basic operations of IFNs, (5) and (6) can be
proven easily, which is omitted here. ■
Theorem 1.3 (Lei and Xu 2015c) If a1 ¼ðl1 ; m1 Þ, a2 ¼ðl2 ; m2 Þ and a3 ¼ðl3 ; m3 Þ,
which satisfy the condition S ða1 Þ
S ða2 Þ
S ða3 Þ, then
(1) ða1 a2 Þða2 a3 Þ ¼ a1 a3 .
(2) ða1 a3 Þða2 a3 Þ ¼ a1 a2 .
(3) ða1 a2 Þ ða2 a3 Þ ¼ a1 a3 .
Proof Based on the operational laws IFNs, the equation of (1) can be calculated as:
ða1 a2 Þða2 a3 Þ ¼ ð1 ð1 l1 Þð1 l2 Þ; m1 m2 Þð1 ð1 l2 Þð1 l3 Þ; m2 m3 Þ

ð1 l2 Þð1 l3 Þ ð1 l1 Þð1 l2 Þ m1 m2
¼ ;
ð1 l2 Þð1 l3 Þ m2 m3

1 l1 m1 l1 l3 m1
¼ 1 ; ¼ ;
1 l3 m3 1 l3 m3
¼ a1 a3
In the same way, the proof of (2) can be processed as follows:

l1 l3 m1 l2 l3 m2
ða1 a3 Þða2 a3 Þ ¼ ; ;
1 l3 m 3 1 l3 m 3
l1 l3 l2 l3 !
1l3 1l3 m1 m3 l1 l2 m1
¼ ; ¼ ;
1 l1l
2 l3
m3 m2 1 l2 m 2
3
¼ a1 a2
Moreover, the equation of (3) can be proved as follows:

l1 l2 m1 l2 l3 m2
ða1 a2 Þ ða2 a3 Þ ¼ ; ;
1 l2 m2 1 l m3
3
l1 l2 l l3 m1 m2 l1 l3 m1
¼ 1 1 1 2 ; ¼ ;
1 l2 1 l3 m2 m3 1 l3 m3
¼ a1 a3
The proof of Theorem 1.3 is completed. ■

Theorem 1.4 (Lei and Xu 2015c) If a1 ¼ðl1 ; m1 Þ, a2 ¼ðl2 ; m2 Þ and a3 ¼ðl3 ; m3 Þ,
which satisfy the condition S ða1 Þ
S ða2 Þ
S ða3 Þ, then we have
(1) ða1 a2 Þ ða2 a3 Þ ¼ a1 a3 .
(2) ða1 a3 Þ ða2 a3 Þ ¼ a1 a2 .
(3) ða1 a2 Þ ða2 a3 Þ ¼ a1 a3 .
Proof Firstly, we prove (1) as follows:

l1 m1 m2 l2 m2 m3
ða1 a2 Þ ða2 a3 Þ ¼ ; ;
l 1 m2 l3 1 m3
2
l1 m1 m2 m2 m3 l1 m1 m3
¼ ;1 1 1 ¼ ;
l3 1 m2 1 m3 l3 1 m3
¼ a1 a3
Next, we can prove the proof of (2):

l 1 m1 m3 l m2 m3
ða1 a3 Þ ða2 a3 Þ ¼ ; 2;
l 3 1 m3 l3 1 m3
m1 m3 m2 m3
!
l1 l3 1m3 1m3 l 1 m1 m2
¼ ; ¼ ;
l3 l2 1 m1m 2 m3
3
l 2 1 m2
¼ a1 a2
Finally, we prove the conclusion (3):
ða1 a2 Þ ða2 a3 Þ ¼ ðl1 l2 ; 1 ð1 m1 Þð1 m2 ÞÞ ðl2 l3 ; 1 ð1 m2 Þð1 m3 ÞÞ

l1 l2 ð1 m2 Þð1 m3 Þ ð1 m1 Þð1 m2 Þ l 1 m1 m3
¼ ; ¼ ;
l2 l3 ð1 m2 Þð1 m3 Þ l 3 1 m3
¼ a1 a3
which completes the proofs. ■
1.3 Order Relations of IFNs
In order to compare and rank IFNs, we introduce the concept of the order relations
of IFNs. Moreover, Sect. 1.2 has pointed out the fact that any IFN can be con-
sidered as one point in two-dimensional plane. However, as we all know, there is
not a satisfying order for two-dimensional points, which explains why we do not
define “x\y” and “x [ y” in the complex numbers field. Hence, we will show
several common order relations of IFNs in this section, which have their own
advantages and disadvantages.
Firstly, we introduce the fundamental knowledge of the order relations as
follows:
An order relation is essentially a special kind of binary relations. Let P be a set
with a binary relation R. The relation R consists of some ordered pairs, these basic
elements of which are both in P. For example, for any two elements p1 and p2
(p1 2 P and p2 2 P), if the order pair ðp1 ; p2 Þ 2 R, then it is denoted by p1 R p2 . In
addition, if the binary relation R satisfies the following three conditions:
(1) (Reflexivity) For any elements p 2 P, there is pRp.
(2) (Antisymmetry) If p1 R p2 and p2 R p1 , then p1 ¼ p2 .
(3) (Transitivity) If p1 R p2 and p2 R p3 , there is p1 R p3 .
then we call the binary relation R as a partial order, and call the set P as a poset.
In addition, if there must be p1 R p2 or p2 R p1 for any two given p1 and p2 (p1 2 P
and p2 2 P), which means that any two elements in P are always comparable, then
we call the partial order R as a total order or a linear order.
1.3 Order Relations of IFNs 15
1.3.1 Three Kinds of Orders of IFNs and Relationships

Among Them
Based on the fundamental knowledge of the order relations, we introduce some

order relations of IFNs:
Let a ¼ðla ; ma Þ and b ¼ lb ; mb be two IFNs. Then, there is a partial order
defined (Deschrijver and Kerre 2001) in the set N that consists of all IFNs:
(1) If la lb and ma mb , then a b.
(2) If la lb and ma mb , then a b.
(3) If la ¼ lb and ma ¼ mb , then a ¼ b.
Obviously, the order relation on N is only a partial order because ð0:4; 0:3Þ and
ð0:5; 0:4Þ is incomparable according to the above order, hence, N is only a poset
here, which can be shown in Fig. 1.13:
From Fig. 1.13, for any IFN b in the shadow area B, there is a b. In addition,
we have a b if only b 2 E.
Next, we introduce a kind of order relations of IFNs based on the score function
and the accuracy function (Chen and Tan 1994; Hong and Choi 2000), which are
respectively defined as sðaÞ ¼ la ma and hðaÞ ¼ la þ ma , for any IFN
a ¼ðla ; ma Þ. According to the two functions s and h, Xu and Yager (2006) proposed
an order of IFNs as follows:
(1) If sðaÞ\sðbÞ, then a\XY b.
(2) If sðaÞ ¼ sðbÞ, then
(a) when hðaÞ ¼ hðbÞ, a ¼XY b.
(b) when hðaÞ\hðbÞ, a\XY b.
Fig. 1.13 Order relation

“”
Fig. 1.14 Order relation “\XY ”
This kind of order relation can be shown in Fig. 1.14.

From Fig. 1.14, for any IFN b in the shadow area A, there is a\XY b. In
addition, we have a [ XY b if only b 2 D. Significantly, we can also know that
“\XY ” is a linear order on the set N.
In what follows, we introduce a novel order relation given by Lei and Xu
(2015c), which plays an important role in this book. It is developed based on the
addition and the subtraction of IFNs, and its specific definition is described as:
Definition 1.6 (Lei and Xu 2015c) If there exists an IFN e, such that a e ¼ b,
then we define that a is less than or equal to b, denoted by aEb. Moreover, if there
is an IFN e meeting a e ¼ b and e 6¼ O, then we define that a is less than or equal
to b, which is denoted by a / b.
According
definition of “E”, we have that aEb , S ðbÞ
S ðaÞ, and
to the
lb la mb
ba ¼ 1la ; ma must be an IFN when aEb. In fact, we can show the order
relation “E” by utilizing Fig. 1.15.
From Fig. 1.15, for any IFN b in the shadow area C, there is aEb. In addition,
we have bEa if only b 2 F.
Next, we will reveal the relationships of these different order relations of IFNs.
Fig. 1.15 Order relation “E”

Fig. 1.16 Comparison of

order relations of IFNs
Firstly, we will put Figs. 1.13, 1.14 and 1.15 together, and give Fig. 1.16 to
manifest the three orders “ ”, “\XY ” and “E”.
According to Fig. 1.16, we can conduct the following conclusions:
(1) C B A, which means that C is a subset of B, and B is a subset of A. It can
be used to prove aEb ) a b ) a\XY b and aEb (a = b (a\= XY b.
(2) There are bEa ) b a ) b\XY a and bEa (b = a (b\
= XY a because of
F
E
D.
1.3.2 Properties of the Order Based on the Operations

of IFNs
Next, we discuss (Lei and Xu 2015c) whether “E” is a partial order on the set N as
follows:
(1) (Reflexivity) Because a O ¼ a, there is aEa.
According to Figs. 1.4 and 1.5, we can get a
S ðaÞ and a
S ðaÞ, which also
means that aEa holds.
(2) (Antisymmetry) If aEb and bEa, which means that there exist two IFNs c1
and c2 that satisfy a c1 ¼ b and b c2 ¼ a, then
a c1 c2 ¼ a ) c 1 c2 ¼ O ) c1 ¼ c2 ¼ O ) a ¼ b
which means that a ¼ b if aEb and bEa.

Fig. 1.17 b
S ðaÞ and b
S ðaÞ
On the other hand, we can get there must be b

S ðaÞ and b
S ðaÞ if aEb and
bEa. Moreover, there is S ðaÞ \ S ðaÞ ¼ fa g. Hence, a must be equal to b,
otherwise it must be contradicted with Fig. 1.17:
(3) (Transitivity) If aEb and bEg, then there are two IFNs c1 and c2 that satisfy
a c1 ¼ b and b c2 ¼ g. Hence, we have
a c1 c2 ¼ b c2 ¼ g
which means that aEg if aEb and bEg.

Similarly, we also can prove this property by utilizing a Fig. 1.18. Due to aEb
and bEg, there is g 2 S ðbÞ and S ðbÞ
S ðaÞ. Hence, we can get g 2 S ðaÞ,
which means that aEg.
Fig. 1.18 g 2 S ðbÞ and

S ðbÞ
S ðaÞ
Fig. 1.19 “E” is a partial

order
The above (1)–(3) show that “E” is a partial order. In addition, “E” is not a total
order on the set N because ð0; 0Þ5ð0:5; 0:5Þ and ð0:5; 0:5Þ5ð0; 0Þ. Obviously, by
Fig. 1.19, b ¼ð0:5; 0:5Þ is the non-change region of a ¼ ð0; 0Þ.
Then we prove several proprieties of “E” below:
Theorem 1.5 (Lei and Xu 2015c) If a1 Eb1 and a2 Eb2 , then a1 a2 Eb1 b2 .
Proof Since a1 Eb1 and a2 Eb2 , then there must exist two IFNs c1 and c2 satisfying
a1 c1 ¼ b1 and a2 c2 ¼ b2 . Then we have
ða1 a2 Þ ðc1 c2 Þ ¼ ða1 c1 Þ ða2 c2 Þ ¼ b1 b2
which means a1 a2 Eb1 b2 . ■

Theorem 1.6 (Lei and Xu 2015c) If aEb, then there must exist an IFN c, such that
a cEb.
Proof Let c be an IFN, which meets cEba. Based on Theorem 1.5, we have
a cEba a ¼ b
The proof of this theorem is completed. ■

Theorem 1.7 (Lei and Xu 2015c) If cEaEb, then acEbc.
Proof According to aEb, we have OEba. Meanwhile, according to Theorem 1.5,
we can obtain that
acEðacÞ ðbaÞ ¼ bc
The proof is completed. ■

Theorem 1.8 (Lei and Xu 2015c) If aEb, then kaEkb ( k 0).
Proof By Theorem 1.2, there is kðbaÞ ¼ kbka. Hence, ka kðbaÞ ¼ kb,

which means kaEkb. ■
Theorem 1.9 (Lei and Xu 2015c) If 0 k 1 k 2 , then k 1 aEk 2 a.
Proof Theorem 1.2 has proved that ðk 2 k 1 Þa ¼ k 2 ak 1 a. Hence, the equality
k 1 a ðk 2 k 1 Þa ¼ k 2 a holds, which means k 1 aEk 2 a. ■
Theorem 1.10 (Lei and Xu 2015c) If 0 k 1 k 2 and aEb, then k 1 aEk 2 b.
Proof According to Theorem 1.8 and Theorem 1.9, there are k 1 aEk 2 a and
k 2 aEk 2 b. Thus, we can get k 1 aEk 2 b. ■
1.4 Conclusions
In this chapter, we have first introduced the concepts of the fuzzy set and the IFS.
Then, we have shown the definition of the IFNs is actually an ordered pair of
nonnegative real numbers ðl; mÞ for which l þ m 1. In addition, various methods
have been provided to represent the IFNs, including considering the IFNs as some
points in two-dimensional plane and the subintervals of ½0; 1. Moreover, we have
shown some operations of IFNs, namely: addition, subtraction, multiplication,
division, scalar-multiplication and power operation, and analyzed these operations
of IFNs in detail. Based on which, we have shown some geometrical and algebraic
properties of these operations, and defined the concepts of the change region and
the non-change region of IFNs. Finally, three kind of order relations of IFNs and
the relationships among them have been presented. In brief, the main work of this
chapter is to provide a preparation work for studying the intuitionistic fuzzy
calculus.
Chapter 2
Derivatives and Differentials
of Intuitionistic Fuzzy Functions
Calculus, which is an important branch of classical mathematics, is the mathe-

matical study of change. Like the calculus of real numbers and the complex
numbers, the calculus of IFNs is very significant to the theory environment. Thus,
this chapter aims to do work in the calculus in intuitionistic fuzzy environment. To
begin with, we introduce the concept of intuitionistic fuzzy functions (IFFs) (Lei
and Xu 2015b), which are the main research subjects of intuitionistic fuzzy cal-
culus. Then, we study the derivatives and differentials of IFFs in detail.
Firstly, we introduce some topological knowledge of IFNs, which makes
preparations for defining the limit and the continuity of functions of IFNs.
When considering the complex plane C, let a 2 C, r 2 ð0; þ 1Þ, then we call
the set
fz : jz aj\r; z 2 Cg
a neighborhood or r-neighborhood of a, denoted by Uða; rÞ. Moreover, the set
fz : j z aj r g
is called a closed neighborhood of a, denoted by Uða; rÞ.

Hence, we can analogize the definition of neighborhood in the C to get a similar
notion in N, which is described as follows:
Definition 2.1 Let a 2 N and O E e E a. Then we call the set
fX : jX aj / e; X 2 S ðaÞ [ S ðaÞg
a neighborhood or e-neighborhood of a, denoted by Uða; eÞ, where

22 2 Derivatives and Differentials of Intuitionistic Fuzzy Functions
Fig. 2.1 Closed

neighborhood of a
(
X a; if X 2 S ðaÞ
jXaj ¼
a X; if X 2 S ðaÞ
In addition, fX : jXajEe; X 2 S ðaÞ [ S ðaÞg is called a closed neighborhood

eÞ, which is shown in Fig. 2.1.
of a, denoted by Uða;
For any given set E N, if for any e (OEeEa), there always exists an infinity of
IFNs in the set Uða; eÞ \ E, then we call a an accumulation point of E.
2.1 Intuitionistic Fuzzy Functions
In this section, we study the concept of functions related to IFNs (Lei and Xu,
2015b, c, 2016a).
Let E (E N) be a non-empty set, which consists of some IFNs. Then we call
u : E ! N an intuitionistic fuzzy function (IFF) defined in E, which is denoted by
Y ¼ uðXÞ; X 2 E
where is the domain E of u, X is the independent variable, and Y is the dependent

variable. According to the definition of IFF, we assume that an IFF u consists of
two real functions f and g, which are
uðXÞ ¼ ðf ðl; mÞ; gðl; mÞÞ; X ¼ðl; mÞ 2 E
where f and g satisfy the conditions: 0 f ðl; mÞ 1, 0 gðl; mÞ 1 and

0 f ðl; mÞ þ gðl; mÞ 1 for any ðl; mÞ 2 E.
In addition, we call that the IFF u ¼ðf ; gÞ is meaningful at some points
X ¼ðl; mÞ if only f ðl; mÞ and gðl; mÞ of uðXÞ ¼ ðf ðl; mÞ; gðl; mÞÞ meet the above
three inequalities, otherwise u is meaningless at X. If u is meaningful at all points
2.1 Intuitionistic Fuzzy Functions 23
of one set, then we call that u is meaningful in the set. Obviously, the IFF u is
meaningful in its domain E.
Reviewing the knowledge of derivative of real and complex functions, we know
that its definition is the limit value of the expression f ðyyx
Þf ðxÞ
when y ! x, where x
and y are both real or complex numbers.
However,
for the IFNs, there exists a
luðYÞ luðXÞ muðYÞ
question whether uðYÞuðXÞ ¼ 1luðXÞ ; muðXÞ is still an IFN when YX is an
IFN, which means that if uðYÞ will fall into S ðuðXÞÞ when Y falls into S ðXÞ.
Unfortunately, the answer about the question is negative. In order to solve the
problem, the following definition is provided:
Definition 2.2 (Lei and Xu 2015a) Let u ¼ðf ; gÞ be an IFF in the set E, X and Y be
both IFNs in E. If XEY, uðXÞEuðYÞ holds, then we call u a monotonically
increasing IFF.
Based on the concept of monotonically increasing IFF, we know that
!
luðYÞ luðXÞ muðYÞ
uðYÞuðXÞ ¼ ;
1 luðXÞ muðXÞ
must be an IFN when YX is an IFN, which is just the property of IFFs we want.
Considering uðXÞ ¼ kXðX 2 NÞ, we get that uðXÞ ¼ kX is a monotonically
increasing IFF due to that uðYÞuðXÞ ¼ kYkX ¼kðYXÞ.
In addition, the intuitionistic fuzzy weighted aggregation (IFWA) operator was
proposed by Xu (2007), which has the following form:
!
n Y
n Y
n
IFWAx ða1 ; a2 ; . . .; an Þ ¼ xi ai ¼ 1 ð1 lai Þxi ; mx
ai
i
i¼1
i¼1 i¼1
where ai ði ¼ 1; 2; . . .; nÞ is a collection of IFNs, P and xi ði ¼ 1; 2; . . .; nÞ is the

n
weights of ai ði ¼ 1; 2; . . .; nÞ with xi 0 and i¼1 xi ¼ 1. Apparently, the
function IFWAx ða1 ; a2 ; . . .; an Þ is also a monotonically increasing IFF of any ai .
It is worth noting that IFFs studied in this chapter are both assumed as mono-
tonically increasing IFFs. Below we investigate the limit and continuity of IFFs (Lei
and Xu 2015b):
We define u ¼ðf ; gÞ in the set E, and let X 0 be an accumulation point of E
(Maybe X 0 62 E) and a be an intuitionistic fuzzy constant, then for any given e . O,
there always exists an IFN d . O, such that if only X 2 UðX 0 ; dÞ, we have
uðX Þ 2 U ða; eÞ. We call uðX Þ approaches a when X approaches X 0 , which can be
written by
lim uðXÞ¼ a
X!X 0 ;X2E
According to the definition of neighborhood and the limit of IFFs, we can define
the limit of IFFs from another angle:
Fig. 2.2 Continuity of uðX Þ
For any given e . O, there is an IFN d . O, such that when O / jXX0 j / d, we

have juðXÞaj / e. Moreover, we introduce the continuity of IFFs as follows:
Assume that u is an IFF in the set E, X 0 is an accumulation point of E and
X 0 2 E, if lim uðX Þ¼ uðX0 Þ, then we call uðX Þ to be continuous at X 0 . It
X!X 0 ;X2E
means that for any given e . O, there always exists an IFN d . O, such that if only
X 2 UðX 0 ; dÞ, then uðX Þ 2 U ðuðX 0 Þ; eÞ, which can be shown in Fig. 2.2.
Because u is a monotonically increasing IFF, we obtain that if uðX Þ is con-
tinuous at X 0 , then we have the following conclusions:
(1) When X 2 S ðX 0 Þ, lim uðX ÞuðX0 Þ ¼ O, denoted as lim
X!X 0 ;X2E X!X ;X2E
0
uðX ÞuðX0 Þ ¼ O.
(2) When X 2 S ðX 0 Þ, lim uðX 0 ÞuðX Þ ¼ O, denoted as lim
X!X 0 ;X2E X!X ;X2E
0
uðX ÞuðX0 Þ ¼ O.
The above conclusions are similar to the left-continuity and the right-continuity
of real functions: lim f ð xÞ ¼ f ðx0 Þ and limþ f ð xÞ ¼ f ðx0 Þ. Consequently, we can
x!x0 x!x0
define the continuity of IFFs as follows:
For any given e . O, there is an IFN d . O, such that when O / jXX0 j / d, we
have
lim juðX ÞuðX 0 Þj¼ O

X!X 0 ;X2E
Obviously, the functions uðXÞ ¼ kX and IFWAx ða1 ; a2 ; . . .; an Þ are both

continuous.
2.2 Derivatives of Intuitionistic Fuzzy Functions 25
2.2 Derivatives of Intuitionistic Fuzzy Functions
Based on the discussions about the IFFs in the last section, this section studies the
derivatives of IFFs.
By analogizing the definitions of derivatives of real and complex functions,
which are both defined as the limit of f ðyyx
Þf ðxÞ
(that is lim f ðyyx
Þf ðxÞ
), we can give a
y!x
definition about the derivative of IFF as follows:
Definition 2.3 (Lei and Xu 2015b) Let u ¼ðf ; gÞ be a monotonically increasing
IFF defined in the set E, X be an accumulation point of E (Maybe there is X 62 E).
uðYÞuðXÞ
If lim
YX is still an IFN, then we call it the right derivative of u at X,
Y!X ;Y2E
denoted by u0 ðXÞ. Similarly, lim uðXÞuðYÞ

XY is the left derivative if it is an
Y!X ;Y2E
IFN, which can be noted by u0 ðXÞ. In addition, if the left and the right derivatives
are both IFNs and equal to each other, then we call u is derivable at X and
lim juðYÞuðXÞ
jYX j
j
is the derivative of u at X, denoted by duðXÞ dX . If u exists the
Y!X;Y2E
derivative at every point in a set E, then we call that u is derivable in E.
Moreover, another form of the left and right derivatives can be expressed as:
uðYÞuðXÞ uðX DXÞuðXÞ

lim , lim
Y!X YX DX!O DX
and
uðXÞuðYÞ uðXÞuðXDXÞ
lim , lim
Y!X XY DX!O DX
After defining the derivative of u, we give the following theorem:

Theorem 2.1 (Lei and Xu 2015b) Let uðXÞ¼ðf ðl; mÞ; gðl; mÞÞ be a monotonically
increasing IFF in the set E, If u is derivable in E, if and only if u meets:
@f ðl; mÞ @gðl; mÞ 1 l @f ðl; mÞ m @gðmÞ

¼ ¼0 and 0 1
@m @l 1 f ðl; mÞ @l gðl; mÞ @m
Due to fm0 ¼ g0l ¼ 0, uðXÞ can be written as ðf ðlÞ; gðmÞÞ. Then, the derivative of
u has the following form:

duðXÞ juðYÞuðXÞj 1 l df ðlÞ m dgðmÞ
¼ lim ¼ ; 1
dX Y!X jYX j 1 f ðlÞ dl gðmÞ dm
Proof We first prove the right derivative of u, which is lim uðYÞuðXÞ

YX , where
Y!X
Y 2 S ðXÞ.
Assume that Y ¼ ðl þ Dl; m þ DmÞ ¼ ðl0 ; m0 Þ 2 S ðXÞ, which means YDX,
then there is uðYÞ ¼ ðf ðl þ Dl; m þ DmÞ; gðl þ Dl; m þ DmÞÞ ¼ ðf ðl0 ; m0 Þ;
0 0
gðl ; m ÞÞ. Hence, we can get the following derivation process:
duðXÞ uðYÞuðXÞ
¼ lim
dX Y!X YX
ðf ðl0 ; m0 Þ; gðl0 ; m0 ÞÞðf ðl; mÞ; gðl; mÞÞ
¼ lim
l0 !l ðl0 ; m0 Þðl; mÞ
m0 !m
0 0
f ðl ;m Þf ðl;mÞ gðl0 ;m0 Þ
1f ðl;mÞ ; gðl;mÞ
¼ lim 0
l l m0
1l ; m
l0 !l
m0 !m
gðl0 ;m0 Þ m0
!
f ðl0 ; m0 Þ f ðl; mÞ 1 l gðl;mÞ m
¼ lim ; 0
l0 !l 1 f ðl; mÞ l0 l 1 mm
m0 !m
0 1
gðl0 ;m0 Þ 0
f ðl 0 0
; m Þ f ðl; mÞ 1 l mm
¼ @lim A
gðl;mÞ
; lim 0
l0 !l 1 f ðl; mÞ l0 l l0 !l 1 mm
0 m !m 0m !m
Next, we respectively discuss the membership and the non-membership of the

above equation in detail. For the membership part, it can be simplified as:
f ðl0 ; m0 Þ f ðl; mÞ 1 l
lim
0
l !l 1 f ðl; mÞ l0 l
m0 !m
1 l f ðl0 ; m0 Þ f ðl; mÞ
¼ lim
l !l 1 f ðl; mÞ
0 l0 l
m0 !m
0 0
1l f ðl ; m Þ f ðl; m0 Þ f ðl; m0 Þ f ðl; mÞ m0 m
¼ lim þ
l0 !l 1 f ðl; mÞ l0 l m0 m l0 l
m0 !m

1l @f ðl; mÞ @f ðl; mÞ
¼ þ cos h
1 f ðl; mÞ @l @m
and for the non-membership part, it can be calculated as:

gðl0 ;m0 Þ 0
gðl;mÞ mm m gðl0 ; m0 Þ gðl0 ; mÞ
m 0
gðl;mÞ gðl ; mÞ m0
lim m0
¼ lim þ
l0 !l 1 m
l0 !l gðl; mÞ m m0 m m0
m0 !m m0 !m
m gðl0 ; m0 Þ gðl0 ; mÞ
¼ lim
l0 !l gðl; mÞ m m0
m0 !m
m 0 m m
gðl;mÞ gðl ; mÞ gðl;mÞ gðl; mÞ þ gðl;mÞ gðl; mÞ m0
þ
m m0
m gðl ; m Þ gðl0 ; mÞ
0 0
m gðl0 ; mÞ gðl; mÞ l l0
¼ lim 0
þ þ1
l0 !l gðl; mÞ mm gðl; mÞ l l0 m m0
m0 !m

m @gðl; mÞ @gðl; mÞ 1
¼1 þ
gðl; mÞ @m @l cos h
Therefore, the right derivative of u can be written as:

1l @f ðl; mÞ @f ðl; mÞ m @gðl; mÞ @gðl; mÞ 1
u0 ¼ þ cos h ; 1 þ
1 f ðl; mÞ @l @m gðl; mÞ @m @l cos h
In order to make the above expression of u0 not depend on the change of cos h,
we let
@f ðl; mÞ @gðl; mÞ
¼ ¼0
@m @l
Based on the condition fm0 ¼ g0l ¼ 0, we note uðXÞ ¼ ðf ðlÞ; gðmÞÞ. Moreover,
the expression of u0 is

0 uðYÞuðXÞ 1 l df ðlÞ m dgðmÞ
u ¼ lim ¼ ; 1
Y!X YX 1 f ðlÞ dl gðmÞ dm
Similarly, we can also prove u0 , and it can be represented as:

uðXÞuðYÞ 1 l df ðlÞ m dgðmÞ
u0 ¼ lim ¼ ; 1
Y!X XY 1 f ðlÞ dl gðmÞ dm
Hence, if only fl0 and g0m exist, the derivative of u exists and

0 0 0 1 l df ðlÞ m dgðmÞ
u ¼u ¼u ¼ ; 1
1 f ðlÞ dl gðmÞ dm

1l df ðlÞ m dgðmÞ
Moreover, in order to make 1f ðlÞ dl ; 1 gðmÞ dm to be still an IFN, the
inequalities 0 1f ðlÞ dl gðmÞ dm 1 need hold.
Theorem 2.1 gives the criterion to judge whether an IFF is derivable, which is
similar to the “C-R condition” in the complex number field.
In what follows, we introduce several examples (Lei and Xu 2015b):
(1) For uðXÞ ¼ a0 , where a0 is a constant. Then, its derivative duðXÞ

dX ¼ O, which is
df ðxÞ
similar to dx ¼ 0 for the real constant function f ðxÞ ¼ c.
(2) Let uðXÞ ¼ kX, then we have duðXÞ
dX ¼ ðk; 1 kÞ. Specially, for the identity IFF
df ðxÞ
uðXÞ ¼ X, there is duðXÞ dX ¼ E. It is like the situation that dx ¼ 1 when
f ðxÞ ¼ x.
(3) When considering uðai Þ ¼ IFWAx ða1 ; a2 ; . . .; an Þði ¼ 1; 2; . . .; nÞ, there is
dIFWAx ða1 ;a2 ;...;an Þ
dai ¼ ðxi ; 1 xi Þ, which expresses the information of the weights
of the independent variable ai .
For the real functions f ðxÞ ¼ kx þ a and gðxÞ ¼ kx þ b, there is f 0 ðxÞ ¼ g0 ðxÞ.
For IFFs, the similar conclusion (Lei and Xu 2015b) can be obtained by the
following two illustrations:
(4) For uðXÞ ¼ kX a and wðXÞ ¼ kX b, there is duðXÞ dwðXÞ
dX ¼ dX .
duðXÞ
(5) For uðXÞ ¼ K X a and wðXÞ ¼ K X b, there is dX ¼ dwðXÞ
dX .
In the following, the concept of elasticity coefficient, which is usually used in

economics, will be introduced to analyze the derivative of IFFs:
Let y ¼ f ðxÞ be a real function, then we call
Dy dy
y y dy x
Ef ðxÞ ¼ lim ¼ dx ¼
Dx!0 Dx dx y
x x
the elasticity coefficient of f ðxÞ at x.

According to the definition of elasticity coefficient, we get that the elasticity
coefficient actually describes what percentage of the dependent variable y will
change if 1% of x happens. It is just an indicator to depict the sensitivity of y to the
argument x. Moreover, there is also
1 l df ðlÞ l dð1 f ðlÞÞ dl 1 l E1f ðlÞ

¼ ¼
1 f ðlÞ dl 1 f ðlÞ dl dð1 lÞ l E1l
and
m dgðmÞ EgðmÞ
1 ¼1
gðmÞ dm Em
Thus, the expression of the derivative of uðXÞ can be written (Lei and Xu
2015b) as

duðXÞ 1 l df ðlÞ m dgðmÞ E1f ðlÞ EgðmÞ
¼ ;1 ¼ ;1
dX 1 f ðlÞ dl gðmÞ dm E1l Em
If we understand the IFF u : ðl; mÞ ! ðf ðlÞ; gðmÞÞ by utilizing its interval form
introduced in Chap. 1, that is, u : ½m; 1 l ! ½gðmÞ; 1 f ðlÞ , then we can get that
the derivative of uðXÞ at X essentially depicts that the reaction extent of intervals
endpoints of ½gðmÞ; 1 f ðlÞ when the endpoints of ½m; 1 l change (Lei and Xu
2015b).
Below we introduce the notion of the compound IFFs, and the chain rule of
derivatives:
Definition 2.4 (Lei and Xu 2015a) Let uðXÞ be an IFF defined in a set B, and XðtÞ
be an IFF in a set A. Then, the set G is a non-empty of A that satisfies XðtÞ 2 B for
any t 2 G, which means
G ¼ ftjt 2 A; XðtÞ 2 Bg 6¼ ø
For any t 2 G, according to the corresponding relation X, there is an IFN XðtÞ
belonging to B. Then according to the corresponding relation u, there is an IFN
uðXÞ. Hence, we define an IFF u
X in the set G as a compound IFF, which is
ðu
XÞðtÞ ¼ uðXðtÞÞ; t 2 G
where X is called the middle variable.

It needs to point out that if there are no special instructions, both the IFFs u and
X in a compound IFF u
X are monotonically increasing IFFs.
Theorem 2.2 (Lei and Xu 2015a) Let uðXðtÞÞ be a compound IFF. If the
derivatives of u and X both exist, then the derivative of uðXðtÞÞ also exists, and
duðXðtÞÞ duðXðtÞÞ dXðtÞ

¼
dt dXðtÞ dt
Proof We prove it in two different ways:

Method 1. Firstly, we can give a proof based on the definition of derivative of IFFs.
For the right derivative, there is
duðXðtÞÞ uðXðt DtÞÞuðXðtÞÞ

¼ lim
dt Dt!O Dt

uðXðt DtÞÞuðXðtÞÞ Xðt DtÞXðtÞ
¼ lim
Dt!O Xðt DtÞXðtÞ Dt
uðXðt DtÞÞuðXðtÞÞ Xðt DtÞXðtÞ
¼ lim lim
Dt!O Xðt DtÞXðtÞ Dt!O Dt
duðXðtÞÞ dXðtÞ
¼
dXðtÞ dt
Similarly, we can get the same conclusion for the left derivative.
Method 2. We study its membership and non-membership, respectively. The proof
is shown as follows:

Assume that uðaÞ ¼ fu ðla Þ; gu ðma Þ and XðtÞ ¼ ðfX ðlt Þ; gX ðmt ÞÞ are the IFFs of
the independent a ¼ ðla ; ma Þ and t ¼ ðlt ; mt Þ, respectively, and uðXðtÞÞ ¼

fu ðfX ðlt ÞÞ;gu ðgX ðmt ÞÞ is a compound IFF, then

duðXðtÞÞ 1 lt dfu ðfX ðlt ÞÞ mt dgu ðgX ðmt ÞÞ
¼ ;1
dt 1 fu ðfX ðlt ÞÞ dlt gu ðgX ðmt ÞÞ dmt

1 lt dfX ðlt Þ 1 fX ðlt Þ dfu ðfX ðlt ÞÞ
¼ ;
1 fX ðlt Þ dlt 1 fu ðfX ðlt ÞÞ dfX ðlt Þ

mt dgX ðmt Þ gX ðmt Þ dgu ðgX ðmt ÞÞ
1 1 1
gX ðmt Þ dmt gu ðgX ðmt ÞÞ dgX ðmt Þ

1 lt dfX ðlt Þ mt dgX ðmt Þ
¼ ;1
1 fX ðlt Þ dlt gX ðmt Þ dmt

1 fX ðlt Þ dfu ðfX ðlt ÞÞ gX ðmt Þ dgu ðgX ðmt ÞÞ
;1
1 fu ðfX ðlt ÞÞ dfX ðlt Þ gu ðgX ðmt ÞÞ dgX ðmt Þ
duðXðtÞÞ dXðtÞ
¼
dXðtÞ dt
which completes the proof of the theorem.

According to Theorem 2.2, the derivative of the compound function is also equal
to the product of the derivatives of the components in intuitionistic fuzzy calculus,
which is the same as one of the traditional calculus. Furthermore, we provide some
examples (Lei & Xu, 2015a) as follows:
(1) Let u 1 ðaÞ ¼ ka C, u2 ðaÞ ¼ ka, u3 ðaÞ ¼ a C and u4 ðaÞ ¼ a be four IFFs,
where 0 k 1. Then, according to the derivatives of IFFs, there are
du 1 ðaÞ du2 ðaÞ du3 ðaÞ du4 ðaÞ
da ¼ ðk; 1 kÞ, da ¼ ðk; 1 kÞ, da ¼ E and da ¼ E. Moreover,
we can conduct the following conclusions:
(a) If u 1 ðaÞ ¼ ka C is expressed as u 1 ðu4 ðaÞÞ ¼ ku4 ðaÞ C, then there is
du1 ðaÞ du1 ðu4 ðaÞÞ

¼
da da
du1 ðu4 ðaÞÞ du4 ðaÞ du1 ðaÞ du4 ðaÞ
¼ ¼
du4 ðaÞ da da da
¼ ðk; 1 kÞ E ¼ ðk; 1 kÞ
(b) If u 1 ðaÞ ¼ ka C is written as u1 ðu3 ðaÞÞ ¼ u3 ðaÞ C, then
du 1 ðaÞ du1 ðu3 ðaÞÞ du1 ðu3 ðaÞÞ du3 ðaÞ

¼ ¼ ¼ ðk; 1 kÞ
da da du3 ðaÞ da
(2) Considering uðaÞ ¼ ka C and IFWAx ða1 ; a2 ; . . .; an Þ ¼ ni¼1 xi ai , then

when regarding a1 as its independent variable, the derivative of
uðIFWAx ða1 ; a2 ; . . .; an ÞÞ can be calculated:
duðIFWAx ða1 ; a2 ; . . .; an ÞÞ
da1
duðIFWAx ða1 ; a2 ; . . .; an ÞÞ dIFWAx ða1 ; a2 ; . . .; an Þ
¼
dIFWAx ða1 ; a2 ; . . .; an Þ da1
¼ ðk ; 1 kÞ ðk 1 ; 1 k 1 Þ
¼ ðk k 1 ; 1 k k 1 Þ
n
What’s more, we know uðIFWAx ða1 ; a2 ; . . .; a n ÞÞ ¼ kki ai C, then
i¼1
duðIFWAx ða1 ; a2 ; . . .; an ÞÞ
¼ ðkk1 ; 1 kk1 Þ
da1
Before introducing the properties of derivatives of IFFs, we first show several

concepts as follows:
Definition 2.5 (Lei and Xu 2015b) For any given IFN a ¼ ðl; mÞ, we define
UðaÞ ¼ l, VðaÞ ¼ m and pðaÞ ¼ 1 l m.
Based on Definition 2.5, we get UðaÞ ¼ 0:1, VðaÞ ¼ 0:6 and pðaÞ ¼ 0:3 for
a ¼ ð0:1; 0:6Þ. In addition, any IFN a ¼ ðl; mÞ can be denoted by
a ¼ ðUðaÞ; VðaÞÞ. Similarly, any IFF uðXÞ ¼ ðf ðlÞ; gðmÞÞ can be written as
uðXÞ ¼ ðUðuðXÞÞ; VðuðXÞÞÞ.
Theorem 2.3 (Lei and Xu 2015a) Let uðXÞ and ui ðXÞði ¼ 1; 2; ; nÞ be n þ 1
derivable IFFs, then

n P
n
0
P
n
0
(1) d
dX ui ðXÞ ¼ U ðu i ðXÞÞ; 1 ð1 V ðu i ðXÞÞÞ , where u0 i ðXÞ
i¼1 i¼1 i¼1
dui ðXÞ
represents the derivative of ui ðXÞ, which is dX .
(2) If ui ðXÞEuj ðXÞð1 i; j nÞ, then
d
uj ðXÞui ðXÞ ¼ U u0 j ðXÞ U ðu0 i ðXÞÞ; 1 V ðu0 i ðXÞÞ V u0 j ðXÞ
dX
(3) dX ðkuðXÞÞ
d
¼ ðk; 1 kÞ duðXÞ
dX , where 0 k 1.
Proof We can prove the conclusion (1) based on the derivatives of IFFs. When
i ¼ 1; 2, there is
u1 ðXÞ u2 ðXÞ ¼ ðf1 ðlÞ þ f2 ðlÞ f1 ðlÞf2 ðlÞ; g1 ðmÞg2 ðmÞÞ

2
Hence, dX ui ðXÞ ¼ dX
d d
ðf1 ðlÞ þ f2 ðlÞ f1 ðlÞf2 ðlÞ; g1 ðmÞg2 ðmÞÞ. Then, its
i¼1
membership and non-membership degrees are respectively discussed as follows:
dX ðf1 ðlÞ þ f2 ðlÞ f1 ðlÞf2 ðlÞ; g1 ðmÞ

d
(a) Considering the membership degree of
g2 ðmÞÞ, we have
1l 0 X 2
f1 ðlÞð1 f2 ðlÞÞ þ f20 ðlÞð1 f1 ðlÞÞ ¼ U ðu0 i ðXÞÞ
ð1 f1 ðlÞÞð1 f2 ðlÞÞ i¼1
(b) For its non-membership degree, there is
m 0 Xn
1 g1 ðmÞg2 ðmÞ þ g1 ðmÞg02 ðmÞ ¼ 1 V ðu0 i ðXÞÞ
g1 ðmÞg2 ðmÞ i¼1
On the basis of (a) and (b), we get that the conclusion (1) holds when i ¼ 1; 2.
Moreover, according to the mathematical induction, it is easy to prove that it also
holds when i ¼ 1; 2; ; n.
For the conclusion (2), we also investigate its membership and non-membership
degrees, respectively, whose proofs are similar to those of the conclusion (1), and
are omitted here.
In what follows, we prove the conclusion (3) by utilizing the chain rule of
derivatives:
d ðkuðXÞÞ d ðkuðXÞÞ duðXÞ duðXÞ

¼ ¼ ðk; 1 kÞ
dX duðXÞ dX dX
Thus, the proof of this theorem is completed.

2.3 Differentials of Intuitionistic Fuzzy Functions 33
2.3 Differentials of Intuitionistic Fuzzy Functions
Let uðXÞ be an IFF, and the function value uðX0 Þ is derivable at X 0 . If we want to
get the function values of some points near X 0 , such as uðX 0 DXÞ at X0 DX, it
is usual to acquire the approximation of uðX 0 DXÞ instead of the precise value of
uðX 0 DXÞ in practical applications. Hence, we will focus on the methods to
calculate the approximate of IFFs in this chapter.
For a monotonically increasing IFF uðXÞ, DY ¼ uðX 0 DXÞuðX0 Þ or
uðX 0 DXÞ ¼ uðX0 Þ DY, then we only need to handle the approximation of DY
in order to get the expression uðX0 DXÞ. Obviously, DY is related to DX. To
facilitate the calculation, DY will be replaced by a simple IFF depending on DX.
Now we give a definition of differential of IFFs as follows:
Definition 2.6 (Lei and Xu 2015b) Let uðXÞ be a derivable IFF. If we note a tiny
change of X as DX, then
duðXÞ
duðXÞ ¼ DX
dX
is called the differential of uðXÞ. In addition, dX ¼ E DX ¼ DX due to dX dX ¼ E

for the identity IFF uðXÞ ¼ X. Therefore, the differential can be rewritten as:
duðXÞ
duðXÞ ¼ dX
dX
Based on the differential of uðXÞ, the following theorem is provided to reveal

the relationship between the increment of u (DuðXÞ) and its differential (duðXÞ):
Theorem 2.4 (Lei and Xu 2015b) Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ be a monotonically
increasing IFF, and also be derivable, then
duðXÞ
uðYÞuðXÞ
ðYXÞ
dX

Dl
where XEY. If we note DX ¼ YX ¼ 1l ; 1 Dm
m and Du ¼ uðYÞuðXÞ,
then the equality can be actually expressed as Du du, which satisfies the fol-
lowing conditions:
UðDuÞ UðduÞ VðDuÞ VðduÞ

lim ¼ 0; lim ¼0
Dl!0 Dl Dm!0 Dm
Proof If XEY, X and Y are expressed as ðl; mÞ and ðl0 ; m0 Þ, respectively, then

0 0 l0 l m0
DX ¼ YX ¼ ðl ; m Þðl; mÞ ¼ ;
1l m
and there is

duðXÞ 1 l df ðlÞ m dgðmÞ
¼ ; 1
dX 1 f ðlÞ dl gðmÞ dm
Hence, we have

duðXÞ l0 l df ðlÞ m m0 dgðmÞ
du ¼ DX ¼ ; 1
dX 1 f ðlÞ dl gðmÞ dm
Furthermore
duðXÞ
uðXÞ DX
dX
l0 l df ðlÞ l0 l df ðlÞ m m0 dgðmÞ
¼ f ðlÞ þ f ðlÞ ; gðmÞ 1
1 f ðlÞ dl 1 f ðlÞ dl gðmÞ dm

df ðlÞ dgðmÞ
¼ f ðlÞ þ ðl0 lÞ ; gðmÞ þ ðm0 mÞ
dl dm
¼ ðf ðlÞ þ ðf ðl0 Þ f ðlÞ þ oðl0 lÞÞ; gðmÞ þ ðgðm0 Þ gðmÞ þ oðm0 mÞÞÞ
¼ ðf ðl0 Þ þ oðl0 lÞ; gðm0 Þ þ oðm0 mÞÞ
FðYÞ
Hence, uðYÞuðXÞ duðXÞ

dX ðYX Þ holds, which satisfies the conditions:
UðDuÞUðduÞ VðDuÞVðduÞ
lim Dl ¼ 0 and lim Dm ¼ 0.
Dl!0 Dm!0
Theorem 2.4 describes the fact that du closely approximates to Du by com-
paring the membership degrees (UðDuÞ and UðduÞ) and the non-membership
degrees (VðDuÞ and VðduÞ) of du and Du, respectively. Moreover, this theorem
manifests that the differences UðDuÞ UðduÞ and VðDuÞ VðduÞ are the
infinitesimals of higher order of Dl and Dm, respectively. However, we doubt
whether the following hold:
duðXÞ
uðYÞuðXÞ ¼ ðYX Þ e
dX
or
duðXÞ
uðYÞuðXÞ ¼ ðYX Þ e
dX
Fig. 2.3 The relationship

between du and Du
which is Du ¼ du e or Du ¼ du e. In order to investigate this problem,

Fig. 2.3 is provided:
Because UðDuÞ UðduÞ and VðDuÞ VðduÞ are respectively the infinitesi-
mals of higher order of Dl and Dm, hence, we know that du must fall into a
rectangular area like D in Fig. 2.2. If du is located in D \ S ðDuÞ, then
Du ¼ du e. In addition, Du ¼ du e if only du falls into D \ S ðDuÞ.
Generally, there is not always Du ¼ du e or Du ¼ du e.
In the following, we analyze the differentials of IFFs by utilizing the elasticity
coefficient. The formula in Theorem 2.4 can be transformed into the interval form
(Lei & Xu, 2015b) as follows:
duðXÞ
uðYÞuðXÞ ðYXÞ
0 dX 0
f ðl Þ f ðlÞ gðm0 Þ 1 l df ðlÞ m dgðmÞ l l m0
, ; ; 1 ;
1 f ðlÞ gðmÞ 1 f ðlÞ dl gðmÞ dm 1l m
0 0
f ðl Þ f ðlÞ gðmÞ gðm0 Þ 1 l df ðlÞ m dgðmÞ l l m m0
, ; ; ;
1 f ðlÞ gðmÞ 1 f ðlÞ dl gðmÞ dm 1l m
Therefore, it is easy to obtain that the formula in Theorem 2.4 reveals how to
estimate the relative increment of ½f ðlÞ; 1 gðmÞ when
h the relative increment
i of
½l; 1 m and the range of the elasticity coefficient 1f ðlÞ dl ; gðmÞ dm are both
known.
Two examples (Lei and Xu 2015b) are provided to show Theorem 3.4 below:
(1) Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ ¼ kX (0\k 1), then f ðlÞ ¼ 1 ð1 lÞk and
gðmÞ ¼ mk . In addition, there is
!
duðXÞ 1l k1
¼ kð1 lÞ ; 1k ¼ ðk; 1 kÞ
dX ð1 lÞk
Thus, according to Theorem 3.4, there is uðX DXÞuðXÞ ðk; 1 kÞ DX.

Moreover, by the operational law of IFNs: kða1 a2 Þ ¼ ka1 ka2 ,
uðX DXÞuðXÞ ¼ kðX DXÞkX ¼ k DX holds.
Without loss of generality, we assume k ¼ 0:2 and DX ¼ ð0:02; 0:97Þ, then
ðk; 1 kÞ DX ¼ ð0:004; 0:994Þ and k DX ¼ ð0:0040324; 0:9939267Þ
Obviously, replacing k DX by ðk; 1 kÞ DX is approximate.

(2) Assume that there are four IFNs a 1 ¼ ð0:3; 0:4Þ, a 2 ¼ ð0:2; 0:5Þ, a 3 ¼
ð0:1; 0:2Þ and a 4 ¼ ð0:3; 0:4Þ, and their weights are respectively 0.2, 0.4, 0.1
and 0.3, then we have
!
Y
4 Y
4
IFWAx ða1 ; a2 ; a3 ; a4 Þ ¼ 1 ð1 laj Þxj ; mx
aj
j
j¼1 j¼1

¼ 1 ð1 0:3Þ0:2 ð1 0:2Þ0:4 ð1 0:1Þ0:1

ð1 0:3Þ0:3 ; 0:40:2 0:50:4 0:20:1 0:40:3
¼ ð0:243; 0:408Þ
If some decision makers think that their assessment a 1 ¼ ð0:3; 0:4Þ has an
estimated
error, and want to give the value again. Let the new value be
a0 1 ¼ l01 ; m01 , then
(1) If assuming a1 Ea01 , which means that there exists an IFN b 1 such that
a0 1 ¼ a 1 b 1 , and a0 1 ¼ ð0:4; 0:3Þ, then b1 ¼ a01 a1 ¼ ð0:143; 0:75Þ.
Hence, there is
IFWAx ða0 1 ; a2 ; a3 ; a4 Þ IFWAx ða1 ; a2 ; a3 ; a4 Þ ðx1 ; 1 x1 Þ ða0 1 a1 Þ

¼ ð0:243; 0:408Þ ð0:2; 0:8Þ ð0:143; 0:75Þ
¼ ð0:265; 0; 388Þ
(2) When a0 1 Ea 1 , which means that there is b2 such that a0 1 ¼ a 1 b2 , and
a0 1 ¼ ð0:2; 0:5Þ, then b2 ¼ a 1 a0 1 ¼ ð0:125; 0:8Þ. Hence, there is
IFWAx ða0 1 ; a2 ; a3 ; a4 Þ IFWAx ða1 ; a2 ; a3 ; a4 Þðx1 ; 1 x1 Þ ða1 a0 1 Þ

¼ ð0:243; 0:408Þð0:2; 0:8Þ ð0:125; 0:8Þ
¼ ð0:224; 0:425Þ
The following theorem shows that the situation holds when Du ¼ du:
Theorem 2.5 (Lei and Xu 2015b) Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ be an IFF, which
satisfies
d 2 f ðlÞ d 2 gðmÞ
¼ ¼0
dl2 dm2
when XEY, then
duðXÞ
uðYÞuðXÞ ¼ ðYXÞ
dX
Based on the proof of Theorem 2.4, it is easy to prove Theorem 2.5. So it is
omitted here. Next, we give an example (Lei and Xu 2015b) to illustrate this
theorem:
Let a0 ¼ ðl0 ; m0 Þ, a ¼ ðl; mÞ and X ¼ ðlX ; mX Þ be three IFNs, and uðaÞ ¼
a0 a be an IFF. Because f ðlÞ ¼ l0 l and gðmÞ ¼ m þ m0 m0 m satisfy
d 2 f ðlÞ d 2 gðmÞ
¼ ¼0
dl2 dm2
According to Theorem 2.5, we have
duðaÞ
uða0 Þ ¼ uðaÞ ða0 aÞ
da
In addition, by using the formula of derivative of IFFs, we have

duðaÞ 1l m
¼ a0 ; 1
da 1 l0 l m þ m0 m0 m
Hence, there is

1l m
a0 ða XÞ ¼ a0 a a0 X ; 1
1 l0 l m þ m0 m0 m
Now we use the following (1)–(3) to test the above equality:

(1) a0 ¼ a X ¼ ðl þ lX llX ; mmX Þ:
(2) duðaÞ X ¼ ðlX ; mX Þ ðl0 ; m0 Þ 1 l ; 1 m
da 1 l0 l m þ m0 m0 m

l0 l0 l m0
¼ ðlX ; mX Þ ;
1 l0 l m þ m0 m0 m

lX l0 lX l0 l m0 mX m0
¼ ; mX þ
1 l0 l m þ m0 m0 m m þ m0 m0 m
duðaÞ duðaÞ
(3) uðaÞ X ¼ a0 a X
da da
¼ ðl0 l þ l0 lX l0 llX ; m0 þ mmX m0 mmX Þ
Furthermore,
uða0 Þ ¼ a0 ða XÞ ¼ ðl0 l þ l0 lX l0 llX ; m0 þ mmX m0 mmX Þ
Hence, uða0 Þ ¼ uðaÞ duðaÞ 0

da ða aÞ holds, which is

1l m
a0 ða XÞ ¼ a0 a a0 X ; 1
1 l0 l m þ m0 m0 m
which indicates a0 ða XÞ 6¼ a0 a a0 X holds.

In the last section, we have studied the chain rule of derivatives of the compound
IFFs. It is natural to obtain the form invariance of differential in intuitionistic fuzzy
calculus. Next, we show the important property (Lei and Xu 2015a) as follows:
Let uðwðtÞÞ be a compound IFF, which consists of two IFFs Y ¼ uðXÞ and
X ¼ wðtÞ. If both Y 0 X and X 0 t exist, then Y 0 t ¼ Y 0 X X0 t .
If considering the middle variable X as the independent variable of Y, then
dY ¼ Y 0 X dX
On the other hand, if we do not consider X as the independent variable of Y, but

t, then
dY ¼ Y 0 t dt
However, if we replace Y 0 t with Y 0 X X0 t and notice that X 0 t dt is actually

dX, then we have the following derivational process:
dY ¼ Y 0 t dt ¼ Y 0 X X 0 t dt ¼ Y 0 X dX
which is just the form invariance of differential in intuitionistic fuzzy calculus.

Based on these analyses, we know that the differential form remains unchanged
when the original variable of IFFs is replaced by a new variable. The only differ-
ence is that dX will not be regarded as the increment DX but the differential of
X ¼ wðtÞ when we consider t as the independent variable of Y.
2.4 Conclusions
In this chapter, we have mainly studied the derivatives and differentials of IFFs.
Firstly, we have defined the concept of IFF, and pointed out that the IFFs are the
main research object in intuitionistic fuzzy calculus. In addition, we have also given
the definitions of monotonically increasing IFFs and continuous IFFs. By taking the
limit value of difference quotients of IFFs, we have defined the derivatives of IFFs.
After that, we have studied some important proprieties of derivatives, like the chain
rule of derivatives of the compound IFFs. In addition, this chapter has also inves-
tigated the derivatives of IFFs by introducing the notion of elasticity coefficient.
Based on the derivatives of IFFs, we have defined the differentials of IFFs, and
proven the relationship between the increment of u (DuðXÞ) and its differential
(duðXÞ). Finally, we have revealed the form invariance of differential in intu-
itionistic fuzzy calculus.
Chapter 3
Integrals of Intuitionistic Fuzzy Functions
Based on the derivatives of intuitionistic fuzzy functions (IFFs), this chapter first
introduce its inverse operation, which is the indefinite integrals of IFFs, and then
investigates the properties of the indefinite integrals of IFFs. In addition, this
chapter deliberates on the definite integrals of IFFs. The Newton-Leibniz formula in
intuitionistic fuzzy environment, which is the fundamental theorem of intuitionistic
fuzzy calculus, will be provided to manifest the important relationship between the
indefinite integrals and the definite integrals of IFFs.
3.1 Indefinite Integrals of Intuitionistic Fuzzy Functions
After acquiring the derivatives of IFFs, it is natural to make further efforts to study
their inverse operations, which are the indefinite integrals. Firstly, we make some
discussions about the primitive functions (Lei and Xu 2015c) of IFFs below:
Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ be an IFF. In order to get its primitive function
UðXÞ ¼ ðFðlÞ; GðmÞÞ, which satisfies dUdXðXÞ ¼ uðXÞ, we need to solve two ordi-
nary differential equations:
8 8 Z
> 1 l d FðlÞ >
> f ðlÞ
>
< 1 FðlÞ dl ¼ f ðlÞ >
< FðlÞ ¼ 1 c 1 exp dl
1l
) Z
>
> m d GðmÞ >
> 1 gðmÞ
: 1 ¼ gðmÞ >
: GðmÞ ¼ c 2 exp dm
GðmÞ d m m
where c 1 and c2 are two integral constants, which are both real numbers such that
UðXÞ is an IFF. In other words, c 1 and c2 should make the following (1)–(3) hold.
42 3 Integrals of Intuitionistic Fuzzy Functions
n R o
f ðlÞ
(1) 0 1 c1 exp 1l dl 1.
nR o
(2) 0 c2 exp 1gðmÞ
m dm 1.
n R o nR o
f ðlÞ 1gðmÞ
(3) 0 1 c1 exp 1l dl þ c 2 exp m dm 1.
By the above differential equations, we know that U ðXÞ has the following form
(Lei and Xu 2015c):
Z Z
f ðlÞ 1 gðmÞ
UðXÞ ¼ 1 c1 exp dl ; c2 exp dm
1l m
R
which is the indefinite integral of uðXÞ, denoted by uðXÞdX.
Now we present
R the following derivational process to demonstrate whether the
derivative of uðXÞdX is certainly uðXÞ:
n R o nR o
R d 1 c 1 exp f ðlÞ
; c 2 exp 1gðmÞ
dU ðXÞ d uðXÞdX 1l dl m dm
¼ ¼
dX dX n R o dX nR o
0 f ðlÞ
1
ð1 lÞc1 exp 1l dl f ðlÞ m c 2 exp 1gðmÞ
m dm 1 gðmÞ
¼@ n R o ; 1 nR o A
f ðlÞ
c1 exp 1l dl 1l c 2 exp 1gðmÞ dm m
m
¼ ðf ðlÞ; gðmÞÞ ¼ uðXÞ
which proves that UðXÞ is just a primitive function of uðXÞ.

Theorem 3.1 (Lei and Xu 2015c) Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ be an IFF. Then the
primitive function UðXÞ of uðXÞ must have the following form:
Z Z
f ðlÞ 1 gðmÞ
UðXÞ ¼ 1 c1 exp dl ; c2 exp dm
1l m
where c 1 and c2 are two integral constants.

Proof
Firstly,
n according
R f ðlÞ o
to the above o we get that UðXÞ with the form
nR analysis,
1 c1 exp 1l dl ; c2 exp 1gðmÞ
m dm must be a primitive function of
uðXÞ.
Moreover, for the uniqueness of solution of ordinary differential equations, all
primitive functions of uðXÞ must have the above-mentioned form. In fact, for any
two primitive functions UðXÞ and WðXÞ (UðXÞ 6¼ WðXÞ), the only difference
between them is that the integral constants of U ðXÞ are different from ones of
WðXÞ. ■
3.1 Indefinite Integrals of Intuitionistic Fuzzy Functions 43
Theorem 3.2 (Lei and Xu 2015c) Let UðXÞ and WðXÞ be two IFFs, and
Z Z
f ðlÞ 1 gðmÞ
U ðXÞ ¼ 1 c 1 exp dl ; c 2 exp dm
1l m
then there exist two real numbers k 1 and k 2 , such that

Z Z
f ðlÞ 1 gðmÞ
WðXÞ ¼ 1 k 1 c1 exp dl ; k2 c2 exp dm
1l m
Based on Theorem 3.1, we can easily obtain the proof of Theorem 3.2, which is
omitted here.
Theorem 3.3 (Lei and Xu 2015c) If there are C2 EUðXÞ and dUdXðXÞ ¼ uðXÞ, then
we have
dU ðXÞ dðUðXÞC1 Þ dðUðXÞC2 Þ

¼ ¼
dX dX dX

Proof Let C1 ¼ lC1 ; mC1 , then
Z Z
f ðlÞ 1 gðmÞ
UðXÞ C1 ¼ 1 b1 exp dl ; b 2 exp dm l C 1 ; mC 1
1l m
Z Z
f ðlÞ 1 gðmÞ
¼ 1 b 1 1 lC1 exp dl ; b 2 mC1 exp dm
1l m

Denoting the constants b1 1 lC1 and b 2 mC1 as const1 and const2 , respec-
tively, then
Z Z
f ðlÞ 1 gðmÞ
U ðXÞ C1 ¼ 1 const1 exp dl ; const2 exp dm
1l m
Hence, UðXÞ C1 can be expressed as one of the forms of primitive functions

of uðXÞ, which means dðU ðXÞC
dX
1Þ
¼ uðXÞ. In addition, we can prove dUdXðXÞ ¼
dðU ðXÞC2 Þ
dX in the same way. ■
Next, we present some properties of indefinite integrals of IFFs:
R
Theorem 3.4 (Lei and Xu 2015a) If there is U ðXÞ ¼ uðXÞdX, then
Z
uðXðtÞÞX 0 ðtÞdt ¼ UðXðtÞÞ
where uðXðtÞÞX 0 ðtÞ represents that uðXðtÞÞ dXðtÞ

dt .
Proof We prove it by using the chain rule of derivatives of the compound IFFs.
Because
dU ðXðtÞÞ dU ðXðtÞÞ dXðtÞ

¼
dt dXðtÞ dt
Z
dU ðXðtÞÞ dXðtÞ d
) ¼ uðXÞdX ¼ uðXðtÞÞ X0 ðtÞ
dt dt dX
Hence, UR ðXðtÞÞ must be the primitive functions of uðXðtÞÞ X 0 ðtÞ, which

0
means that uðXðtÞÞX ðtÞdt ¼ UðXðtÞÞ holds. ■
Theorem 3.5 (Lei and Xu 2015a) Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ and ui ðXÞ ¼
ðfi ðlÞ; gi ðmÞÞ (i ¼ 1; 2; . . .; n) be n þ 1 derivable IFFs, then
R R
(1) ðk; 1 kÞ uðXÞdX ¼ k uðXÞdX, where 0 k 1:
n
R P Pn n R
(2) fi ðlÞ; 1 ð1 gi ðmÞÞ dX ¼ ðfi ðlÞ; gi ðmÞÞdX:
i¼1
R i¼1 i¼1 R R
(3) ðf1 ðlÞ f2 ðlÞ; 1 ðg2 ðmÞ g1 ðmÞÞÞdX ¼ ðf1 ðlÞ; g1 ðmÞÞdX ðf2 ðlÞ; g2 ðmÞÞdX:
Proof We utilize the chain rule of derivatives to prove (1) below: Since
Z
d
k uðXÞdX ¼ ðk; 1 kÞ uðXÞ
dX
R R
then we can get that ðk; 1 kÞ uðXÞdX ¼ k uðXÞdX holds. Moreover, we
can also prove it by using the calculating formula of indefinite integrals of IFFs:
Z Z Z
f ðlÞ 1 gðmÞ
k uðXÞdX ¼ k 1 c1 exp dl ; c2 exp dm
1l m
Z Z
f ðlÞ 1 gðmÞ
¼ 1 c1 exp k
k
dl ; c2 exp
k
k dm
1l m
Z Z
kf ðlÞ 1 ð 1 k þ gðmÞ ð1 kÞgðmÞÞ
¼ 1 ck1 exp dl ; ck2 exp dm
1l m
Z
¼ ðk; 1 kÞ uðXÞdX
In addition, another method called the substitution rule for indefinite integrals
can be provided to prove (1), which is actually introduced in Theorem 3.4, i.e.,
Z Z Z Z
ðk; 1 kÞ uðXÞdX ¼ ðk; 1 kÞd uðXÞdX ¼k uðXÞdX
All in all, the conclusion (1) holds. Similarly, the proofs of (2) and (3) can be
conducted in the same manner, which are omitted here. ■
3.2 Definite Integrals of Intuitionistic Fuzzy Functions 45
3.2 Definite Integrals of Intuitionistic Fuzzy Functions
At the beginning of this section, we review the integrals of the complex functions:
Let C be a simple curve in the complex plane C, which means that the curve is
smooth or piecewise smooth, and let f ðzÞ ¼ uðx; yÞ þ ivðx; yÞ be a continuous
function in C, where uðx; yÞ and vðx; yÞ are called the real part and the imaginary
part of f ðzÞ, respectively. Then in order to define the integral of f ðzÞ along the curve
C, we introduce the following steps:
(1) Dividing the simple curve. We first interpolate some break points, namely z0 ,
z1 , z2 , . . ., zn1 , zn ¼ z, into the simple curve C. Then C will be divided into lots
_ _ _
of small arcs z0 z1 , z1 z2 , . . ., zn1 zn , where zk (k ¼ 0; 1; . . .; n) are arranged from
_
z0 to z, which is shown in Fig. 3.2a (Lei and Xu 2016a).
(2) Making the product. From every arc zk zk þ 1 , we take a value fk ¼ nk þ igk to
get the product f ðfk Þðzk þ 1 zk Þ, which is actually
½uðnk ; gk Þ þ ivðnk ; gk Þ½ðxk þ 1 xk Þ þ iðyk þ 1 yk Þ
(3) Calculating the sum. We add all f ðfk Þðzk þ 1 zk Þ(k ¼ 0; 1; . . .; n 1) toge-
P
ther to get the sum n1
i¼1 f ðfk Þðzk þ 1 zk Þ, that is,
Xn1
i¼1
½uðnk ; gk Þ þ ivðnk ; gk Þ½ðxk þ 1 xk Þ þ iðyk þ 1 yk Þ
Pn1 Pn1
which can also be represented as
hP i¼1 uðnk ; gk Þðxk þ 1 xk Þ i¼1 ivðnk ; gk Þ
n1 Pn1
ðyk þ 1 yk Þ þ i i¼1 vðnk ; gk Þðxk þ 1 xk Þ þ i¼1 uðnk ; gk Þðyk þ 1 yk Þ :
(4) Taking the limit. When the number

of break points zk increases infinitely, and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
satisfies the condition: max jzk þ 1 zk j ¼ ðxk þ 1 xk Þ2 þ ðyk þ 1 yk Þ2
i
Pn1 Pn1
! 0, in addition, these sum i¼1 uðnk ; gk Þðxk þ 1 xk Þ, i¼1 vðnk ; gk Þ
Pn1 Pn1
ðyk þ 1 yk Þ, vðn ;
i¼1 R k k g Þðx k þ 1
R x k Þ and Ri¼1 uðn ;
k k g Þðy k þ 1 R yk Þ exist
their limit values C uðx; yÞdx, C vðx; yÞdy, C vðx; yÞdx and C uðx; yÞdy,
respectively, then the limit of the expression:
Xn1 Xn1
i¼1
uðnk ; gk Þðxk þ 1 xk Þ i¼1
vðnk ; gk Þðyk þ 1 yk Þ
hXn1 Xn1 i
þi i¼1
vðnk ; gk Þðxk þ 1 xk Þ þ i¼1
uðnk ; gk Þðyk þ 1 yk Þ
R R
is defined as C uðx; yÞdx vðx; yÞdy þ i C vðx; yÞdx þ uðx; yÞdy, and we call it the
R
integral of f ðzÞ along to C, denoted by C f ðzÞdz.
3.2.1 Generating Ways of Definite Integrals of IFFs
The simple curves are very significant in studying the integrals of complex func-
tions. Similarly, in the research process of integrals of IFFs, it is necessary to define
a novel kind of curves called intuitionistic fuzzy integral curves (IFICs), which are
introduced as:
Definition 3.1 (Lei and Xu 2015c) Assume that there is a curve I linking between a
and b (a and b are both IFNs) that can be written as a bijective mapping
= : ½0; L ! I, where L represents the arc length of I from a to b, and this mapping
satisfies: =ð0Þ ¼ a and =ðLÞ ¼ b. If there always be =ðt1 ÞE=ðt2 Þ for
0 t1 t2 L, then we call the curve I an intuitionistic fuzzy integral curve (IFIC).
Furthermore, we show several IFICs in Fig. 3.1 (Lei and Xu 2016a).
In Fig. 3.1, all curves I i (i ¼ 1; 2; . . .; 5) are IFICs according to the definition of
IFICs in Definition 3.1. It is worth noticing that the order relation “E” is a linear
order in any IFIC because there is always aEb or bEa if only both a and b belong
to the same IFIC I.
Now we define the integrals of IFFs based on an IFIC (Lei and Xu 2016a). It is
different from the integrals of complex functions, which are developed based on the
simple curves (as shown in Fig. 3.2a). Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ be an IFF defined in
an IFIC I, which links between the starting point a and the end point b, then the
integral of uðXÞ along I is defined by the following process:
(1) Dividing the IFIC. By interpolating some break points (IFNs), such as
a ¼ h0 , h1 , h2 , . . ., hn1 , hn ¼ b, into the IFIC I, we can divide I into many
_ _ _
small arcs ah1 , h1 h2 , . . ., hn1 b, where hk (k ¼ 0; 1; . . .; n) are arranged from a
to b, which is shown in Fig. 3.2b (Lei and Xu 2016a).
Fig. 3.1 IFICs linking

between a and b
Fig. 3.2 A simple curve and an IFIC
_
(2) Making the product. From every arc hk hk þ 1 , we take a value ni ¼ li ; mgk

to get the product uðnk Þ ðhk þ 1 hk Þ, which is actually
l
i þ 1 l i mi þ 1
f ðlni Þ; gðmni Þ ;
1 li mi
(3) Calculating the sum. We add all uðnk Þ ðhk þ 1 hk Þ (k ¼ 0; 1; . . .; n 1)

i¼1 uðnk Þ ðhk þ 1 hk Þ, that is
together to get the sum n1
l
i þ 1 l i mi þ 1
n1 f ðlni Þ; gðmni Þ ;
i¼1
1 li mi
(4) Taking the limit. When the number of break points hk increases infinitely, and
meets the condition: hk þ 1 hk ! O (k ¼ 0; 1; . . .; n 1), if both the mem-
bership part and the non-membership part of n1 i¼1 uðnk Þ ðhk þ 1 hk Þ have
their own limits and are respectively equal to the real numbers U and V, and
ðU; V Þ is an IFN, then we call ðU; V Þ the limit value of
i¼1 uðnk Þ ðhkRþ 1 hk Þ, and define it as the integral of uðXÞ along I, which
n1
can be noted by I uðXÞdX.
R
It is acknowledged that the integrals of the complex functions C f ðzÞdz can be
Rz
denoted by z0 f ðzÞdz, where z0 and z are respectively the starting point and the end
R
point of C, because C f ðzÞdz does not depend on the specific integral path but the
extreme points of C. It means that if onlyRthe starting point
R and the end point of C1
are same as the ones of C2 , there must be C1 f ðzÞdz ¼ C2 f ðzÞdz although C1 6¼ C2 .
In the following, we show that there is a similar conclusion in intuitionistic fuzzy

calculus:
Theorem 3.6 (Lei and Xu 2015c, 2016a) Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ be an IFF
defined in an IFIC I, which links between the starting point a and end point b, then
0 8 9 8 mb 91
Z < Z b f ðlÞ >
> l
= <Z 1 gðmÞ =
B C
uðXÞdX ¼ @1 exp dl ; exp dm A
I >
: 1l > ; : m ;
la ma
Proof According to the definition of integral of IFF, we have

Z
k
uðXÞdX ¼ lim ðuðni Þ DX i Þ
I DX 1 ;DX2 ;...;DX k !O i¼1
n
Y Yn !
f ðlni Þ 1 gðmni Þ
¼ 1 lim 1 ðl li Þ ; lim 1þ ðmi þ 1 mi Þ
d!0
i1
1 li i þ 1 d!0
i¼1
mi
( ) ( )!
ð1Þ X n
f ðlni Þ X n
1 gðmni Þ
¼ 1 exp lim ðli þ 1 li Þ ; exp lim ðmi þ 1 mi Þ
d!0
i¼1
1 li d!0
i¼1
mi
( ) ( )!
ð2Þ X n
f ðlni Þ Xn
1 gðmni Þ
¼ 1 exp lim ðli þ 1 li Þ ; exp lim ðmi þ 1 mi Þ
d!0
i¼1
1 lni d!0
i¼1
mni
0 8 9 8 9 1
< Z b f ðlÞ >
> l
= <Z b 1 gðmÞ =
m
B C
¼ @1 exp dl ; exp dm A
>
: 1l > ; : m ;
la ma
Firstly, we prove the non-membership degree of the equality (1) below:

1gðmni Þ
If we denote xi ¼ mi ðmi þ 1 mi Þ, then the equality (1) is equivalent to the
following equality:
( )
Y
n X
n
lim ð1 þ xi Þ ¼ exp lim xi
d!0 d!0
i¼1 i¼1

P
n P
n
which also can be expressed as exp lim lnð1 þ xi Þ ¼ exp lim xi .
d!0 i¼1 d!0 i¼1
BecauseP the exponential function P is continuous, we only need to prove
limd!0 ni¼1 lnð1 þ xi Þ ¼ limd!0 ni¼1 xi , and the specific proof of which is shown
as:
when DX1 ; DX 2 ; . . .; DXk ! O, there will be d ! 0, n ! 1, and each xi
approaches zero. For lim lnð1xþ xÞ ¼ 1, we get that for any e [ 0, there must exist a
x!0

neighborhood dðeÞ such that lnð1xþi xi Þ 1 e holds if only jxi 0j dðeÞ. Hence,

1gðm Þ 1gðm Þ
if noting mi ni ¼ mi ni ¼ M (mi 6¼ 0), then we have jðmi þ 1 mi Þj dðeÞ M .
Because the endless subdivision makes each mi þ 1 mi approaches zero, there must
be a positive integer N 0 (N 0 2 N þ ), which represents the number of small arcs, such

0
that jðmi þ 1 mi Þj dðeÞ
M holds for any 1 i N . Hence, for any given e [ 0, there

must exist N (N N 0 ) such that lnð1xþi xi Þ 1 e (1 i N), which means that
ð1 þ eÞxi lnð1 þ xi Þ ð1 eÞxi , where xi 0
Then we add the N inequalities together to get
X
N X
N X
N
ð1 þ eÞ xi lnð1 þ xi Þ ð1 eÞ xi
i¼1 i¼1 i¼1
P1 P1
If we let e ! 0, then i¼1 lnð1 þ xi Þ ¼ i¼1 xi , which also means that
( )
Y
n X
n
lim ð1 þ xi Þ , exp lim xi
d!0 d!0
i¼1 i¼1
Similarly, we prove the equality (2), which is
X
n
1 gðmn Þ X
n
1 gðmn Þ
lim i
ðmi þ 1 mi Þ ¼ lim i
ðmi þ 1 mi Þ
d!0
i¼1
mi d!0
i¼1
mn i
1gðmni Þ 1gðmni Þ m ni
Firstly, let ai ¼ mi ðmi þ 1 mi Þ and bi ¼ m ni ðmi þ 1 mi Þ, then ai
bi ¼ mi . In
addition, due to the endless subdivision makes each mi þ 1 mi ! 0, there
must
exist
N 0 2 N þ (N 0 is the number of small arcs), such that jmi þ 1 mi j e mb for any
given e [ 0. Hence, we have a positive integer N (N N 0 ) such that

mn mi jmi þ 1 mi j e mb ejmi j
i

which is abii 1 e for any 1 i N, then ð1 þ eÞbi ai ð1 eÞbi , where bi 0.
We add these above inequalities to get the following sum:
X
N X
N X
N
ð1 þ eÞ bi ai ð1 eÞ bi
i¼1 i¼1 i¼1
P1 P1
If letting e ! 0, then i¼1 ai ¼ i¼1 bi , which is just
X
n
1 gðmn Þ X
n
1 gðmn Þ
lim i
ðmi þ 1 mi Þ ¼ lim i
ðmi þ 1 mi Þ
d!0
i¼1
mi d!0
i¼1
mn i
The proofs of the membership parts in (1) and (2) are similar, and thus, we omit
them here. In brief, the equality
0 8 9 8 mb 91
Z < Z b f ðlÞ >
> l
= < Z 1 gðmÞ =
B C
uðXÞdX ¼ @1 exp dl ; exp dm A
I >
: 1 l >
; : m ;
la ma
holds, which completes the proof of this theorem. R ■

Based on Theorem 3.6, we can denote the integral I uðXÞdX of the IFF uðXÞ
Rb
by a uðXÞdX since it is only related to the starting point a and the end point b of I.
Rb
Specially, a uðXÞdX ¼ O when a¼b.
What’s more, we introduce another way to define the definite integrals of IFFs
(Lei and Xu 2015c):
Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ be an IFF defined in the set fX : a E X E bg, which can
be denoted by ½a; b . The form of the set ½a; b is similar to the closed interval ½a; b
in the real number axis. Hence, we can analogize the definite integrals of real
functions to give the similar one of IFFs:
(1) We first interpolate some IFNs hi (i ¼ 1; 2; . . .; n) between a and b, which
means that a ¼ h1 Eh2 E. . .Ehn ¼ b. Then, ½a; b is divided into some smaller
“intervals” d i ¼ ½hi ; hi þ 1 (i ¼ 1; 2; . . .; k), and Ddi ¼ hi þ 1 hi . When the
break points increase infinitely, all Dd i ¼ hi þ 1 hi will approach O. The
process is shown in Fig. 3.3 (Lei and Xu 2015c):
When we interpolate an IFN h in the set ½a; b in (a) of Fig. 3.3, ½a; b is divided
into two parts ½a; h and ½h; b that is shown in (b). Similarly, we continue to put the
break points hm and hn in the ½a; h and ½h; b , respectively. Then ½a; h will be
divided into ½a; hm and ½hm ; h . In addition, ½h; b will be replaced with ½h; hn and
½hn ; b , which are shown in (c). In the same way, Fig. (c) will change into (d), and
end in (e) when the interpolating points increase infinitely. Moreover, according to
the definition of ½ ; , it is clear to get that the curve in (e) is an IFIC introduced in
Definition 3.1.
(2) We choose randomly an IFN ni from d i ¼ ½hi ; hi þ 1 , that
is, ni 2 ½hi ; hi þ 1 ,
and calculate its function value uðni Þ ¼ f ðlni Þ; gðmni Þ . After that, we can
get uðni Þ Dd i (i ¼ 1; 2; . . .; k).
(3) The k IFNs uðni Þ Dd i (i ¼ 1; 2; . . .; k) will be added together to get their
sum ki¼1 ðuðni Þ Dd i Þ.
(4) By interpolating some IFNs between a and b infinitely, if the limit value of
ki¼1 ðuðni Þ Dd i Þ exists, then we call it the definite integral of uðXÞ in ½a; b .
In addition, we can get a fact that the limit value of ki¼1 ðuðni Þ Dd i Þ does not
depend on the extreme points of ½a; b but the random choices of points in the above
Rb
(2) based on Theorem 3.6. Hence, we can denote the limit value by a uðXÞdX.
Fig. 3.3 Another way for introducing the definite integral
Obviously, the definite integral of the IFF defined by the above (1)–(4) is com-
pletely equivalent to the one developed based on the concept of IFICs. Of course,
there is also an essential difference between the two ways in defining the definite
integrals of IFFs, one of which is based on IFICs and similar to the integral
R Rb
C f ðzÞdz of the complex functions, while another is more like the integral a f ðxÞdx
of the real function f ðxÞ in a closed interval.
It is necessary to point out that in the process (2) mentioned above, which is to
randomly
choose
an IFN ni from d i ¼ ½hi ; hi þ 1 to calculate uðni Þ ¼
f ðlni Þ; gðmni Þ , we suppose that the IFF uðXÞ is meaningful for any X 2 ½hi ; hi þ 1 .
In what follows, we will investigate the situation where there exist some mean-
ingless points in the set ½a; b , which is consisted of the lower and upper limits of
the integral of an IFF.
Because the function values of uðXÞ at meaningless point are not IFNs, the
Rb
operation uðXÞ dX is invalid and the integral a uðXÞ dX may be invalid. In
Rb
such a case, we are curious about whether a uðXÞ dX is feasible. According to
the definition of the definite integral of the IFF and its calculating
formula in
Theorem 3.6, the value of lim ki¼1 ðuðni Þ Ddi Þ will keep unchanged
Dd1 ;Dd2 ;...;Ddk !O
when we change the function values of uðXÞ at the point set fn1 ; n2 ; n3 ; . . .; nn ; . . .g
that is a countable set. The limit value does not change just because the two integral
R l f ðlÞ Rm
values lab 1l dl and mab 1gðmÞ m dm will not change even though the function values
of
n f ðlÞ and gðmÞ have
o changed in the corresponding countable sets

ln1 ; ln2 ; ln3 ; . . .; lnn ; . . . and mn1 ; mn2 ; mn3 ; . . .; mnn ; . . . .
Hence, if only there are intuitionistic fuzzy integral curves linking the points a
and b, each of which only includes countably meaningless points of the integrand
uðXÞ, then the definition of the definite integral of uðXÞ in ½a; b is feasible.
In the following, we discuss two situations about the meaningless points (Lei
and Xu 2015c):
Situation 1. If the integrand uðXÞ is meaningless only at a limit number of discrete
points as shown in Fig. 3.4 (Lei and Xu 2015c), then the definite integral of uðXÞ in
½a; b is feasible so long as aEb, which is labeled as the shadow region S. For any
b 2 S, there is at least one IFIC linking a and b, and the meaningless points in the
Rb
IFIC is limited. Hence, a uðXÞ dX can be calculated by the formula in
Theorem 3.6.
Situation 2. If uðXÞ is meaningless in a region, which is noted as D in the Fig. 3.5
(Lei and Xu 2015c) and Fig. 3.6 (Lei and Xu 2015c), then the upper limit b of
Rb Rb
a uðXÞ dX must fall in the shaded area to make a uðXÞ dX be feasible. If
the upper limit b is not in the shaded area, then there is no any IFIC linking a and b
that only includes the limited number of meaningless points, which let the definition
of the definite integral of uðXÞ in ½a; b be invalid.
In brief, we can get that the integrand, the lower and the upper limits of an
Rb
integral a uðXÞ dX can affect whether the definition of the definite integral of
uðXÞ in ½a; b is invalid.
Fig. 3.4 Discrete

meaningless points
Fig. 3.5 The region consists

of meaningless points
Fig. 3.6 D consists of

meaningless points
3.2.2 Properties of Definite Integrals of IFFs
In this subsection, we show some important properties of the definite integral of

IFFs.
Theorem 3.7 (Lei and Xu 2015c) Let uðXÞ ¼ ðf ðlÞ; gðmÞÞ be an IFF, and
0 k 1, then
Zb Zb
ðk; 1 kÞ uðXÞ dX ¼ k uðXÞ dX
a a
Proof According to the formula in Theorem 3.6, we can prove the equality as:
0 8 9 8 mb 91
Zb >
< Z b kf ðlÞ >
l
= < Z 1 ð1 k þ gðmÞ ð1 kÞgðmÞÞ =
B C
ðk; 1 kÞ uðXÞdX ¼ @1 exp dl ; exp dm A
>
: 1l > ; : m ;
a la ma
0 0 8 91 k 0 8 91 k 1
Z >
< >
=
lb
<Z b 1 gðmÞ =
m
B B f ðlÞ C C
¼ @1 @exp dl A ; @exp dm A A
>
: 1l > ; : m ;
la ma
0 8 9 8 mb 91
< Z b f ðlÞ >
> l
= < Z 1 gðmÞ =
B C
¼ k@1 exp dl ; exp dm A
>
: 1l > ; : m ;
la ma
Zb
¼k uðXÞ dX
a
which completes the proof of the theorem. ■

Rb Rb
In addition, when uðXÞ ¼ E, ðk; 1 kÞ E dX ¼ ðk; 1 kÞdX ¼ kðb aÞ
a a
Rb
holds. Specially, E dX ¼ 1
ðb aÞ ¼ b a, and if we let a ¼ O, then
a
Zb
ðk; 1 kÞ dX ¼ k ðb OÞ ¼ k b
O
which reveals the fact when 0 k 1, the scalar multiplication operation (k b) of

the IFNs (the multiplication between IFNs and nonnegative real numbers) can be
replaced by the countless addition “” and multiplication “” of IFNs because the
integral of IFFs consists of the infinite “” and “”. It means that “k b” can be
developed based on “” and “” when 0 k 1.
According to Theorem 3.7, we can discover that the common aggregation
operator IFWAx ða1 ; a2 ; . . .; an Þ is essentially the integral of a piecewise IFF, which
is shown below (Lei and Xu 2015a).
In Chap. 2, we have introduced that
n
IFWAx ða1 ; a2 ; . . .; an Þ ¼ xi ai
i¼1
If we denote b0 ¼ O and bi þ 1 ¼ bi ai þ 1 , then we can introduce the following

piecewise IFF (Lei and Xu 2015a):
8
>
> ðx1 ; 1 x1 Þ; when b 0 EXEb 1 ; ;
>
< ðx 2 ; 1 x 2 Þ; when b 1 EXEb 2 ;
uðXÞ ¼ ..
>
> .
>
:
ðx n ; 1 x n Þ; when bn1 EXEbn ;
which means uðXÞ ¼ ðxi ; 1 x i Þ when b i1 E X E b i (1 i n). Then we know

the following equality holds:
Zbn
uðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ
O
where b n ¼ ni¼1 ai .
Rb
In addition, based on O ðk; 1 kÞ dX ¼ k b, when 0 k 1, we can prove the
following two conclusions:
(1) kða bÞ ¼ ka kb:
(2) k 1 a k 2 a ¼ ðk 1 þ k 2 Þa:
The proofs of (1) and (2) can be shown as follows (Lei and Xu 2015a):
Za Zb
ka kb ¼ ðk; 1 kÞdX ðk; 1 kÞdX
O O
0 8 l 9 8 m 91
< Za k = <Z a k =
: 1l ; : m ;
0 1
0 8 lb 9 8 mb 91
< Z k = <Z
k =A
@1 exp dl ; exp dm
: 1l ; : m ;
0 1
0 8 l 9 8 lb 9
< Z a
= < Z =
k k
¼ @1 exp dl exp dl ;
: 1l ; : 1l ;
0 0
8 m 9 8 mb 91
<Z a k = <Z k =
exp d m ; exp dm A
: m ; : m ;
1 1
0 8 9 8 ma mb 91
> Þð1l bÞ >
<
1ð1l
Za = < Z
B k k =C
>
: 1l > ; : m ;
0 1
¼ ka kb
Za Za
k 1a k 2a ¼ ðk1 ; 1 k1 ÞdX ðk2 ; 1 k2 ÞdX
O O
0 8 l 9 8m 91
< Za k = <Z a k =
¼ @1 exp dm A
1 1
dl ; exp
: 1l ; : m ;
0 1
0 8 l 9 8m 91
< Za k = <Z a k =
@1 exp dm A
2 2
dl ; exp
: 1l ; : m ;
0 1
0 8 l 9 8 l 9
< Za k = < Za k =
¼ @1 exp
1 2
dl exp dl ;
: 1l ; : 1l ;
0 0
8 m 9 8 m 91
<Z a k = <Z a k =
dm A
1 2
exp d m exp
: m ; : m ;
1 1
0 8 l 9 8m 91
< Z a k þk = <Z a 1 ð1 ðk þ k ÞÞ =
¼ @1 exp dm A
1 2 1 2
dl ; exp
: 1l ; : m ;
0 1
¼ ðk 1 þ k 2 Þa
which completes the proofs of the conclusions (1) and (2).

Pn Pn
Theorem 3.8 (Lei and Xu 2015c) If i¼1 fi ðlÞ; 1 i¼1 ð1 gi ðmÞÞ is still an
IFF, then
Zb ! Zb
X
n X
n
n
fi ðlÞ; 1 ð1 gi ðmÞÞ dX ¼ ðfi ðlÞ; gi ðmÞÞdX
i¼1
i¼1 i¼1
a a
Proof Based on the integrals of IFFs and the operations of IFNs, we have
Zb !
X
n X
n
fi ðlÞ; 1 ð1 gi ðmÞÞ dX
i¼1 i¼1
a
0 8 9 8 91
> > P
n
> Pn
>
< Zlb>
= > Zmb ð1 gi ðmÞÞ >
fi ðlÞ
< >
=C
B
¼B@ 1 exp i¼1
dl ; exp i¼1
dm C
A
>
> 1l >
> >
> m >
>
: la ; : ma ;
0 8 9 8 mb 91
Yn < Z b f ðlÞ >
> l
= Y n < Z ð1 g ðmÞÞ =
B i i C
i¼1
>
: 1 l >
; i¼1 : m ;
la ma
Zb
n
¼ ðfi ðlÞ; gi ðmÞÞdX
i¼1
a
which completes the proof of this theorem. ■

Theorem 3.9 (Lei and Xu 2015c) If u1 ðXÞ ¼ ðf1 ðlÞ; g1 ðmÞÞ and u2 ðXÞ ¼
ðf2 ðlÞ; g2 ðmÞÞ be two IFFs and u1 ðXÞ u2 ðXÞ, which means f1 ðlÞ f2 ðlÞ and
g1 ðmÞ g2 ðmÞ, then
Zb Zb
u 1 ð XÞ dX u 2 ð XÞ dX
a a
Proof According to the definition of “ ” in Chap. 1, we need to prove the fol-

lowing inequalities:
n R o n R o
l 1 ðlÞ l 2 ðlÞ
(1) 1 exp l b f1l dl 1 exp l b f1l dl :
nR a
o nR oa
mb 1g1 ðmÞ mb 1g2 ðmÞ
(2) exp ma m dm exp ma m dm :
Rb Rb
which are obvious, and therefore, the inequality a u 1 ð XÞ dX a u 2 ð XÞ dX
holds. ■
Theorem 3.10 (Lei and Xu 2015c) Let a, b and c be three IFNs, which satisfy
a E b E c, then
Zb Zc Zc
u ð XÞ dX u ð XÞ dX ¼ u ð XÞ dX
a b a
Proof Based on the formula of integral of the IFF, we can easily get
Zb Zc
u ðXÞdX uðXÞ dX
a b
0 8 9 8 mb 91
>
< Zlb >
= < Z
B f ðlÞ 1 gðmÞ =C
>
: 1l > ; : m ;
la ma
0 8 9 8 91
< Z c f ðlÞ >
> l
= < Z c 1 gðmÞ >
> m
=
B C
@1 exp dl ; exp dm A
>
: 1l > ; >
: m >
;
lb mb
0 8 0 19
>
< Zlb Zlc >
B B f ðlÞ f ðlÞ C=
¼ @1 exp @ dl þ dlA ;
>
: 1l 1l >
;
la lb
8 91
>
<Z b 1 gðmÞ
m Zmc >
1 gðmÞ =C
exp dm þ dm A
>
: m m >
;
ma mb
0 8 9 8 mc 91
< Z c f ðlÞ >
> l
= < Z 1 gðmÞ =
B C
>
: 1l > ; : m ;
la ma
Zc
¼ uðXÞdX
a
Furthermore, we can prove the conclusion according to the definition of integral

as follows:
Rb
As mentioned before, the integral of a u ð XÞ dX is acquired by infinitely
dividing an IFIC I a b linking between a and b. In addition, we divide an IFIC I b c ,
Rc
which links between b and c, to get b u ð XÞ dX. Then, by connecting the end point
b of I a b to the starting point b of I b c , we can get a new curve. Moreover, according
to the definition of IFIC and a E b E c, we get that the new curve is still an IFIC,
which is denoted by I a c . Hence, we can define the integral of IFFs along I a c below:
Zc Z
uðXÞdX ¼ uðXÞdX
Ia c
a

k
¼ lim ðuðni Þ DXi Þ
DX1 ;DX 2 ;
;DX k !O i¼1

ð1 Þ j k
¼ lim ðuðni Þ DX i Þ ðuðni Þ DXi Þ
DX1 ;DX2 ;
;DXk !O i¼1 i¼j þ 1

ð2 Þ j
¼ lim ðuðni Þ DXi Þ
DX1 ;DX2 ;
;DXj !O i¼1

k
lim ðuðni Þ DXi Þ
DX j þ 1 ;DXj þ 2 ;
;DXk !O i¼j þ 1
Z Z
¼ uðXÞdX u ðXÞdX
Ia b I bc
Zb Zc
¼ u ðXÞdX uðXÞdX
a b
In the above process, the equality (1) holds because the addition “” meets the
associative law. In addition, when considering the equality (2), we assume that DXt
_
(1 t j) actually represents the difference between ht þ 1 and ht , which are just the
_
_
extreme points of the arc hk hk þ 1 , and let all hk hk þ 1 (1 t j) be included in the
IFIC I a b , and hk hk þ 1 ( j þ 1 t k) in I a c . In summary, Theorem 3.10 holds. ■
3.3 Fundamental Theorem of Intuitionistic

Fuzzy Calculus
After getting the indefinite integrals and the definite integrals of IFFs, this section
shows their relationship. Inspired by the traditional mathematical analysis, we give
the following definition:
Definition 3.2 (Lei and Xu 2015c) Let u ðdÞ be an IFF, and a be an intuitionistic
fuzzy constant. If there exists a variable X, which satisfies a E X, then we call
ZX
UðXÞ ¼ u ðdÞ dd
a
a definite integral of IFF with the variable upper limit (VUL-IFF).

RX
Theorem 3.11 (Lei and Xu 2015c) If a u ðdÞ dd is a VUL-IFF, then
0 X 1
Z
d @
u ðdÞ dd A ¼ uðXÞ
dX
a
RX
which means that a u ðdÞ dd is a primitive function of uðXÞ.
Proof By the definitions of the derivative and the definite integral of IFFs, we have
0 X 1 0 8 9 8 m 91
Z >
< ZlX f ðlÞ >= < Z X 1 gðmÞ =
d @ d B C
uðdÞdd A ¼ @1 exp dl ; exp dm A
dX dX >
: 1l > ; : m ;
a la ma
0 ( ) ( ) 1
RX
l
f ðlÞ RmX 1gðmÞ
Bð1 lX Þ exp 1l dl mX exp m dm C
B la f ðlX Þ ma 1 gðmX ÞC
¼B
B ( ) ;1 ( ) C
C
@ lRX f ðlÞ 1 lX RmX 1gðmÞ mX A
exp 1 - l dl exp m dm
la ma
¼ ðf ðlX Þ; gðmX ÞÞ
¼ uðXÞ

Theorem 3.12 (Lei and Xu 2015c) Let a, b 1 and b 2 be three IFNs, which meet
RX
a E b 1 E b 2 , and UðXÞ ¼ a u ðdÞ dd be a VUL-IFF, then
Zb 1 Zb 2
u ðdÞ dd E u ðdÞ dd
a a
Proof By Theorem 3.6 and Theorem 3.10, we have
Zb 2 Zb 1 Zb 2
u ðdÞ dd u ðdÞ dd ¼ u ðdÞ dd
a a b1
Rb Rb
which is still an IFN. It means that there is a 1 u ðdÞ dd E a 2 u ðdÞ dd , which
indicates Uðb 1 Þ E Uðb 2 Þ if only b 1 E b 2 . Hence, UðXÞ is a monotonically
increasing IFF, which is a very important property of UðXÞ in studying the
derivatives and the differentials of IFFs. ■
Theorem 3.12 manifests an essential fact that for any given IFF u ðdÞ, which is
not necessarily a monotonically increasing IFF, we can structure a monotonically
3.3 Fundamental Theorem of Intuitionistic Fuzzy Calculus 61
RX
increasing IFF UðXÞ ¼ a u ðdÞ dd by u ðdÞ. For example, it can be transferred
into uðXÞ ¼ kX C and uðXÞ ¼ kX C, which are both monotonically increas-
ing IFFs. Moreover, these two monotonically increasing IFFs are actually the
RX
primitive functions of u ðdÞ ¼ ðk; 1 kÞ, that is a ðk; 1 kÞ dd.
Below we present the fundamental theorem of intuitionistic fuzzy calculus,
which is the Newton-Leibniz formula.
Theorem 3.13 (Lei and Xu 2015c) Let WðXÞ be a primitive function of uðXÞ, then
Zb
u ðXÞ dX ¼ WðbÞWðaÞ
a
Proof By the definition of the definite integral of the IFF, let X ¼ ðlX ; mX Þ be a
RX
variable, a ¼ ðla ; ma Þ be an intuitionistic fuzzy constant, and UðXÞ ¼ a u ð dÞ dd
be a VUL-IFF, then
0 8 9 8 m 91
ZX < Zl X f ðlÞ >
> = < Z X 1 gðmÞ =
B C
UðXÞ ¼ u ðdÞ dd ¼ @1 exp dl ; exp dm A
>
: 1l > ; : m ;
a la ma
Thus, UðXÞ is a primitive function of uðXÞ. Moreover, due to that WðXÞ is a

primitive function of uðXÞ, then it has the following form:
0 8 9 8 m 91
>
< Zl X >
= < ZX
WðXÞ ¼ @1 k1 exp dl ; k 2 exp dm A
>
: 1l > ; : m ;
la ma
and then
0 8 9 8m 91
>
< Zl X >
= < ZX
WðXÞWðaÞ ¼ @1 k1 exp dl ; k2 exp dm A
>
: 1l > ; : m ;
la ma
ð1 k 1 ; k 2 Þ
ZX
¼ U ðXÞ ¼ uðdÞ dd
a
Rb
In addition, if we let X ¼ b, then U ðbÞ ¼ a u ðdÞ dd ¼ WðbÞWðaÞ, which is
just the Newton-Leibniz formula of intuitionistic fuzzy calculus. ■
Next, we give several examples below (Lei and Xu 2015c):
(1) Let uðXÞ ¼ ðx ; 1 xÞ, and 0 x 1, then
8Z
> x
>
< dl ¼ x ln ð1 lÞ þ c1
1l
Z ) WðXÞ ¼ ð1 ~c1 ð1 lÞx ; ~c2 vx Þ
>
> x
: dm ¼ x ln m þ c2
m
Z Þ
ð0:5;0:5
uðXÞdX ¼ ð1 ~c1 ð1 0:5Þx ; ~c2 0:5x Þð1 ~c1 ð1 0Þx ; ~c2 1x Þ

ð0; 1Þ
¼ ð1 0:5x ; 0:5x Þ ¼ ð1 ð1 0:5Þx ; 0:5x Þ ¼ xð0:5; 0:5Þ
(2) Let uðXÞ ¼ X, then we have
8Z
> l
>
< dl ¼ ð1 lÞ ln ð1 lÞ þ c1
1l
Z ) WðXÞ
>
> 1m
: dm ¼ ln m m þ c2
m
1l m
¼ 1 ~c1 ; ~c2
expf1 lg expf mg
Z Þ
ð1=2;1=2
1=2 1=2 1 1
uðXÞdX ¼ 1 ~c1 ; ~c2 1 ~c1 ; ~c2
expf1=2g expf1=2g e e
ð0; 1Þ

1 pffiffiffi 1 pffiffiffi
¼ 1 e; e
2 2
and
Z Þ
ð1=3;1=3
2=3 1=3 1 1
uðXÞdX ¼ 1 ~c1 ; ~c2 1 ~c1 ; ~c2
expf2=3g expf1=3g e e
ð0; 1Þ

2 1 1 2
¼ 1 exp ; exp
3 3 3 3
Furthermore, we utilize the Newton-Leibniz formula to prove the following

theorem:
3.3 Fundamental Theorem of Intuitionistic Fuzzy Calculus 63
Theorem 3.13 (Lei and Xu 2015a) Let XðtÞ be a derivable IFF, and satisfy
XðaÞ ¼ a and XðbÞ ¼ b. Then
Zb Zb
uðXÞdX ¼ uðXðtÞÞX 0 ðtÞdt
a a
where uðXðtÞÞX 0 ðtÞ represents uðXðtÞÞ dXðtÞ

dt .
Proof Based on the Newton-Leibniz formula, there is
Zb
u ðXÞ dX ¼ WðbÞWðaÞ
a
where W is a primitive function of u. Denoting UðtÞ ¼ WðXðtÞÞ, then we can use

the chain rule of intuitionistic fuzzy derivative to calculate the following equations:
dUðtÞ dWðXðtÞÞ dXðtÞ dXðtÞ

¼ ¼ uðXðtÞÞ
dt dXðtÞ dt dt
Hence, UðtÞ is also a primitive function of uðXðtÞÞX 0 ðtÞ, and thus,
Zb
uðXðtÞÞX 0 ðtÞdt ¼ UðbÞUðaÞ
a
Moreover, since UðtÞ ¼ WðXðtÞÞ, XðaÞ ¼ a and XðbÞ ¼ b, we have
UðbÞUðaÞ ¼ WðXðbÞÞWðXðaÞÞ ¼ WðbÞWðaÞ
Rb Rb
Hence, uðXÞdX ¼ WðbÞWðaÞ ¼ UðbÞUðaÞ ¼ uðXðtÞÞX 0 ðtÞdt
a a
3.4 Application of the Definite Integrals of IFFs
In this section, we apply the definite integrals of IFFs to aggregate information or

data in the intuitionistic fuzzy environment.
3.4.1 Aggregating Operator Based on the Definite Integrals

of IFFs
Before using the definite integrals of IFFs, it is necessary to make further expla-
nations about it. Firstly, we review the definite integrals of real functions. For
example, how to obtain the distance S when the velocity vðtÞ, which depends on the
moment t, is it known from t0 to t1 ? By the related knowledge of mathematical and
physics, we can get that the distance S is equal to the definite integral of vðtÞ in the
Rt
interval ½t0 ; t1 , that is, S ¼ t01 vðtÞdt. The vðtÞdt of the integral represents the
infinitesimal with respect to the increment of time (dt). Hence, if we want to utilize
Rb
the integral a uðX ÞdX better, it is necessary to understand the special infinitesimal
uðX ÞdX. To this end, we study the multiplication “” between IFNs, whose
properties will be shown below:
Property 3.1 (Lei and Xu 2015c) Let DX, a ¼ ðl; mÞ, and Da0 ¼ ðl0 ; m0 Þ be three
IFNs, then
(1) E DX¼DX:
(2) O DX ¼ O:
(3) If a0 a, that is l0 l and m0 m, then a0 DX a DX:
(4) Specially, if a ¼ ðl; 1 lÞ, then a DX ¼ l DX (ðDX ! OÞ:
By these properties mentioned above, the parameter h of b h DX can be
considered as the support level of b DX. h can be explained as the decision
makers (DMs)’ views whether b should add the increment DX. Obviously, if the
DMs fully agree with b DX, then h will be equal to E and there is
b E DX ¼ b DX. In the case where the DMs are totally against b DX, h
should be taken as O, and b O DX ¼ b. Moreover, thanks to (3) of Property
3.1, there is b h 1 DX b h 2 DX if only U ðh 1 Þ U ðh 2 Þ and
V ðh 1 Þ V ðh 2 Þ.
In what follows, we show several corresponding properties of the definite
integral of IFFs according to Property 3.1.
Property 3.2 (Lei and Xu 2015c).
Rb
(1) a E dd ¼ ba.
Rb
(2) a O dd ¼ O.
(3) If u 1 ðXÞ ¼ ðf1 ðlÞ; g1 ðmÞÞ and u 2 ðXÞ ¼ ðf2 ðlÞ; g2 ðmÞÞ satisfy u 1 ðXÞ u 2 ðXÞ,
which means f1 ðlÞ f2 ðlÞ and g1 ðmÞ g2 ðmÞ, then
Zb Zb
u 1 ðXÞ dX u 2 ðXÞ dX
a a
3.4 Application of the Definite Integrals of IFFs 65
(4) Let u ð XÞ ¼ ðf ðlÞ; gðmÞÞ be an IFF, and 0 k 1, then
Zb
ðk ; 1 kÞ dX ¼ k ðb aÞ
a
Based on the above properties of the integrals of IFFs, we can use the integrals
of IFFs to aggregate IFNs, which is introduced as follows (Lei and Xu 2015c):
Assume that there are n DMs, who are numbered from 1 to n, and they want to
give assessments for an object with the IFNs ai ¼ ðl i ; m i Þ (i ¼ 1; 2; . . .; n) (as
shown in Fig. 3.7 (Lei and Xu 2015c)), where ai is the assessment provided by the
i-th DM. Then, we introduce the following symbols: lmax ¼ maxi fl g,
lmin ¼ mini fl g, mmax ¼ maxi fmg, mmin ¼ mini fmg, b ¼ ðlmax ; mmin Þ and
½O; b ¼ fXjOEXEbg. It is convenient for us to discuss if all assessments are put
into a set. However, it is possible that the DMs give the same assessment (IFN),
which conflicts with the property of a set that cannot have two same elements.
Hence, the following method has been proposed to solve the issue:
Suppose that the i-th DM gives his/her assessment as ðl i ; m i Þ, we define a basic
element ði; l i ; m i Þ of a new set ASS(Lei and Xu 2015c), which is
ASS ¼ fði; li ; m i Þji 2 f1; 2; . . .; ngg
Furthermore, we define two subsets !l and "m of ASS(Lei and Xu 2015c),

which are with respect to l and m:
!l ¼ fði; q; rÞjq [ l; and ði; q; rÞ 2 ASSg
Fig. 3.7 The assessments of

the DMs
each element of which is on the right side of q ¼ l.
"m ¼ fði; q; rÞjr m; and ði; q; rÞ 2 ASSg
whose elements are on top of r ¼ m.

Then we let the function j j : a finite set ! N be the number of the elements,
which are contained in the finite set. Its domain and codomain are the class of all
finite sets and the set of natural numbers, respectively. For example, if we let a finite
set Set ¼ fa; b; c; d; e; f g, then jSetj ¼ 6. Next, we define an IFF with respect to the
variable X ¼ ðl; mÞ as follows (Lei and Xu 2015c):

! l j "m j
CountðXÞ ¼ ðf ðlÞ; gðmÞÞ ¼ ;
jASSj jASSj

where !l and j "m j express a certain extent whether the DMs support to add an

increment to X ¼ ðl; mÞ. If !l is very great, it shows that the majority of the DMs
thinks that l of X is too small, and it should be increased. Moreover, if j "m j is very
m of X is so small and it should not continue to
great, then it reveals that they agree
descend. As has been shown, !l and j "m j contain some information of the
distribution of assessments.
In what follows,
we will provide some analyses to the IFF CountðXÞ. It is clear
that 0 !l =jASSj 1 and 0 j "m j=jASSj 1 hold for any X ¼ ðl; mÞ 2 N.

However, 0 ð !l =jASSjÞ þ ðj "m j=jASSjÞ 1 doesn’t always establish for any
X ¼ ðl; mÞ 2 N, which means that CountðXÞ is meaningless at some points.
Fortunately, there exists a special IFIC I Ob linking between O and b, which is
marked by the dotted line in Fig. 3.8 (Lei and Xu 2015c).
Fig. 3.8 A special IFIC

linking between O and b
It is easy to get that the IFF CountðXÞ is meaningful for every point X 2 I Ob . It
means that we can calculate the definite integral of CountðXÞ along I Ob , that is,
Rb

O CountðXÞdX. Because both !l =jASSj and j "m j=jASSj of the integrand
Rb
CountðXÞ are the piecewise continuous real functions, O CountðXÞdX can be
Rb
easily acquired, therefore, we can utilize O CountðXÞdX to aggregate the
assessments.
3.4.2 Properties of Aggregation Operator Built

by the Definite Integrals of IFFs
In the following, we study the basic properties of the above-mentioned aggregation

method to verify its validity and usefulness, and investigate whether
Rb
O CountðXÞdX satisfies the fundamental properties of the aggregation operators
concerning idempotency, boundedness, monotonicity.
Theorem 3.14 (Idempotency) (Lei and Xu 2015c) If all assessments given by the
DMs are equal to a, then
Zb
CountðXÞdX ¼ a
O
Proof All assessments equaling to a can be shown in Fig. 3.9 (Lei and Xu 2015c).
Fig. 3.9 All assessments are

the same
Based on the figure, it is easy to obtain that CountðXÞ is a piecewise IFF:

8
< E when O E X / c;
CountðXÞ ¼ ð0; 0Þ; when c E X / a;
:
O; when X ¼ a:
According to the definition of the definite integral of the IFF, we can discover that
the integral value will remain unchanged when the function values of IFFs at the
limit numbers of points change. Hence, let CountðXÞ ¼ ð0; 0Þ when X ¼ a, then
Zb Zc Za
CountðXÞdX ¼ CountðXÞdX CountðXÞdX
O O c
Zc Za
¼ EdX ð0; 0ÞdX ¼ a
O c

Theorem 3.15 (Boundedness) (Lei and Xu 2015c) Let ðlmax ; mmin Þ and
ðlmin ; mmax Þ be denoted by a and a þ , respectively, then
Zb

a CountðXÞdX a þ
O
Proof In order to discuss conveniently and visually, we provide a figure (Lei and
Xu 2015c) at first:
According to Fig. 3.10, we can get the piecewise IFF as follows:
8
>
> E; when OEX / c1 ;
>
> j !l j
>
> c1 E X / c2 ;
< jASSj ; 0 ;
> when

CountðXÞ ¼ j ! l j j "m j
>
>
> jASSj ; jASSj ; when c2 E X / c3 ;
>
>
>
>
: 0; j "m j ; when c3 E X E a þ :
jASSj
Fig. 3.10 All assessments

are in the area
½lmin ; lmax ½mmin ; mmax
If we let c1 ¼ ðl1 ; m1 Þ, c2 ¼ ðl2 ; m2 Þ and c3 ¼ ðl3 ; m3 Þ, then

8
> 1; when 0 l \ l1 ;
!l >
<
¼ j !l j ; when l1 l \ l 3 ;
jASSj >
> jASSj
:
0; when l3 l 1:
and
8
>
>0; when m2 \m 1 ;
>
<
j "m j j "m j
¼ ; when mmin \m m2 ;
jASSj > > jASSj
>
:
1 ; when 0 m mmin :
Based on the calculating formula of the definite integral of the IFF, we can get
0 8 9 8m 91
Zb < 1
Zlmax
! l = < Zmin 1 j " j
=
CountðXÞdX ¼ @1 exp dl ; exp 1 m
dm A
: 1 l jASSj ; : m jASSj ;
O 0 1
0 8 9
>
< Zl1 1 Zl3 >
B 1 ! l =
¼ @1 exp dl dl ;
>
: 1l 1 l jASSj > ;
0 l1
8 m 9 1
<Z 2 1 Zmmin =
1 j "m j
exp dm þ 1 dm A
: m m jASSj ;
1 m2
In addition, the following inequalities hold:
8 9
>
< Zl1 1 Zl3 >
1 !l =
l1 1 exp dl dl
>
: 1l 1 l jASSj >;
0 l1
ð1Þ 8 9
>
< Zl1 1 Zl3 >
=
1
1 exp dl dl ¼ l3 :
>
: 1l 1l > ;
0 l1
8 m 9 8 m 9
<Z 2 Zmmin <Z 2 1 Zmmin =
1 1 = 1 j "m j
ð2Þ mmin exp dm þ dm exp dm þ 1 dm m2 :
: m m ; : m m jASSj ;
1 m2 1 m2
Moreover, for l1 ¼ lmin , l3 ¼ lmax , m2 ¼ mmax , a ¼ ðl1 ; m2 Þ and

þ
a ¼ ðl3 ; mmin Þ, we get the inequality below:
0 8 9
>
< Zl1 1 Zl3 >
B 1 !l =
a @1 exp dl dl ;
>
: 1l 1 l jASSj >;
0 l1
8 m 9 1
<Z21 Zmmin =
1 j "m j
exp dm þ 1 dm A a þ
: m m jASSj ;
1 m2
Rb
which is equivalent to a O CountðXÞdX a þ . ■
Theorem
3.16 (Monotonicity) (Lei and Xu 2015c) If we let Count1 ðXÞ ¼

ð !l 1 =jASSj; j"m j1 =jASSjÞ and Count2 ðXÞ ¼ ð !l 2 =jASSj; j"m j2 =jASSjÞ be two

IFFs, which meet !l 1 =jASSj "l 2 =jASSj and j"m j1 =jASSj j"m j2 =jASSj, then
Zb Zb
Count1 ðXÞdX Count2 ðXÞdX
O O
whose proof of Theorem 3.16 is omitted here.

These basic proprieties (idempotency, boundedness and monotonicity) of
Rb
O CountðXÞdX show that it is feasible and reasonable to be considered as an
aggregation method. Based on which, we introduce the process to aggregate the
Rb
IFNs by O CountðXÞdX in detail (Lei and Xu 2015c):
Step 1. According to the assessments provided by the DMs, we obtain the

minimum value of the non-membership degree, mmin , and the maximal
, and let b be ðlmax ; mmax Þ.
lmax
value of the membership degree,
Step 2. Construct two real functions !l =jASSj and j"m j=jASSj to depict the
distribution
of assessments, and let CountðXÞ be equal to
ð !l =jASSj; j"m j=jASSjÞ.
Step 3. Acquire the aggregated value of the assessments by calculating the
Rb
definite integral of CountðXÞ along I Ob , which is O CountðXÞdX.
In the following, we introduce an example to illustrate the aggregating process
Rb
by O CountðXÞdX (Lei and Xu 2015c):
Assume that five DMs give their assessments as ð0:2; 0:3Þ, ð0:4; 0:5Þ, ð0:6; 0:3Þ,
ð0:2; 0:1Þ and ð0:3; 0:3Þ, then we have the following process:
Step 1. It is easy to obtain that jASSj ¼ 5, and the frequency of the assessments
(Lei and Xu 2015c)
can be shown in (Table 3.1)
Step 2. The functions !l =jASSj and j"m j=jASSj can be expressed as:
8
> 1; when 0 l \ 0:2;
> >
!l >
< 0:6; when 0:2 l \ 0:3;
¼ 0:4; when 0:3 l \ 0:4;
jASSj >>
> 0:2; when 0:4
> l \ 0:6;
:
0; when 0:6 l 1;
and
8
>
> 0; when 0:5 \ m 1;
j "m j <
0:2; when 0:3 \ m 0:5;
¼
jASSj > > 0:8; when 0:1 \ m 0:3;
:
1; when 0 m 0:1:
Table 3.1 The frequency of The value of l 0.2 0.3 0.4 0.6
the assessments
Frequency 0.4 0.2 0.2 0.2
The value of m 0.1 0.3 0.5
Frequency 0.2 0.6 0.2
Step 3. Calculate the definite integral by
Zb
CountðXÞdX
O
0 8 0:6 9 8 0:1 91
< Z 1 !l = <Z 1 j " j
=
¼ @1 exp dl ; exp 1 m
dm A
: 1 l jASSj ; : m jASSj ;
0 1
Therefore, the aggregated value is approximately equal to ð0:3598; 0:2667Þ.

As we all know, the weight information is very obbligato and important for the
common aggregation operators, such as the IFWA operator, the IFWG operator,
etc. However, by the definite integrals of IFFs, the IFNs is aggregated by combining
the information of positions of assessments, which doesn’t need the weights of the
IFNs. Hence, the definite integrals of IFFs are totally dependent on the original data,
which can naturally reduce the influence of the DMs’ subjectivity to the final
aggregated results.
3.5 Conclusions
R
In this chapter, we have first studied the indefinite integral uðXÞdX of the IFF
uðXÞ, which is actually the inverse operation of derivative introduced in Chap. 2.
Then we have investigated some of its basic properties. Moreover, we have defined
Rb
the definite integral a uðXÞdX of the IFFs uðXÞ in two different ways, one of
R
which is based on the notion of IFIC that is similar to the integral C f ðzÞdz of the
complex function f ðzÞ along a simple curve C; and another way to obtain the
Rb Rb
definite integral a uðXÞdX of the IFF uðXÞ is more like the integral a f ðxÞdx of
the real function f ðxÞ in a closed interval ½a; b. By defining the definite integral of
RX
the IFF with the variable upper limit a uðdÞdd, we have discovered the rela-
tionship between the indefinite integrals and the definite integrals of IFFs, which is
essentially the fundamental theorem in intuitionistic fuzzy calculus
(Newton-Leibniz formula). Finally, we have successfully applied the definite
integrals of IFFs to aggregate information and data in intuitionistic fuzzy
environment.
Chapter 4
Aggregation Operations of Continuous
Intuitionistic Fuzzy Information
In this chapter, we focus on a problem about how to aggregate the IFNs spreading
all over an area, which means that each point in a two-dimensional plane to be
aggregated is an IFN that we want to aggregate (Fig. 4.1b; Lei and Xu 2016a).
Until now, lots of aggregation techniques have been proposed for dealing with a
limited number of IFNs that take the form of discrete information (Fig. 4.1a; Lei
and Xu 2016a). However, sometimes it not only needs to deal with the discrete
IFNs, but also needs to solve the problems with continuous intuitionistic fuzzy
information in our real life, which likes that we study the discrete-type random
variables, as well as the continuous-type random variables in the probability theory
and the mathematical statistics. Hence, it is meaningful to give a method to cope
with continuous intuitionistic fuzzy information.
In order to get the method to aggregate the continuous information in intu-
itionistic fuzzy environment, we first give some definitions and theorems as follows
(Lei et al. 2015):
Suppose that there is an area D, which satisfies D N, then we call it a region of
IFNs, which can be shown in Fig. 4.2 (Lei et al. 2015).
If we define a real non-negative function f in D as f : D ! R þ [ f 0 g, then the
function can be denoted by f ðl; mÞ (ðl; mÞ 2 D). Moreover, the two-dimensional
point ðl; mÞ 2 D can also be represented as an IFN, therefore, we can denote f ðl; mÞ
by f ðXÞ (X 2 D). It should be noticed that f ðXÞ is not an IFF but a real function
because its function value is not an IFN but a real number.
Theorem 4.1 (Lei et al. 2015) Let a, a 1 , a 2 , a 3 and a 4 be five IFNs, then
(1) If k 1 ; k 2 0 and k 1 [ k 2 , then k 1 a k 2 a and ak 1 ak 2 .
(2) If k 0 and a 1 a 2 , which means that la 1 la 2 and ma 1 ma 2 , then ka 1 ka 2
and a k1 a k2 .
(3) If a 1 a 3 and a 2 a 4 , then a 1 a 2 a 3 a 4 and a 1 a 2 a 3 a 4 .
It is easy to prove this theorem based on the operations of IFNs and the defi-
nition of “ ”. Hence, the proof is omitted here.
74 4 Aggregation Operations of Continuous Intuitionistic …
Fig. 4.1 Discrete information and continuous information
Fig. 4.2 A region of IFNs
4.1 Method Dealing with Continuous Intuitionistic Fuzzy

Information
In order to aggregate the continuous intuitionistic fuzzy information, we utilize the

methods and the thinking of the mathematical analysis, like the concepts of limit
and integral.
Let D be a region of IFNs, and f ðXÞ be a non-negative function defined in
D. Denoting a ¼ ðinf a2D f la g; supa2D f ma gÞ and a þ ¼ ðsupa2D f la g;
inf a2D f ma gÞ, then the method can be shown as follows (Lei et al. 2015):
4.1 Method Dealing with Continuous Intuitionistic Fuzzy Information 75
Step 1. Dividing the region D. We first divide the region of IFNs D into k parts:
8l i 2 ½inf a 2D fla g; supa2D fla g; inf a 2D fla g ¼ l0 \l1 \ \ln ¼ supa2D fla g
8mi 2 ½inf a2D fma g; supa 2D fma g; inf a2D fma g ¼ l0 \l1 \ \ln ¼ supa 2D fma g
Then we can denote the k small regions by di ; i ¼ 0 ; 1 ; . . .; k and let
d¼ max f su pf d ðx; yÞ : x; y 2 di gg
0 i k
Step 2. Making the product. We randomly choose fi ¼ ðni ; gi Þ from di to acquire

its function value f ðfi Þ, which is shown in Fig. 4.3. Next, we multiply fi ¼ ðni ; gi Þ
by the real value f ðfi ÞDdi , where Ddi is the area value of di . Then we can get an IFN
f ðfi Þfi Ddi , which is f ðni ; gi Þðni ; gi ÞDdi .
Step 3. Calculating the sum. We add k IFNs f ðfi Þfi Ddi (i ¼ 0 ; 1 ; . . . ; k) together
k
to acquire f ðfi Þfi Ddi
i¼1
Step 4. Taking the limit. By subdividing D infinitely, we get the limit of
k
lim f ðfi Þfi Ddi .
d!0 i¼1
k
Definition 4.1 (Lei et al. 2015) If the limit value of lim f ðfi Þfi Ddi exists, then
d!0 i¼1
we
RR call it the integral aggregating value of the region D, and denote it as
D f ðXÞX dd:
In what follows, we study the lower and the upper Darboux sums of the integral
aggregating value of D. In every small region di , we have the following conclusions
(Lei et al. 2015):
(1) For any a 2 di , there must be ai0 a ai1 , where ai0 and ai1 are respectively the
maximum and minimum IFNs in the region di based on the order relation “ ”,
which is shown in Fig. 4.4.
i
(2) Let fsup i
and finf be the supremum and the infimum of f ðXÞ in di , respectively.
Then according to Theorem 4.1, we can get
Fig. 4.3 Dividing the region

and randomly choosing a
point
Fig. 4.4 The maximum and

the minimum IFNs in the
region
i i
finf a0 Ddi f ðfi Þfi Ddi fsup
i
ai1 Ddi
and
k k k
finf
i i
a0 Ddi f ðfi Þfi Ddi fsup
i
ai1 Ddi
i¼1 i¼1 i¼1
k
Definition 4.2 (Lei et al. 2015) Let finf
i i
a0 Ddi be the lower Darboux sum of the
i¼1
k
integral aggregating value of D, and fsup
i
ai1 Ddi be the upper Darboux sum.
i¼1
Theorem 4.2 (Lei et al. 2015) If the limit value of the lower Darboux sum is equal
k
to one of the upper Darboux sum, then lim f ðfi Þfi Ddi exists.
d!0 i¼1
k
Proof By the squeeze theorem, there is a guarantee that lim f ðfi Þfi Ddi exists
d!0 i¼1
n n
and is equal to lim i i
finf a0 Ddi and lim i
fsup ai1 Ddi .
d!0 i¼1 d!0 i¼1
RR
Next, we give the calculating formula of D f ðXÞX dd as follows:
Theorem 4.3 (Lei et al. 2015) Let D be a region RR of IFNs, and f ðXÞ be a
non-negative real function in D, then the value of D f ðXÞX dd is also an IFN,
which can be expressed as:
0 8 9 8 91
ZZ < ZZ = < ZZ =
f ðXÞX dd ¼ @1 exp f ðl; mÞ ln ð1 lÞ dd ; exp f ðl; mÞ ln m dd A
: ; : ;
D D D
Proof In order to prove this theorem, we divide D into the following two categories
at first:
4.1 Method Dealing with Continuous Intuitionistic Fuzzy Information 77
Fig. 4.5 Two kinds of

regions of IFNs
(1) D \ f m ¼ 0 g ¼
, such as D1 in Fig. 4.5.
(2) D \ f m ¼ 0 g 6¼
, such as D2 in Fig. 4.5.
For the first category, we prove Theorem 4.3 when D meets
D \ f m ¼ 0 g ¼
, which means that D belongs to the category such like D1 . By
the calculating formula of the IFWA operator, we have
!
n Y
n
f ðni ;gi Þ Ddi
Y
n
f ðni ;gi Þ Ddi
f ðni ; gi Þðni ; gi ÞDdi ¼ 1 ð1 ni Þ ; gi
i¼1
i¼1 i¼1
then
!
n Y
n
f ðni ;gi Þ Ddi
Y
n
f ðn ;g Þ Dd
lim f ðni ; gi Þðni ; gi ÞDdi ¼ 1 lim ð1 ni Þ ; lim gi i i i
d!0 i¼1 d!0 d!0
i¼1 i¼1
The membership degree and the non-membership degree of the above expres-
sion can be respectively investigated as:
( )
Y
n
f ðn ;g Þ Dd
X
n
ð1Þ lim gi i i i ¼ lim exp f ðni ; gi Þ ln gi Ddi
d!0 d!0
i¼1 ( i¼1 )
Xn
¼ exp lim f ðni ; gi Þ ln gi Ddi
d!0
8 i¼1 9
<ZZ =
¼ exp f ðl; mÞ ln m dd
: ;
D
where gi ði ¼ 1; 2; . . .; nÞ is not equal to zero for D \ f m ¼ 0 g ¼

.
( )
Y
n
f ðni ;gi Þ Ddi
X
n
ð2Þ 1 lim ð1 ni Þ ¼ 1 lim exp f ðni ; gi Þ ln ð1 ni ÞDdi
d!0 d!0
i¼1 i¼1
( )
X
n
¼ 1 exp lim f ðni ; gi Þ ln ð1 ni ÞDdi
d!0
i¼1
8 9
<ZZ =
¼ 1 exp f ðl; mÞ ln ð1 lÞdd
: ;
D
where ni ði ¼ 1; 2; . . .; nÞ is not equal to one since D \ f m ¼ 0 g ¼

. Thus,
ZZ
n
f ðXÞX dd ¼ lim f ðni ; gi Þðni ; gi ÞDdi
d!0 i¼1
D
0 8 9 8 91
<ZZ = <ZZ =
¼ @1 exp f ðl; mÞ ln ð1 lÞ dd ; exp f ðl; mÞ ln m dd A
: ; : ;
D D
In what follows, we analyze the situation that likes D 2 , which is

D \ f m ¼ 0 g 6¼
. This situation
RR contains some singular pointsRR in D.
Because the integrands of D f ðl; mÞ ln ð1 lÞ dd and D f ðl; mÞ ln m dd are
meaningless when l ¼ 1 and m ¼ 0, the two definite integrals are improper inte-
grals, and the results of which may be negative infinity, when integrals are diver-
gence. What’s more, we study the four possible cases, which are shown in
Table 4.1 (Lei et al. 2015).
In Table 4.1, there exists an impossible situation (Lei et al. 2015), listed as
follows: RR RR
For 0 l þ m 1, we know that D f ðl; mÞ ln m dd D f ðl; mÞ ln ð1 lÞ dd.
RR RR
Hence, it is impossible that D f ðl;
RR
mÞ ln m dd [ 1 when
RRD
f ðl; mÞ
ln ð1 lÞdd ! 1, which means that D f ðl; mÞ ln mdd [ 1 and D f ðl; mÞ
ln ð1 lÞdd ! 1 do not occur simultaneously.
In the following, the other three situations can be discussed as:
RR RR
(1) If both D f ðl; mÞ ln ð1 lÞdd and D f ðl; mÞ ln mdd are convergent, then
0 8 9 8 91
ZZ < ZZ = < ZZ =
: ; : ;
D D D
Table 4.1 Four possible situations
RR RR
f ðl; mÞ ln m dd [ 1
D f ðl; mÞ ln mdd ! 1
D
RR RR RR RR
f ðl; mÞ ln ð1 lÞdd [ 1 f ðl;mÞ ln ð1lÞ dd f ðl;mÞ ln m dd f ðl;mÞ ln ð1lÞ dd
D 1e D ;e D 1e D ;0
RR
D f ðl; mÞ ln ð1 lÞdd ! 1 An impossible situation (1, 0)
4.1 Method Dealing with Continuous Intuitionistic Fuzzy Information
79
RR RR
(2) When f ðl; mÞ ln ð1 lÞ dd is convergent and D f ðl; mÞ ln m dd is diver-
D
RR
gent, we define exp f ðl; mÞ ln m dd ¼ e1 ¼ 0, then
D
0 8 9 1
ZZ < ZZ =
f ðXÞX dd ¼ @1 exp f ðl; mÞ ln ð1 lÞ dd ; 0A
: ;
D D
RR RR
(3) If both D f ðl; mÞ ln ð1 lÞ dd and D f ðl; mÞ ln mdd are divergent, then 1

RR RR
exp f ðl; mÞ ln ð1 lÞ dd ¼ 1 e 1 ¼ 1 and exp f ðl; mÞ ln m dd ¼
D RR D
e1 ¼ 0, which means D f ðXÞX dd ¼ E ¼ ð1; 0Þ.
In summary, the following expression always holds:
0 8 9 8 91
ZZ < ZZ = < ZZ =
: ; : ;
D D D
Now we prove that it is still an IFN. Significantly, 0 l 1, 0 m 1 and

0 l þ m 1 since ðl; mÞ is an IFN. Then, according to the properties of the expo-
nential functions and the logarithmic functions, we can get the following conclusions:
RR
(1) 0 1 exp D f ðl; mÞ lnð1 lÞdd 1.
RR
(2) 0 exp D f ðl; mÞ ln mdd 1.
RR RR
(3) 0 1 exp D f ðl; mÞ ln ð1 lÞdd þ exp D f ðl; mÞ ln mdd 1.
RR
Hence, the result of D f ðXÞX dd is still an IFN.
RR k
In addition, we know D f ðXÞX dd ¼ lim f ðfi Þfi Ddi . It is clear to get that
RR d!0 i¼1
D f ðXÞX dd must be an IFN for the closure of the operations (ka and ) of IFNs,
which means that ka and a b are still IFNs if only a and b are IFNs, and k is a
real number lying in the interval [0, 1]. ■
4.2 Properties of Integral Aggregating Value
In this section, we study the properties of the integral aggregating value of some
regions.
S T
Theorem 4.4 (Lei et al. 2015) Let D ¼ ni¼1 Di , and Di Dj ¼
if i 6¼ j, and
f ðXÞ be a non-negative real function, then
4.2 Properties of Integral Aggregating Value 81
ZZ ZZ ZZ ZZ
f ðXÞX dd ¼ f ðXÞX dd f ðXÞX dd f ðXÞX dd
D D1 D2 Dn
RR n RR
which means that D f ðXÞX dd ¼ Di f ðXÞX dd.
i¼1
Proof Firstly, when k ¼ 2, we can prove the theorem below:

ZZ ZZ RR RR
f ðl;mÞ lnð1lÞdldm f ðl;mÞ ln mdldm
f ðXÞX dd f ðXÞX dd ¼ 1 e D1 ; e D1
D1 D2
RR RR
1e D2
; e D2
RR RR
¼ 1e D1 [ D2
; e D1 [ D2
ZZ
¼ f ðXÞX dd
D1 [ D2
Suppose that when k ¼ n 1, the theorem holds, which means

ZZ ZZ
n1
f ðXÞX dd ¼ f ðXÞX dd
i¼1
[ ni ¼11 Di Di
Then, when k ¼ n, we have

ZZ ZZ ZZ
f ðXÞX dd f ðXÞX dd f ðXÞX dd
D1 D2 Dn
ZZ ZZ
n1
¼ f ðXÞX dd f ðXÞX dd
i¼1
Di Dn
RR RR
[ n 1 Di [ n 1 Di
¼ 1e i ¼ 1 ; e i ¼ 1
RR RR
1 e Dn ; e Dn
RR RR
¼ 1e D ; e D
ZZ
¼ f ðXÞX dd
D

Theorem 4.5 (Lei et al. 2015) If there are two regions of IFNs, D1 and D2 , which
satisfy D2 D1 , then
ZZ ZZ ZZ
f ðXÞX dd f ðXÞX dd ¼ f ðXÞX dd
D1 D2 D1 D2
where D1 D2 ¼ fXjX 2 D1 ; X 62 D2 g.
Proof Based on the subtraction of the IFNs, we can calculate the expression as:
ZZ ZZ
f ðXÞX dd f ðXÞX dd
D1 D2
0 RR RR RR 1
f ðl;mÞ lnð1lÞdldm f ðl;mÞ lnð1lÞdldm f ðl;mÞ ln mdldm
e D2
e D1
e D1
¼@ RR ; RR A
e D2
e D2
RR RR
¼ 1e D1 D2
; e D1 D2
ZZ
¼ f ðXÞX dd
D1 D2

Theorem 4.6 (Lei et al. 2015) Let f ðXÞ and gðXÞ be two non-negative functions of
D, and x 1 0, x 2 0, then
ZZ ZZ ZZ
ðx 1 f ðXÞ þ x 2 gðXÞÞXdd ¼ x 1 f ðXÞXdd x 2 gðXÞXdd
D D D
Proof By the formula of integral aggregating value, we have

ZZ
ðx 1 f ðXÞ þ x 2 gðXÞÞXdd
D
RR RR
ðx 1 f ðl;mÞ þ x 2 gðl;mÞÞ lnð1lÞdldm ðx f ðl;mÞ þ x gðl;mÞÞ ln mdldm
¼ 1e D ; e D 1 2
RR RR
x f ðl;mÞ lnð1lÞdldm x 2 gðl;mÞ lnð1lÞdldm
¼ 1e 1 D e D ;
RR RR
x f ðl;mÞ ln mdldm x 2 gðl;mÞ ln mdldm
e 1 D e D
RR RR
f ðl;mÞ lnð1lÞdldm x 1 f ðl;mÞ ln mdldm x 1
¼ 1 ðe D Þ ; ðe D Þ
RR RR
gðl;mÞ lnð1lÞdldm x2 gðl;mÞ ln mdldm x2
1 ðe D Þ ; ðe D Þ
RR RR
¼ x1 1 e D ; e D
RR RR
gðl;mÞ lnð1lÞdldm gðl;mÞ ln mdldm
x2 1 e D ; e D
ZZ ZZ
¼x1 f ðXÞXdd x 2 gðXÞXdd
D D
Hence, Theorem 4.6 holds. ■

In addition, if there are n non-negative functions of D, which are fi ðXÞ ði ¼
1; 2; . . .; nÞ and x i 0 i ¼ 1; 2; . . .; n then
ZZ ! ZZ
X
n
n
x i fi ðXÞ Xdd ¼ x i fi ðXÞXdd
i¼1
i¼1
D D
Theorem 4.7 (Lei et al. 2015) Let f ðXÞ and gðXÞ be two non-negative functions of
D, which meet f ðXÞ gðXÞ, then we can know that f ðXÞ gðXÞ is still a
non-negative function of D, in addition, we have
ZZ ZZ ZZ
ðf ðXÞ gðXÞÞX dd ¼ f ðXÞX dd gðXÞX dd
D D D
Proof By the subtraction of the IFNs, we have

ZZ ZZ
f ðXÞX dd gðXÞX dd
D D
RR RR
¼ 1e D ; e D
RR RR
1e D ; e D
RR RR RR !
gðl;mÞ lnð1lÞdldm f ðl;mÞ lnð1lÞdldm f ðl;mÞ ln mdldm
e D e D e D
¼ RR ; RR
e D e D
RR RR
ðf ðl;mÞgðl;mÞÞ lnð1lÞdldm ðf ðl;mÞgðl;mÞÞ ln mdldm
¼ 1e D ; e D
ZZ
¼ ðf ðXÞ gðXÞÞX dd
D
which shows that the theorem holds. ■

Theorem 4.8 (Lei et al. 2015) Let p [ 1, 1=p þ 1=q ¼ 1, f ðXÞ and gðXÞ be two
non-negative functions, then
ZZ ZZ ZZ
f p ðXÞ g q ðXÞ
f ðXÞgðXÞXdd Xdd Xdd
p q
D D D
Specially, if p ¼ q ¼ 2, then
ZZ RR RR
f 2 ðXÞXdd g2 ðXÞXdd
f ðXÞgðXÞXdd D D
2
D
Proof Because f ðXÞ 0 and gðXÞ 0 for any X 2 D, then according to Young
inequality, we can get
f p ðXÞ g q ðXÞ
f ðXÞgðXÞ þ
p q
and
f p ðXÞ g q ðXÞ
f ðXÞgðXÞ ln m ln m þ ln m
p q
Hence, there is
ZZ ZZ ZZ
f p ðXÞ g q ðXÞ
f ðXÞgðXÞ ln mdd ln mdd þ ln mdd
p q
D D D
and then
RR RR f p ðXÞ
RR g q ðXÞ
f ðXÞgðXÞ ln mdd p ln mdd q ln mdd
e D e D e D
In the same way, we also can get

RR RR f p ðXÞ
RR g q ðXÞ
f ðXÞgðXÞ ln ð1lÞdd p ln ð1lÞdd q ln ð1lÞdd
1e D 1e D e D
Therefore, the following inequality holds:

ZZ ZZ ZZ
f p ðXÞ g q ðXÞ
f ðXÞgðXÞXdd Xdd Xdd
p q
D D D
The proof of this theorem is completed. ■

Theorem 4.9 (Lei et al. 2015) Let f ðXÞ be a continuous non-negative function,
then
ZZ
f ðXÞXdd ¼ O , f ðXÞ ¼ 0
D
Proof For f ðXÞ is continuous, we have

ZZ RR RR
f ðl;mÞ ln ð1lÞ dd f ðl;mÞ ln m dd
f ðXÞXdd ¼ 1 e D ; e D ¼O
D
RR RR
,1e D ¼ 0 and e D ¼1
, f ðXÞ ¼ 0
which shows that Theorem 4.9 holds. ■

Theorem 4.10 (Lei et al. 2015) Let f ðXÞ and gðXÞ be two non-negative functions
of D, which satisfy f ðXÞ gðXÞ for any X 2 D, then
ZZ ZZ
f ðXÞXdd gðXÞXdd
D D
Proof By the properties of the exponential functions and the logarithmic functions,
we can get
RR RR
f ðl;mÞ ln ð1lÞ dd g ðl;mÞ ln ð1lÞ dd
e D
(1) 1 RR RR 1 e D .
f ðl;mÞ ln m dd g ðl;mÞ ln m dd :
(2) e D e D
RR RR
which means that D f ðXÞXdd D gðXÞXdd. ■
Theorem 4.11 (Lei et al. 2015) Let D 2 be a subarea of D 1 , which means D 2 D 1 ,

and the function f ðXÞ be a non-negative functions of D 1 , then
ZZ ZZ
f ðXÞXdd f ðXÞXdd
D1 D2
Proof Since
RR RR
f ðl;mÞ ln m dd f ðl;mÞ ln m dd
(1) e D1
RR e D2 RR.
f ðl;mÞ ln ð1lÞ dd f ðl;mÞ ln ð1lÞ dd
(2) 1 e D1
1 e D2
.
Then, we can get
ZZ RR RR
f ðXÞXdd ¼ 1 e D1 ; e D1
D1
RR RR ZZ
1 e D2 ; e D2 ¼ f ðXÞXdd
D2
4.3 Application of the Integral Aggregating Value
At the beginning of this section, we put forward a question, which is how to

aggregate all points in D, which is shown in Fig. 4.1. Below we apply the notion of
the integral aggregating value to build an aggregating operator:
RR of IFNs. If f ðl; mÞ ði.e. f ðXÞÞ is a

Definition 4.3 (Lei et al. 2015) Let D be a region
non-negative real function in D, and satisfies D f ðl; mÞ dd ¼ 1, then we call it a
weight density function of D.
Theorem 4.12 (Lei et al. 2015) Let D be a region of IFNs, and f ðXÞ be a weight
density function. Then, we have an aggregation operator as follows:
0 8 9 8 91
ZZ < ZZ = < ZZ =
: ; : ;
D D D
which is called an intuitionistic fuzzy P integral aggregating (IFIA) operator.

Specially, if f ðXÞ ¼ dðX; a1 ; a2 ; . . .; an Þ ¼ nk ¼ 1 xk dðX ak Þ, where
4.3 Application of the Integral Aggregating Value 87
Fig. 4.6 The region of the

IFNs is only a point
(1) X ak of dðX ak Þ does not represent the difference X ak between two

IFNs, which has been mentioned in Sect. 4.1, but only depicts whether X is
equal to ak . Moreover, dðX ak Þ is defined as:

0; when X 6¼ ak
dð X ak Þ ¼
þ 1; when X ¼ ak
RR
which satisfies that R 2 dðX ak Þdd 1.
Pn
(2) k¼1 x k ¼ 1 and x k 0 ðk ¼ 1; 2; . . .; nÞ
then the IFIA operator reduces to the intuitionistic fuzzy weighted averaging
(IFWA) operator.
Proof According to the definition of dðX; a1 ; a2 ; . . .; an Þ, we can get

0; when X 62 fa1 ; a2 ; . . .; an g
(1) dðX; a1 ; a2 ; . . .; an Þ ¼ .
þ 1; when X 2 fa1 ; a2 ; . . .; an g
RR Pn
(2) R2 dðX; a1 ; a2 ; . . .; an Þ dd ¼ k¼1 xk ¼ 1.
(3) If gðXÞ is a continuous function, then
ZZ X
n
gðXÞ dðX; a1 ; a2 ; . . .; an Þ dd ¼ x k gðak Þ
k¼1
R2
Hence, we know that

ZZ RR RR
f ðXÞX dd ¼ 1 e D ; e D
D
0 P
n P
n 1 !
x i ln ð1l i Þ x i ln m i Y
n Y
n
¼ @1 e i ¼ 1 ; e i¼1 A¼ 1 xi
ð1 li Þ ; mx
i
i
i¼1 i¼1
¼ IFWAx ða1 ; a2 ; . . .; an Þ
which shows that the IFIA operator reduces to the IFWA operator when
f ðXÞ ¼ dðX; a1 ; a2 ; . . .; an Þ. ■
In the following, we discuss some basic properties of the IFIA operator,
including idempotency, boundedness, and monotonicity:
Theorem 4.13 (Idempotency) (Lei et al. 2015) Let D be a region of IFNs, which is
shown in Fig. 4.6 (Lei et al. 2015), then the result of the IFIA operator belongs
to D.
Proof By Fig. 4.6 (Lei et al. 2015), we can get the weight density function as:

0; when X¼
6 P
f ðXÞ ¼ dðX; PÞ ¼
þ 1; when X¼P
RR RR
and R2 dðX; PÞ dd ¼ N dðX; PÞ dd ¼ 1. In addition, there are
ZZ ZZ
dðX; PÞ lnð1 lÞdd ¼ lnð1 l 0 Þ and dðX; PÞ ln m dd ¼ ln m0
R2 R2
(a) A shift vector (b) A moderate shift vector
Fig. 4.7 Shift vector of D

Hence, we have
ZZ
f ðXÞXdd ¼ 1 e ln ð1l 0 Þ ; e ln m 0 ¼ ðl 0 ; m 0 Þ ¼ P 2 D
D
which shows that P belongs to D. ■

Theorem 4.14 (Boundedness) (Lei et al. 2015) Let D be a region of IFNs, and
f ðXÞ be a weight density function, then
ZZ

a f ðXÞXdd a þ
D
where a ¼ ðinf a2D f la g; supa2D f ma gÞ and a þ ¼ ðsupa2D f la g; inf a2D f ma gÞ.
Proof Let a ¼ ðl0 ; m0 Þ and a þ ¼ ðl1 ; m1 Þ, then

RR RR
f ðl;mÞ ln ð1lÞ dldm ln ð1l 1 Þ f ðl;mÞ dldm
e D
(1) 1RR RR 1 e D ¼ 1 eln ð1l 1 Þ ¼ l 1 .
f ðl;mÞ ln m dldm ln m 1 f ðl;mÞ dldm
(2) e D e D ¼ m 1.
RR
Hence, we have D f ðXÞXdd a þ . In addition, we can get:
RR RR
f ðl;mÞ ln ð1lÞ dldm ln ð1l 0 Þ f ðl;mÞ dldm
e D
(1) 1RR RR 1 e D ¼ 1 eln ð1l 0 Þ ¼ l 0 .
f ðl;mÞ ln m dldm ln m f ðl;mÞ dldm
(2) e D e 0 D ¼ m0 .
RR
which means that D f ðXÞXdd a .
Fig. 4.8 Three regions of

IFNs
RR
In brief, the inequalities a D f ðXÞXdd a þ hold. ■
Before showing the monotonicity of the IFIA operator, we introduce the concept
of shift vector of a region below.
Definition 4.4 (Lei et al. 2015) Let D be a region of IFNs. Given a vector
Vec ¼ ðM; NÞ, we define D0 ¼ D þ Vec as follows:
D0 ¼ N \ fðl þ M; m þ NÞjðl; mÞ 2 Dg
and Vec is called a shift vector of D. If fðl þ M; m þ NÞjðl; mÞ 2 DgN, then we

call it a moderate shift vector (Fig. 4.7) (Lei et al. 2015).
Theorem 4.15 (Monotonicity) (Lei et al. 2015) Given a moderate shift vector
Vec ¼ ðM; NÞ of D, let D0 ¼ D þ Vec and f ðl þ M; m þ NÞ ¼ f ðl; mÞ, then
(1) If M 0 and N 0, then
ZZ ZZ
D0 D
ZZ ZZ
D0 D
(a) Round region (b) Rectangular region
Fig. 4.9 Common regions of IFNs

Fig. 4.10 Mean weight

density function
(
l0 ¼ l þ M
Proof By the definition of moderate shift vectors, we know that .
m0 ¼ m þ N
Hence, there is
RR RR RR
f ðl0 ;m0 Þ ln ð1l0 Þdl0 dm0 f ðl;mÞ ln ð1lMÞdldm f ðl;mÞ ln ð1lÞdldm
e
(1) 1RR D0 = 1e D 1e D .
RR RR
f ðl0 ;m0 Þ ln m0 dldm f ðl;mÞ lnðm þ NÞdldm f ðl;mÞ ln mdldm
(2) e D0 ¼e D e D .
which means that
RR RR ! RR RR !
f ðl0 ;m0 Þ lnð1l0 Þdl0 dm0 f ðl0 ;m0 Þ ln m0 dldm f ðl;mÞ lnð1lÞdldm f ðl;mÞ ln m dldm
1e D0 ; e D0 1e D ; e D
RR RR
Thus, D0 f ðXÞXdd D f ðXÞXdd. Similarly, (2) can be proved. ■
In what follows, we give an example (Lei et al. 2015):
Suppose that there are three regions of IFNs, namely: D1 , D2 and D3 , and their
weights density functions are f1 ðXÞ, f2 ðXÞ and f3 ðXÞ, respectively, which are shown
in Fig. 4.8 (Lei et al. 2015).
Assume that f1 ðXÞ, f2 ðXÞ and f3 ðXÞ are the uniform distribution functions, which
are fi ðXÞ ¼ 1=jDi j ði ¼ RR
1; 2; 3Þ, where jDi j ði 2 1; 2; 3Þ represents the area value of
Di ði 2 1; 2; 3Þ . Then, D1 f1 ðXÞXdd can be calculated as follows:
RR
f1 ðl;mÞ ln m dldm m4 ð ln m4 1Þm3 ð ln m3 1Þ
(1) e D1
¼e m4 m3
.
RR
f1 ðl;mÞ lnð1lÞ dldm ð1l1 Þð ln ð1l1 Þ1Þð1l2 Þð ln ð1l2 Þ1Þ
(2) 1 e D1 ¼1e ð1l1 Þð1l2 Þ
.
Hence, there is
ZZ ð1l1 Þð ln ð1l1 Þ1Þð1l2 Þð ln ð1l2 Þ1Þ m4 ð ln m4 1Þm3 ð ln m3 1Þ

f1 ðXÞXdd ¼ 1 e ð1l1 Þð1l2 Þ
; e m4 m3
D1
Similarly, we have
ZZ ð1l3 Þð ln ð1l3 Þ1Þð1l4 Þð ln ð1l4 Þ1Þ m6 ð ln m6 1Þm5 ð ln m5 1Þ

f2 ðXÞXdd ¼ 1e ð1l3 Þð1l4 Þ
; e m6 m5
D2
and
ZZ ð1l5 Þð ln ð1l5 Þ1Þð1l6 Þð ln ð1l6 Þ1Þ

m2 ð ln m2 1Þm1 ð ln m1 1Þ
f3 ðXÞXdd ¼ 1 e ð1l5 Þð1l6 Þ
; e m2 m1
D3
Generally, the region of IFNs D may be of various shapes. The regions are
usually perfectly round because the Euclidean distance is most commonly used. For
all IFNs, the Euclidean distance of which to the given a ¼ ðl0 ; m0 Þ are less than or
equal to a constant d, then we can get the region of IFNs is perfectly round
(Fig. 4.9a; Lei et al. 2015), which can be expressed by
n o
D ¼ ðl; mÞj ðl l0 Þ2 þ ðm m0 Þ2 d
In addition, if we study the membership and the non-membership part of IFNs,

respectively, then it is more possible to research the set (as shown in Figs. 4.9b (Lei
et al. 2015).
D ¼ fðl; mÞj jl l0 j a ; jm m0 j b ; ðl0 þ aÞ þ ðm0 þ bÞ 1g
In what follows, we introduce three typical weight density functions:

(1) The mean weight density function (Fig. 4.10; Lei et al. 2015):
(2) The cone weight density function (Fig. 4.11; Lei et al. 2015):
(3) The normal distribution weight density function (Fig. 4.12; Lei et al. 2015):
A practical application of the IFIA operator can be provided as (Lei et al. 2015):
The application is about aggregating the assessments of Beijing given by the
whole citizens of Beijing, which are expressed by IFNs. Clearly, it is difficult to
aggregate these data by utilizing the traditional aggregation operators, like the
IFWA operators, due to that collecting all data is almost impossible. It shows that
Fig. 4.11 Cone weight

density function
Fig. 4.12 Normal

distribution weight density
function
the traditional aggregation techniques seem to be powerless for dealing with a large
number of data. Hence, we need some novel operators to solve this kind of
problem.
Utilizing the statistical methods to acquire the sampling data of intuitionistic

fuzzy information is a good choice. Then we acquire the distribution function by the
sample data, which depicts the proportion that every value has in the overall data.
The higher proportion one value has, the greater weight it gets. Assume that there
are n sampling IFNs, if we aggregate the n IFNs by the IFWA operator, then there
will not be a satisfying result. We know that it is perfectly possible that these nearby
values of the n sampling IFNs also appear in the overall data, however, aggregating
by IFWA operator cannot take these nearby values into account. While choosing
the IFIA operator to aggregate data can contain the information of the nearby
values, which maps the actual situation.
To this end, we can introduce a method to aggregate a large number of IFNs as
follows (Lei et al. 2015):
Step 1. Sample all decision information at random, then we can get the distribution
function and the distribution area based on these sampling data.
Step 2. After getting the distribution function and the distribution area, let the
distribution function be the weight density function f ðXÞ, and the distribution area
D be the region of IFNs.
Step 3. Use the IFIA operator to aggregate decision information.
Based on the above steps, we provide an application example (Lei et al. 2015) to
present the process below:
Suppose that there are a large number of decision makers (DMs), who need to
assess an object with IFNs. Before aggregating the incredible amount of intu-
itionistic fuzzy data, we first make an assumption that these DMs aren’t able to give
the very accurate assessments, which is obviously reasonable. For example, the
great majority of DMs cannot explain why the assessment is not the IFN
ð0:29; 0:41Þ but ð0:3; 0:4Þ. However, we should notice that people are generally
able to distinguish between 20% (i.e., 0.2) and 30% (i.e., 0.3), and they know which
one of them (0.2 and 0.3) is more close to their assessments. Hence, we divide the
whole interval ½0; 1 into many small intervals, the length of which are all 0.05, like
½0:2; 0:25Þ and ½0:25; 0:3. In the process of statistics, we only care which one of
small intervals the DMs’ assessment will fall into, and do not need to know the
concrete data. In order to get the aggregated value, the specific steps are shown as
follows:
(1) We first represent these assessments of DMs as two-dimension points in the
lm plane, then give the statistical data about the number of the IFN data
falling into the small square (the area of every small square is 0:05 0:05),
which is shown in Fig. 4.13 (Lei et al. 2015):
In Fig. 4.13, the volume of each cube represents the frequency value, which falls
into the small square, and the sum of all volumes is 1.
(2) Utilizing the interpolation function makes the figure smoother. In probability
theory, we infinitely subdivide these small squares to get the probability
density functions. However, because these DMs are not able to give the very
Fig. 4.13 Histogram of frequency
Fig. 4.14 Interpolation function
accurate assessments as mentioned earlier, the method of dividing these small

squares infinitely is infeasible. Hence, in order to make the weight density
function smoother, we decide to use the interpolation functions. Then, the
smooth weight density function is shown as (Lei et al. 2015) (Fig. 4.14):
(3) Based
RR on the interpolation function, we can calculate the numerical solution of
D f ðXÞXdd. Moreover, the results of ten times numerical experiments are
presented in Table 4.2 (Lei et al. 2015):
From 4.2, we can get that the result is approximately equal to ð0:4; 0:2Þ, which
appears between the peak values of the two weight density functions in Fig. 4.15
(Lei et al. 2015). It shows that the IFIA operator certainly depicts the distribution
situation and effectively aggregates information of the sampled population.
In addition, we know that the aggregating method by the IFIA operator is
feasible and stable. When the information (or data) obeys a distribution, the
Table 4.2 Numerical Number Results Number Results

solutions
1 ð0:4355; 0:1846Þ 6 ð0:4331; 0:1839Þ
2 ð0:4347; 0:1825Þ 7 ð0:4382; 0:1780Þ
3 ð0:4386; 0:1788Þ 8 ð0:4431; 0:1819Þ
4 ð0:4385; 0:1824Þ 9 ð0:4344; 0:1810Þ
5 ð0:4366; 0:1785Þ 10 ð0:4364; 0:1776Þ
Fig. 4.15 The lm plane
numerical solution of the IFIA operator is relatively fixed even though the given
sampled data is different for each time.
4.4 Conclusions
The main purpose of this chapter is to solve the problem how to aggregate con-
tinuous intuitionistic fuzzy information or data. By analogizing the process of
getting the definite integral of real numbers, we have deduced the integral aggre-
gating value of some regions of IFNs. Then, its calculating formula and the
properties have been investigated, and these good properties have shown that the
integral aggregating value has many common characteristics with the definite
integral of real numbers. Based on the concept of integral aggregating value, we
have utilized them to acquire a novel aggregation technique (the IFIA operator). By
taking a special function dðX; a1 ; a2 ; . . .; an Þ, we have gotten that the IFIA operator
4.4 Conclusions 97
is a continuous form of the IFWA operator, which also means that the IFWA
operator is a discrete formula of the IFAI operator. In addition, we have analyzed
the idempotency, boundedness and monotonicity of the IFIA operator, which show
the validity of the IFIA operator. Finally, we have shown some common regions
and weight density functions, and solved the practical problems by utilizing the
novel aggregation operator.
Chapter 5
Relationships Among IFWA Operator,
IFIA Operator and Definite Integrals
of IFFs
Chapter 5 first reveals the relationship between the IFIA operator introduced in
Chap. 4 and the definite integrals of IFFs defined in Chap. 3, where the IFIA
operator is used to aggregate continuous intuitionistic fuzzy information. Even
though the motivation and the purpose of proposing the IFIA operator and the
integral of IFFs are completely different, this chapter builds the bridge between the
two different concepts, which manifests that the IFIA operator is actually the def-
inite integral of a special IFF. Moreover, we will also show the relationship between
the IFIA operator and the IFWA operator from another perspective, which is dif-
ferent from one given in Chap. 4, which declares that the IFIA operator is the
continuous form of the IFWA operator.
5.1 IFWA Operator and Definite Integrals of IFFs
Firstly, we extend the IFF CountðXÞ given in Chap. 4 as follows (Lei and Xu
2016b):
There are n IFNs ai ¼ ðli ; mi Þ (i ¼ 1; 2; . . .; n), which satisfy ai 6¼ aj if only i 6¼ j,
P
and their weights are respectively xi (i ¼ 1; 2; . . .; n) with ni¼1 xi ¼ 1. Then we
give the following two real functions:
X X
RðlÞ ¼ xi ; TðmÞ ¼ xi
li [ l mi m
where RðlÞ reveals the sum of the weights xi whose membership degrees of the
corresponding ai are greater than the given l. In addition, TðmÞ counts the sum of
these weights xi whose non-membership degrees of the corresponding ai are
greater than or equal to the given m.
We utilize these two real functions to build (Lei and Xu 2016b) a novel IFF
ðRðlÞ; TðmÞÞ, which is expressed by LðXÞ. According to the discussion of
100 5 Relationships Among IFWA Operator, IFIA Operator …
CountðXÞ, we can get the integral of LðXÞ along the curve, which has been
introduced in Fig. 3.5, still exists. Next, we show the relationship between the
IFWA operator and the integral of LðXÞ along the curve in the following theorem.
Theorem 5.1 (Lei and Xu 2016b) If there are n IFNs ai ¼ ðli ; mi Þ ( i ¼ 1; 2; . . .; n),
which satisfy ai 6¼ aj if only i 6¼ j, and their weights are respectively xi (
P
i ¼ 1; 2; . . .; n) with ni¼1 xi ¼ 1, then
Zb
LðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ
O
where O ¼ ð0; 1Þ and b ¼ ðlmax ; mmin Þ.

Proof Let U ¼ fli j1 i ng and V ¼ fmi j1 i ng be two given sets, then we
have jUj n and jVj n because there may be some repeated elements in U and V.
Hence, we can rank li (i ¼ 1; 2; . . .; n) and mi (i ¼ 1; 2; . . .; n) as
lð1Þ \lð2Þ \ \lðjUjÞ and mð1Þ \mð2Þ \ \mðjVjÞ , respectively. After analysis,
we know that LðXÞ consists of two real piecewise continuous functions:
8
> 1; 0 l\lð1Þ ;
>
>
>
> k 1 ; lð1Þ l\lð2Þ ;
<
RðlÞ ¼ k 2 ; lð2Þ l\lð3Þ ;
>
> ..
>
> .
>
:
0; l ðjUjÞ l 1;
and
8
> 0; mðjVjÞ \m 1;
>
>
>
> k ; m ðjVj1Þ \m mðjVjÞ ;
< 1
TðmÞ ¼ k 2 ; m ðjVj2Þ \m mðjVj1Þ ;
>
> ..
>
> .
>
:
1; 0 m mð1Þ ;
P P
where ki ¼ lj [ lðiÞ xj and ki ¼ mj [ mðjVjiÞ xj . Hence, we have
X X X
ki ki þ 1 ¼ xj xj ¼ xj
lj [ lðiÞ lj [ lði þ 1Þ lj ¼lði þ 1Þ
X X X
k i þ 1 ki ¼ xj xj ¼ xj
mj [ mðjVji1Þ mj [ mðjVjiÞ mj ¼mðjVjiÞ
According to the calculating formula of the definite integral of the IFF, if we let
lð0Þ ¼ 0 and mðjVj þ 1Þ ¼ 1, then
5.1 IFWA Operator and Definite Integrals of IFFs 101
0 8 u 9 8m 91
Zb < Zmax RðlÞ = < Zmin 1 TðmÞ =
LðXÞdX ¼ @1 exp dl ; exp dm A
: 1l ; : m ;
O 0 1
0 8 0 19 8 0 191
>
< jUj1
X Z
lði þ 1Þ >
= >
<jVj1
X Z
mðjVjiÞ >
=
B B 1 C B 1 C C
¼ @1 exp @ki dlA ; exp @ð1 ki Þ dmA A
>
: i¼0 1l >
; >
: i¼0 m >
;
lðiÞ mðjVji þ 1Þ
0 ! 1
Y mðjVjiÞ 1ki
k
Y 1 lði þ 1Þ i jVj1
jUj1
¼ @1 ; A
i¼0
1 lðiÞ i¼0
mðjVji þ 1Þ
!
Y
jUj1 ki ki þ 1 jVYj1
ki þ 1 ki
¼ 1 1 lði þ 1Þ ; mðjVjiÞ
i¼0 i¼0
P P !
Y
jUj1
lj ¼lði þ 1Þ
xj Y
jVj1
mj ¼m xj
¼ 1 1 lði þ 1Þ ; mðjVjiÞ ðjV jiÞ
i¼0 i¼0
¼ IFWAx ða1 ; a2 ; . . .; an Þ
which completes the proof of this theorem.

We know that any IFN can be represented as a point in the two-dimensional
plane. The IFWA operator can aggregate some discrete IFNs when the weights of
these IFNs are known, which likes the situation where the joint distributions of the
two-dimensional discrete random variables are known in probability theory and
mathematical statistics. However, the joint distribution can uniquely determine the
marginal distributions, while the marginal distributions cannot give a unique joint
distribution. Hence, more information is needed in the joint distribution than in the
marginal distributions.
Moreover, Theorem 5.1 shows that the integral of LðXÞ can aggregate data only
by utilizing the marginal distributions (RðlÞ and TðmÞ), while the IFWA operator
must use the information in the joint distribution to the aggregating data.
Rb
Consequently, O LðXÞdX is superior to the IFWA operator because the integral
can acquire the same result based on less information (Lei and Xu 2016b).
In what follows, we give an example (Lei and Xu 2016b) to illustrate the
theorem:
Suppose that there are three IFNs ð0:3; 0:4Þ, ð0:1; 0:2Þ and ð0:5; 0:2Þ, and their
weights are 0.3, 0.3 and 0.4, respectively (as shown in Table 5.1) (Lei and Xu
2016b):
On the one hand, we use the IFWA operator to aggregate these IFNs as follows:
Table 5.1 All data and their Values of l 0.1 0.3 0.5
weights
Weights 0.3 0.3 0.4
Values of m 0.2 0.4
Weights 0.7 0.3
Values of IFNs ð0:1; 0:2Þ ð0:3; 0:4Þ ð0:5; 0:2Þ
Weights 0.3 0.3 0.4
IFWAx ða1 ; a2 ; . . .; an Þ ¼ 0:3ð0:1; 0:2Þ 0:3ð0:3; 0:4Þ 0:4ð0:5; 0:2Þ

¼ 1 0:90:3 0:70:3 0:50:4 ; 0:20:7 0:40:3
On the other hand, we analyze the IFF LðXÞ, which consists of RðlÞ and TðmÞ,
and is shown below:
8
> 1; 0 l\0:1; 8
>
< < 0; 0:4\m 1;
0:7; 0:1 l\0:3;
RðlÞ ¼ ; TðmÞ ¼ 0:3; 0:2\m 0:4;
>
> 0:4; 0:3 l\0:5; :
: 1; 0\m 0:2:
0; 0:5 l\1;
Then we can obtain the integral by the calculating formula:

0 8 9 8m 91
Zb < Zumax = < Zmin 1 TðmÞ =
RðlÞ
LðXÞdX ¼ @1 exp dl ; exp dm A
: 1l ; : m ;
O 0 1
P P !
jUj
Y xj Y jVj
lj ¼lðiÞ
mj ¼m xj
¼ 1 1 lðiÞ ; mðjVjiÞ ðjVjiÞ
i¼1 i¼1

¼ 1 0:90:3 0:70:3 0:50:4 ; 0:20:7 0:40:3
Rb
Hence, we know that O LðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ holds.
Moreover, according to Chap. 3, we denote b0 ¼ O and bi þ 1 ¼ bi ai þ 1 , then
(Lei and Xu 2015a):
Zbn
PiðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ
O
where b n ¼ ni¼1 ai ; PiðXÞ ¼ ðx i ; 1 x i Þ when b i1 EXEb i (1 i n). It also

shows that the IFWA operator is actually the definite integral of the IFF PiðXÞ.
Below we show three methods to aggregate IFNs, all of which have the forms of
definite integral of a specific IFF and closed connection with the IFWA operator:
Given n IFNs ai ¼ ðli ; mi Þ (i ¼ 1; 2; . . .; n), which satisfy ai 6¼ aj if only i 6¼ j,
P
and their weights are xi (i ¼ 1; 2; . . .; n), respectively, which meet ni¼1 xi ¼ 1.
Method 1. The n IFNs can be depicted in Fig. 5.1 (Lei and Xu 2015c).
Obviously, this random scattering of sites of the n IFNs seem to be desperately
Rb
random. We can utilize the integral O LðXÞdX to aggregate them, where O ¼
ð0; 1Þ and b ¼ ðlmax ; mmin Þ, and the aggregated value is just equal to the result of
IFWAx ða1 ; a2 ; . . .; an Þ.
Fig. 5.1 n IFNs
Method 2. Denoting that b0 ¼ O and bi þ 1 ¼ bi ai þ 1 , we know the piecewise

IFF PiðXÞ can be expressed as (Lei and Xu 2015a):
8
>
> ðx 1 ; 1 x 1 Þ; when b 0 EXEb 1 ;
>
< ðx 2 ; 1 x 2 Þ; when b 1 EXEb 2 ;
PiðXÞ ¼ ..
>
> .
>
:
ðx n ; 1 x n Þ; when bn1 EXEbn ;
Then there still be
Zbn
PiðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ
O
where b n ¼ ni¼1 ai . This situation is shown in Fig. 5.2.

By Fig. 5.2, we know
O ¼ b0 Eb1 Eb2 E Ebn1 Ebn ¼ ni¼1 ai
which are sorted regularly in the figure.

Method 3. Arrange the n IFNs ai ¼ ðli ; mi Þ (i ¼ 1; 2; . . .; n) with the ascending
order of their weights, which means
Fig. 5.2 The piecewise IFF

PiðXÞ
að1Þ ; að2Þ ; að3Þ ; að4Þ ; að5Þ ; að6Þ ; . . .; aðn2Þ ; aðn1Þ ; aðnÞ
satisfy the condition: their corresponding weights of IFNs meet
xð1Þ xð2Þ xð3Þ xð4Þ xð5Þ xð6Þ xðn2Þ xðn1Þ xðnÞ
i1
Let K 0 ¼ O and K i þ 1 ¼ K i aðniÞ , which indicates that K i ¼ aðnjÞ , then
j¼0
O ¼ K 0 EK 1 EK 2 E EK n1 EK n ¼ ni¼1 aðni þ 1Þ ¼ ni¼1 ai ¼ bn
Moreover, let the weight of K i (i 2 f0; 1; 2; . . .; ng) be equal to xðni þ 1Þ

xðniÞ (let xð0Þ ¼ 0 and xðn þ 1Þ ¼ 1), which is larger than or equal to zero (as
shown in Fig. 5.3), then we can get the IFF LðXÞ in this case as follows:
8
>
>
> x ðnÞ ; 1 x ðnÞ ; when K 0 EXEK 1 ;
< x ðn1Þ ; 1 x ðn1Þ ; when K 1 EXEK 2 ;
LðXÞ ¼ ..
>
> .
>
:
x ð1Þ ; 1 x ð1Þ ; when K n1 EXEK n ;
Next, we analyze the situation when K0 E X E K1 , and other situations are

similar to it. If K0 E X E K1 , due to

RðlÞ ¼ x ðnÞ x ðn1Þ þ x ðn1Þ x ðn2Þ þ x ðn2Þ x ðn3Þ

þ þ x ð2Þ x ð1Þ þ x ð1Þ x ð0Þ
then, we get RðlÞ ¼ x ðnÞ x ð0Þ ¼ xðnÞ 0 ¼ xðnÞ when K0 E XE K1 .

In addition, there is also TðmÞ ¼ 1 x ðnÞ . So we get
Fig. 5.3 The situation of

Method 3

LðXÞ ¼ x ðnÞ ; 1 x ðnÞ ; when K0 EXEK1
Similarly, we can conclude

LðXÞ ¼ x ðniÞ ; 1 x ðniÞ ; when Ki EXEKi þ 1
Based on the expressions of LðXÞ in this case, it is easy to acquire
ZK n Zb n
LðXÞdX ¼ LðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ
O O
Based on Method 3, we can know that for any given n IFNs ai ¼ ðli ; mi Þ
(i ¼ 1; 2; . . .; n) with the weights xi (i ¼ 1; 2; . . .; n), these IFNs can be transformed
into Ki ði ¼ 0; 1; 2; . . .; nÞ with the new weights xK i ¼ xðni þ 1Þ xðniÞ (let
xð0Þ ¼ 0 and xðn þ 1Þ ¼ 1) (i ¼ 1; 2; . . .; n), then the following conclusion is
established:
IFWA xK i ðK 0 ; K 1 ; K 2 . . .; K n Þ ¼ IFWAx ða1 ; a2 ; . . .; an Þ
because there is
n
IFWA xK i ðK 0 ; K 1 ; K 2 . . .; K n Þ ¼ xðni þ 1Þ xðniÞ K i
i¼0

¼ xðn þ 1Þ xðnÞ K 0 xðnÞ xðn1Þ K 1 xðn1Þ xðn2Þ K 2

xð2Þ xð1Þ K n1 xð1Þ xð0Þ K n

¼ xðn þ 1Þ xðnÞ O xðnÞ xðn1Þ aðnÞ xðn1Þ xðn2Þ aðnÞ aðn1Þ

xðn2Þ xðn3Þ aðnÞ aðn1Þ aðn2Þ

xð2Þ xð1Þ aðnÞ aðn1Þ aðn2Þ að2Þ

xð1Þ xð0Þ aðnÞ aðn1Þ aðn2Þ að2Þ að1Þ

¼ xðnÞ xðn1Þ þ xðn1Þ xðn2Þ þ xðn2Þ xðn3Þ þ þ xð2Þ xð1Þ þ xð1Þ xð0Þ aðnÞ

xðn1Þ xðn2Þ þ xðn2Þ xðn3Þ þ þ xð2Þ xð1Þ þ xð1Þ xð0Þ aðn1Þ

xðn2Þ xðn3Þ þ xð2Þ xð1Þ þ xð1Þ xð0Þ aðn2Þ

xð2Þ xð1Þ þ xð1Þ xð0Þ að2Þ xð1Þ xð0Þ að1Þ
¼ xðnÞ aðnÞ xðn1Þ aðn1Þ xðn2Þ aðn2Þ xð2Þ að2Þ xð1Þ að1Þ
¼ IFWAx ða1 ; a2 ; . . .; an Þ
Based on the three aggregating methods to collect IFNs, we can get the fol-
lowing equalities:
ZK n Zbn Zb
LðXÞdX ¼ PiðXÞdX ¼ LðXÞdX ¼ IFWA xK i ðK 0 ; K 1 ; K 2 ; . . .; K n Þ
O O O
¼ IFWAx ða1 ; a2 ; . . .; an Þ
which reveals that the IFWA can be represented as three different forms of integrals
of IFFs, and there are five ways to aggregate any given IFNs.
Considering the above-mentioned example, which supposes that there are three
IFNs a1 ¼ ð0:3; 0:4Þ, a2 ¼ ð0:1; 0:2Þ and a3 ¼ ð0:5; 0:2Þ, and their weights are
x1 ¼ 0:3, x2 ¼ 0:3 and x3 ¼ 0:4, respectively, then
Zb

LðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ ¼ 1 0:90:3 0:70:3 0:50:4 ; 0:20:7 0:40:3
O
RK
In what follows, we aggregate these IFNs by utilizing O n LðXÞdX and
R bn R bn
O PiðXÞdX. Firstly, O PiðXÞdX can be used to deal with these data as follows:
Let b0 ¼ O and b1 ¼ b0 a1 ¼ ð0:3; 0:4Þ, b2 ¼ b1 a2 ¼ ð0:3; 0:4Þ

ð0:1; 0:2Þ, and b3 ¼ b2 a3 ¼ ð0:3; 0:4Þ ð0:1; 0:2Þ ð0:5; 0:2Þ. According to
the definition of PiðXÞ, there is
8
< ð0:3; 1 0:3Þ; when b 0 E X E b 1 ;
PiðXÞ ¼ ð0:3; 1 0:3Þ; when b 1 E X E b 2 ;
:
ð0:4; 1 0:4Þ; when b2 E X E b3 ;
Then we can get
Zb3 Zb1 Zb2 Zb3

PiðXÞdX ¼ PiðXÞdX PiðXÞdX PiðXÞdX
O O b1 b2
Zb1 Zb2 Zb3

¼ ð0:3; 1 0:3ÞdX ð0:3; 1 0:3ÞdX ð0:4; 1 0:4ÞdX
O b1 b2
¼ 0:3ðb1 OÞ 0:3ðb2 b1 Þ 0:4ðb3 b2 Þ

¼ 0:3a1 0:3a2 0:4a3 ¼ 1 0:90:3 0:70:3 0:50:4 ; 0:20:7 0:40:3
¼ IFWAx ða1 ; a2 ; . . .; an Þ
RK
Moreover, we can utilize O n LðXÞdX to aggregate them.
Firstly, we arrange a1 ¼ ð0:3; 0:4Þ, a2 ¼ ð0:1; 0:2Þ and a3 ¼ ð0:5; 0:2Þ accord-
ing to the ascending order of their weights, and then we get
að1Þ ¼ a1 ; að2Þ ¼ a2 ; að3Þ ¼ a3
for x1 x2 x3 . Let K 0 ¼ O, K 1 ¼ a3 , K 2 ¼ a3 a2 and K 3 ¼ K 2 a1 . The

weights of K 0 , K 1 , K 2 and K 3 are 0.6, 0.1, 0 and 0.3, respectively. By the special
IFF LðXÞ, we can get
8 8
< 0:4; UðK 0 Þ l\UðK 1 Þ; < 0:6; VðK 1 Þ\m VðK 0 Þ;
RðlÞ ¼ 0:3; UðK 1 Þ l\UðK 2 Þ; and TðmÞ ¼ 0:7; VðK 2 Þ\m VðK 1 Þ;
: :
0:3; UðK 2 Þ l\UðK 3 Þ; 0:7; VðK 3 Þ\m VðK 2 Þ;
Therefore, we can obtain that

8
< x ð3Þ ; 1 x ð3Þ ¼ ð0:4; 1 0:4Þ; when K 0 E X E K 1;
LðXÞ ¼ x ð2Þ ; 1 x ð2Þ ¼ ð0:3; 1 0:3Þ; when K 1 E X E K 2;
:
x ð1Þ ; 1 x ð1Þ ¼ ð0:3; 1 0:3Þ; when K 2 E X E K 3;
where x ðiÞ ¼ x i (i ¼ 1; 2; 3), and thus, we have

ZK 3 ZK 1 ZK 2 ZK 3
LðXÞdX ¼ LðXÞdX LðXÞdX LðXÞdX
O K0 K1 K2
ZK 1 ZK 2 ZK 3

¼ x ð3Þ ; 1 x ð3Þ dX x ð2Þ ; 1 x ð2Þ dX x ð1Þ ; 1 x ð1Þ dX
K0 K1 K2
ZK 1 ZK 2 ZK 3
¼ ðx 3 ; 1 x 3 ÞdX ðx 2 ; 1 x 2 ÞdX ðx 1 ; 1 x 1 ÞdX
K0 K1 K2
¼ 0:4a3 0:3a2 0:3a1

¼ IFWAx ða1 ; a2 ; . . .; an Þ
Furthermore, we test the aggregated value by using the IFWA operator to cope
with K0 , K1 , K2 and K3 as follows:
IFWA xKi ðK0 ; K1 ; K2 ; K3 Þ ¼ 0:6K0 0:1K1 0K2 0:3K3

¼ 0:6O 0:1a3 0ða3 a2 Þ 0:3ða3 a2 a1 Þ
¼ 0:6O 0:4a3 0:3a2 0:3a1
¼ IFWAx ða1 ; a2 ; . . .; an Þ
In general, the following equalities hold:
ZK n Zbn Zb
LðXÞdX ¼ PiðXÞdX ¼ LðXÞdX ¼ IFWA xK i ðK 0 ; K 1 ; K 2 ; K n Þ
O O O
¼ IFWAx ða1 ; a2 ; . . .; an Þ
5.2 IFIA Operator and Definite Integrals of IFFs

Rb
Theorem 5.1 points out that O LðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ, which shows the
relationshipRRbetween the IFWA operator and the integral of LðXÞ. In addition, we
know that D f ðXÞXdd has a close connection with the IFWA operator, which
shows that the IFIA operator is a continuous
RR form of the IFWA operator. Hence, in
order to acquire the integral form of D f ðXÞXdd, we need to extend LðXÞ to its
continuous generalization.
Let D be a region of IFNs shown in Fig. 5.4 (Lei and Xu 2016b), and PðXÞ be a
non-negative real function:Then we define two subsets l and m of D as follows
(Lei and Xu 2016b):
5.2 IFIA Operator and Definite Integrals of IFFs 109
Fig. 5.4 Continuous

generalization of LðXÞ
l ¼ fðq; rÞjðq; rÞ 2 D and q lg
is a subset of D, whose elements are on the right side of q ¼ l and belong to the
region D, and
m ¼ fðq; rÞjðq; rÞ 2 D and r mg
is also a subset of D, whose elements are on top of r ¼ m and belong to the

region D.
Based on l and m , we can define two real functions RðlÞ and T ðmÞ below
(Lei and Xu 2016b):
ZZ ZZ
RðlÞ ¼ Pðq; rÞdqdr and T ðmÞ ¼ Pðq; rÞdqdr
l m
are the double integrals of Pðq; rÞ located in the regions l and m , respectively,
which are similar to RðlÞ and TðmÞ mentioned in the previous section. Obviously,
both RðlÞ and T ðmÞ are the continuous functions, which are shown in Fig. 5.5 (Lei
and XuRR 2016b).
If D PðXÞdd ¼ 1, then we can build an IFF LðXÞ ¼ ðRðlÞ; T ðmÞÞ by using
RðlÞ and T ðmÞ likes LðXÞ. In the same way, we also can analyze the integral of
LðXÞ along to the curve shown in Fig. 5.6 (Lei and Xu 2016b).
By the formula of the definite integral of IFFs, we get
(a) P( X )
(b) ( ) (c) ( )
Fig. 5.5 PðXÞ, RðlÞ and T ðmÞ
0 0 1 0 11
Zb Zlb Zmb
RðlÞ A 1 T ðmÞ AA
LðXÞdX ¼ @1 exp@ dl ; exp@ dm
1l m
O 0 1
Rb Rb
It is clear that LðXÞdX and
LðXÞdX are on the basis of the same idea, and
Rb O
Rb O
O LðXÞdX is a continuous form of O LðXÞdX.

In the following, we can get some conclusions (Lei and Xu 2016b):
Rb Rb
(1) O LðXÞdX is a continuous form of O LðXÞdX.
Rb
(2) O LðXÞdX ¼ IFWAx ða1 ; a2 ; . . .; an Þ.
Fig. 5.6 An IFIC of LðXÞ
RR
(3) IFWAx ða1 ; a2 ; . . .; an Þ is the discrete form of D PðXÞXdd.
Rb RR
Now we reveal the relationship between O LðXÞdX and D PðXÞXdd(Lei and
RR
Xu 2016b). In fact, D PðXÞXdd consists of countless Pðni ; gi Þðni ; gi ÞDdi according
to the definition of the IFIA operator:
ZZ
k
PðXÞXdd ¼ lim Pðni ; gi Þðni ; gi ÞDdi
d!0 i¼1
D
In addition, we know that every Pðni ; gi Þðni ; gi ÞDdi can be transformed into an
integral form:
ðZ
ni ;gi Þ
Pðni ; gi Þðni ; gi ÞDdi ¼ ðPðni ; gi ÞDdi ; 1 Pðni ; gi ÞDdi Þ dX

O
Hence, we can get the following expression:

0 1
ZZ ðZ
ni ;gi Þ
B k C
PðXÞXdd ¼ lim @ ðPðni ; gi ÞDdi ; 1 Pðni ; gi ÞDdi Þ dX A
d!0 i¼1
D O
According to Theorem 3.8, if the k upper and lower limits of the integrals are the
same, then we can combine these integrals into one integral. Hence, we need to find
a common upper limit of integrals, which is denoted as X.
Fig. 5.7 A common upper

limit X
By the definition of the integral of IFFs, there is a E b, where a and b are

Rb
respectively the upper and lower limits of a uðXÞdX. Hence, the common upper
limit X must meet the following condition:
ðni ; gi ÞEX; for any ðni ; gi Þ 2 D
According to the definition of S , we can get that the X in Fig. 5.7 must satisfy
ðni ; gi ÞE X for any ðni ; gi Þ 2 D, because it can fall into S of all IFNs ðni ; gi Þ
(i ¼ 0; 1; . . .; k) which belong to D. Hence, X in Fig. 5.7 (Lei and Xu 2016b) is a
common upper limit.
Moreover, we can also get that: (1) b E X, where b ¼ lsup ; minf ; (2) Any IFN
n is a common upper limit if only X E n, which means that X is not the only one
common upper limit.
In what follows, we define an IFF lðni ;gi Þ ðXÞ of X, for any given IFN ðni ; gi Þ. For

lðni ;gi Þ ðXÞ ¼ Uðlðni ;gi Þ ðXÞÞ; Vðlðni ;gi Þ ðXÞÞ , we first define its membership part as
(Lei and Xu 2016b):
8
< Pðni ; gi ÞDdi ; 0 UðXÞ ni ;
U lðni ;gi Þ ðXÞ ¼
:
0; ni \UðXÞ UðXÞ ;
and the non-membership part is defined (Lei and Xu 2016b) as:

8
< 1 Pðni ; gi ÞDdi ; gi VðXÞ 1;
V lðni ;gi Þ ðXÞ ¼
:
1; VðXÞ VðXÞ\gi ;
The region
Fig. 5.8 defined
by U lðni ;gi Þ ðXÞ and

V lðni ;gi Þ ðXÞ

U lðni ;gi Þ ðXÞ and V lðni ;gi Þ ðXÞ provide the function values in the shadow region
of Fig. 5.8 (Lei and Xu 2016b).

According to U lðni ;gi Þ ðXÞ and V lðni ;gi Þ ðXÞ , we can get lðni ;gi Þ ðXÞ as:
8
< ðPðni ; gi ÞDdi ; 1 Pðni ; gi ÞDdi Þ; OEXEðni ; gi Þ;
lðni ;gi Þ ðXÞ ¼
:
O; ðni ; gi Þ / XEX:
which defines the function values in the shadow region in Fig. 5.9 (Lei and Xu 2016b).
Obviously, the region defined by lðni ;gi Þ ðXÞ is contained in which defined by

U lðni ;gi Þ ðXÞ and V lðni ;gi Þ ðXÞ by Figs. 5.8 and 5.9. Hence, the real functions

U lðni ;gi Þ ðXÞ and V lðni ;gi Þ ðXÞ can uniquely determine the IFF lðni ;gi Þ ðXÞ, rather
than vice versa.
Next, we study the integral of lðni ;gi Þ ðXÞ as follows:
ZX ðZ
ni ;gi Þ ZX
lðni ;gi Þ ðXÞdX ¼ lðni ;gi Þ ðXÞdX lðni ;gi Þ ðXÞdX
O O ðni ;gi Þ
ðZ
ni ;gi Þ ZX
¼ ðPðni ; gi ÞDdi ; 1 Pðni ; gi ÞDdi ÞdX OdX
O ðni ;gi Þ
ðZ
ni ;gi Þ
¼ ðPðni ; gi ÞDdi ; 1 Pðni ; gi ÞDdi ÞdX

O
Fig. 5.9 The region defined

by lðni ;gi Þ ðXÞ
These different upper limits ðni ; gi Þ (i ¼ 1; 2; . . .; k) of the integrals have been

transformed into X, and then we can combine these integrals into one integral based
on Theorem 3.8.
Theorem 5.2 (Lei and Xu 2016b) Let D be a region
RR of IFNs, and PðXÞ be a
non-negative real function of D, which satisfies D PðXÞdd ¼ 1. If we note

b ¼ supðl;mÞ2D flg; inf ðl;mÞ2D fmg , then
ZZ Zb ZZ ZZ !
PðXÞXdd ¼ PðXÞdd; PðXÞdd dX
l m
D O
RR Rb
which reveals the relationship between D and O as:
ZZ Zb
PðXÞXdd ¼ LðXÞdX
D O
Proof We first let X be an IFN, which

RR satisfies ðni ; gi Þ E X for any ðni ; gi Þ 2 D,
then according to the definition of D
ZZ
PðXÞXdd
D
k
¼ lim Pðni ; gi Þðni ; gi ÞDdi
d!0 i¼1
0 1
ðZ
ni ;gi Þ
k B C
¼ lim @ ðPðni ; gi ÞDdi ; 1 Pðni ; gi ÞDdi Þ dX A
d!0 i¼1
O
ZX ZX X !
k
k X
k
¼ lim lðni ;gi Þ ðXÞdX ¼ lim U lðni ;gi Þ ðXÞ ; 1 1 V lðni ;gi Þ ðXÞ dX
d!0 i¼1 d!0
i¼1 i¼1
O O
ZX !
X
k X
k
¼ lim U lðni ;gi Þ ðXÞ ; 1 lim 1 V lðni ;gi Þ ðXÞ dX
d!0 d!0
i¼1 i¼1
O
ZX !
X X
¼ lim U lðni ;gi Þ ðXÞ þ U lðni ;gi Þ ðXÞ ;
d!0
ni l ni \l
O
!!
X X
1 lim 1 V lðni ;gi Þ ðXÞ þ 1 V lðni ;gi Þ ðXÞ dX
d!0
gi m gi \m
ZX ! !!
X X
¼ lim Pðni ; gi ÞDdi þ 0 ; 1 lim 0 þ Pðni ; gi ÞDdi dX
d!0 d!0
ni l gi \m
O
Zb ZZ ZZ !
¼ PðXÞdd; PðXÞdd dX
l m
O
ZX ZX
¼ ðRðlÞ; T ðmÞÞdX ¼ LðXÞdX
O O
Since b E X and LðXÞ ¼ O when b E X E X, then we have
ZZ ZX Zb ZX Zb
PðXÞXdd ¼ LðXÞdX ¼ LðXÞdX OdX ¼ LðXÞdX
D O O b O
which completes the proof of this theorem. RR

Theorem 5.2 reveals that the IFIA operator D PðXÞXdd is actually the definite
integral of the IFF LðXÞ. Moreover, a figure is provided to show the relationships
among different operators (Lei and Xu 2016b) below (Fig. 5.10):
In what follows, we give two examples (Lei and Xu 2016b) to test the con-
clusion in Theorem 5.2. Let D be a region of IFNs, which is shown
RR in Fig. 5.11 (Lei
and Xu 2016b), and PðXÞ ¼ 4 for any X 2 D. Obviously, D PðXÞdd ¼ 1.
Fig. 5.10 Relationship

among different operators
Fig. 5.11 Figure of the

example
By the formula of the IFIA operator, we have

0 8 9 8 91
ZZ < ZZ = < ZZ =
PðXÞXdd ¼ @1 exp 4 ln ð1 lÞ dd ; exp 4 ln mdd A
: ; : ;
D D D
2 1
¼ 1 ;
e 2e
Moreover
8
< 0; 1=2\m 1;
RðlÞ ¼ 1 2l and T ðmÞ ¼
:
1 2m; 0 m 1=2:
Hence, the IFF LðXÞ can be expressed as:

8
< ð1 2l; 0Þ; OEX / ð1=2; 1=2Þ;
LðXÞ ¼
:
ð1 2l; 1 2mÞ; ð1=2; 1=2ÞEXEð1=2; 0Þ;
Then we can obtain

0 8 1=2 9 8 91
Zb < Z 1 2l = >
<Z1=2 1 Z0 >
=
B C
LðXÞdX ¼ @1 exp dl ; exp dm þ 2dm A
: 1l ; >
: m >
;
O
0 1 1=2
2 1
¼ 1 ;
e 2e
RR Rb
So PðXÞXdd ¼ O LðXÞdX holds.
D
Another example (Lei and Xu 2016b) is provided as follows:
Suppose that there are many students who need to assess their English teacher
with IFNs, in which the membership degree given by a student represents to which
extent he/she likes the teacher, and the non-membership degree shows to which
extent he/she dislikes the teacher. All assessments given by the students can be
regarded as the continuous intuitionistic fuzzy information distributed in a region D.
Let
RR PðXÞ be the density function of the sampled assessments, where
D PðXÞdd ¼ 1, which indicates the weights of every IFN X in D. In addition, the
sampled data is shown in Fig. 5.12 (Lei and Xu 2016b).
In the previous sections, we have known that there are two methods to aggregate
the sampled data:
Method 1. Because the sampled data (IFNs) can be represented as two-dimension
points in the lm plane, and the statistical data about the number of the IFNs which
fall into the small squares (the area of every small square is 0:05 0:05) can also be
Fig. 5.12 The sampled data

Fig. 5.13 The histogram of frequency
acquired, we can draw the histogram of frequency below (Lei and Xu 2016b)
(Fig. 5.13):
Next, we can use the interpolation functions to make the histogram of frequency
smoother. Then, the smoother weight function can be shown as (Lei and Xu
2016b):
Based on the interpolation function
RR shown in Fig. 5.14 (Lei and Xu 2016b), we
can get the numerical solution of D PðXÞXdd:
Fig. 5.14 The interpolation function

Table 5.2 The values of l l 0.00 0.05 0.10 0.15 0.20

and RðlÞ
RðlÞ 1.0000 0.9799 0.9301 0.8435 0.7411
l 0.55 0.60 0.65 0.70 0.75
RðlÞ 0.3270 0.2237 0.1264 0.0553 0.0219
0.25 0.30 0.35 0.40 0.45 0.50
0.6429 0.5598 0.5122 0.4801 0.4501 0.3995
0.80 0.85 0.90 0.95 1.00
0.0039 0.0004 0 0 0
Table 5.3 The values of m m 1.00 0.95 0.90 0.85 0.80

and T ðmÞ
T ðmÞ 0 0 0 0 0
m 0.45 0.40 0.35 0.30 0.25
T ðmÞ 0.3613 0.4445 0.5114 0.5714 0.6511
0.75 0.70 0.65 0.60 0.55 0.50
0.0013 0.0064 0.0304 0.0737 0.1483 0.2512
0.20 0.15 0.10 0.05 0.00
0.7501 0.8543 0.9323 0.9799 1.0000
ZZ
PðXÞXdd ¼ ð0:4337; 0:3017Þ
D
and some conclusions can be got as:

(1) About 43% of students likes their English teacher, and around 30% dislikes the
teacher.
(2) If all students are concentrated in a mixed one, then each student should give
the approximate IFN ð0:4; 0:3Þ as his/her assessment.
Method 2. Firstly, we acquire the two real functions RðlÞ and T ðmÞ based on the
sampled data. All the values of RðlÞ and T ðmÞ are listed in Tables 5.2 and 5.3 (Lei
and Xu 2016b), respectively.
Based on the interpolation functions, RðlÞ and T ðmÞ can be presented in
Fig. 5.15 (Lei and Xu 2016b).
According to the calculating formula of the integral of IFF, we can get
Rb RR
O LðXÞdX ¼ ð00:4313; 00:3004Þ. Hence, we know that D PðXÞXdd is roughly
Rb
equivalent to O LðXÞdX.
( ) ( )
Fig. 5.15 RðlÞ and T ðmÞ
5.3 IFWA Operator and IFIA Operator
Chapter 4 has proved that the IFWA operator is the discrete form of the IFIA
operator; Contrarily, we can call that the IFIA operator is the continuous form of the
IFWA operator. In addition, this section will build the relationship between the
IFWA operator and the IFIA operator from another perspective.
Firstly, the concept of Archimedean t-conorm and t-norm are introduced as
follows:
Definition 5.1 (Klir and Yuan 1995) A function s : ½0; 1
½0; 1
! ½0; 1
is called
a t-norm if it satisfies the following conditions:
(1) sð1; xÞ ¼ x, for all x.
(2) sðx; yÞ ¼ sðy; xÞ, for all x and y.
(3) sðx; sðy; zÞÞ ¼ sðsðx; yÞ; zÞ, for all x, y and z.
(4) If x x0 and y y0 , then sðx; yÞ sðx0 ; y0 Þ.
Definition 5.2 (Klir and Yuan 1995) A function s : ½0; 1

½0; 1
! ½0; 1
is called
a t-conorm if it satisfies the following conditions:
(1) sð0; xÞ ¼ x, for all x.
(2) sðx; yÞ ¼ sðy; xÞ, for all x and y.
(3) sðx; sðy; zÞÞ ¼ sðsðx; yÞ; zÞ, for all x, y and z.
(4) If x x0 and y y0 , then sðx; yÞ sðx0 ; y0 Þ.
Definition 5.3 (Klir and Yuan 1995) A t-norm sðx; yÞ is called Archimedean
t-norm if it is continuous and sðx; xÞ\x for all x 2 ð0; 1Þ. An Archimedean t-norm
is called a strict Archimedean t-norm if it is strictly increasing in each variable for
x; y 2 ð0; 1Þ.
5.3 IFWA Operator and IFIA Operator 121
Definition 5.4 (Klir and Yuan 1995) A t-conorm sðx; yÞ is called an Archimedean
t-conorm if it is continuous and sðx; xÞ [ x for all x 2 ð0; 1Þ. An Archimedean
t-conorm is called a strict Archimedean t-conorm if it is strictly increasing in each
variable.
Klement and Mesiar (2005) proved that a strict Archimedean t-conorm can be
expressed via its additive generator h as sðx; yÞ ¼ h1 ðhðxÞ þ hðyÞÞ, where h is a
strictly increasing continuous function, and similarly, applied to its dual t-norm
sðx; yÞ ¼ g1 ðgðxÞ þ gðyÞÞ with hðtÞ ¼ gð1 tÞ.
Later on, some operational laws of IFNs based on Archimedean t-norm and
t-conorm were given:

Definition 5.5 (Xia et al. 2012a) Let a ¼ðla ; ma Þ and b ¼ lb ; mb be two IFNs,
then

(1) a b ¼ h1 ðhðla Þ þ hðlb ÞÞ; g1 ðgðma Þ þ gðmb ÞÞ .

(2) a b ¼ g1 ðgðla Þ þ gðlb ÞÞ; h1 ðhðma Þ þ hðmb ÞÞ .
(3) ka ¼ ðh1 ðkhðla ÞÞ; g1 ðkgðma ÞÞÞ; k 0.
(4) ak ¼ ðg1 ðkgðla ÞÞ; h1 ðkhðma ÞÞÞ; k 0.

Definition 5.6 (Lei and Xu 2016) Let a ¼ðla ; ma Þ and b ¼ lb ; mb be two IFNs,
then
(1) The subtraction based on Archimedean t-conorm and t-norm:

h1 ðhðlb Þ hðla ÞÞ; g1 ðgðmb Þ gðma ÞÞ ; if 0 hðlb Þ hðla Þ gðmb Þ gðma Þ ;
ba ¼
O; otherwise:
(2) The division based on Archimedean t-conorm and t-norm:

g1 ðgðlb Þ gðla ÞÞ; h1 ðhðmb Þ hðma ÞÞ ; if 0 hðmb Þ hðma Þ gðlb Þ gðla Þ ;
b a ¼
E; otherwise:
Next, we analyze the subtraction and the division based on Archimedean

t-conorm and t-norm. If 0 hðlb Þ hðla Þ gðmb Þ gðma Þ, then

(1) 0 h1 hðlb Þ hðla Þ 1.

(2) 0 g1 gðmb Þ gðma Þ 1.

(3) 0 g1 gðmb Þ gðma Þ þ h1 hðlb Þ hðla Þ ¼ 1 h1 gðmb Þ gðma Þ

þ h1 hðlb Þ hðla Þ 1:

which shows that b a¼ h1 ðhðlb Þ hðla ÞÞ; g1 ðgðmb Þ gðma ÞÞ is still an IFN
when 0 hðlb Þ hðla Þ gðmb Þ gðma Þ. In addition, there is also

ðb aÞ a ¼ h1 ðhðlb Þ hðla ÞÞ; g1 ðgðmb Þ gðma ÞÞ h1 ðhðla Þ; g1 ðgðma Þ

¼ h1 ðhðlb Þ; g1 ðgðmb Þ ¼ b
which indicates that the subtraction is just the inverse operation of the addition in
Definition 5.5.
Actually, if there are gðtÞ ¼ ln t, hðtÞ ¼ ln ð1 tÞ, g1 ðtÞ ¼ et and h1 ðtÞ ¼
1 et in Definitions 5.5 and 5.6, then these operations based on Archimedean
t-norm and t-conorm reduce (Xu 2013a) to the corresponding ones in Chap. 1.
5.3.1 Integral Aggregating Value Based on Archimedean

T-Norm and T-Conorm
RR
If we handle the process of obtaining D f ðXÞXdd in Definitions 1.3 and 1.4 of
Chap. 4 with replacing the addition and the scalar multiplication of IFNs by the
operational laws based on Archimedean t-norm and t-conorm in Definition 5.5, then
we have the following theorem:
Theorem 5.3 (Lei and Xu 2016) Let D be a region of IFNs, and f ðXÞ( f ðl; mÞ) be a
non-negative function
RR of D. Then the value acquired by following
RR the steps in the
process of defining D f ðXÞXdd, can be defined as ATS D f ðXÞXdd. In addi-
RR
tion, ATS D f ðXÞXdd is still an IFN, and can be expressed as:
0 0 1 0 11
ZZ ZZ ZZ
ATS f ðXÞXdd ¼ @h1 @ f ðl; mÞ h ðlÞ ddA; g1 @ f ðl; mÞ gðmÞ ddAA
D D D
RR
Proof According to the process of defining D f ðXÞXdd and the operations in
Definition 5.5, we can get
ZZ
ATS f ðXÞXdd
D

k
¼ lim ATS f ðni ; gi Þðni ; gi ÞDdi
d!0 i¼1
! !!
X k X
k
¼ h1 lim f ðni ; gi Þhðni ÞDdi ; g1 lim f ðni ; gi Þgðgi ÞDdi
d!0 d!0
i¼1 i¼1
0 0 1 0 11
ZZ ZZ
¼ @h1 @ f ðl; mÞ h ðlÞ ddA; g1 @ f ðl; mÞ gðmÞ ddAA
D D
RR
Next, we prove that ATS D f ðXÞXdd is still an IFN. For hðtÞ ¼ gð1 tÞ, and
g is a strictly decreasing function defined in the interval ½0; 1
, then h is a strictly
increasing function, and we can get
0 1
ZZ
0 h1 @ f ðl; mÞ h ðlÞ ddA 1
D
and
0 1
ZZ
1 @
0g f ðl; mÞ gðmÞ ddA 1
D
In addition, we have
0 1 0 1
ZZ ZZ
0 h1 @ f ðl; mÞ h ðlÞ ddA þ g1 @ f ðl; mÞ gðmÞ ddA
D D
0 1 0 1
ZZ ZZ
h1 @ f ðl; mÞ h ðlÞ ddA þ g1 @ f ðl; mÞ gð1 lÞ ddA
D D
0 1 0 1
ZZ ZZ
¼ h1 @ f ðl; mÞ h ðlÞ ddA þ 1 h1 @ f ðl; mÞ hðlÞ ddA ¼ 1
D D
RR
Hence, we can obtain that ATS D f ðXÞXdd is still an IFN.
1 t
Especially, if gðtÞ ¼ ln t, which
RR means that hðtÞ ¼ ln
RRð1 tÞ, g ðtÞ ¼ e
1 t
and h ðtÞ ¼ 1 e , then ATS D f ðXÞXdd reduces to D f ðXÞXdd, which is
introduced in Chap. 4:
0 8 98 91
ZZ < ZZ =< ZZ =
f ðXÞXdd ¼ @1 exp f ðl; mÞ ln ð1 lÞ dd ; exp f ðl; mÞ ln m dd A
: ; : ;
D D D
RR
Some properties of ATS f ðXÞXdd can be given as:
D
S
Theorem 5.4 (Lei and Xu 2016) Let D ¼ ni¼1 Di , which satisfies Di \ Dj ¼ ø
when i 6¼ j, and f ðXÞ be a non-negative, then
0 1
ZZ ZZ
n
ATS f ðXÞXdd ¼ @ATS f ðXÞXddA
i¼1
D Di
Proof According to Definition 5.5, we have

0 1
ZZ
n
@ATS f ðXÞXddA
i¼1
Di
0 0 1 0 11
ZZ ZZ
n
¼ @h1 @ f ðl; mÞ hðlÞ ddA; g1 @ f ðl; mÞ gðmÞ ddAA
i¼1
Di Di
0 0 1 0 11
n ZZ
X n ZZ
X
i¼1 i¼1
Di Di
0 0 1 0 11
ZZ ZZ
ZZD D
¼ ATS f ðXÞXdd
D
which completes the proof of Theorem 5.4.

Theorem 5.5 (Lei and Xu 2016) If there are two regions of IFNs D1 and D2 , which
meet D2 D1 , then
0 1 0 1
ZZ ZZ ZZ
@ATS f ðXÞXddA@ATS f ðXÞXddA ¼ ATS f ðXÞXdd
D1 D2 D1 D2
Proof For hðlÞ ¼ gð1 lÞ gðmÞ, then

ZZ ZZ ZZ
0 f ðl; mÞ hðlÞ dd f ðl; mÞ hðlÞ dd f ðl; mÞ gðmÞ dd
ZZ
D1 D2 D1
f ðl; mÞ gðmÞ dd
D2
By the calculating formula of the subtraction in Definition 5.6, we can get

0 1 0 1
ZZ ZZ
@ATS f ðXÞXddA@ATS f ðXÞXddA
D1 D2
0 0 1 0 11
ZZ ZZ
D1 D1
0 0 1 0 11
ZZ ZZ
@h1 @ f ðl; mÞ hðlÞ ddA; g1 @ f ðl; mÞ gðmÞ ddAA
D2 D2
0 0 1 0 11
ZZ ZZ ZZ ZZ
¼ @h1 @ f ðl; mÞ hðlÞ dd f ðl; mÞ hðlÞ ddA; g1 @ f ðl; mÞ gðmÞ dd f ðl; mÞ gðmÞ ddAA
D1 D2 D1 D2
0 0 1 0 11
ZZ ZZ
D 1 D 2 D 1 D 2
ZZ
¼ ATS f ðXÞXdd
D1 D2

Theorem 5.6 (Lei and Xu 2016) If there are n non-negative functions of D(fi ðXÞ (
i ¼ 1; 2; . . .; n)), andP
x i 0 ( i ¼ 1; 2; . . .; n) are respectively the weights of fi ðXÞ(
i ¼ 1; 2; . . .; n) with ni¼1 x i ¼ 1, then
! 0 1
ZZ X
n ZZ
n
ATS x i fi ðXÞ Xdd ¼ x i @ATS fi ðXÞXddA
i¼1
i¼1
D D
Proof By these operational laws of IFNs based on Archimedean t-norm and

t-conorm, we have
0 1
ZZ
n
x i @ATS fi ðXÞXddA
i¼1
D
0 0 1 0 11
ZZ ZZ
n
¼ x i @h1 @ fi ðl; mÞ hðlÞ ddA; g1 @ fi ðl; mÞ gðmÞ ddAA
i¼1
D D
0 0 ! 1 0 ! 11
ZZ X
n ZZ X
n
¼ @h @
1
x i fi ðl; mÞ hðlÞ ddA; g 1 @
x i fi ðl; mÞ gðmÞ ddAA
i¼1 i¼1
D D
ZZ !
X
n
¼ ATS x i fi ðXÞ Xdd
i¼1
D
The proof is completed.

Theorem 5.7 (Lei and Xu 2016) Let f1 ðXÞ and f2 ðXÞ be two non-negative
functions defined in D; which satisfy f1 ðXÞ f2 ðXÞ; and thus, f1 ðXÞ f2 ðXÞ is still a
non-negative functions, then
0 1 0 1
ZZ ZZ ZZ
ATS ðf1 ðXÞ f2 ðXÞÞX dd ¼ @ATS f1 ðXÞX ddA@ATS f2 ðXÞX ddA
D D D
Proof According to the corresponding operations, we have

0 1 0 1
ZZ ZZ
@ATS f1 ðXÞX ddA@ATS f2 ðXÞX ddA
D D
0 0 1 0 11
ZZ ZZ
¼ @h1 @ f1 ðl; mÞ hðlÞ ddA; g1 @ f1 ðl; mÞ gðmÞ ddAA
D D
0 0 1 0 11
ZZ ZZ
@h1 @ f2 ðl; mÞ hðlÞ ddA; g1 @ f2 ðl; mÞ gðmÞ ddAA
D D
0 0 1 0 11
ZZ ZZ ZZ ZZ
¼ @h 1 @
f1 ðl; mÞ hðlÞ dd f2 ðl; mÞ hðlÞ ddA; g 1 @
f1 ðl; mÞ gðmÞ dd f2 ðl; mÞ gðmÞ ddAA
D D D D
ZZ
¼ ATS ðf1 ðXÞ f2 ðXÞÞX dd
D

Theorem 5.8 (Lei and Xu 2016) Let p [ 1, 1=p þ 1=q ¼ 1, f1 ðXÞ and f2 ðXÞ be two
non-negative functions defined in D, then
0 1 0 1
ZZ ZZ ZZ q
f1p ðXÞ f ðXÞ
ATS f1 ðXÞ f2 ðXÞXdd @ATS XddA @ATS 2
XddA
p q
D D D
Especially, if we let p ¼ q ¼ 2, then

RR 2 RR 2
ZZ ATS f1 ðXÞXdd ATS f2 ðXÞXdd
D D
ATS f1 ðXÞ f2 ðXÞXdd
2
D
Proof Because f1 ðXÞ and f2 ðXÞ are two non-negative functions defined in D, which
indicate that f1 ðXÞ 0 and f2 ðXÞ 0 for any X 2 D, by the Young inequality, we
have
f1p ðXÞ f2q ðXÞ

f1 ðXÞf2 ðXÞ þ
p q
In addition, since gðmÞ 0, then
f1p ðXÞ f q ðXÞ

f1 ðXÞf2 ðXÞgðmÞ gðmÞ þ 2 gðmÞ
p q
Hence,
ZZ ZZ ZZ
f1p ðXÞ f2q ðXÞ
f1 ðXÞf2 ðXÞgðmÞdd gðmÞdd þ gðmÞdd
p q
D D D
Moreover, because g1 ðmÞ is also a strictly decreasing function, then

0 1 0 1
ZZ ZZ ZZ
f1p ðXÞ f2q ðXÞ
g1 @ gðmÞdd þ gðmÞddA g1 @ f1 ðXÞf2 ðXÞgðmÞddA
p q
D D D
Similarly, we can prove

0 1 0 1
ZZ ZZ ZZ
f1p ðXÞ f2q ðXÞ
h1 @ f1 ðXÞf2 ðXÞhðlÞddA h1 @ hðlÞdd þ hðlÞddA
p q
D D D
Therefore, we can get

0 1 0 1
ZZ ZZ ZZ q
f1p ðXÞ f ðXÞ
ATS f1 ðXÞf2 ðXÞXdd @ATS XddA @ATS 2
XddA
p q
D D D
The proof is completed.

Theorem 5.9 (Lei and Xu 2016) Let f ðXÞ be a continuous non-negative function,
then
ZZ
ATS f ðXÞX dd ¼ O , f ðXÞ ¼ 0
D
Proof
00 1 0 11
ZZ ZZ ZZ
ATS f ðXÞX dd ¼ @h1 @ f ðl; mÞ h ðlÞ ddA; g1 @ f ðl; mÞ gðmÞ ddAA ¼ O
D D D
0 1 0 1
ZZ ZZ
, h1 @ f ðl; mÞ h ðlÞ ddA ¼ 0 and g1 @ f ðl; mÞ gðmÞ ddA ¼ 1
D D
ZZ ZZ
, f ðl; mÞ h ðlÞ dd ¼ 0 and f ðl; mÞ gðmÞ dd ¼ 0
D D
, f ðl; mÞ ¼ 0
which completes the proof.

Theorem 5.10 (Lei and Xu 2016) Let D 2 be a subregion of D 1 , which means that
D 2 D 1 , and f ðXÞ be a non-negative function defined in D 1 , then
ZZ ZZ
ATS f ðXÞX dd ATS f ðXÞX dd
D1 D2
The proof is omitted here.
5.3.2 IFIA Operator Based on Archimedean T-Norm

and T-Conorm
According to the integral aggregating value based on Archimedean t-norm and

t-conorm, we can get the corresponding IFIA operator:
Theorem 5.11 RR (Lei and Xu 2016) When the non-negative real function f ðXÞ
satisfies that D f ðXÞdd ¼ 1, then we call

k
ATSIFIAðD; f ðXÞÞ ¼ lim f ðni ; gi Þðni ; gi ÞDdi
d!0 i¼1
0 0 1 0 11
ZZ ZZ
¼ @h1 @ f ðl; mÞ h ðlÞ ddA; g1 @ f ðl; mÞ gðmÞ ddAA
D D
an intuitionistic fuzzy integral averaging (ATS-IFIA) operator based on

Archimedean t-conorm and t-norm. In addition, if f ðXÞ ¼ dðX; a1 ; a2 ; . . .; an Þ in-
troduced in Chapter 4, then the ATS-IFIA operator reduces to the IFWA operator
based on Archimedean t-conorm and t-norm (ATS-IFWA):
n
ATSIFWAx ða1 ; a2 ; . . .; an Þ ¼ xi ai
i¼1 ! !!
X
n X
n
1 1
¼ h xi hðlai Þ ; g xi gðmai Þ
i¼1 i¼1
Proof According to the definition of dðX; a1 ; a2 ; . . .; an Þ, we have

ZZ
ATSIFIAðD; dðX; a1 ; a2 ; . . .; an ÞÞ ¼ ATS dðX; a1 ; a2 ; . . .; an ÞX dd
D
0 0 1 0 11
ZZ ZZ
¼ @h1 @ dðX; a1 ; a2 ; . . .; an ÞhðlÞddA; g1 @ dðX; a1 ; a2 ; . . .; an ÞgðmÞddAA
D D
! !!
X
n X
n
¼ h1 xi hðlai Þ ; g1 xi gðmai Þ
i¼1 i¼1
¼ ATSIFWAx ða1 ; a2 ; . . .; an Þ

In the following, we will show several basic properties of the ATS-IFIA
operator:
Theorem 5.12 (Idempotency) (Lei and Xu 2016) Let D be the region of IFNs,
which only includes one point (an IFN) D ¼ f a0 g. Then the aggregated value by
utilizing the ATS-IFIA operator is equal to a0 .
Proof We can get that the weight density function has the following form:

0; when X 6¼ a0
f ðXÞ ¼ dðXÞ ¼
þ 1; when X ¼ a0
RR RR
and R2 dðXÞ dldm ¼ N dðXÞ dl dm ¼ 1.
RR RR
Because R2 dðXÞ lnð1 lÞdl dm ¼ lnð1 l 0 Þ and R2 dðXÞ ln m dl dm ¼ ln m0 ,
we can obtain
ZZ
ATS f ðXÞXdd ¼ a0
D

Theorem 5.13 (Boundedness) (Lei and Xu 2016) Let D be the region of IFNs, and
f ðXÞ be a weight density function defined in D. Then we can get
ZZ
a ATS f ðXÞXdd a þ
D
where a ¼ ðinf a2D f la g; supa2D f ma gÞ and a þ ¼ ðsupa2D f la g; inf a2D f ma gÞ.

Proof It is easy to get
RR RR
(1) h1 D f ðl; mÞ h ðlÞ dd h1 D f ðl; mÞ h ðsupf la gÞ dd ¼ supa2D f la g.
RR RR
(2) g1 D f ðl; mÞ gðmÞ dd g1 D f ðl; mÞ gðinf f ma gÞ dd ¼ inf a2D f ma g:
RR
(3) Hence, there is ATS D f ðXÞXdd a þ . Similarly, it can be obtained that
RR RR
ATS D f ðXÞXdd a . Thus, a ATS D f ðXÞXdd a þ holds.
Theorem 5.14 (Monotonicity) (Lei and Xu 2016) Give a moderate shift vector
Vec ¼ ðM; NÞ of D, which is a region of IFNs, and let D0 ¼ D þ Vec and
f ðl þ M; m þ NÞ ¼ f ðl; mÞ, then
ZZ ZZ
ATS f ðXÞXdd ATS f ðXÞXdd
D0 D
ZZ ZZ
ATS f ðXÞXdd ATS f ðXÞXdd
D0 D
Proof By the ranking rule of “ ”, if M 0 and N 0, then

!
1
RR RR
(a) h f1 ðl ; m Þ h ðl Þdd h1
0 0 0
f2 ðl; mÞh ðlÞdd :
D0 D
!
1
RR 0 0 0 1
RR
(b) g f1 ðl ; m Þgðm Þ dd g f2 ðl; mÞgðmÞdd :
D0 D
RR RR
which indicate that ATS D0 f ðXÞXdd ATS D f ðXÞXdd holds. In the same
way, we can prove the conclusion (2).
Theorem 5.11 clearly reveals the relationship between the IFIA operator and the
IFWA operator, which indicates that the IFIA operator is the continuous form of the
IFWA operator by comparing the calculating forms of the ATS-IFIA and
ATS-IFWA operators:
! !!
X
n X
n
1 1
h xi hðlai Þ ; g xi gðmai Þ
i¼1 i¼1
and
0 0 1 0 11
ZZ ZZ
@h1 @ f ðl; mÞ h ðlÞ ddA; g1 @ f ðl; mÞ gðmÞ ddAA
D D
5.4 Conclusions
Chapter 5 has mainly investigated the relationships among definitions and concepts
proposed in the previous chapters. Furthermore, a figure is provided to manifest the
relationships among the IFWA operator, the IFIA operator and the integrals of IFFs:
By Fig. 5.16, there are closed connections among the IFWA operator, the IFIA
operator and the definite integral of IFFs. The IFWA operator is actually the
Fig. 5.16 Relationships

among the IFWA operator,
the IFIA operator and the
integral of IFFs
integrals of several specific IFFs, like LðXÞ and PiðXÞ. In addition, The IFIA
operator is actually the integral of LðXÞ. Significantly, the IFIA operator is the
continuous form of the IFWA operator, which means that the IFWA operator is the
discrete form of the IFIA operator. The relationships among them are presented as
the main work of this chapter.
Chapter 6
Complement Theory of Intuitionistic
Fuzzy Calculus
In this chapter, we first introduce the complement operator of IFNs, which actually
interchanges the membership degree and the non-membership degree of an IFN.
Then we study the intuitionistic fuzzy calculus based on the complement operator to
obtain a parallel theory framework, which is the complement theory of intuitionistic
fuzzy calculus. It consists of the complement of derivative, differential, indefinite
integral, definite integral of IFFs and some aggregating operators introduced in
previous chapters. Moreover, we will investigate the relationship between the
original conclusions and their complements in detail.
Firstly, the definition of complement of an IFN is presented as follows:
Definition 6.1 (Xia et al. 2012b). Let a ¼ ðla ; ma Þ be an IFN, then we call the IFN
ðma ; la Þ the complement of a, and denote it by a.
In what follows, we study the content of previous chapters based on the concept
of complement of IFN.
6.1 Complements of Fundamental Knowledge

Related to IFNs
According to the fundamental knowledge of IFNs, we can get the following

conclusions:
Theorem 6.1 (Lei and Xu 2016a).
(1) a b ¼ a b:
(2) ka ¼ ak ; k 0:
(3) S ðaÞ ¼ S ð
aÞ:
(4) ba ¼ b a:
134 6 Complement Theory of Intuitionistic Fuzzy Calculus
(5) S ðaÞ ¼ S ðaÞ:

(6) S ka ¼ S a k ; k 0:

where S ðaÞ is the set X : X 2 S ðaÞ , but it does not represent
fX : X 62 S ðaÞ; X 2 Hg ðS ðaÞHÞ, which is essentially the complement of a set
in the usual sense of set theory. Similarly, S ðaÞ and S ka are also not the traditional
complement sets in the set theory.
Proof It is easy to prove the conclusion (1) and the conclusion (2) according to the
operational laws of IFNs. Hence, their proofs are omitted here. The conclusion
(3) can be proven as follows:
Suppose that any IFN X belongs to the set S ðaÞ, which means that there must
exists an IFN e satisfying a e ¼ X. Hence, we have
) X ¼ X ¼ a e ¼ a e ) X 2 S ð
ae¼X aÞ
So we can get S ðaÞS ðaÞ. In the same way, we can prove S ðaÞ S ðaÞ.
Therefore, S ðaÞ ¼ S ðaÞ can be easily deducible.
In addition, we can prove the conclusion (3) by Fig. 6.1 (Lei and Xu 2016a).
According to Fig. 6.1 (Lei and Xu 2016a), we can get that the images of S ðaÞ
and S ðaÞ are exactly symmetrical when considering l ¼ m as the symmetry axis.
Hence, there is ðm; lÞ 2 S ðaÞ for any ðl; mÞ 2 S ðaÞ, and vice versa. Then we can
obtain that S ðaÞ ¼ S ðaÞ holds.
We can prove the conclusion (4) via the following two situations:
mb 1
lb
(a) When 0 ma 1
la 1 holds, it yields

lb
la mb mb lb
la
ba¼ ;
and b a¼ ;
1
la m a m a 1
la
a holds.
Hence, b a ¼ b
Fig. 6.1 S ðaÞ and S ðaÞ

6.1 Complements of Fundamental Knowledge Related to IFNs 135
mb 1
lb
(b) When 0 ma 1
la 1 does not hold, we have
ba¼O and

b a¼E
a still holds.
So b a ¼ b
Based on the above (a) and (b), we can get that b a ¼ b a always holds.
For the conclusion (5), it can be proven by the following two methods:
(a) For any a0 2 S ðaÞ, there is an IFN e meeting a 0 ¼ a e. Then a0 ¼ a0 ¼

a e ¼ a e, and hence, a0 2 S ðaÞ. We know S ðaÞS ð aÞ. In the same
manner, we also get S ðaÞ S ðaÞ. Hence, S ðaÞ ¼ S ð
aÞ.
(b) It is easy to get the equality of (5) by the images of “S ðaÞ” and “S ð aÞ”,
which are shown in Fig. 6.2 (Lei and Xu 2016a).
The proof of the conclusion (6) is similar to those of (3) and (5), and thus, it is
omitted here. All in all, the proofs of the conclusions (1)–(6) are completed. ■
In addition, we can know that the order “E” is based on the operations “” and
“” (as presented in Chap. 1). Then we introduce another order “E ” based on “”
and “”.
Definition 6.2 (Lei and Xu 2015c) If there is an IFN e, such that a e ¼ b, then b
is less than or equal to a, denoted by bE a. If there is an IFN e, such that a e ¼ b
and e 6¼ E, then b is less than a, denoted by b / a.
We can prove that “E ” is also an order relation in the set N, and the proof is
omitted here. In addition, we can also denote “E” by “E ” in order to distinguish
between “E” and “E ” clearer. Next, we show the relationship between “E ” and
“E ” as follows:

Theorem 6.2 (Lei and Xu 2016a) If aE b, then bE a.
Proof According to the definition of the order “E ”, if aE b, then there exists an
IFN e, which satisfies a e ¼ b, and thus, we get e such that Therefore,
a e ¼ b.
Fig. 6.2 S ðaÞ and S ðaÞ

a. In the same way, we can get aE b if bE

there is bE
a. Hence, Theorem 6.2.
holds. ■
6.2 Complements of Derivatives and Differentials of IFFs
In what follows, we first study an issue about for the complement uðXÞ of IFF
uðXÞ, whether there exists uðXÞ ¼ u ðXÞ ? It is clear that uðXÞ ¼ ðf ðlÞ; gðmÞÞ¼
ðXÞ should be equal to ðgðlÞ; f ðmÞÞ instead of ðgðmÞ; f ðlÞÞ.
ðgðmÞ; f ðlÞÞ, while u
Hence, uðXÞ 6¼ u ðXÞ. After analysis, we can get the following conclusion:
Theorem 6.3 (Lei and Xu 2016a).

ðXÞ
uðXÞ ¼ u
Specially, for a given compound IFF ðu wÞðtÞ (or uðwðtÞÞ), which means that u
and w are both IFFs and t is independent variable, then

wðtÞ ¼ u
uðwðtÞÞ ¼ u tÞ
wð

That is, ðu wÞðtÞ ¼ ðu wÞðtÞ ¼ u ðtÞ.
w
Proof Let u ¼ ðf ; gÞ be an IFF of the variable X ¼ ðl; mÞ, i.e., uðXÞ ¼ ðf ðlÞ; gðmÞÞ,
then uðXÞ ¼ ðf ðlÞ; gðmÞÞ ¼ ðgðmÞ; f ðlÞÞ, which is actually the IFF u ¼ ðg; f Þ with
¼ ðm; lÞ, and thus, uðXÞ ¼ u
respect to X Therefore, it is easy to prove the
ðXÞ.
result about the compound IFFs. ■
Below we introduce a new kind of IFFs called the second monotonically
increasing IFFs:
Definition 6.3 (Lei and Xu 2015a) If bE a, then uðbÞE uðaÞ. We call the IFF u
a second monotonically increasing IFF.
Obviously, the second monotonically increasing IFFs is defined based on the
order relation “E ”, however, it is different from the monotonically increasing IFFs
introduced in Chap. 2, which is developed based on “E ”. In the following, we
show the relationship between the monotonically increasing IFFs and the second
monotonically increasing IFFs:
Theorem 6.4 (Lei and Xu 2016a) If u is a monotonically increasing IFF with respect

must be a second monotonically increasing IFF with respect to X.
to X, then u
Proof. In fact, if u is a monotonically increasing IFF, which means that
uðYÞE uðZÞ if YE Z, then we can get that uðZÞE uðYÞ (or u u
ðZÞE if
ðYÞ)
■
ZE Y. Hence, u must be a second monotonically increasing IFF with respect to X.
6.2 Complements of Derivatives and Differentials of IFFs 137
6.2.1 Complements of Derivatives of IFFs
Based on the complements of IFFs, we will study the complements of the

derivatives of IFFs as follows:
According to the complement operator and the definition of derivative of the
IFF:

duðXÞ juðYÞ uðXÞj 1
l df ðlÞ m dgðmÞ
¼ lim ¼ ;1
dX Y!X jY Xj 1
f ðlÞ dl gðmÞ dm
we can define the complement of derivative of the IFF uðXÞ below (Lei and Xu
2016a):

duðXÞ juðYÞ uðXÞj 1
l df ðlÞ m dgðmÞ
¼ lim ¼ ;1
dX Y!X jY Xj 1
f ðlÞ dl gðmÞ dm
juðYÞ uðXÞj
Firstly, the expression lim jY Xj can be analyzed in the following cases:
Y!X
(1) When Y ! X , and because u is a monotonically increasing IFF, then in this

case, we have

juðYÞ uðXÞj uðYÞ uðXÞ ðYÞ
u Y
lim ¼ lim ¼ lim
Y!X jY X j Y!X YX Y!X ðXÞ
u X
If we let b ¼Y, a ¼X and w¼u, then

ðYÞ
u Y wðbÞ b
lim
Y!X ðXÞ
u ¼ Y!X
X lim
wðaÞ

a
Moreover, Y ! X means that Y infinitely approaches X and satisfies XE Y.

Since b ¼Y, and a ¼X, then we know that b infinitely approaches a and
b ¼YE X¼ a, which can be denoted by b ! a . Hence, we can get

wðbÞ b wðbÞ b
lim ¼ lim
Y!X wðaÞ a b!a wðaÞ a
(2) When Y ! X , there is

juðYÞuðXÞj uðXÞuðYÞ ðXÞ X
u
lim ¼ lim ¼ lim
Y!X jYXj Y!X XY Y!X ðYÞ
u Y
Similarly, we let b ¼Y, a ¼X and w¼u. In addition, Y ! X means YE X,

hence, there is a ¼XE Y¼ b, which is denoted by b ! a . Then we have

uðXÞ X wðaÞ a
lim ¼ lim
Y!X uðYÞ Y b!a wðbÞ b
According to the above (1) and (2), we can give the definition of the complement
of derivative of the IFF:
Definition 6.4 (Lei and Xu 2015b) Let uðXÞ be a second monotonically increasing
IFF, if the value

uðYÞ

Y

b
b=a; if b 2 S ðaÞ

where
¼
lim
Y!X
uðXÞ

X
; a a=b; if b 2 S ðaÞ
Y2S ðXÞ [ S ðXÞ
is still an IFN, then we call it the complement of derivative of uðXÞ, denoted by

luðXÞ
lX .
Theorem 6.5 (Lei and Xu 2016a) If the complement of derivative of uðXÞ ¼

ðf ðlÞ; gðmÞÞ exists, then

luðXÞ l df ðlÞ 1
m dgðmÞ Ef ðlÞ E1
gðmÞ
¼ 1
; ¼ 1
;
lX f ðlÞ dl 1
gðmÞ dm El E1
m
This theorem can be proven based on the operational laws of IFNs, which is
omitted here. By Theorem 6.5, we can get
luðXÞ lX0 luðXÞ lX k

¼ ¼ E; ¼ ¼ ð1
k; kÞ
lX lX lX lX
where X 0 is a constant IFN.

Theorem 6.6 (Lei and Xu 2016a)

uðXÞ
duðXÞ l
¼
dX lX
Proof It is easy to prove the theorem based on the definition of the complement of
derivative of the IFF, and by Theorem 6.5, we have

uðXÞ
l m dgðmÞ 1
l df ðlÞ
¼
lX
1
;
gðmÞ dm 1
f ðlÞ dl
¼ ðgðmÞ; f ðlÞÞ, then

ðXÞ
since u

duðXÞ 1
l df ðlÞ m dgðmÞ uðXÞ
l
¼ ;1
¼
dX 1
f ðlÞ dl gðmÞ dm lX
holds. The proof is completed. ■

In what follows, we analyze the chain rule of the complements of derivatives
(Lei and Xu 2016a):
We know that for a compound IFF uðwðtÞÞ,
duðwðtÞÞ duðwðtÞÞ dwðtÞ

¼
dt dwðtÞ dt
Then we can get
duðwðtÞÞ duðwðtÞÞ dwðtÞ

¼
dt dwðtÞ dt
uðwð
tÞÞ
For the left-hand side of the equality, there is duðdtwðtÞÞ ¼ luðwðtÞÞ l
lt ¼ lt .
Meanwhile, for the right-hand side of the equality, we have

duðwðtÞÞ dwðtÞ l
u wð tÞ tÞ
lwð
¼
dwðtÞ dt
lwðtÞ lt
Hence, we can obtain

u wð
l tÞ u wð
l tÞ tÞ
lwð
¼

lt lwðtÞ lt
¼ Y and t ¼ k, then
¼ /, w
If we let u
l/ðYðkÞÞ l/ðYðkÞÞ lYðkÞ

¼
lk lYðkÞ lk
which is just the chain rule of the complement of derivatives.

Some properties of the complement of derivative can be presented as follows:
luk ðXÞ
(1) lX ¼ ð1
k ; kÞ luðXÞ
lX ; where 0 k 1
lði¼1 ui ðXÞÞ
n Pn P n
lui ðXÞ lui ðXÞ
(2) lX ¼ 1
1
U lX ; V lX .
i¼1 i¼1
(3) If ui ðXÞE uj ðXÞ, then

l ui ðXÞuj ðXÞ luj ðXÞ lui ðXÞ lui ðXÞ luj ðXÞ
¼ 1
U
U ;V
V
lX lX lX lX lX
Proof For the conclusion (1), we have
d duðXÞ
ðkuðXÞÞ ¼ ðk; 1
kÞ
dX dX
Therefore, we can get
d duðXÞ
ðkuðXÞÞ ¼ ðk; 1
kÞ
dX dX

Then the left-hand side of equality is equal to lXl u , and the right-hand side
k ðXÞ
uðXÞ
ðk; 1
kÞ duðXÞ l
dX is equal to ð1
k; kÞ lX . Hence, this conclusion holds.
For the conclusion (2), we notice that

Xn X n !
d n d d
ui ðXÞ ¼ U u ðXÞ ; 1
1
V u ðXÞ
dX i¼1
i¼1
dX i i¼1
dX i
X n X n !
l n
d d
, u ðXÞ ¼ 1
1
V u ðXÞ ; U u ðXÞ
lX i¼1 i i¼1
dX i i¼1
dX i
X n X n !
l n
¼ 1
d d
, u ðXÞ 1
U u ðXÞ ; V u ðXÞ
lX i¼1 i i¼1
dX i i¼1
dX i
X n X n !
l n
¼ 1
l l
, u ðXÞ 1
U u ðXÞ ; V u ðXÞ
lX i¼1 i i¼1
lX i i¼1
lX i
which completes the proof of (2).

For the conclusion (3), we know that if uk ðXÞEul ðXÞ, which means
u
l ðXÞE
u then
k ðXÞ,

d ðul ðXÞuk ðXÞÞ dul ðXÞ duk ðXÞ duk ðXÞ dul ðXÞ
¼ U
U ;1
V
V
dX dX dX dX dX
uk ðXÞ
l ðXÞ
lðu Þ
duk ðXÞ

dul ðXÞ

dul ðXÞ

duk ðXÞ

, ¼ 1
V
V ;U
U
lX dX dX dX dX
! !! ! !!

l ðXÞ
lðu
uk ðXÞÞ duk ðXÞ dul ðXÞ dul ðXÞ duk ðXÞ
, ¼ 1
U
U ; V
V
lX dX dX dX dX

l ðXÞ
lðu uk ðXÞÞ uk ðXÞ
l uk ðXÞ
l uk ðXÞ
l uk ðXÞ
l
, ¼ 1
U
U ; V
V
lX lX lX lX lX
The proof of the conclusion (3) is completed. ■
6.2.2 Complements of Differentials of IFFs
In this subsection, we introduce the complements of differentials of IFFs based on

the complements of derivatives.
Definition 6.5 (Lei and Xu 2015b) If the complement of derivative of u exists, and
we denote rX ¼ X 0 X, then we call
luðXÞ
luðXÞ ¼ rX
lX
the complement of differential of u. In addition, since lX ¼ O rX ¼ rX, then

luðXÞ can also be represented by
luðXÞ
luðXÞ ¼ lX
lX
By the definition of “duðXÞ” and “luðXÞ”, it is easy to obtain the following

theorem:
duðXÞ ¼ l
uðXÞ
Proof For
duðXÞ l
uðXÞ
duðXÞ ¼ ¼ l
DX ¼ rX
uðXÞ
dX lX
we can get that duðXÞ ¼ l holds.

uðXÞ ■
Theorem 6.9 (Lei and Xu 2015b) If the complement of derivative of w, then
lwðaÞ
wðbÞwðaÞ ðbaÞ
la
If we denote rw ¼ wðbÞwðaÞ, then
UðrwÞ
UðlwÞ VðrwÞ
VðlwÞ
lim ¼ 0 and lim ¼0
Dl!0 Dl Dm!0 Dm
Proof According the differential formula, we can get
duðXÞ
uðYÞ uðXÞ ðYXÞ
dX
u
ðYÞ
the left-hand side of which is equal to u while its right-hand side is just
ðXÞ,

uðXÞ
l
ðY X Þ, then
lX

uðXÞ
l
uðXÞ
ðYÞ
u ðY X Þ
lX
If we denote b ¼Y, a ¼X and w ¼u, then
lwðaÞ
wðbÞwðaÞ ðbaÞ
la
which completes the proof. ■

The following example (Lei and Xu 2015b) can be provided to verify the
conclusion in Theorem 6.8:
Suppose that uðXÞ ¼ X k ð0 k 1Þ, which means that f ðlÞ ¼ lk and
gðmÞ ¼ 1
ð1
mÞk . In addition, we have
!
luðXÞ 1
l k
1
¼ 1
k; kð1
lÞ ¼ ð1
k; kÞ
lX ð1
lÞk
By Theorem 6.8, we can get
uðX DXÞ X DX
ð1
k; kÞ ¼ ð1
k; kÞ DX
uðXÞ X
In addition, according to the operational laws of IFNs, we can also get

ðX DXÞk ¼ Xk ðDX Þk . Hence
uðX DXÞ ðX DXÞk

¼ ¼ ðDXÞk
uðXÞ Xk
Assume that DX ¼ ð0:91; 0:04Þ and k ¼ 0:5, we can get that ð1

k; kÞ DX ¼
ð0:955; 0:02Þ and ðDX Þk ¼ ð0:9539392; 0:0202041Þ. Hence, ð1
k; kÞ DX is
close to ðDXÞk .
In what follows, we analyze the form invariance of the complement of differ-
ential (Lei and Xu 2016a):
¼u
If there are a compound IFF Y tÞÞ, which consists of two IFFs Y
ðwð ¼u
ðXÞ

and X ¼ wðtÞ, then
(1) According to dY ¼ dX
dY
dX and dY ¼ dY
dt dt, we get

¼ lY lX
lY and ¼
lY
lY
lt

lX lt
(2) With dY
dt ¼ dX
dY
dX
dt and dX ¼ dt dt, we know
dX
lY lY lX
lX
¼ and ¼
lX lt
lt lX lt lt
and then
lY lX
lY
lY
¼
lY
lt ¼ lt ¼ lX

lt
lX lt lX

which is just the form invariance of the complement of differential of the IFF Y.
In brief, this section has mainly studied the complements of derivatives and
differentials of IFFs, and acquired the following conclusions (Lei and Xu 2016a):

ðXÞ.
(1) uðXÞ ¼ u

(2) duðXÞ
dX ¼ lulðXXÞ.
(3) duðXÞ ¼ l
uðXÞ.
6.3 Complements of Integrals of IFFs
In this section, we investigate the complements of integrals of IFFs from the aspects
of the indefinite integrals and the definite integrals of IFFs.
6.3.1 Complements of Indefinite Integrals of IFFs
Based on the definition of the complement of derivative, as the inverse operation of

it, the complement of indefinite integral of the IFF can be discussed:
Definition 6.6 (Lei and Xu 2015c) If there is an IFF U ðXÞ, which satisfies
l U ðXÞ
lX ¼ uðXÞ¼ ðf ðlÞ; gðmÞÞ, then it should have the following form:
Z Z
1
f ðlÞ gðmÞ
U ðXÞ ¼ c 1 exp dl ; 1
c 2 exp
dm
l 1
m
which can be called the complement of indefinite integral of the IFF, and denoted
by o uðXÞ lX.
Z
uðXÞdX ¼ o u lX
ðXÞ
Proof Below we will prove the conclusion in two different ways, one of which is
shown as:
R
For any IFF UðXÞ 2 uðXÞdX, we have dUðXÞ dX ¼ uðXÞ based on the definition
of the indefinite integral of the IFF. With dUðXÞ lUðXÞ

ðXÞ,
dX ¼ uðXÞ, we get lX ¼u
R
which means that UðXÞ 2 o u ðXÞ lX. Hence, there is uðXÞdX o u

ðXÞlX. In the
R
same way, we can also know that uðXÞdX o u
ðXÞlX. In summary, we can
R
know uðXÞdX ¼ o uðXÞlX.
Another proof method of this theorem is to utilize the formulas of the indefinite
integrals of IFFs. For any antiderivative UðXÞ, which meets dUðXÞ
dX ¼ uðXÞ, it must
have the following form:
Z Z
f ðlÞ 1
gðmÞ
UðXÞ ¼ 1
c 1 exp
dl ; c 2 exp dm ð1Þ
1
l m
where c 1 and c2 are two integral constants, which are real numbers such that UðXÞ
is an IFF. Then we denote A as the following set:

Z Z

f ðlÞ 1
gðmÞ
A¼ hc1 ; c2 i
UðXÞ ¼ 1
c 1 exp
dl ; c 2 exp dm is an IFF
1
l m
where hc1 ; c2 i is only a two dimensional vector, but not an IFN.

Moreover, for any antiderivative W ðXÞ satisfying l W ðXÞ it has
ðXÞ,
¼u
lX
Z Z
¼ 1
gðmÞ f ðlÞ
W ðXÞ c3 exp dm ; 1
c 4 exp
dl ð2Þ
m 1
l

where c 3 and c 4 are two integral constants, which are real numbers such that W ðXÞ
is an IFF. Then we let

Z Z

1
gðmÞ f ðlÞ
B¼ hc3 ; c4 i
W ðXÞ
¼ c3 exp dm ; 1
c 4 exp
dl is an IFF
m 1
l
where hc1 ; c2 i is also a two dimensional vector, but not an IFN.

Hence, we can obtain that for any given UðXÞ, it has the form of the equality (1),
and yields
6.3 Complements of Integrals of IFFs 145
Z Z
1
gðmÞ f ðlÞ
UðXÞ ¼ c 2 exp dm ; 1
c 1 exp
dl
m 1
l
Obviously, we can get hc2 ; c1 i 2 B according to the definition of the sets A and
R
B, and thus, lUlXðXÞ

which means that uðXÞdX o u
ðXÞ,
¼u lX.
ðXÞ

Furthermore, any IFF W ðXÞ satisfying l W ð

XÞ
¼u has the form of the
ðXÞ
lX
equality (2):
Z Z
¼ f ðlÞ 1
gðmÞ
W ðXÞ 1
c 4 exp
dl ; c3 exp dm
1
l m
R
Since hc4 ; c3 i 2 A, then we get dWdXðXÞ ¼ uðXÞ, that is, o u

lX
ðXÞ uðXÞdX.
R R
Then, o uðXÞ lX uðXÞdX, hence, we get that uðXÞdX ¼ o uðXÞ lX.
■
Theorem 6.11 (Lei and Xu 2015a, 2016b)
o wðYðkÞÞ lYðkÞ
lk
lk ¼ WðYðkÞÞ
Proof According to the substitution rule of the indefinite integral, which is

Z
uðXðtÞÞX 0 ðtÞdt ¼ UðXðtÞÞ
We have
Z Z
0 dXðtÞ
uðXðtÞÞX ðtÞdt ¼ uðXðtÞÞ dt ¼ UðXðtÞÞ
dt

) o uðXð
tÞÞ lXðtÞ lt ¼ UðXð
tÞÞ
lt
R
where UðXÞ ¼ uðXÞdX ¼ o uðXÞ lX.
If we let u ¼ w, X ¼ Y, t ¼ k and U ¼ W, then
o wðYðkÞÞ lYðkÞ
lk
lk ¼ WðYðkÞÞ

Theorem 6.12 (Lei and Xu 2015c, 2016a)
k
(1) o ðð1
k; kÞ uðXÞÞlX ¼ o uðXÞlX ; where 0 k 1.

Pn Pn
(2) o 1
ð1
fi ðlÞÞ; gi ðmÞ lX ¼ o ðfi ðlÞ; gi ðmÞÞlX.
n
i¼1 i¼1 i¼1

(3) oð1
ðf2 ðlÞ
f1 ðlÞÞ; g1 ðmÞ
g2 ðmÞÞlX ¼ oðf1 ðlÞ; g1 ðmÞÞlX oðf2 ðlÞ; g2 ðmÞÞlX:
Proof We will prove them based on the corresponding conclusion of the indefinite
integrals of IFFs.
R R
Firstly, since ðk; 1
kÞ uðXÞdX ¼ k uðXÞdX, then we get
Z
ðk; 1
kÞ uðXÞdX ¼ o ðk; 1
kÞ uðXÞlX
¼ o ð1
k; kÞ u X
ðXÞl
Moreover
Z Z k
k
k uðXÞdX ¼ uðXÞdX ¼ o u ðXÞl
X
k
Then we get that o ðð1
k; kÞ uðXÞÞlX ¼ o uðXÞlX holds.
For the conclusion (2), we have
Z ! Z
X
n X
n
n
hi ðlÞ; 1
ð1
ki ðmÞÞ dY ¼ ðhi ðlÞ; ki ðmÞÞdY
i¼1
i¼1 i¼1
!
X
n X
n
,o o ðhi ðlÞ; ki ðmÞÞ lY
n
hi ðlÞ; 1
¼
ð1
ki ðmÞÞ lY
i¼1
i¼1 i¼1
!
X
n X
n
,o 1
oðki ðmÞ; hi ðlÞÞ lY
n
ð1
ki ðmÞÞ; ¼
hi ðlÞ lY
i¼1
i¼1 i¼1
If we denote X ¼ Y, f ¼ k and g ¼ h, then U(X) ¼ V(Y), V(X) ¼ U(Y) and the

above equation is equivalent to the following form:
!
X
n X
n
o oðfi ðlÞ; gi ðmÞÞ lX
n
1
ð1
fi ðlÞÞ; gi ðmÞ lX ¼
i¼1
i¼1 i¼1
which completes the proof of (2).

For the conclusion (3), we know that
Z Z Z
ðh1 ðlÞ
h2 ðlÞ; 1
ðk2 ðmÞ
k1 ðmÞÞÞdY ¼ ðh1 ðlÞ; k1 ðmÞÞdY ðh2 ðlÞ; k2 ðmÞÞdY
, o ðh1 ðlÞ
h2 ðlÞ; 1
ðk2 ðmÞ
k1 ðmÞÞÞ lY
¼ o ðh1 ðlÞ; k1 ðmÞÞ lY
o ðh2 ðlÞ; k2 ðmÞÞ lY

, oð1
ðk2 ðmÞ
k1 ðmÞÞ; h1 ðlÞ
h2 ðlÞÞ lY ¼ oðk1 ðmÞ; h1 ðlÞÞ lY
oðk2 ðmÞ; h2 ðlÞÞ lY

If we denote X ¼ Y, f ¼ k and g ¼ h, then
oð1
ðf2 ðlÞ
f1 ðlÞÞ; g1 ðmÞ
g2 ðmÞÞlX ¼ oðf1 ðlÞ; g1 ðmÞÞlX oðf2 ðlÞ; g2 ðmÞÞlX
which completes the proof of (3). ■
6.3.2 Complements of Definite Integrals of IFFs
Before introducing the complement of definite integral of the IFF, we first study the
complements of IFICs.
According to the concept of “IFIC” based on the order relation “E ”, we can
define a kind of similar curves related to “E ”.
Definition 6.7 (Lei and Xu 2015c) Let J be a curve linking between a and b that
can be written as a bijective mapping J : ½0; L ! J, where L is the length from a to
b. This mapping satisfies Jð0Þ ¼ a and JðLÞ ¼ b. If Jðs2 ÞE Jðs1 Þ for
0 s1 s2 L, then we call J a second intuitionistic fuzzy integral curve (II-IFIC),
and several II-IFICs can be shown in Fig. 6.3 (Lei and Xu 2016a):
Theorem 6.13 (Lei and Xu 2016a) If I is an IFIC linking a and b, then I is an

II-IFIC linking a and b.
Proof We will prove it by Fig. 6.4 (Lei and Xu 2016a).
By utilizing the concept of “II-IFIC”, we can define the complement of definite
integral of the IFF as follows (Lei and Xu 2016a):
(1) Dividing the II-IFIC. By interpolating some break points a ¼ h0 ,
h1 ; h2 ; . . .; hn
1 , hn ¼ b, we can divide the II-IFIC J into several smaller arcs
_ _ _
ah1 , h1 h2 , . . ., hn
1 b, and these points hk ðk ¼ 0; 1; . . .; nÞ are arranged from a
to b.
Fig. 6.3 II-IFICs

Fig. 6.4 IFIC I and its

complement I
_
(2) Making to sum. From every small arc hk hk þ 1, we take an IFN nk ¼ lni ; mni

to get the value uðnk Þ ðhk þ 1 hk Þ, which can be represented as:

l i þ 1 mi þ 1
mi
f ðlni Þ; gðmni Þ ;
li 1
mi
(3) Calculating the product. We combine all uðnk Þ ðhk þ 1 hk Þ

i¼1 ðuðnk Þ
(k ¼ 0; 1; . . .; n
1) by using multiplication to get the product n
1
ðhk þ 1 hk ÞÞ.
(4) Taking the limit. If the number of the break points hk (k ¼ 0; 1; . . .; n
1)
increases infinitely, and meets hk þ 1 hk ! E (k ¼ 0; 1; . . .; n
1), then the
limits of the membership and non-membership parts of n
1 i¼1 ðuðnk Þ
ðhk þ 1 hk ÞÞ are equal to U and V, respectively, and ðU; V Þ is an IFN. In this
case, we call ðU; V Þ the limit of the expression in Step (3), and define it as the
integral of uðXÞ along the II-IFIC J, denoted by oJ uðXÞ lX.
Theorem 6.14 (Lei and Xu 2015c) oJ uðXÞ lX only depends on the two endpoints
of the II-IFIC J. In addition, there is
0 8 9 8 mb 91
< Z b 1
f ðlÞ >
> l
= < Z gðmÞ =
oJ uðXÞ lX ¼ B @exp
> l
dl ; 1
exp
> :
m
dm A
;
C
: ; 1
la ma
The proof of Theorem 6.12 is similar to one of Theorem 3.6, so it is omitted here.
Because the complement of definite integral oJ uðXÞ lX is only related to two
endpoints of the II-IFIC J, hence oJ uðXÞ lX can be denoted by oba uðXÞ lX.
Specially, oa uðXÞ lX ¼ E.
a
Next, we reveal that there is a closed connection between the definite integral of
the IFF and its complement:
Zb
b
uðXÞdX ¼ oa lX
ðXÞ
u
a
Proof We prove it in two different ways:

Rb
Firstly, by the definitions of uðXÞdX, we can get
a
Zb
k
uðXÞdX ¼ lim ðuðni Þ Ddi Þ
Dd1 ;Dd2 ;;Ddk !O i¼1
a

k k
¼ lim ðuðni Þ Ddi Þ ¼ lim u ni Þ r
ð di
r
d1 ;r
d2 ;;r
dk !E i¼1 rd1 ;rd2 ;;rdk !E i¼1

b
¼ oa lX
ðXÞ
u
In addition, this theorem can be proven in another way: For the left-hand side of
the equality, we have
0 8 mb 9 8 91
Zb < Z 1
gðmÞ = < Z b f ðlÞ >
> l
=
B C
uðXÞdX ¼ @exp dm ; 1
exp
dl A
: m ; >
: 1
l > ;
a ma la
for the right-hand side of the equality, we get

0 8 mb 9 8 91

b < Z 1
gðmÞ = < Z b f ðlÞ >
> l
=
oa ðXÞ
u ¼B
lX @exp dm ; 1
exp
dl A
C
: m ; >
: 1
l >
;
ma la
Hence, Theorem 6.13 holds. ■

Moreover, according to the equality in Theorem 6.13, we can derive some
Rb
results of a uðXÞdX to acquire some corresponding conclusions of oa uðXÞ lX
b
(Lei and Xu 2016a):

Rb Rb
(1) Since a ðk ; 1
kÞ uðXÞ dX ¼ k a uðXÞ dX, then we get

!k
b b
oa lX
ðXÞ
ð1
k ; kÞ u ¼ oa lX
ðXÞ
u
k
Specially, oE ð1
k ; kÞ lX
¼ b when a ¼ E and u ¼ O, which mean a ¼
b
ðXÞ
O and u ðXÞ ¼ E, respectively. It also reveals that “b ” can be developed by “”
k
and “” of IFNs when 0 k 1.

R b Pn Pn n Rb
(2) By a i¼1 fi ðlÞ; 1
i¼1 ð1
gi ðmÞÞ dX ¼ a ðfi ðlÞ; gi ðmÞÞdX, we
i¼1
have

b Xn Xn
b
oa o
n
1
ð1
gi ðmÞÞ; fi ðlÞ l ¼
X
ðgi ðmÞ; fi ðlÞÞ lX
i¼1 i¼1 i¼1 a
Rb Rb
(3) According to a uðXÞdX ¼ a uðXðtÞÞX0 ðtÞdt, we get
b
b
oa lX
ðXÞ
u ¼ oa u ðXð
tÞÞ lXðtÞ lt
lt
Rb Rc Rc
(4) It follows from a uðXÞdX b uðX ÞdX ¼ a uðXÞdX that

b c c
oa lX
ðXÞ
u o

lX
ðXÞ
u ¼ oa u ðXÞ
lX

b
a.
where c E bE
Rb
(5) Based on a uðX ÞdX ¼ WðbÞ WðaÞ, we have

b
oa lX
ðXÞ
u Þ Wð
¼ Wðb aÞ
Furthermore, the following conclusions can be proven on the basis of

k
o b
E
¼ b
ð1
k; kÞ lX (Lei and Xu 2015a):
(1) ak bk ¼ ða bÞk ; where 0 k 1:

(2) ak 1 ak 2 ¼ ak 1 þ k 2 ; where 0 k 1 ; k 2 1.
The processes of obtaining the two conclusions are:
a b
ak bk ¼ o ð1
k; kÞlX oð1
k; kÞlX
E E
0 8l 9 8 m 91 0 8 l b 9 8 mb 91
<Z a k = < Za k = <Z k = < Z k =
¼ @exp dl ; 1
exp
d m A @exp dl ; 1
exp
dm A
: l ; : 1
m ; : l ; : 1
m ;
1 0 1 0
0 8l 9 8 lb 9 8 m 9 8 mb 91
<Z a k = <Z k = < Za k = < Z k =
¼ @exp dl exp dl ; 1
exp
d m exp
dm A
: l ; : l ; : 1
m ; : 1
m ;
1 1 0 0
0 8 9 8 91
< Za b k =
l l >
< Z a Þð1
m b Þ
1
ð1
m >
=
B k C
¼ @exp dl ; 1
exp
dm A
: l ; >
: 1
m >
;
1 0
¼ ða bÞk
a a
ak 1 ak 2 ¼ oð1
k 1 ; k 1 ÞlX o ð1
k 2 ; k 2 ÞlX
E E
0 8l 9 8 m 91 0 8 l 9 8 m 91
<Z a k = < Za k = <Z a k = < Za k =
¼ @exp d m A @exp dm A
1 1 2 2
dl ; 1
exp
dl ; 1
exp
: l ; : 1
m ; : l ; : 1
m ;
1 0 1 0
0 8l 9 8l 9 8 m 9 8 m 91
<Z k a
= <Z ka
= < Z ka
= < Z k a
=
¼ @exp dm A
1 2 1 2
dl exp dl ; 1
exp
d m exp
: l ; : l ; : 1
m ; : 1
m ;
1 1 0 0
0 8l 9 8 m 91
<Z 1
ð1
ðk þ k ÞÞ =
a
< Z k þk
a
=
¼ @exp dm A
1 2 1 2
dl ; 1
exp
: l ; : 1
m ;
1 0
¼ ak 1 þ k 2
6.4 Complements of Aggregation Operators
In this section, we introduce the complements of the IFWA operator and the IFIA
operator. Furthermore, we give the integral forms of their complements, which
indicates that the complements of the operators can be represented as some definite
integrals of IFFs.
6.4.1 Complements of IFWA Operator and IFIA Operator
Firstly, we introduce the IFWA operator and the IFWG operator (Xu and Yager
2006, 2007), which are commonly to be used for aggregating the discrete intu-
itionistic fuzzy information (data):

n Q
n
xi Q
n
(1) (IFWA) IFWAx ða1 ; a2 ; . . .; an Þ ¼ xi ai ¼ 1
ð1
lai Þ ; . mx
ai
i
i¼1
n i¼1 i¼1
n Q xi Q
n
(2) (IFWG) IFWGx ða1 ; a2 ; . . .; an Þ ¼ axi ¼
i
lai ; 1
ð1
mai Þxi .
i¼1 i¼1 i¼1
Pn
where xi 0 ði ¼ 1; 2; . . .; nÞ and i¼1 xi ¼ 1.
Obviously, there exists a closed connection between the IFWA operator and the
IFWG operator:
a1 ;
IFWAx ða1 ; a2 ; . . .; an Þ ¼ IFWGx ð a2 ; . . .;
an Þ
As we know, the IFWA operator can be represented as a definite integral of a

piecewise continuous IFF in Chap. 3. The IFWG operator can also be expressed in a
similar way (Lei and Xu 2015a):
If we denote b0 ¼ E and bi þ 1 ¼ bi ai þ 1 , and introduce a piecewise IFF uðXÞ
as:
8
>
> ð1
x 1 ; x 1 Þ; when b1 E X E b0 ;
>
< ð1
x 2 ; x 2 Þ; when b2 E X E b 1 ;
uðXÞ ¼ ..
>
> .
>
:
ð1
x n ; x n Þ; when bn E X E bn
1 ;
which means that uðXÞ ¼ ð1

x i ; x i Þ when bi E X E bi
1 (1 i n). Then
we can get the following equality:
bn
o uðXÞlX ¼ IFWGx ða1 ; a2 ; . . .; an Þ
E
where bn ¼ ni¼1 ai , which shows that the IFWG operator is also the definite
integral of a piecewise IFF. RR
As we have discussed before, D PðXÞXdd is the continuous form of the IFWA
RR
operator when D PðXÞdd ¼ 1. Here, we give another method to aggregate the
continuous intuitionistic fuzzy information, which is the continuous form of the
IFWG operator (Lei and Xu 2016a):
Step 1. Dividing the region D into k sub-regions, which are di ði ¼ 1; 2; ; kÞ,
respectively.
Step 2. Choosing an IFN ðni ; gi Þ from the sub-region di ð1 i kÞ randomly, and
making the produces ðni ; gi ÞPðni ;gi ÞDdi ð1 i kÞ, where Ddi is the area of
the i-th sub-region.
Step 3. Calculating the product of ðni ; gi ÞPðni ;gi ÞDdi ð1 i k Þ, which is
ki¼1 ðni ; gi ÞPðni ;gi ÞDdi .
Step 4. Taking the limit lim k ðni ; gi ÞPðni ;gi ÞDdi
i¼1
d!0
6.4 Complements of Aggregation Operators 153
Denote the limit lim ki¼1 ðni ; gi ÞPðni ;gi ÞDdi by

d!0
o oD XPðXÞdd , then it has the fol-
lowing form:
0 8 9 8 91
<ZZ = <ZZ =
oDo XPðXÞdd ¼ @exp: Pðl; mÞ ln ldd ; 1
exp
; :
Pðl; mÞ lnð1
mÞdd A
;
D D
When PðXÞ ¼ dðXÞ, this limit is essentially the continuous form

RR of the IFWG
operator. In what follows, we reveal the relationship between D PðXÞXdd and
o oD XPðXÞdd . At first, we present an expression about the link between the IFWA
operator and the IFWG operator (Xia et al. 2012b):
n n
IFWAx ða1 ; a2 ; ; an Þ ¼ xi ai ¼ ax a1 ;
i ¼ IFWGx ð
i
a2 ; ;
an Þ
i¼1 i¼1
The following steps are provided to acquire the complement of the IFWA
operator (Lei and Xu 2016a):
(1) Finding the basic components of IFWAx ða1 ; a2 ; ; an Þ. Because
IFWAx ða1 ; a2 ; ; an Þ ¼ ni¼1 xi ai , we can get that its basic elements are
xi ai (i ¼ 1; 2; . . .; n). The IFWA operator combines these xi ai ði ¼ 1; 2; . . .; nÞ
with the addition “” of IFNs.
(2) Acquired the complements of these basic components. In order to get the
complement of IFWAx ða1 ; a2 ; ; an Þ, we first need to get these complements
of the basic elements xi ai (ði ¼ 1; 2; . . .; nÞ), which are essentially axi
i
i ¼ 1; 2; . . .; n. It needs to point out that ai and

ai have the same weight xi .
(3) Combining these complements of the basic components with the multipli-
cation “ ” of IFNs. After getting all complements of the basic elements,
namely: ax i ði ¼ 1; 2; . . .; nÞ, we assemble them with “” to get i¼1
i n
axi ,
i
which is just IFWGx ða1 ; a2 ; ; an Þ.

RR
In the same manner, we can get the relationship between D PðXÞXdd and
o oD XPðXÞdd by following the above steps (Lei and Xu 2016a):
RR
(1) Finding the basic components of D PðXÞXdd. Because
ZZ
PðXÞXdd ¼ lim ki¼1 Pðni ; gi Þðni ; gi ÞDdi
d!0
D
then we can get its basic components Pðni ; gi Þðni ; gi ÞDdi .

(2) Getting the complements of these basic components. The complement of the
Pðni ;gi ÞDdi
basic component Pðni ; gi Þðni ; gi ÞDdi is ð ni ; gi Þ , which is
ðgi ; ni ÞPðni ;gi ÞDdi .
(3) Combining these complements of the basic components with the multipli-
cation “ ”. After getting all ðgi ; ni ÞPðni ;gi ÞDdi , we can acquire the limit
k
lim ðgi ; ni ÞPðni ;gi ÞDdi .
d!0 i¼1
k
Based on these steps, it is clear that lim ðgi ; ni ÞPðni ;gi ÞDdi can be expressed as
d!0 i¼1
o oD ðl; mÞPðl;mÞdd , which is o oD XPðXÞdd . In addition, the following theorem can be

obtained:
ZZ
PðXÞXdd ¼ o o X
PðXÞdd
D
D
RR
which is the relationship between “ D ” and “ o oD ”.
According to the above analysis, it is easy to prove Theorem 6.16, and thus, it is
omitted here.
In the following, we define two novel concepts, namely: the complement of a
region of IFNs, D, and the complement of a weight function PðXÞ for further
RR
analysis of the relationship between “ D ” and “ o oD ”.
Definition 6.8 (Lei and Xu 2016a) Let D be a region of IFNs, then we call
D ¼ faja 2 D g or D ¼ faja 2 D g the complement of D, which can be represented
in Fig. 6.5 (Lei and Xu 2016a).
In order to acquire the complement of IFWAx ða1 ; a2 ; ; an Þ, we need to obtain
every complement ai of ai (1 i n). Similarly, it is necessary to get the com-
plement D of D when aggregating the continuous intuitionistic fuzzy information.
Fig. 6.5 Complement of a

region of IFNs
In what follows, we define the complement of a weight function:

Definition 6.9 (Lei and Xu 2016a) Let PðXÞ (or Pðl; mÞ) be a weight function of D.
Then if a weight function QðYÞ defined in D, which satisfies that QðYÞ ¼ PðXÞ if

only Y ¼ X, then we call QðYÞ defined in D the complement of PðXÞ in D, denoted
P
by P. and P can be shown in Fig. 6.6 (Lei and Xu 2016a).
By the concept of PðXÞ, we can get the weight information of X for any X 2 D.
Meanwhile, the weight of X(X 2 D) is actually equal to PðXÞ( X 2 D), which shows

that the weight PðXÞ is just equal to the weight PðXÞ
of X in D of X in D. It likes
that ai and ai have the same weight xi when we analyze the conclusion
IFWAx ða1 ; a2 ; ; an Þ ¼ IFWGx ða1 ; a2 ; ; an Þ.
By virtue of D and PðXÞ we can get
in D,
lim ðgi ; ni ÞPðni ;gi ÞDdi ¼ o o X PðXÞdd

k
d!0 i¼1
D
RR
Hence, we have the following relationship between “ D ” and “o oD ”:
ZZ
oDo XPðXÞdd ¼ PðXÞXdd ¼ o o X PðXÞdd

D
D
and PðXÞ,
Proof According to the definitions of D we know that
oo XPðXÞdd

¼ ooX
D
PðXÞdd
RR
and there exists o oD XPðXÞdd ¼ D PðXÞXdd. Hence, Theorem 6.15 holds.
Fig. 6.6 Complement of a weight function PðXÞ

Moreover, another proof can be given as follows:

0 8 9 8 91
<ZZ = <ZZ =
oo XPðXÞdd

¼ @exp
:
lÞ ln mdd ; 1
exp
Pðm;
; :
lÞ lnð1
lÞdd A
Pðm;
;
D
D D
0 8 9 8 91
<ZZ = <ZZ =
¼ @exp Pðl; mÞ ln mdd ; 1
exp Pðl; mÞ lnð1
lÞdd A
: ; : ;
D D
ZZ
¼ PðXÞXdd
D
and
0 8 9 8 91
<ZZ = <ZZ =
oDo X PðXÞdd
¼ @exp
:
Pðl; mÞ ln mdd ; 1
exp
; :
Pðl; mÞ lnð1
lÞdd A
;
D D
ZZ
¼ PðXÞXdd
D
RR
Therefore, o oD XPðXÞdd ¼ D PðXÞXdd ¼ o oD X
Pðl;mÞdd
holds. ■
RR
For the symmetry between “ D ” and “o oD ”, we can get a similar conclusion:
ZZ ZZ

PðXÞXdd ¼ o o XPðXÞdd ¼ PðXÞX dd
D

D D
The proof of Theorem 6.18 is similar to that of Theorem 6.17, which is omitted
here.
6.4.2 Integral Forms of Complements of IFWA Operator

and IFIA Operator
Rb
a uðXÞdX ¼ oa uðXÞ lX
and IFWAx ða1 ; a2 ; ; an Þ ¼ IFWGx
b
Due to
ða1 ; a2 ; ; an Þ, we can transform IFWGx ða1 ; a2 ; ;
an Þ as (Lei and Xu 2016a):
a1 ;
IFWGx ða1 ; a2 ; ; an Þ ¼ IFWAx ð a2 ; ;
an Þ
Rb
Based on O LðXÞdX ¼ IFWAx ða1 ; a2 ; ; an Þ, we can obtain
Fig. 6.7 II-IFIC linking

between E and c
Zb
X
LðXÞd ¼ IFWAx ða1 ; a2 ; ;
an Þ
O
and then
c Zb
o
E

LðXÞ lX ¼ X
LðXÞd ¼ IFWAx ða1 ; a2 ; ;
an Þ ¼ IFWGx ða1 ; a2 ; ; an Þ
O
Furthermore, the following theorem can be provided:

Theorem 6.19 (Lei and Xu 2016b) Let ai ¼ ðli ; mi Þ ( i ¼ 1; 2; ; n) be n IFNs,
which satisfy ai 6¼ aj if only i 6¼ j, and their weights are xi ( i ¼ 1; 2; ; n),
P
respectively, which meet ni¼1 xi ¼ 1, then we have that the integral of LðXÞ along
the II-IFIC in Fig. 6.7 is equal to the aggregated value by using the IFWG operator:
c
o
E
LðXÞ lX ¼ IFWGx ða1 ; a2 ; ; an Þ
where E ¼ð1 ; 0Þ and c ¼ ðlmin ; mmax Þ. The II-IFIC is shown in the following figure
(Lei and Xu 2015c) (Fig. 6.8).
Proof If we let U ¼ fli j1 i ng and V ¼ fmi j1 i ng be two given sets, then
there are jUj n and jVj n since there may be some repeated elements in U and
V. Hence, we can rank li ði ¼ 1; 2; ; nÞ and mi ði ¼ 1; 2; ; nÞ as
Fig. 6.8 Common upper

limit H
lð1Þ \lð2Þ \ \lðjUjÞ and mð1Þ \mð2Þ \ \mðjVjÞ , respectively. Then we can
know that LðXÞ consists of two real piecewise continuous functions:
8
> 1; 0 l\lð1Þ ;
>
>
> k 1;
> lð1Þ l\lð2Þ ;
<
RðlÞ ¼ k 2 ; lð2Þ l\lð3Þ ;
>
> ..
>
> .
>
:
0; l ðjUjÞ l 1;
and
8
> 0; mðjVjÞ \m 1;
>
>
>
< k 1 ; mðjVj
1Þ \m mðjVjÞ ;
>
TðmÞ ¼ k 2 ; mðjVj
2Þ \m mðjVj
1Þ ;
>
> ..
>
> .
>
:
1; 0 m mð1Þ ;
P P
where ki ¼ lj [ lðiÞ xj and ki ¼ mj [ mðjVj
iÞ xj . Hence, we have
X X X
ki
ki þ 1 ¼ xj
xj ¼ xj
lj [ lðiÞ lj [ lði þ 1Þ lj ¼lði þ 1Þ
X X X
k i þ 1
ki ¼ xj
xj ¼ xj
mj [ mðjVj
i
1Þ mj [ mðjVj
iÞ mj ¼mðjVj
iÞ
According to the calculating formula of the definite integral of the IFF, if we

denote mð0Þ ¼ 0 and lðjUj þ 1Þ ¼ 1, then
0 8 l 9 8 91
c < Zmin 1
RðlÞ = < Zmmax
TðmÞ =A
o
E
LðXÞ lX ¼ @exp
: l
dl ; 1
exp
; :
dm
1
m ;
1 0
0 8 0 19
>
<jUj
1 Z
lðjUj
iÞ >
B XB 1 C=
¼ @exp @ð1
kjUj
i Þ dlA ;
>
: i¼0 l >
;
lðjUj
i þ 1Þ
8 0 19 1
>
< jVj
1
XB Z
mði þ 1Þ >
=
1 C C
1
exp
@ jVj
i
k dm A A
>
: i¼0 1
m >
;
mðiÞ
0 !1
kjUj
i 1
Y
jUj
1
lðjUj
iÞ Y 1
mði þ 1Þ kjVj
i
jVj
1
¼@ ; 1
A
i¼0
l ðjUj
i þ 1Þ i¼0
1
m ðiÞ
!
Y
jUj
1 kjUj
i
1
kjUj
i Y
jVj
1
kjV j
i
kjVj
i
1
¼ lðjUj
iÞ ; 1
1
mði þ 1Þ
i¼0 i¼0
P P !
Y
jUj
1
lj ¼lðjUj
iÞ
xj Y
jVj
1
mj ¼m xj
¼ lðjUj
iÞ ; 1
1
mði þ 1Þ ði þ 1Þ
i¼0 i¼0
¼ IFWGx ða1 ; a2 ; ; an Þ
which completes the proof. ■

It is easy to get that there is a contradiction between the two results:
c
o
E

and
c
o
E
It is natural to raise a question: Which one is right? In what follows, we show

that they are both correct.
After restudying the special IFF LðXÞ ¼ ðRðlÞ; TðmÞÞ, we can get
!
X X
LðXÞ ¼ ðRðlÞ; TðmÞÞ ¼ xi ; xi
li [ l mi m
Then
! !
X X X X
LðXÞ ¼ xi ; xi ¼ xi ; xi
¼ LðXÞ
li [ l mi m mi m li [ l
In addition, according to the complements of IFFs, we can get (Lei and Xu

2016a)
XÞ
LðXÞ ¼ Lð
¼ Lð
Then we get that LðXÞ XÞ
holds, which means L ¼ L,
and thus,
c
o
E

LðXÞ lX ¼ IFWGx ða1 ; a2 ; ; an Þ and
c
o
E
are both correct.

The same analyses about the IFF LðXÞ can be given as (Lei and Xu 2016a):
Rb RR
Based on a uðXÞdX ¼ oa u and o oD XPðXÞdd ¼ PðXÞXdd,
lX
b
ðXÞ D we get
ZZ
oDo X PðXÞdd
¼
PðXÞXdd

D
RR Rb
Since D PðXÞXdd ¼ O LðXÞdX, then there must be
ZZ Zb

PðXÞXdd ¼ X
LðXÞd

D O
Thus, we can get
ZZ Zb c
oDo X PðXÞdd
¼
PðXÞXdd ¼ ¼
LðXÞd X o
E
LðXÞ lX

D O
o oD XPðXÞdd ¼ d!0
n
Moreover, as we know lim ðni ; gi Þ Pðn ;g ÞDd , then according to
i i i
i¼1
the properties of the complement of integral, we get
oDo XPðXÞdd ¼ d!0

n
lim ðni ; gi Þ Pðn ;g ÞDd i i i
i¼1
ðni ;gi Þ
o
n
¼ lim ð1
Pðni ; gi ÞDdi ; Pðni ; gi ÞDdi Þ ld
d!0 i¼1 E
For more investigation, we need to find the common upper limit “H” of D based
on the order “E ”:
HE ðni ; gi Þ; for any ðni ; gi Þ 2 D
By the multiplication and the division regions given in Chap. 1, we can get that
R ðn ;g Þ
the common upper limit “H” of “oE i i ” is different from “X” of “ O i i ” in
ðn ;g Þ
Chap. 5, where the common upper limit “H” is shown in the following figure (Lei
and Xu 2016a):
For any given IFN ðni ; gi Þ, if we define the following IFF with respect to X:

ð1
Pðni ; gi ÞDdi ; Pðni ; gi ÞDdi Þ; ðni ; gi ÞE XE E;
gðni ;gi Þ ðXÞ ¼
E; HE X / ðni ; gi Þ;
then
H ðni ;gi Þ H
o gðni ;gi Þ ðXÞ lX ¼ o gðni ;gi Þ ðXÞ lX o gðni ;gi Þ ðXÞ lX
E E ðni ;gi Þ
ðni ;gi Þ H
¼ o ð1
Pðni ; gi ÞDdi ; Pðni ; gi ÞDdi Þ lX o E lX
E ðni ;gi Þ
ðni ;gi Þ
¼ o
E
ð1
Pðni ; gi ÞDdi ; Pðni ; gi ÞDdi Þ lX
Based on these preparations, we can get the following theorem, which reveals
the relationship between “o oD ” and “oa ”.
b
Theorem 6.20 (Lei and Xu 2016b) Let D be a region of RR IFNs, and PðXÞ be a
non-negative real function of D, which satisfies that D PðXÞdd ¼ 1, and
c
c ¼ inf ðl;mÞ2D flg; supðl;mÞ2D fmg . Then o o X PðXÞdd ¼ o LðXÞ lX.
D E
Proof According to the definition of o oD , we can get

oDo XPðXÞdd ¼ d!0

k
lim ðni ; gi Þ Pðn ;g ÞDd i i i
i¼1
ðni ;gi Þ
o
k
¼ lim ð1
Pðni ; gi ÞDdi ; Pðni ; gi ÞDdi Þ lX
d!0 i¼1 E
H
o gðn ;g Þ ðXÞ lX
k
¼ lim i i
d!0 i¼1 E
!
H X
k X
k
¼ lim o
d!0 E
1
ð1
Uðgðni ;gi Þ ðXÞÞÞ; Vðgðni ;gi Þ ðXÞÞ lX
i¼1 i¼1
!
H X
k X
k
¼ o
E
1
lim
d!0
ð1
Uðgðni ;gi Þ ðXÞÞÞ; lim
d!0
Vðgðni ;gi Þ ðXÞÞ lX
i¼1 i¼1
!
H X X
¼ o
E
1
lim
d!0
ð1
Uðgðni ;gi Þ ðXÞÞÞ þ ð1
Uðgðni ;gi Þ ðXÞÞÞ ;
ni l ni \l
!!
X X
lim Vðgðni ;gi Þ ðXÞÞ þ Vðgðni ;gi Þ ðXÞÞ lX
d!0
gi m gi \m
! !!
H X X
¼ o
E
1
lim 0 þ
d!0
ð1
Uðgðni ;gi Þ ðXÞÞÞ ; lim
d!0
Vðgðni ;gi Þ ðXÞÞ þ 0 lX
ni \l gi m
ZZ ZZ !
H
¼ o Pðq; rÞdqdr; Pðq; rÞdqdr lX
E l m
H H
¼ o ðRðlÞ; T ðmÞÞ lX ¼ Eo
E
LðXÞ lX
c
¼ o LðXÞ lX
E

In what follows, we provide an example (Lei and Xu 2016b) to verify the
theorem:
Let D be a region of IFNs as shown in Fig. 6.9 (Lei and Xu 2016b), and
PðXÞ
RR ¼ 4 be a non-negative real function of D, which satisfies that
D PðXÞdd ¼ 1.
Fig. 6.9 The figure of an

example
With o oD XPðXÞdd , we can get

0 8 9 8 91
< ZZ = < ZZ =
oDo XPðXÞdd ¼ @exp 4
:
ln l dd ; 1
exp 4
; :
ln ð1
mÞ dd A
;
D D
1 2
¼ ; 1
2e e
Moreover, the two real functions are presented as:

0; 1=2\l 1;
RðlÞ ¼ and T ðmÞ ¼ 1
2m
1
2l; 0 l 1=2;
Then the special IFF LðXÞ can be represented as:

ð0 ; 1
2mÞ; ð1=2; 1=2Þ / XE E;
LðXÞ ¼
ð1
2l ; 1
2mÞ; ð0; 1=2ÞE XE ð1=2; 1=2Þ;
Thereby, we get
0 8 9 8 1=2 91
c >
<Z 1
1=2 Z0 >
= < Z 1
2m =
o LðXÞ lX ¼ B
@exp
> l
dl þ 2dl ; 1
exp
> :
m
dm A
;
C
E : ; 1
1
1=2 0
1 2
¼ ;1
2e e
and then,
c
oDo XPðXÞdd ¼ Eo LðXÞ lX
From the above discussion, we obtain that there exist two contradictory state-
ments apparently:
c c
oDo XPðXÞdd ¼ Eo LðXÞ lX and oDo XPðXÞdd ¼ Eo LðXÞlX
Next, we illustrate that the two qualities both hold (Lei and Xu 2016a):

RR RR
Since LðXÞ ¼ ðRðlÞ; T ðmÞÞ ¼ m Pðq; rÞdqdr; l Pðq; rÞdqdr ¼ LðXÞ,
then we get LðXÞ ¼ LðXÞ. Moreover, LðXÞ ¼ LðXÞ for the complement of the IFF
LðXÞ, then we obtain LðXÞ ¼ LðXÞ, which means L ¼ L. Therefore, the above two
qualities hold.
6.5 Conclusions
In this chapter, we have studied the complement theory of intuitionistic fuzzy

calculus based on the complement operator. The specific contents of the comple-
ment theory include some knowledge related to IFNs, the derivatives, differentials,
indefinite integrals, and definite integrals in the intuitionistic fuzzy calculus. In
addition, we have proven that there are some closed connections between intu-
itionistic fuzzy calculus and its complement theory. Moreover, we have investigated
the relationships among a few aggregation operators, like the IFWA operator, the
RR
IFWG operator, D and o oD . Finally, we have verified an important fact that
L ¼ L and L ¼ L. Actually, we have managed to reveal the fact that any state-
ment or conclusion in the intuitionistic fuzzy calculus must have a counterpart in its
complement theory. As we know from the previous chapters, the integrals of IFFs,
RR
D and o oD are usually used to build the aggregation operations to deal with
continuous or a large number of intuitionistic fuzzy numbers (or information).
However, according to their relationships revealed in this chapter, we have dis-
covered that it is unnecessary to simultaneously use two related operations (such as
RR
D and o oD ) to aggregate information, because one of the two related operations
and the complement operator of IFNs can fully realize the work of another
aggregation operation.
6.5 Conclusions 165
Fig. 6.10 The organizational structure of Chap. 6
In order to show the content of this chapter more clearly, now we draw the
diagram of this chapter’s organizational structure in Fig. 6.10, which shows that
Chap. 6 actually has proposed the complement theory of intuitionistic fuzzy cal-
culus and has built the relationship between the calculus of IFFs and its complement
theory.
References
Atanassov, K. (1986). Intuitionistic fuzzy set. Fuzzy Sets and Systems, 20, 87–96.
Atanassov, K. (2012). On intuitionistic fuzzy sets theory. Berlin, Heidelberg: Springer.
Atanassov, K., Pasi, G., & Yager, R. R. (2005). Intuitionistic fuzzy interpretations of multicriteria
multiperson and multi-measurement tool decision making. International Journal of Systems
Science, 36, 859–868.
Beliakov, G., Bustince, H., Goswami, D. P., Mukherjee, U. K., & Pal, N. R. (2011). On averaging
operators for Atanassov’s intuitionistic fuzzy sets. Information Sciences, 181, 1116–1124.
Beliakov, G., Bustince, H., James, S., Calvo, T., & Fernandez, J. (2012). Aggregation for
Atanassov’s intuitionistic and interval valued fuzzy sets: The median operator. IEEE
Transactions on Fuzzy Systems, 20, 487–498.
Bustince, H., Herrera, F., & Montero, J. (2008). Fuzzy sets and their extensions: Representation,
aggregation and models. Berlin: Springer.
Chen, S. M., & Tan, J. M. (1994). Handling multicriteria fuzzy decisionmaking problems based on
vague set theory. Fuzzy Sets and Systems, 67, 163–172.
Chen, T. Y. (2011). A comparative analysis of score functions for multiple criteria decision
making in intuitionistic fuzzy settings. Information Sciences, 181, 3652–3676.
De, S. K., Biswas, R., & Roy, A. R. (2000). Some operations on intuitionistic fuzzy sets. Fuzzy
Sets and Systems, 114, 477–484.
De, S. K., Biswas, R., & Roy, A. R. (2001). An application of intuitionistic fuzzy sets in medical
diagnosis. Fuzzy Sets and Systems, 117, 209–213.
Deschrijver, G., & Kerre, E. E. (2001). On the relationship between some extensions of fuzzy set
theory. Fuzzy Sets and Systems, 117, 209–213.
Hong, D. H., & Choi, C. H. (2000). Multicriteria fuzzy decision-making problems based on vague
set theory. Fuzzy Sets and Systems, 114, 103–113.
Hung, W. L., & Yang, M. S. (2004). Similarity measures of intuitionistic fuzzy sets based on
Hausdorff distance. Pattern Recognition Letters, 25, 1603–1611.
Khatibi, V., & Montazer, G. A. (2009). Intuitionistic fuzzy set vs. fuzzy set application in medical
pattern recognition. Artificial Intelligence in Medicine, 47, 43–52.
Klement, E. P., & Mesiar, R. (2005). Logical, algebraic, analytic, and probabilistic aspects of
triangular norms. New York: Elsevier.
Klir, G., & Yuan, B. (1995). Fuzzy sets and fuzzy logic: Theory and applications. Upper Saddle
River: Prentice Hall.
Lei, Q., & Xu, Z. S. (2015). Chain and substitution rules of intuitionistic fuzzy calculus. IEEE
Transactions on Fuzzy Systems, 24(3), 519–529.
Lei, Q., & Xu, Z. S. (2015). Derivative and differential operations of intuitionistic fuzzy numbers.
International Journal of Intelligent Systems, 30(4), 468–498.
Lei, Q., & Xu, Z. S. (2015). Fundamental properties of intuitionistic fuzzy calculus.
Knowledge-Based Systems, 76, 1–16.
168 References
Lei, Q., Xu, Z. S., Bustince, H., & Burusco, A. (2015). Definite integrals of Atanassov’s
intuitionistic fuzzy information. IEEE Transactions on Fuzzy Systems, 23(5), 1519–1533.
Lei, Q., & Xu, Z. S. (2016a). A unification of intuitionistic fuzzy calculus theories based on
subtraction derivatives and division derivatives. IEEE Transactions on Fuzzy Systems.
doi: 10.1109/TFUZZ.2016.2593498.
Lei, Q., & Xu, Z. S. (2016). Relationships between two types of intuitionistic fuzzy definite
integrals. IEEE Transctions on Fuzzy Systems, 24(6), 1410–1425.
Lei, Q., Xu, Z. S., Bustince, H., & Fernandez, J. (2016). Intuitionistic fuzzy integrals based on
Archimedean t-conorm and t-norm. Information Sciences, 327, 57–70.
Li, D. F. (2014). Decision and game theory in management with intuitionistic fuzzy sets. Berlin,
Heidelberg: Springer.
Liang, Z. Z., & Shi, P. F. (2003). Similarity measures on intuitionistic fuzzy sets. Pattern
Recognition Letters, 24, 2687–2693.
Szmidt, E., & Kacprzyk, J. (2004). A similarity measure for intuitionistic fuzzy sets and its
application in supporting medical diagnostic reasoning. Lecture Notes in Computer Science,
3070, 388–393.
Vlachos, K. I., & Sergiadis, G. D. (2007). Intuitionistic fuzzy information applications to pattern
recognition. Pattern Recognition Letters, 28, 197–206.
Wei, G. W. (2010). GRA method for multiple attribute decision making with incomplete weight
information in intuitionistic fuzzy setting. Knowledge-Based Systems, 23, 243–247.
Xia, M. M., & Xu, Z. S. (2010). Generalized point operators for aggregating intuitionistic fuzzy
information. International Journal of Intelligent Systems, 25, 1061–1080.
Xia, M. M., Xu, Z. S., & Zhu, B. (2012). Generalized intuitionistic fuzzy Bonferroni means.
International Journal of Intelligent Systems, 27, 23–47.
Xia, M. M., Xu, Z. S., & Zhu, B. (2012). Some issues on intuitionistic fuzzy aggregation operators
based on Archimedean t-conorm and t-norm. Knowledge-Based Systems, 31, 71–88.
Xu, D. W., & Xu, Z. S. (2013). A spectral clustering algorithm based on intuitionistic fuzzy
information. Knowledge-Based Systems, 53, 20–26.
Xu, Z. S., & Yager, R. R. (2006). Some geometric aggregation operators based on intuitionistic
fuzzy sets. International Journal of General Systems, 35, 417–433.
Xu, Z. S. (2007). Intuitionistic fuzzy aggregation operations. IEEE Transactions on Fuzzy Systems,
15, 1179–1187.
Xu, Z. S., Chen, J., & Wu, J. J. (2008). Clustering algorithm for intuitionistic fuzzy sets.
Information Sciences, 178, 3775–3790.
Xu, Z. S., & Yager, R. R. (2009). Intuitionistic and interval-valued intuitionistic fuzzy preference
relations and their measures of similarity for the evaluation of agreement within a group. Fuzzy
Optimization and Decision Making, 8(2), 123–139.
Xu, Z. S., & Cai, X. Q. (2010). Nonlinear optimization models for multiple attribute group
decision making with intuitionistic fuzzy information. International Journal of Intelligent
Systems, 25, 489–513.
Xu, Z. S. (2012). Linguistic decision making: Theory and methods. Beijing: Science Press, and
Berlin, Heidelberg: Springer.
Xu, Z. S., & Cai, X. Q. (2012). Intuitionistic fuzzy information aggregation: Theory and
applications. Beijing: Science Press, and Berlin, Heidelberg: Springer.
Xu, Z. S. (2013). Intuitionistic fuzzy aggregation and clustering. Berlin, Heidelberg: Springer.
Xu, Z. S. (2013). Intuitionistic fuzzy preference modeling and interactive decision making. Berlin,
Heidelberg: Springer.
Yager, R. R., & Kacprzyk, J. (1997). The ordered weighted averaging operators: Theory and
applications. Norwell, MA: Kluwer.
Yu, D. J., & Shi, S. S. (2015). Researching the development of Atanassov intuitionistic fuzzy set:
Using a citation network analysis. Applied Soft Computing, 32, 189–198.
Zadeh, L. A. (1965). Fuzzy sets. Information and Control, 8, 338–356.
Zhao, H., Xu, Z. S., Ni, M. F., & Liu, S. S. (2010). Generalized aggregation operators for
intuitionistic fuzzy sets. International Journal of Intelligent Systems, 25, 1–30.

Qian Lei, Zeshui Xu - Intuitionistic Fuzzy Calculus-Springer (2017)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Qian Lei, Zeshui Xu - Intuitionistic Fuzzy Calculus-Springer (2017)

Uploaded by

Copyright:

Available Formats

Qian Lei Zeshui Xu

Intuitionistic Fuzzy Calculus

ISSN 1434-9922 ISSN 1860-0808 (electronic)

© Springer International Publishing AG 2017

This Springer imprint is published by Springer Nature

Nanjing, China Qian Lei

1 Basic Concepts Related to Intuitionistic Fuzzy Numbers . . . . . . . . . . 1

Intuitionistic fuzzy calculus is investigated by utilizing intuitionistic fuzzy numbers

1.1 Introduction to Intuitionistic Fuzzy Numbers

Fig. 1.1 Representation of

Fig. 1.2 Expressing a0 as the subintervals of ½0; 1

The above-mentioned two representations of an IFN play important roles in the

1.2 Basic Operations Between Intuitionistic Fuzzy

Fig. 1.3 Addition and multiplication between a and b

1.2.1 Geometrical Analysis of the Operations of IFNs

In addition, for a set fa  xjx 2 Ng, then b 2 fa  xjx 2 Ng. Hence,

Fig. 1.4 Addition region

According to the deﬁnition of S  ðaÞ, we have the corresponding notion of the

Fig. 1.5 Subtraction region

Fig. 1.6 Addition region of

Fig. 1.7 Multiplication

Fig. 1.8 Division region

Fig. 1.9 Multiplication

Fig. 1.10 The image of the

Fig. 1.11 The image of the

Fig. 1.12 The change region

1.2.2 Algebraic Analysis of the Operations of IFNs

In this subsection, we will give some algebraic properties of IFNs.

Similarly, we can get ða  bÞk 1 ¼ ak 1  bk 1 . Considering the conclusion (4), we

and we have the following process:

On the other hand, there is also

Hence, in this case, k 1 ðb  aÞ ¼ k 1 b  k 1 a holds.

ða1  a2 Þða2  a3 Þ ¼ ð1 ð1 l1 Þð1 l2 Þ; m1 m2 Þð1 ð1 l2 Þð1 l3 Þ; m2 m3 Þ

In the same way, the proof of (2) can be processed as follows:

Moreover, the equation of (3) can be proved as follows:

The proof of Theorem 1.3 is completed. ■

Proof Firstly, we prove (1) as follows:

Next, we can prove the proof of (2):

Finally, we prove the conclusion (3):

ða1  a2 Þ  ða2  a3 Þ ¼ ðl1 l2 ; 1 ð1 m1 Þð1 m2 ÞÞ  ðl2 l3 ; 1 ð1 m2 Þð1 m3 ÞÞ

which completes the proofs. ■

1.3 Order Relations of IFNs

1.3.1 Three Kinds of Orders of IFNs and Relationships

Based on the fundamental knowledge of the order relations, we introduce some

Fig. 1.13 Order relation

Fig. 1.14 Order relation “\XY ”

This kind of order relation can be shown in Fig. 1.14.

Fig. 1.15 Order relation “E”

Fig. 1.16 Comparison of

1.3.2 Properties of the Order Based on the Operations

which means that a ¼ b if aEb and bEa.

On the other hand, we can get there must be b

which means that aEg if aEb and bEg.

Fig. 1.18 g 2 S  ðbÞ and

Fig. 1.19 “E” is a partial

ða1  a2 Þ  ðc1  c2 Þ ¼ ða1  c1 Þ  ða2  c2 Þ ¼ b1  b2

which means a1  a2 Eb1  b2 . ■

The proof of this theorem is completed. ■

acEðacÞ  ðbaÞ ¼ bc

The proof is completed. ■

Proof By Theorem 1.2, there is kðbaÞ ¼ kbka. Hence, ka  kðbaÞ ¼ kb,

Calculus, which is an important branch of classical mathematics, is the mathe-

a neighborhood or r-neighborhood of a, denoted by Uða; rÞ. Moreover, the set

is called a closed neighborhood of a, denoted by Uða; rÞ.

In addition, for a set fa xjx 2 Ng, then b 2 fa xjx 2 Ng. Hence,

According to the deﬁnition of S ðaÞ, we have the corresponding notion of the

Similarly, we can get ða bÞk 1 ¼ ak 1 bk 1 . Considering the conclusion (4), we

Hence, in this case, k 1 ðb aÞ ¼ k 1 b k 1 a holds.

ða1 a2 Þða2 a3 Þ ¼ ð1 ð1 l1 Þð1 l2 Þ; m1 m2 Þð1 ð1 l2 Þð1 l3 Þ; m2 m3 Þ

ða1 a2 Þ ða2 a3 Þ ¼ ðl1 l2 ; 1 ð1 m1 Þð1 m2 ÞÞ ðl2 l3 ; 1 ð1 m2 Þð1 m3 ÞÞ

Fig. 1.18 g 2 S ðbÞ and

ða1 a2 Þ ðc1 c2 Þ ¼ ða1 c1 Þ ða2 c2 Þ ¼ b1 b2

which means a1 a2 Eb1 b2 . ■

acEðacÞ ðbaÞ ¼ bc

Proof By Theorem 1.2, there is kðbaÞ ¼ kbka. Hence, ka kðbaÞ ¼ kb,

In addition, fX : jXajEe; X 2 S ðaÞ [ S ðaÞg is called a closed neighborhood

denoted by u0 ðXÞ. Similarly, lim uðXÞuðYÞ

uðYÞuðXÞ uðX DXÞuðXÞ

Proof We ﬁrst prove the right derivative of u, which is lim uðYÞuðXÞ

duðXðtÞÞ uðXðt DtÞÞuðXðtÞÞ

(a) If u 1 ðaÞ ¼ ka C is expressed as u 1 ðu4 ðaÞÞ ¼ ku4 ðaÞ C, then there is

(b) If u 1 ðaÞ ¼ ka C is written as u1 ðu3 ðaÞÞ ¼ u3 ðaÞ C, then

(2) Considering uðaÞ ¼ ka C and IFWAx ða1 ; a2 ; . . .; an Þ ¼ ni¼1 xi ai , then

u1 ðXÞ u2 ðXÞ ¼ ðf1 ðlÞ þ f2 ðlÞ f1 ðlÞf2 ðlÞ; g1 ðmÞg2 ðmÞÞ

is called the differential of uðXÞ. In addition, dX ¼ E DX ¼ DX due to dX dX ¼ E

which is Du ¼ du e or Du ¼ du e. In order to investigate this problem,

Thus, according to Theorem 3.4, there is uðX DXÞuðXÞ ðk; 1 kÞ DX.

ðk; 1 kÞ DX ¼ ð0:004; 0:994Þ and k DX ¼ ð0:0040324; 0:9939267Þ

Obviously, replacing k DX by ðk; 1 kÞ DX is approximate.

IFWAx ða0 1 ; a2 ; a3 ; a4 Þ IFWAx ða1 ; a2 ; a3 ; a4 Þ ðx1 ; 1 x1 Þ ða0 1 a1 Þ

IFWAx ða0 1 ; a2 ; a3 ; a4 Þ IFWAx ða1 ; a2 ; a3 ; a4 Þðx1 ; 1 x1 Þ ða1 a0 1 Þ

uða0 Þ ¼ a0 ða XÞ ¼ ðl0 l þ l0 lX l0 llX ; m0 þ mmX m0 mmX Þ

Hence, uða0 Þ ¼ uðaÞ duðaÞ 0

which indicates a0 ða XÞ 6¼ a0 a a0 X holds.

However, if we replace Y 0 t with Y 0 X X0 t and notice that X 0 t dt is actually

dU ðXÞ dðUðXÞC1 Þ dðUðXÞC2 Þ