Fuzzy Mathematics An Introduction for Engineers

and Scientists 2nd Edition Professor John N.

Mordeson pdf download

https://ebookgate.com/product/fuzzy-mathematics-an-introduction-
for-engineers-and-scientists-2nd-edition-professor-john-n-
mordeson/

Get Instant Ebook Downloads – Browse at https://ebookgate.com

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Water Technology Third Edition An Introduction for

Environmental Scientists and Engineers N. F. Gray

https://ebookgate.com/product/water-technology-third-edition-an-
introduction-for-environmental-scientists-and-engineers-n-f-gray/

ebookgate.com

Mathematical Physics Applied Mathematics for Scientists

and Engineers 2nd Edition Bruce R. Kusse

https://ebookgate.com/product/mathematical-physics-applied-
mathematics-for-scientists-and-engineers-2nd-edition-bruce-r-kusse/

ebookgate.com

Management for Engineers Scientists and Technologists 2nd

Edition John V. Chelsom

https://ebookgate.com/product/management-for-engineers-scientists-and-
technologists-2nd-edition-john-v-chelsom/

ebookgate.com

Chaotic vibrations an introduction for applied scientists

and engineers 1st Edition Francis C. Moon

https://ebookgate.com/product/chaotic-vibrations-an-introduction-for-
applied-scientists-and-engineers-1st-edition-francis-c-moon/

ebookgate.com
Fuzzy Mathematics in Economics and Engineering 1st Edition
Professor James J. Buckley

https://ebookgate.com/product/fuzzy-mathematics-in-economics-and-
engineering-1st-edition-professor-james-j-buckley/

ebookgate.com

Transport modeling for environmental engineers and

scientists 2nd ed Edition Clark

https://ebookgate.com/product/transport-modeling-for-environmental-
engineers-and-scientists-2nd-ed-edition-clark/

ebookgate.com

Discrete Mathematics For Computer Scientists And

Mathematicians 2nd Edition Joe Mott

https://ebookgate.com/product/discrete-mathematics-for-computer-
scientists-and-mathematicians-2nd-edition-joe-mott/

ebookgate.com

Basic Electronics for Scientists and Engineers Eggleston

D.L.

https://ebookgate.com/product/basic-electronics-for-scientists-and-
engineers-eggleston-d-l/

ebookgate.com

An Analog Electronics Companion Basic Circuit Design for

Engineers and Scientists 1st Edition Scott Hamilton

https://ebookgate.com/product/an-analog-electronics-companion-basic-
circuit-design-for-engineers-and-scientists-1st-edition-scott-
hamilton/
ebookgate.com
Fuzzy Mathematics
Studies in Fuzziness and Soft Computing
Editor-in-chief
Prof. Janusz Kacprzyk
Systems Research Institute
Polish Academy of Sciences
ul. Newelska 6
01-447 Warsaw, Poland
E-mail: [email protected]
http://www.springer.de/cgi-binlsearch_book.pl ?series=2941

Further Volumes of this series can Vol. 68. A. Kandel, M. Last and H. Bunke (Eds.)
Data Mining and Computational Intelligence, 2001
be found at our homepage. ISBN 3-7908-1371-0
Vol. 58. D.S. Malik and J.N. Mordeson Vol. 69. A. Piegat
Fuzzy Discrete Structures, 2000 Fuzzy Modeling and Control, 2001
ISBN 3-7908-1335-4 ISBN 3-7908-1385-0
Vol. 59. T. Furuhashi, S. Tano and H.-A. Jacobsen Vol. 70. W. Pedrycz (Ed.)
(Eds.) Granular Computing, 2001
Deep Fusion of Computational ISBN 3-7908-1387-7
and Symbolic Processing, 2001
Vol. 71. K. Leiviskli (Ed.)
ISBN 3-7908-1339-7
Industrial Applications of Soft Computing, 2001
Vol. 60. K.J. Cios (Ed.) ISBN 3-7908-1388-5
Medical Data Mining and Knowledge Discovery, Vol. 72. M. Mares
2001 Fuzzy Cooperative Games, 2001
ISBN 3-7908-1340-0 ISBN 3-7908-1392-3
Vol. 61. D. Driankov and A. Saffiotti (Eds.)
Fuzzy Logic Techniques for Autonomous Vehicle Vol. 73. Y. Yoshida (Ed.)
Navigation, 2001 Dynamical Aspects in Fuzzy Decision, 2001
ISBN 3-7908-1341-9 ISBN 3-7908-1397-4

Vol. 62. N. Baba and L. C. Jain (Eds.) Vol. 74. H.-N. Teodorescu, L.c. Jain and
Computational Intelligence in Games, 2001 A. Kandel (Eds.)
Hardware Implementation of Intelligent Systems,
ISBN 3-7908-1348-6
2001
Vol. 63. O. Castillo and P. Melin ISBN 3-7908-1399-0
Soft Computing for Control of Non-Linear
Vol. 75. V. Loia and S. Sessa (Eds.)
Dynamical Systems, 2001
Soft Computing Agents, 2001
ISBN 3-7908-1349-4
ISBN 3-7908-1404-0
Vol. 64. I. Nishizaki and M. Sakawa Vol. 76. D. Ruan, J. Kacprzyk and M. Fedrizzi (Eds.)
Fuzzy and Multiobjective Games for Conflict Soft Computing for Risk Evaluation and
Resolution, 2001 Management, 2001
ISBN 3-7908-1341-9 ISBN 3-7908-1406-7
Vol. 65. E. Orlowska and A. Szalas (Eds.) Vol. 77. W. Liu
Relational Methods for Computer Science Propositional, Probabilistic and Evidential
Applications, 2001 Reasoning, 2001
ISBN 3-7908-1365-6 ISBN 3-7908-1414-8
Vol. 66. R.J. Howlett and L.c. Jain (Eds.) Vol. 78. U. Seiffert and L. C. Jain (Eds.)
Radial Basis Function Networks 1, 2001 Self-Organizing Neural Network, 2001
ISBN 3-7908-1367-2 ISBN 3-7908-1417-2
Vol. 67. R.J. Howlett and L.c. Jain (Eds.) Vol. 79. A. Osyczka
Radial Basis Function Networks 2, 2001 Evolutionary Algorithms for Single and
ISBN 3-7908-1368-0 Multicriteria Design Optimization, 2001
ISBN 3-7908-1418-0
 John N. Mordeson
Premchand S. Nair

Fuzzy Mathematics
An Introduction for Engineers and Scientists

Second Edition

With 30 Figures
and 9 Tables

Springer-Verlag Berlin Heidelberg GmbH

Professor John N. Mordeson
Director
Center for Research in Fuzzy Mathematics
and Computer Science
Creighton University
2500 California Plaza
Omaha, Nebraska 68178
USA
[email protected]

Associate Professor Premchand S. Nair

Department of Mathematics and Computer Science
Creighton University
2500 California Plaza
Omaha, Nebraska 68178
USA
[email protected]

ISSN 1434-9922
ISBN 978-3-7908-2494-0 ISBN 978-3-7908-1808-6 (eBook)
DOI 10.1007/978-3-7908-1808-6
Cataloging-in-Publication Data applied for
Die Deutsche Bibliothek - CIP-Einheitsaufnahme
Mordeson, John N.: Fuzzy mathematics: an introduction for engineers and scientists; with 24 tables I John
N. Mordeson; Premchand S. Nair. - 2. ed. - Heidelberg; New York: Physica-Verl., 2001
(Studies in fuzziness and soft computing; Vol. 20)

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,
1965, in its current version, and permission for use must always be obtained from Springer-Verlag Berlin
Heidelberg GmbH. Violations are liable for prosecution under the German Copyright Law.

© Springer-Verlag Berlin Heidelberg 2001

Originally published by Physica-Verlag Heidelberg New York in 2001
Softcover reprint of the hardcover 2nd edition 200 1

The use of general descriptive names, registered names, trademarks, etc. in this publication does not im-
ply, even in the absence of a specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.
Hardcover Design: Erich IGrchner, Heidelberg
SPIN 10844846 88/2202-5 4 3 2 1 0 - Printed on acid-free paper
FOREWORD

In the mid-1960's I had the pleasure of attending a talk by Lotfi Zadeh

at which he presented some of his basic (and at the time, recent) work on
fuzzy sets. Lotfi's algebra of fuzzy subsets of a set struck me as very nice; in
fact, as a graduate student in the mid-1950's, I had suggested similar ideas
about continuous-truth-valued propositional calculus (inffor "and", sup for
"or") to my advisor, but he didn't go for it (and in fact, confused it with
the foundations of probability theory), so I ended up writing a thesis in a
more conventional area of mathematics (differential algebra). I especially
enjoyed Lotfi's discussion of fuzzy convexity; I remember talking to him
about possible ways of extending this work, but I didn't pursue this at the
time.
I have elsewhere told the story of how, when I saw C.L. Chang's 1968
paper on fuzzy topological spaces, I was impelled to try my hand at fuzzi-
fying algebra. This led to my 1971 paper "Fuzzy groups", which became
the starting point of an entire literature on fuzzy algebraic structures.
In 1974 King-Sun Fu invited me to speak at a U.S.-Japan seminar on
Fuzzy Sets and their Applications, which was to be held that summer in
Berkeley. I wasn't doing any work in fuzzy mathematics at that time, but I
had a long-standing interest in fuzzifying some of the basic ideas of pattern
recognition, so I put together a paper dealing with fuzzy relations on fuzzy
sets, treating them from the viewpoint of fuzzy graphs. This doesn't seem
to have led to a flood of papers on fuzzy graph theory, but the topic does
have important applications, and perhaps the prominence given to it in
this book will lead to renewed activity in this area. [My 1976 paper "Scene
labeling by relaxation operations" (with R.A. Hummel and S.W. Zucker;
IEEE Trans. SMC-6, 420-433) included a treatment of the fuzzy version of
VI FOREWORD

the constraint satisfaction problem, but this too hasn't been widely followed
up.]
Over the past 20 years I have made many excursions into various areas of
fuzzy geometry, digital and otherwise. My 1978 paper "A note on the use of
local min and max operators in digital picture processing" (with Y. Naka-
gawa; IEEE Trans. SMC-8, 632-635) was the first published treatment of
fuzzy mathematical morphology (local max=dilation, local min=erosion).
My 1979 paper "Thinning algorithms for gray-scale pictures" (with C.R.
Dyer; IEEE Trans. PAMI-1, 88-89) presented a fuzzy version of the thin-
ning process in which local min is applied only where it does not weaken
fuzzy connectedness, as defined in my paper "Fuzzy digital topology" (pub-
lished the same year).
The number of papers on fuzzy (digital) geometry grew steadily during
the 1980's. As early as 1984 I was able to survey the subject in my pa-
per "The fuzzy geometry of image subsets" (Pattern Recognition Letters
2, 311-317). A subsequent survey appeared in the Proceedings of the First
IEEE International Conference on Fuzzy Systems (San Diego, March 8-
12, 1992), pp. 113-117. An updated version of this survey was presented
at the opening session of the Joint Conference on Information Sciences in
Wrightsville Beach, NC on September 28,1995 and is to appear in Informa-
tion Sciences. [An early reference on fuzzy geometry, inadvertently omitted
from these surveys is J.G. Brown, "A note on fuzzy sets," Info. Control
18, 1971, 32-39.]
At the end of 1980 I attempted to publish a tutorial paper on fuzzy
mathematics in the American Mathematical Monthly, but the editors ap-
parently didn't like the topic. I then tried the Mathematical Intelligencer;
they eventually published a heavily revised version of that paper ("How
many are few? Fuzzy sets, fuzzy numbers, and fuzzy mathematics" , 2, 1982,
139-143), but most of my discussion of fuzzy mathematical structures was
dropped. The original version was University of Maryland Computer Sci-
ence Technical Report 991, December 1980; it finally saw print in Paul
Wang's book Advances in Fuzzy Theory and Technology I, Bookwrights
Press, Durham, NC, 1993, 1-8.
I was delighted to hear that Profs. Mordeson and Nair were publishing
this book-length introduction to fuzzy mathematics. I hope the book is
successful and stimulates increased interest in the subject.

Azriel Rosenfeld
University of Maryland
PREFACE

We eagerly accepted the invitation of Physica-Verlag to prepare a second

edition of our book. The second edition contains an expanded version of
the first. The first four chapters remain essentially the same. Chapter 5
is expanded to contain the work of Rosenfeld and Klette dealing with the
degree of adjacency and the degree of surroundness. The work of Pal and
Rosenfeld on image enhancement and thresholding by optimization of fuzzy
compactness is also included. Rosenfeld's results on Hausdorff distance be-
tween fuzzy subsets is also included. \¥e expand the geometry of Buckley
and Eslami concerning points and lines in fuzzy plane geometry and in-
clude their new work on circles and polygons in fuzzy plane geometry. In
Chapter 6, we add the latest results on the solution of nonlinear systems
of fuzzy intersection equations of fuzzy singletons.
The book deals with fuzzy graph theory, fuzzy topology, fuzzy geometry,
and fuzzy abstract algebra. The book is based on papers that have appeared
in journals and conference proceedings. Many of the results that appear in
the book are based on the work of Azriel Rosenfeld. The purpose of the
book is to present the concepts of fuzzy mathematics from these areas which
have applications to engineering, science, and mathematics. Some specific
application areas are cluster analysis, digital image processing, fractal com-
pression, chaotic mappings, coding theory, automata theory, and nonlinear
systems of fuzzy equations. The style is geared to an audience more general
than the research mathematician. In particular, the book is written with
engineers and scientists in mind. Consequently, many theorems are stated
without proof and many examples are given. Crisp results of the more ab-
stract areas of mathematics are reviewed as needed, e. g., topology and
abstract algebra. However some mathematical sophistication is required of
viii PREFACE

the reader. Even though the book is not directed solely to mathematicians,
it involves current mathematical results and so serves as a research book
to those wishing to do research in fuzzy mathematics.
In Chapter 1, basic concepts of fuzzy subset theory are given. The notion
of a fuzzy relation and its basic properties are presented. The concept of
a fuzzy relation is fundamental to many of the applications given, e. g.,
cluster analysis and pattern classification. Chapter 1 is based primarily on
the work of Rosenfeld and Yeh and Bang.
Chapter 2 deals with fuzzy graphs. Here again most of the results of this
chapter are based on the work of Rosenfeld and Yeh and Bang. Applications
of fuzzy graphs to cluster analysis and database theory are presented.
Chapter 3 concerns fuzzy topology. We do not attempt to give anywhere
close to a complete treatment of fuzzy topology. There are two books de-
voted entirely to fuzzy topology and as a combination give an extensive
study of fuzzy topology. These two books are by Diamond and Kloeden
and by Liu and Luo. Their exact references can be found at the end of
Chapter 3. In Chapter 3, we review some basic results of topology. We
then feature the original paper on fuzzy topology by C. 1. Chang. For the
remainder of the chapter, we concentrate on results from fuzzy topology
which have applications. These results deal with metric spaces of fuzzy
subsets.
In Chapter 4, we present the work of Rosenfeld on fuzzy digital topology.
An application to digital image processing is given. The chapter also treats
nontopological concepts such as (digital) convexity.
Chapter 5 is on fuzzy geometry. Once again the work of Rosenfeld is
featured. The fuzzy theory developed in this chapter is applicable to pattern
recognition, computer graphics, and imaging processing. We also present
the geometry currently under development by Buckley and Eslami. The
presentation of their geometry is not complete since the book goes to press
before their geometry is completed. We have expanded this and the next
chapter as described above.
Chapter 6 deals with those results from fuzzy abstract algebra which
have known applications. Rosenfeld is the father of fuzzy abstract algebra.
He published only one paper on the subject. However this paper led to
hundreds of research papers on fuzzy algebraic substructures of various
algebraic structures. Here, as in the chapter on fuzzy topology, we review
crisp concepts which are needed for the understanding of the chapter.
Of course the whole notion of fuzzy set theory is due to Lotfi Zadeh.
His classic paper in 1965 has opened up new insights and applications in
a wide range of scientific areas. A large part of Zadeh's orginal paper on
fuzzy sets deals with fuzzy convexity. This notion plays an important role
in this text.

John N. Mordeson
Premchand S. Nair
ACKNOWLEDGMENTS

The authors are grateful to the editorial and production staffs of Physica-
Verlag, especially Janusz Kacprzyk, Martina Bihn, and Gabriele Keidel. We
are indebted to Paul Wang of Duke University for his constant support of
fuzzy mathematics. We are also appreciative of the support of Fr. Michael
Proterra, Dean, Creighton College of Arts and Sciences. We thank Dr.
and Mrs. George Haddix for their support of the center. We also thank Dr.
Mark Wierman for showing us many important features of LaTeX. The first
author dedicates the book to his youngest grandchildren Jessica and David.
The second author is very grateful to his wife, Suseela, and his parents, Mr.
Sukumaran Nair and Ms. Sarada Devi, for supporting his dreams.
CONTENTS

FOREWORD v

PREFACE vii

ACKNOWLEDGMENTS ix

1 FUZZY SUBSETS 1
1.1 Fuzzy Relations . . . . . . . . . . . . . . 6
1.2 Operations on Fuzzy Relations . . . . . 8
1.3 Reflexivity, Symmetry and Transitivity. 11
1.4 Pattern Classification Based on Fuzzy Relations. 12
1.5 Advanced Topics on Fuzzy Relations 16
1.6 References . . . . . . . . . . . . . . . . . . . . . . 20

2 FUZZY GRAPHS 21
2.1 Paths and Connectedness 22
2.2 Clusters . . . . . . . . . . 24
2.3 Cluster Analysis and Modeling of Information Networks 29
2.4 Connectivity in Fuzzy Graphs . 32
2.5 Application to Cluster Analysis 39
2.6 Operations on Fuzzy Graphs . 44
2.7 Fuzzy Intersection Equations . 52
2.8 Fuzzy Graphs in Database Theory 58
2.9 References . . . . . . . . . . . . . . 61

3 FUZZY TOPOLOGICAL SPACES 67

3.1 Topological Spaces . . . . . . . . . 67
3.2 Metric Spaces and Normed Linear Spaces 74
XII CONTENTS

3.3 Fuzzy Topological Spaces 79

3.4 Sequences of Fuzzy Subsets 81
3.5 F -Continuous Functions . . 82
3.6 Compact Fuzzy Spaces . . . 84
3.7 Iterated Fuzzy Subset Systems 85
3.8 Chaotic Iterations of Fuzzy Subsets . 95
3.9 Starshaped Fuzzy Subsets 99
3.10 References . . . . . . . . . . . . . 102

4 FUZZY DIGITAL TOPOLOGY 115

4.1 Introduction....... 115
4.2 Crisp Digital Topology. 115
4.3 Fuzzy Connectedness. 116
4.4 Fuzzy Components. . . 118
4.5 Fuzzy Surroundedness 123
4.6 Components, Holes, and Surroundedness . 124
4.7 Convexity........ 127
4.8 The Sup Projection. . . 128
4.9 The Integral Projection 128
4.10 Fuzzy Digital Convexity 131
4.11 On Connectivity Properties of Grayscale Pictures. 133
4.12 References. . . . . . . . . . . . . . . . . . . . . . . 135

5 FUZZY GEOMETRY 137

5.1 Introduction....................... 137
5.2 The Area and Perimeter of a Fuzzy Subset . . . . . 137
5.3 The Height, Width and Diameter of a Fuzzy Subset 147
5.4 Distances Between Fuzzy Subsets. . . . . . . . . . . 152
5.5 Fuzzy Rectangles . . . . . . . . . . . . . . . . . . . . 155
5.6 A Fuzzy Medial Axis Transformation Based on Fuzzy Disks 158
5.7 Fuzzy Triangles. . . . . . . . . . . . . . . . . . . . . . . . . 163
5.8 Degree of Adjacency or Surroundedness . . . . . . . . . . . 166
5.9 Image Enhancement and Thresholding Using Fuzzy Com-
pactness . . . . . . . . . . . . . . . . . . . . . 181
5.10 Fuzzy Plane Geometry: Points and Lines. . . 189
5.11 Fuzzy Plane Geometry: Circles and Polygons 197
5.12 Fuzzy Plane Projective Geometry. . . . . . . 204
5.13 A Modified Hausdorff Distance Between Fuzzy Subsets. 207
5.14 References. . . . . . . . . . . . . . . . . . . . . . . . . . 214

6 FUZZY ABSTRACT ALGEBRA 219

6.1 Crisp Algebraic Structures. . . . . . . . . . . 219
6.2 Fuzzy Substructures of Algebraic Structures . 233
6.3 Fuzzy Submonoids and Automata Theory . . 238
6.4 Fuzzy Subgroups, Pattern Recognition and Coding Theory 240
 CONTENTS xiii

6.5 Free Fuzzy Monoids and Coding Theory. . . . . . . . .. 245

6.6 Formal Power Series, Regular Fuzzy Languages, and Fuzzy
Automata . . . . . . . . . . . . . . . . . . . . . . . . 252
6.7 Nonlinear Systems of Equations of Fuzzy Singletons 266
6.8 Localized Fuzzy Subrings . . . . . . . . . . . . . . . 272
6.9 Local Examination of Fuzzy Intersection Equations 276
6.10 More on Coding Theory 281
6.11 Other Applications 286
6.12 References. . . . . . . . 287

LIST OF FIGURES 291

LIST OF TABLES 293

LIST OF SYMBOLS 295

INDEX 303
1
FUZZY SUBSETS

In this chapter we explore fuzziness as tool to capture uncertainty. Let S be

a set and let A and B be subsets of S. We use the notation Au B, An B to
denote the union of A and B and intersection of A and B, respectively. Let
B\A denote the relative complement of A in B. The (relative) complement
of A in S, S\A, is sometimes denoted by AC when S is understood.
Let x be an element of S. If x is an element of A, we write x E A;
otherwise we write x ¥ A. We use the notation A ~ B or B :;2 A to denote
that A is a subset of B. If A ~ B, but there exists x E B such that x ¥ A,
then we write A c B or B ~ A and say that A is a proper subset of B. The
cardinality of A is denoted by IAI or card(A). The power set of A, written
g;J(A), is defined to be the set of all subsets of A, i. e., g;J(A) = {UIU ~ A}.
A partition of S is a set P of nonempty subsets of S such that VU, V E P,
either (1) U = V or Un V = 0, the empty set, and (2) S = U U.
UEP
We let N denote the set of positive integers, Z the set of integers, Q the
set of rational numbers, ffi. the set of real numbers and C the set of complex
numbers.
Let X and Y be sets. If x E X and y E Y, then (x, y) denotes the ordered
pair of x with y. The Cartesian cross product of X with Y is defined to
be the set {(x,y)lx E X,y E Y} and is denoted by X x Y. At times we
write X 2 for X x X. In fact, for n E N, n ~ 2, we let xn denote the set
of all ordered n-tuples of elements from X. A relation R of X into Y is a
subset of X x Y. Let R be such a relation. Then the domain of R, written
Dom(R) , is {x E XI:ly E Y such that (x,y) E R} and the image of R,
written Im(R), is {y E YI:lx E X such that (x, y) E R}. If (x, y) E R, we

J. N. Mordeson et al., Fuzzy Mathematics

© Springer-Verlag Berlin Heidelberg 2001
2 1. FUZZY SUBSETS

sometimes write xRy or R(x) = y. If R is a relation from X into X, we say

that R is a relation on X. A relation R on X is called
(i) reflexive ifV'x E X, (x,x) E R;

(ii) symmetric ifV'x,y EX, (x,y) E R implies (y,x) E R;

(iii) transitive if V'x, y, z E X, (x, y) and (y, z) E R implies (x, z) E R.
If R is a relation on X which is reflexive, symmetric, and transitive,
then R is called an equivalence relation. If R is an equivalence relation on
X, we let [x] denote the equivalence class of x with respect to R and so
[x] = {a E XlaRx}. If R is an equivalence relation on X, then {[x]lx E X}
is a partition of X. Also if P is a partition of X and R is the relation on
X defined by V'x,y EX, (x,y) E R if::lU E P such that X,y E U, then R
is an equivalence relation on X whose equivalence classes are exactly those
members of P.
A relation R on X is called antisymmetric if V'x, y E X, (x, y) E Rand
(y, x) E R implies x = y. If R is a reflexive, antisymmetric, and transitive
relation on X, then R is called a partial order on X and X is said to be
partially ordered by R.
Let R be a relation of X into Y and T a relation of Y into a set Z. Then
the composition of R with T, written ToR, is defined to be the relation
{(x, z) E X x ZI::Iy E Y, such that (x, y) E Rand (y, z) E T}.
If f is a relation of X into Y such that Dom(J) = X and V'x, x' E X,
x = x' implies f(x) = f(x'), then f is called a function of X into Y and
we write f : X ----> Y. Let f be a function of X into Y. Then f is some-
times called a mapping and f is said to map X into Y. If V'y E Y, ::Ix E X
such that f(x) = y, then f is said to be onto Y or to map X onto Y. If
V'x, x' E X, f(x) = f(x') implies that x = x', then f is said to be one-to-
one and f is called an injection. If f is a one-to-one function of X onto Y,
then f is called a bijection. If g is a function of Y into a set Z, then the
composition of f with g, go f, is a function of X into Z which is one-to-one
if f and g are and which is onto Z if f is onto Y and g is onto Z. If Im(J)
is finite, the we say that f is finite-valued. We say that an infinite set X is
countable if there exists a one-to-one function of X onto N; otherwise we
call X uncountable.

Fuzzy theory holds that many things in life are matters of degree. A
black and white photo is not just black and white; there are many levels of
gray shades which can be observed in a typical picture. Computer scientists
and engineers have long accepted this fact. As an example, a pixel can have
a brightness value between 0 and 255. The 0 value stands for black, 255
stands for white and every number between 0 and 255 stands for a certain
gray level.
Let S be a set. A fuzzy subset of S is a mapping A : S ----> [0,1]. We
think of A as assigning to each element xES a degree of membership,
 1. FUZZY SUBSETS 3

0::; A(x) ::; 1. Let A be a fuzzy subset of S. We let At = {x E SIA(x) 2: t}

for all t E [0,1]. The sets At are called level sets or t-cuts of A. We let
supp(A) = {x E SIA(x) > O}. We call supp(A) the support of A. The set
of all fuzzy subsets of S is denoted by Jp(S) and is called the fuzzy power
set of S.
Example 1.1 Let S = {a,b,c,d}. Then A = {a,b} is a subset of S. On
the other hand the mapping A : S ~ [0,1] such that A(a) = 1, A(b) =
1,A(c) = O,A(d) = 0 is a fuzzy subset of S. Similarly, B = {a,c,d} is a
subset of S and the mapping B : S ~ [0,1] such that B(a) = 1, B(b) =
0, B(c) = 1, B(d) = 1 is a fuzzy subset of S. We see that corresponding to
a subset X of S, there is always a fuzzy subset X of S with the following
property.

(i) x E X if and only ~f X(x) = 1

(ii) x ~ X if and only if X(x) = 0

On the other hand the mapping 6 : S ~ [0,1] which assigns 6(a)

0.3,6(b) = 0.9,6(c) = 0.4,6(d) = 0.625 is a fuzzy subset of S. Corre-
sponding to the fuzzy subset 6 there are five level subsets of S, as shown
below.

- 1fb,
ct = c,d}
{b,d}
o ::; t ::; 0.3
0.3 <
0.4 <
t ::; 0.4
t ::; 0.625
{b} 0.625 < t ::; 0.9
o 0.9 < t ::; 1.0

FIGURE 1.1 Graphical representation of the fuzzy set 6.

-

I---

I a b c d

We see that a finite-valued fuzzy subset 6 determines a "chain of sub-

sets of S". Conversely, given a finite chain of subsets C1 ~ C2 ~ ... ~
Ck = S, k > 0, there exists a fuzzy subset 6 such that its level sets are
4 1. FUZZY SUBSETS

C 1 , C 2 , ... , C k = S. The construction of 0 can be outlined as follows.

Choose k numbers 0 ::; h < t2 < ... < tk ::; 1. For 8 E S, define 0(8)
as follows:
0(8) = {ti ~f 8 E Ck-i+1 \ Ck-i,i = 1,2, ... ,k -1,
tk 1f 8 E C1 .

Let A be a subset of a set S. Define XA : S -+ [0,1] by XA(a) = 1 if a E A

and XA (a) = 0 if a E S \ A. Then XA is called the characteristic function
of A in S.
Definition 1.1 Let A, B be any two fuzzy subsets of S. Then

(i) A ~ B if A(x) ::; B(x) for all XES,

(ii) A c B if A(?:) ::; B(x) for all xES and there exists at least one xES
such that A(x) < B(x),
(iii) A = B ~f A(x) = B(x) for all xES.
We now proceed to define the union and intersection of fuzzy subsets
as well as the complement of a fuzzy subset. We use the notation V for
supremum and 1\ for infimum.
Definition 1.2 Let A, B be any two fuzzy subsets of S. Then Au B is the
fuzzy subset of S defined by

(A U B)(x) = A(x) V B(x) for all xES

and A n B is the fuzzy subset of S defined by

(A n B)(x) = A(x) 1\ B(x) for all xES.
Definition 1.3 Let A be any fuzzy subset of S. Then AC is the fuzzy subset
of S defined by
k(x) = 1 - A(x) for all xES.

Example 1.2 Let S = {a, b, c, d}. Let A : S -+ [0,1] be such that A(a) =
0.3, A(b) = 0.9, A(c) = 0.4, A(d) = 0.6 and let B : S -+ [0,1] be such that
B(a) = 0.3, B(b) = 0.5, B(e) = 0.7, B(d) = 0.2. Then

I I
S o < t < 0.3 S o : : ; t ::; 0.2
_ {b, e, d} 0.3 < t ::; 0.4 {a, b, e} 0.2 < t ::; 0.3
At = {b, d} 0.4 < t ::; 0.6 and Bt = {b, c} 0.3 < t ::; 0.5
{b} 0.6 < t ::; 0.9 {c} 0.5 < t::; 0.7
o 0.9 < t ::; 1.0 o 0.7 < t ::; 1.0

Now (AUB)(a) = 0.3, (AUB)(b) = 0.9, (AUB)(e) = 0.7, and (AUB)(d) =

0.6. Also (AnB)(a) = 0.3, (AnB)(b) = 0.5, (AnB)(e) = 0.4, (AnB)(d) =
0.2. Thus
 1. FUZZY SUBSETS 5

(AUE)' ~ !
5
{b,c,d}
{b,c}
{b}
0
o ~ t ~ 0.3
0.3
0.6
0.7
0.9
<t
<t
<t
<t
~
~
~
~
0.6
0.7
0.9
1.0
(AnE)' ~ ! 5
{a,b,c}
{b,c}
{b}
0
o ~ t ~ 0.2
0.2
0.3
0.4
0.5
<t
<t
<t
<t
~
~
~
~
0.3
0.4
0.5
1.0

!
Further,

5 o ~ t ~ 0.1
A<~ {a,c,d}
{a, c}
{a}
0
0.1
0.4
0.6
0.7
<t
<t
<t
<t
~
~
~
~
0.4
0.6
0.7
1.0

FIG URE 1.2 Fuzzy subsets A, B, A U B, A n B, Ac, respectively.

If S is a collection of fuzzy subsets of 5, we define the fuzzy subset

i = nC (intersection) of 5 by Vx E 5, i(x) = A{C(x)IC E S} and the
CES
fuzzy subset U = UC (union) of 5 by Vx E 5, U(x) = v{C(x)IC E S}.
CES
6 1. FUZZY SUBSETS

Algebraic properties of fuzzy subset operators can be summarized as

follows. Now Xs(x) = 1 for all xES and X0(x) = 0 for all xES. Let A, 13,
and C be fuzzy subsets of S. Then we have the following properties.

(1) Au 13 = 13 u A (9) AU (13 u C) = (A U B) U C

(2) AnB = BnA (10) An (13 n C) = (A n B) n C
(3) Au X0 = A (11) An (13 U C) = (A n B) U (A n C)
(4) An X0 = X0 (12) Au (13 n C) = (A U B) n (A U C)
(5) Au Xs = Xs (13) (A U B)C = Jic nBC
(6) An Xs = A (14) (A n B)C = Jic U B C
(7) Au A = A (15) (Jic)C = A
(8) An A = A

It is important to note that the properties A n A C = X0 and A U A C = Xs

do not hold in generaL In logic, the former property is known as the law of
contradiction while the latter is known as the law of the excluded middle.
For the interested reader, the bibliography at the end of the chapter
contains many excellent textbooks on fuzzy subsets.

1.1 Fuzzy Relations

Let Sand T be two sets and let A and 13 be fuzzy subsets of Sand T,
respectively. Then a fuzzy relation R from the fuzzy subset A into the fuzzy
subset 13 is a fuzzy subset R of S x T such that R(x, y) :S A(x) I\B(y), "Ix E
Sand yET. In other words, for R to be a fuzzy relation, we require that
the degree of membership of a pair of elements never exceed the degree of
membership of either of the elements themselves. If we think of the elements
as computers and pairs as the communication links between the computers,
this amounts to requiring that the strength of the communication link can
never exceed the strengths of its connecting computers. More generally,
the amount of flow of a quantity from a source to a sink is limited by the
capability to transmit and receive. Also, the restriction R(x, y) :S A(x) 1\
B(y), "Ix E Sand yET allows Rt to be a relation from At into Bt for all
t E [0,1] and for supp(R) to be a relation from supp(A) into supp(B).
There are three special cases of fuzzy relations which are extensively
found in the literature.

case 1: S = T and A = B. In this case, R is said to be a fuzzy relation

on A. Note that R is a fuzzy subset of S x S such that R(x,y) :S
A(x) 1\ A(y).

case 2: A(x) = 1.0 for all xES and B(Y) = 1.0 for all yET. In this case,
R is said to be a fuzzy relation from S into T.
 1.1 Fuzzy Relations 7

case 3: S = T, A(x) = 1.0 for all xES and 13(y) = 1.0 for all yET. In
this case, R is said to be a fuzzy relation on S.

There are many applications in which a fuzzy relation on a fuzzy subset is

quite useful. Further, any result we obtain is clearly true for fuzzy relations
on a set. Therefore, we devote the first three sections of this chapter to fuzzy
relations on a fuzzy subset. The last section of this chapter is devoted to
the study of case 2 and case 3.

Example 1.3 Let S = {a, b, c, d}. Define A(a) = 0.3, A(b) = 0.7, A(c) =
0.4 and A(d) = 0.5. Then A is a fuzzy subset of S. Let R be a fuzzy subset
of S x S defined as follows:

a b c d
a 0.1 0.2 0.0 0.3
b 0.2 0.7 0·4 0.3
c 0.3 0.6 0.2 0·4
d 0.2 0.5 0.3 0.5

Since R(c, b) i A(c) /\ A(b), R is not a fuzzy relation on A. However, if

we redefine R( c, b) = 0.4 then R is a fuzzy relation on A. Alternatively, we
could redefine A(c) = 0.6 and meet the constraint.

Let R be a fuzzy relation on A. Then R is called the strongest fuzzy rela-

tion on A if and only if for all fuzzy relations Q on A, \:Ix, YES, Q(x, y) ~
R(x, y).

Example 1.4 Let S = {a, b, c, d}. Define A(a) = 0.3, A(b) = 0.5, A(c) =
0.7 and A(dl = 0.9. Then A is a fuzzy subset of S. The strongest fuzzy
relation on A is given below.

a b c d
a 0.3 0.3 0.3 0.3
b 0.3 0.5 0.5 0.5
c 0.3 0.5 0.7 0.7
d 0.3 0.5 0.7 0.9

The converse problem may also arise in practice. That is, we know the
strength of the pairs and we want to compute the minimum strength re-
quired for the elements themselves. For a given fuzzy subset R of S x S,
the weakest fuzzy subset A of S on which R is a fuzzy relation is defined
by A(x) = V{R(x,y) V R(y,x)ly E S} for all xES. That is, if 13 is a fuzzy
subset of Sand R is a fuzzy relation on 13, then A <:;:; 13.

Example 1.5 Let S = {a, b, c, d}. Let R be a fuzzy subset of S x S defined

as follows:
8 1. FUZZY SUBSETS

a b c d
a 0.1 0.2 0.0 0.3
b 0.2 0.7 0·4 0.3
c 0.3 0.6 0.2 0·4
d 0.2 0.5 0.3 0.5

The largest value on the first row and the first column is 0.3. Therefore
we define A(a) = 0.3. Similarly, the largest value on the second row and
the second column is 0.7. Therefore we define A(b) = 0.7. By the same
reasoning, A(c) = 0.6 and A(d) = 0.5. Clearly, R is a fuzzy relation on A.
In fact, A is the weakest .fuzzy subset on which R is a fuzzy relation.

Given a fuzzy relation R on a fuzzy set A, it is possible to apply a

threshold value to both of them to obtain a relation Rt on the subset At.

Exrunple 1.6 Let S = {a,b,c,d}. Define A(a) = 0.4, A(b) = 0.7,A(c) =

0.6 and A(d) = 0.5. Then A is a fuzzy subset of S. Let R be a fuzzy subset
of S x S defined as follows:

a b c d
a 0.1 0.2 0.0 0.3
b 0.2 0.7 0·4 0.3
c 0.3 0.6 0.2 0·4
d 0.2 0.5 0.3 0.5

It is clear that R is a fuzzy relation on A. Let us choose t = 0.55 as our

threshold value. Then Rt = {(b, b), (c, b)} is a relation on At = {b, c}.

1.2 Operations on Fuzzy Relations

Consider two computer networks, sharing the same computers, but different
connecting lines. If we have the choice of selecting either of the connecting
lines, the maximum strength possible as well as the minimum strength that
can be guaranteed between any pair of computers is an interesting problem.
Let Rand Q be two fuzzy relations on a fuzzy subset A of S. Then R U Q
is the relation on A defined by (RU Q)(x, y) = R(x, y) V Q(x, y), \:Ix, yES.
Clearly, R U Q represents the maximum strength possible between any
two computers. R n Q is the relation on A defined by (R n Q) (x, y) =
R(x, y) 1\ Q(x, y), \:Ix, yES. The relation R n Q represents the minimum
strength that can be guaranteed between any two pairs in the event of at
most one of the connecting lines are down at any time.
Suppose we want to send a message from computer x to computer z.
Let us say, we prefer to use the "R" network in the geographical area
 1.2 Operations on Fuzzy Relations 9

of the computer x and the "Q" network in the geographical area of the
computer z. Further we are willing to use exactly one computer, say y,
to route our message. In such a situation, we want to find the maximum
strength possible between x and z. Fuzzy theory provides an operation
called composition to provide the answer.

Definition 1.4 Let R : S x T ---; [0, 1] be a fuzzy relation from a fuzzy

subset A of S into a .fuzzy subs..et B of T and Q : T x U ---; 10, 1] be a fuzzy
relation from a fuzzy subset B of T into a fuzzy subset C of U. Define
R 0 Q: S x U ---; [0, 1] by
R 0 Q(x, z) = V{R(x, y)!\ Q(y, z)ly E T}

for all XES, z E U. The fuzzy relation R 0 Q is called the composition of

R with Q.

Note that composition of R with Q is a fuzzy relation from a fuzzy subset

A of S into a fuzzy subset C of U. A closer look at the definition of the
composition operation reveals that RoQ can be computed similar to matrix
multiplication, where the addition is replaced by V and the multiplication
is replaced by !\. Since composition is associative,we use the notation R2
to denote the composition R 0 R, Rk to denote R k- I 0 R, k > 1. Define
ROO(x,y) = V{Rk(X,y)lk = 1,2, ... } for all X,y E S. Finally, it is conve-
nient to define RO(x, y) = a if xi- y and RO(x, y) = .A(x) otherwise, for all
x, y E S.We have introduced three binary operations. We now introduce a
unary operation on a fuzzy relation. Given a fuzzy relation R on a fuzzy
subset .A of S, define the fuzzy relation RC on A by flc(x, y) = 1 - R(x, y)
for all X,y E S.

Definition 1.5 Let R : S x T ---; [0,1] be a fuzzy relation from a fuzzy

subset A of S into a fuzzy subset B of T. Define the fuzzy relation R- I :
T x S ---; [0,1] of B into A by R-I(y, x) = R(x, y) for all (y, x) E T x S.

Example 1.7 Let S = {a,b,c,d}. Define A(a) = 0.4, A(b) = 0.7,A(c) =

0.6 and A(d) = 0.5. Then A is a fuzzy subset of S. Let Rand Q be fuzzy
relations defined on A as follows:

a b c d a b c d
a 0.1 0.2 0.0 0.3 a 0.3 0.1 0.3 0·4
R: b 0.2 0.7 0·4 0.3 Q: b 0·4 0.5 0.3 0.5
c 0.3 0.6 0.2 0·4 c 0.2 0·4 0.6 0.3
d 0.2 0.5 0.3 0.5 d 0.3 0.3 0·4 0.2

Then the relations R U Q, R nQ and R 0 Q are as below:

10 1. FUZZY SUBSETS

a b c d a b c d
a 0.3 0.2 0.3 0·4 a 0.1 0.1 0.0 0.3
RUQ: b 0·4 0.7 0·4 0.5 RnQ: b 0.2 0.5 0.3 0.3
c 0.3 0.6 0.6 0·4 c 0.2 0·4 0.2 0.3
d 0.3 0.5 0·4 0.5 d 0.2 0.3 0.3 0.2

a b c d a b c d
a 0.3 0.3 0.3 0.2 a 0.9 0.8 1.0 0.7
RoQ: b 0·4 0.5 0·4 0.5 RC: b 0.8 0.3 0.6 0.7
c 0·4 0.5 0·4 0.5 c 0.7 0·4 0.8 0.6
d 0·4 0.5 0·4 0.5 d 0.8 0.5 0.7 0.5

Algebraic Properties of Fuzzy Relation Operators

Let T, P, Rand Q be any four fuzzy relations on a fuzzy subset A of a set
S. Then we have the following.

1. RUQ=QUR
2. RnQ=QnR
3. R= (Rc)C
4. P U (RU Q) = (p U R) U Q
5. P n (R n Q) = (p n R) n Q
6. Po (R 0 Q) = (p 0 R) 0 Q
7. Pn(RuQ)=(PnR)U(PnQ)
8. P U (Rn Q) = (p U R) n (p U Q)
9. (RUQ)c=QcnRc
10. (R n Q)C = QC U RC
11. For all t E [0,1]' (R U Q)t = Rt U Qt
12. For all t E [0,1]' (R n Q)t = Rt n Qt
13. For all t E [0,1], (RoQ)t ;2 RtoQt and if S is finite, (RoQ)t = RtoQt
14. If T s:;; Rand P s:;; Q then T UPs:;; R U Q
15. If T s:;; Rand P s:;; Q then Tn P s:;; R n Q
16. If T s:;; Rand P s:;; Q then ToP s:;; R0 Q
 1.3 Reflexivity, Symmetry and Transitivity 11

1.3 Reflexivity, Symmetry and Transitivity

Throughout this section R and Q are fuzzy relations on a fuzzy subset A of
S. As we have seen, it is quite natural to represent a fuzzy relation in the
form of a matrix. In this section, we shall use the matrix representation of
a fuzzy relation to explain the properties of a fuzzy relation. In particular,
we shall use the term "diagonal" to represent the principal diagonal of the
matrix.

Reflexivity
We call R reflexive if R(x, x) = A(x) for all xES. If R is reflexive, then
R(x, y) :S Aix) 1\ A(Y) :S A(x) = R(x, x) and it follows that "any diagonal
element of R is larger than or equal to any element in its row". Similarly,
"any diagonal element is larger than or equal to any element in its column" .
Conversely, given a fuzzy relation R on A such that "any diagonal element is
larger than or equal to any element in its row and column" , define a fuzzy
subset B of S as B(xL = R(x, x), \Ix E S. T~en B is th~ weakest fuzzy
sy.bset of S such that R is a fuzzy relation on B. Further, R is reflexive on
B.
Fuzzy reflexive relations have some interesting algebraic properties. Let
R and Q be any two fuzzy relations on a fuzzy subset A of S. Then we
have the following.

1. If R is reflexive, Q~ Q0 Rand Q~ R 0 Q.
2. If R is reflexive, R ~ R2.

3. If R is reflexive, RO ~ R ~ R2 ~ R3 ~ ... ~ ROO.

4. If R is reflexive, RO(x,x) = R(x, x) = R2(x,x) = R 3(x,x) = ... =

ROO (x, x) = A(x), \Ix E S.
5. If Rand Q are reflexive, so is R 0 Qand Q0 R.

6. If R is reflexive, for any 0 :S t :S 1, Rt is a reflexive relation on At.

Symmetry
We call R symmetric if R( x, y) = R(y, x), for all x, yES. In other words,
R is symmetric if the matrix representation of R is symmetric (with respect
to the diagonal). Let Rand Q be any two fuzzy relations on a fuzzy subset
A of S. Then we have the following.
1. If Rand Q are symmetric, then so is RoQ if and only if RoQ = QoR.
2. If R is symmetric, then so is every power of R.
12 1. FUZZY St:BSETS

3. If R is symmetric, then for any 0 'S t 'S 1, Rt is a symmetric relation

OIl AI.

Transitivity
We call R transitive if R2 ~ R. It may be noted that Rco is transitive for
any fuzzy relation R. Let P, Rand Q be any three fuzzy relations on a
fuzzy subset A of S. Then we have the following.

1. If R is transitive, then so is every power of k

2. If R is transitive and P ~ R, Q ~ R, then Po Q~ k

3. If R is transitive, Q is reflexive and Q ~ k then R Q= Q R = k

0 0

4. If R is reflexive and transitive, then R2 = R.

5. If R is reflexive and transitive, then RO ~ R = R2 = R3 = ... = Roo.

6. If Rand Q are transitive and R Q= Q R, then R Qis transitive.

0 0 0

7. If R is symmetric and transitive, then R(x, y) 'S R(x, x) and R(y, x) 'S
R(x,x), for all :1;,y E S.
8. If R is transitive, then for any 0 'S t 'S 1, Rt is a transitive relation
on At.

A fuzzy relation R on S which is reflexiYe, symmetric, and transitive is

called a fuzzy equivalence relation on S.

1.4 Pattern Classification Based on Fuzzy

Relations
Let S be a set of patterns. A class~fication fuzzy relation R on S is a fuzzy
relation satisfying the following two conditions.

Cl R(x, x) = 1 for all xES.

C2 R(x, y) = R(y, x) for all x, yES.

Note that the condition C1 states that a pattern .r: is identical with itself.
Thus the relation is reflexive. Condition C2 means that any relation used
to classify patterns has to be symmetric. Since R is reflexive, R ~ R2 ~
R3 ~ ... ~ R=. Note that Roo is a fuzzy equivalence relation. So for any
O'S t 'S 1, (R=)t is an equivalence relation on S. Let pt be the partition of
S induced by the equivalenf'e relAtion (Roor. The next result holds since
(ROO)t is transitive.
 1.4 Pattern Classification Based on Fuzzy Relations 13

Lemma 1.1 For all x, y, z E 5,

ROO (x, z) ;:=: ROO (x, y) /\ ROO(y, z) . •

Theorem 1.2 If ROO (x, y) =I- 1, for all x, y E 5 such that x =I- y, then
p(x, y) = 1- ROO(x, y) satisfies the axioms of distance. That is, "ix, y, z E 5,

(i) p(x, y) = 0 if and only if x = y,

(ii) p(x, y) = p(y, x),

(iii) p(x, z) :S p(x, y) + p(y, z).

I?roof. We show only_property (iii}. By Lemm~ 1.1, ROO(x, z) ;:=: ROO(x,y)/\
ROO(y, z). Thus 1 + ROO(x, z) ;:=: ROO(x, y) + ROO(y, z). The desired result
now follows easily. •

Xl X2 x3 x4 x5
Xl 1.0 0.8 0.0 0.1 0.2
X2 0.8 1.0 0·4 0.0 0.9
X3 0.0 0·4 1.0 0.0 0.0
X4 0.1 0.0 0.0 1.0 0.5
X5 0.2 0.9 0.0 0.5 1.0

Now Roo = R3 is given by

Xl x2 x3 x4 x5
Xl 1.0 0.8 0·4 0.5 0.8
X2 0.8 1.0 0·4 0.5 0.9
X3 0·4 0·4 1.0 0·4 0·4
X4 0.5 0.5 0·4 1.0 0.5
X5 0.8 0.9 0·4 0.5 1.0

and we have the following partitions of 5:

pO = pO.3 = {{Xl, x2, x3, X4,X5}}
p0.45 = {{Xl, X2, X4,X5}, {X3}}
pO.55 = {{Xl, x2, X5}, {X4}' {X3}}
pO.85 = {{Xd,{X2,X5},{X4},{X3}}
ploO = {{xd, {X2}, {X5}, {X4}, {X3}}
Thus there are many partitions possible and depending upon the level of
detail, one could class~fy the patterns based on equivalence relations. Note
that if s ;:=: t, then pB is a refinement of pt.
14 1. FUZZY SUBSETS

Experimental Result
We now present an experiment done by Tamura, Higuchi and Tanaka [8].
Portraits obtained from 60 families were used in their experiment, each
of which were composed of between four and seven members. The rea-
son why they chose the portraits is that even if parents do not possess
a facial resemblance, they may be connected through their children, and
consequently they could classify the portraits into families. First, they di-
vided the 60 families into 20 groups, each of which was composed of 3
families. Each group was, on the average, composed of 15 members. The
portraits of each group were presented to a different student to give the
values of the subjective similarity R( x, y) between all pairs on a scale of
1 to 5. The reason why they used the 5 rank representation instead of a
continuous value representation is that it had been proved that a human
being cannot distinguish into more than 5 ranks. Twenty students joined
in this experiment. An example of the experiment for one group with 16
portraits is shown in Table 1.1 and Table 1.2. The first column of the ta-
bles gives the portrait number. In Table 1.1, the 5 rank representations are
converted to values in [0, 1], namely 0.2, 0.4, 0.6, and 1.0. In this example,
the number of patterns is sufficiently large that they can not be classified
by inspection. The classification requires the calculation of Roo. Since the
levels of the subjective values are different according to individuals, the
threshold was determined in each group as follows. As they lowered the
threshold, the number of classes decreased. Hence, under the assumption
that the number of classes c to be classified was known to be 3, while low-
ering the threshold they stopped at the value which divided the patterns
into 3 classes and some nonconnected patterns. However, as in the present
case, when some R(x, y) take the same value, sometimes there is no thresh-
old by which the patterns are divided into just c given classes. In such a
case, they made it possible to divide them into just c classes by stopping
the threshold at the value where the patterns are divided into less than c
classes and separating some connections randomly that have a minimum
R( x, y) of connections that have the stronger relation than the threshold.
The correctly classified rates, the misclassified rates, and the rejected rates
of 20 groups were within the range of 50-94 percent, 0-33 percent, and 0-
33 percent, respectively, and they obtained the correctly classified rate 75
percent of the time, the misclassified rate 13 percent, and the rejected rate
12 percent as the averages of the 20 groups. Here, since the classes made
in this experiment have no label, they calculated these rates by making a
one-to-one correspondence between 3 families and 3 classes, so as to have
the largest number of correctly classified patterns.
Thus Tamura et al. [8] have studied pattern classification using subjec-
tive information and performed experiments involving classification of por-
traits. The method of classification proposed here is based on the proce-
dure of finding a path connecting 2 patterns. Therefore, this method may
 1.4 Pattern Classification Based on Fuzzy Relations 15

be combined with nonsupervised learning and may also be applicable to

information retrieval and path detection [8].

TABLE 1.1 Representation of subjective similarities 2 R.

3 7 10 11 12 13 1~ 15 16

0
0 0
0 0 .4
0 .8 0 0
.5 1 0 .2 .2 0 1
7 0 .S 0 0 .4 0
8 .4 .2 .2 .5 1 0 .8 0 1
.4 0 .8 .4 .2 .4 0
10 0 0 .2 .2 0 0 .2 0 .2 1
11 0 .5 1 .2 .2 0 0 .8 0 .4 .2
12
13
0
.8
0
0
.2
.2
."., 0
0
0
.4
0
0
0
.4
.4
0
.S
0
0
0 0 1
14 0 .8 0 .2 .4 0 .8 0 .2 .2 .6 0 0 1
15 0 0 .4 .8 0 .2 0 0 .2 0 0 .2 .2 0
16 .6 0 0 .2 .2 .8 0 .4 0 0 0 .4 .2 0

TABLE 1.2 The relation2 Roo.

3 7 10 11 12 13 14 15 16

.4 1
.4
.5
.4
A .,1
.4 .8 .4 A
6 .6 .4 .4 .5 .4 1
7
8
.4
.6 .,
.8
.,
.4 .4
.5
.8
.4
.4
.8
1
.4
9 .5 .4 .4 .8 .4 .5 .4 .5
10
11 ., ., .,
.5 .4
.8
.4
.4
.8
.4
.4

.,
.8
.5
.4
.4
.8
.5
.4 .,.8 .4
12 .5 .8
., .5 .4 .5 .8 .8 .4
13
I'
.8
.4
.4
.8 .,
.4 .5
.4 .8
.6
.4
.4
.8
.6
.4
.5
.4 .,
.5 .4
.8 .,.5 .,1
.,
15 .5 .4 .4 .8 .4 .5 .4 .5 .8 .8 .4 .8 .5
16 .6 .4 .4 .5 .4 .8 .4 .8 .5 .5 .4 .5 .6 .4 .5

Table 1.2 displays the subjective similarities among family members. The
portraits of family members for family one were 1, 6, 8, 13, and 16 with
portraits 6 and 13 being of parents and 1, 8, and 16 being of the children.
The portraits offamily members for family two were 2,5,7,11, and 14, with
portraits 5 and 11 being of parents and 2, 7, and 14 being of the children.
For family three portrait 4 was that of a grand mother, portraits 10 and 15
of parents and 3, 9, and 12 of the children. The rate of correct classification
was 15/16, the rate of misclassification was 0/16, and the rate of rejection
was 1/16. In this experiment the portrait 3 was rejected.

lThis value was converted from 0.6 to 0.5 for division into just three classes.
2Reprinted from: Tamura, S., Higuchi, S., and Tanaka, K., Pattern Classification
Based on Fuzzy Relations, IEEE Trans. SI\IC-1, 61-66, 1971.
16 1. FUZZY SUBSETS

1.5 Advanced Topics on Fuzzy Relations

In this section we will show that the concept of a similarity relation intro-
duced by Zadeh [11] is derivable in much the same way as an equivalence
relation. Throughout this section we shall be dealing with a fuzzy relation
on a set. The results in this section are from [9].

Definition 1.6 Let R be a fuzzy relation on a set S. We define the follow-

ing notions:

(i) R is E-reflexive if V XES, R(x, x) 2 E, where E E [0,1].

(ii) R is irreflexive ifVx E S, R(x, x) = O.

(iii) R is weakly reflexive if for all x, y in S and for all E E [0,1] ,R(x, y)
= E:::} R(x, x) 2 E.

Remark 1 Note that the definition of a reflexive relation as a 1-reflexive

relation coincides with the definition of a refiexive relation in Section 1.3.

Lemma 1.3 If R is a fuzzy relation from S into T, then the fuzzy relation
R 0 R- 1 is weakly reflexive and symmetric . •

Let R be a weakly reflexive and symmetric fuzzy relation on S. Define a

family of non-fuzzy subsets pH as follows:
pH = {K ~ SI(30 < E :::; l)(Vx E S)[x E K {:} (Vx' E K)[R(x, x') >
El]}.
We note that if we let
P€H = {K ~ SI(Vx E S)[x E K {:} (Vx' E K)[R(x, x') 2 El]},
then we see thatE1:::; E2 :::} P€~ =;< P€~ where "=;<" denotes a covering
relation, i.e., every element in P€~ is a subset of an element in P€~.
A subset J of S is called E-complete with respect to R if V x, x' E
J, R(x, x') 2 E. A maximal E-complete set is one which is not properly
contained in any other E-complete set.

Lemma 1.4 pH is the family of all maximal E-complete sets with respect
to R for 0 :::; E :::; 1. •

Lemma 1.5 Whenever R(x, x') > 0, there is some E-complete set K E pH
such that {x, x'} ~ K. •

We remark here that sometimes a subclass of pH, satisfying the condition

of Lemma 1.5, will cover the set S. For example, let R be the fuzzy relation
on S = {a, b, c, d, e, J} given by the following matrix.
 1.5 Advanced Topics on Fuzzy Relations 17

a b c d e f
a 1.0 0.3 0.4 0.0 0.4 0.3
b 0.3 1.0 0.2 0.3 0.0 0.4
c 0.4 0.2 l.0 0.3 0.5 0.0
d 0.0 0.3 0.3 1.0 0.0 0.0
e 0.4 0.0 0.5 0.0 l.0 0.0
f 0.3 0.4 0.0 0.0 0.0 1.0
We see that the family consisting of the three maximal complete sets
{a, b, j}, {b, c, d}, and {a, c, e} satisfy the condition of Lemma 1.5, but it
does not contain the maximal complete set {a, b, c}. Note for example that
ta, b, j} is :naximal E'-cornplete V 0 < E' S E where E = 0.3 since R(a, d) =
= RU,c) = O. We have
R(b, e)
Ff = {{a},{b},{c},{d},{e},{fn
Fd\ = {{c,e},{a},{b},{d},{f}}}
Fd~4 ={{a,c,e},{b,j},{dn
Ff3 ={{a,b,j},{a,c,e},{b,d},{c,d}}
en.
FOR2 = {fa, b, c}, {b, c, d}, {a, b, j}, {a,c,
\Ve also note that R is not transitive: R.(b, c) = .2 i .3 = (.3/\.4) V (1/\
.2) V (.2/\ 1) V (.3/\ .3) (0/\ .5) V (.4/\ 0) = V{R.(b, y) /\ R(y, c) lyE S} =
R 0 R(b, c).
In the next two results X0 is the characteristic function of I/J in S x S.
Lemma 1.6 If R Ie X0 is a weakly r-efiexive and symmetric fuzzy r-elation
on S, then ther-e exists a set T and a fuzzy r-elation P from S into T such
that R = P 0 P-l .•

Combining Lemmas 1.3 and 1.6, we haye the following result.

Theorem 1.7 A fuzzy relation R Ie X0 on a set S is weakly r-eflexive and
symTnetTic ~f and only ~f ther-e is a set T and a fuzzy relation P fTOm S into
T such that R = PoP-I .•

It may be noted given a weakly reflexive and symmetric fuzzy relation

R Ie X0 on a set S, the set T and the fuzzy relation P from S into T can be
constructed as follows. Define T as the set {K*IK E FR}. In other words,
T is a set having the same cardinality as FR and assume that there is a
one-to-one mapping from FR onto T. Now P from S into T is defined as
follows:
p(x, K*) = {a if x E ~,a is the largest number such that K E Fl;.
o otherwIse.
For the rest of this section, we shall use the notation q) R to clenote the
fuzzy relation P defined above.
18 l. FUZZY SUBSETS

Definition 1.7 A cover C on a set S is a family of subsets Si,i E I, of S

such that UiEI Si = S, where I is a nonempty index set.

Definition 1.8 Let R be a fuzzy relation from S into T. We define the

following, where E E [0, 1]:

(i) R is E-determinate if for each XES, there exists at most one yET
such that R(x,y) 2: E.

(ii) R is E-productive if for each XES, there exists at least one yET
such that R(x, y) 2: E.

(iii) R is an E-function if it is both E-determinate and E-productive.

Lemma 1.8 If R is an E-reflexive fuzzy relation on S, then ¢R is E-produc

tive and for each E' :S E, F{!- is a cover of S . •

In the sequel, we use the term productive (determinate, reflexive, func-

tion) for I-productive (I-determinate, I-reflexive, I-function).

Corollary 1.9 If R is reflexive, then ¢R is productive and each FER (0 <

E:S 1) is a cover of S . •

The following result is a consequence of Theorem 1.7 and Corollary 1.9.

Corollary 1.10 R is reflexive and symmetric relation on S if and only if

there is a set T and a productive fuzzy relation P from S into T such that
R= PoP-l .•
Lemma 1.11 Let R be a weakly reflexive, symmetric and transitive fuzzy
relation on S, and let ¢'R denote the relation ¢R whose range is restricted to
FER. That is, ¢'R equals CPR on S x {K*IK EFf} . Then for each 0 < E :S 1,
¢'R is E-determinate and the elements of FER are pairwise disjoint . •

Definition 1.9 A similarity relation R on S is a fuzzy relation on S which

is reflexive, symmetric and transitive. R is called an E-similarity relation if
it is E-reflexive for some 0 < E :S 1, symmetric, and transitive.

A similarity relation on S is merely a fuzzy equivalence relation on S.

Since clearly reflexivity implies weak reflexivity, we have the following
consequence of Lemmas 1.8 and 1.11.

Corollary 1.12 If R is a similarity relation on S, then for each 0 < E:S 1,

FER is a partition of S . •
 1.5 Advanced Topics on Fuzzy Relations 19

Note that Corollary 1.12 says that every similarity relation R can be
represented as U aRc", where Ra is the equivalence relation induced by the
a
partition Fl;. Indeed, it was pointed out by Zadeh [11 J that if the Ra, 0 <
a ::::: 1, are a nested sequence of distinct equivalence relations on S, with al
> a2 if and only if Ral S R a2 , RaJ,. is none~npty and the domain of Ral is
equal to the domain of R a2 , then R =U aRa is a similarity relation on S,
a
where
aRa (x if (x, y ~ E Ra
) = {a
,y 0 otherwIse.
The following result, which is a straightforward consequence of Theo-
rem 1. 7 and Corollary 1.12, yields another characterization of a similarity
relation.

Theorem 1.13 A relation R is an E-similarity (0 < E ::::: 1) relation on a

set S if and only if there is another set T and an E-function P from S into
T such that R = P 0 p-l .•

Example 1.9 Let R be the fuzzy relation on S = {a, b, c, d, e, J} given by

the following matrix, M k

a b c d e f
a 1.0 0.5 0.5 0.2 0.2 0.2
b 0.5 1.0 0.5 0.2 0.2 0.2
c 0.5 0.5 1.0 0.2 0.2 0.2
d 0.2 0.2 0.2 1.0 0·4 0·4
e 0.2 0.2 0.2 0·4 1.0 0·4
f 0.2 0.2 0.2 0·4 0·4 1.0

Now M i/ = M k Thus R is transitive. Clearly, R is reflexive and sym-

metric. We have
Ff = {{a},{b},{c},{d},{e},{f}}
FJ?5 = {{a,b,c},{d},{e},{f}}
FJ?4 = {{a,b,c},{d,e,J}}
FJ?2= {X}.
Let E = 0.4. Then the function P : X x {K*IK EFJ?4} ----+ [0,1],
E -
is defined as follows: P(0,{a,b,c}*) = P(b,{a,b,c}*) = P(c,{a,b,c}*) =
0.5, P( d, {d, e, J} *) = P( e, {d, e, J} *) = P(j, {d, e, J} *) = 0.4, and
p(x, K*) = 0 otherwise.
20 1. FUZZY SUBSETS

1.6 References
1. Dubois, D. and Prade, H., Fuzzy Sets and Systems: Theory and Appli-
cations, Mathematics in Science and Engineering, Vol. 144, Academic
Press, Inc., Orlando, Florida, 1980.

2. Kandel, A., Fuzzy Mathematical Techniques with Applications, Addi-

son-Wesley Pub. Co. 1986.

3. Kaufmann, A., Introduction to the Theory of Fuzzy Sets, Vol. 1, Aca-

demic Press, Inc., Orlando, Florida, 1973.

4. Klir, G.J., U. St. Clair, U.H., and Yuan, B., Fuzzy Set Theory, Foun-
dations and Applications, Prentice Hall, Upper Saddle River, N.J.,
1997.
5. Klir, G.J. and Folger, T.A., Fuzzy Sets, Uncertainty and Information,
Prentice Hall, Englewood Cliffs, N.J., 1988.

6. Klir, G.J. and Yuan, B., Fuzzy Sets and Fuzzy Logic: Theory and
Applications, Prentice Hall, Upper Saddle River, N.J., 1995.

7. Rosenfeld, A., Fuzzy graphs. In: L. A. Zadeh, K S. Fu and M.

Shimura, Eds., Fuzzy Sets and Their Applications, Academic Press,
New York, 77-95, 1975.

8. Tamura, S., Higuchi, S., and Tanaka, K, Pattern Classification Based

on Fuzzy Relations, IEEE Trans. SMC-l, 61--66, 1971.

9. Yeh, R. T. and Bang, S.Y., Fuzzy graphs, fuzzy relations, and their
applications to cluster analysis. In: L. A. Zadeh, K S. Fu and M.
Shimura, Eds., Fuzzy Sets and Their Applications, Academic Press,
New York, 125-149, 1975.

10. Zadeh, L. A., Fuzzy sets as a basis for a theory of possibility, Fuzzy
Sets and Systems, 1:3-28, 1978.

11. Zadeh, L.A., Similarity relations and fuzzy orderings, Information

Sci., 3:177-200, 1971.

12. Zimmermann, H.J., Fuzzy Set Theory and Its Applications, Second
Edition, Kluwer Academic Publishers, Boston, Dordrecht/London,
1991.
2
FUZZY GRAPHS

Any relation R ~ S x S on a set S can be regarded as defining a graph

with vertex set S and edge set R. That is, a graph is a pair (S, R), where S
is a set and R is a relation on S. Similarly, any fuzzy relation R on a fuzzy
subset A of a set S can be regarded as defining a weighted graph, or fuzzy
graph, where the edge (x, y) E S x S has weight or strength R(x, y) E [0,1].
In this chapter, we shall use graph terminology and introduce fuzzy analogs
of several basic graph-theoretical concepts. For simplicity, we will consider
only undirected graphs through out this chapter, except in Section 2.3.
Therefore, all fuzzy relations are symmetric and all edges are regarded as
unordered pairs of vertices, except for Section 2.3. We abuse notation by
writing (x, y) for an edge in an undirected graph (S, R), where x, yES.
(We need not consider loops, that is, edges of the form (x, x); we can
assume, if we wish, that our fuzzy relation is reflexive.) Formally, a fuzzy
graph G = (S, A, R) is a nonempty set S together with a pair of functions
A: S ---> [0,1] and R : S x S ---> [0,1] such that for all x, y in S, R(x, y) ::;
A(x) f\ A(y). We call A the fuzzy vertex set of G and R the fuzzy edge set
of G, respectively. Note that R is a fuzzy relation on A. We will assume
from now on that the underlying set is S and that it is finite. Therefore,
for the sake of notational convenience, we may omit S for the rest of our
discussion and use the abbreviated notation G = (A, R). Thus in the most
general case, both vertices and edges have membership value. However, in
the special case where A(x) = 1, for all XES, edges alone have fuzzy
membership. So, in this case, we use the abbreviated notation G = (S, R).
The fuzzy graph H = (B, T) is called a partial fuzzy subgraph of G = (A, R)
if B ~ A and T ~ k Similarly, the fuzzy graph H = (P, B, T) is called

J. N. Mordeson et al., Fuzzy Mathematics

© Springer-Verlag Berlin Heidelberg 2001
22 2. FUZZY GRAPHS

a fuzzy subgraph of G = (S,A,R) induced by P if P ~ S,E(x) = A(x)

for all x E P and T(x,y) = R(x,y) for all x,y E P. Whenever there is no
confusion, for the sake of simplicity, we call H a fuzzy subgraph of G. It is
worth noticing that a fuzzy subgraph (P, 13, T) of a fuzzy graph (S, A, R)
is in fact a special case of a partial fuzzy subgraph obtained as follows.

° ifxES\P
{ A(x) ifxEP
E(x)

° y) S
{ R(x,y) (x,y) if E P xP
T(x,y)
if (x, E x S \ P x P

Thus a fuzzy graph can have only one fuzzy subgraph corresponding
to a given subset P of S. Hence we shall use the notation (PI to denote
the fuzzy subgraph of G induced by P. For any threshold t,
At = {x E SIA(x) 2: t} and Rt = {(x,y) E S x SIR(x,y) 2: t}. Since
°: ;
t ::; 1,

Rix,7j)::; A(x) I\A(Y) for all X,y E ~ we have Rt r::_At x At, so that
(At, Rt) is a graph with the vertex set At and edge set Rt for all t E [0,1] .
Proposition 2.1 Let G = (A, R) be a fuzzy graph. If
(At,Rt) is a subgraph of (AU, RU) . •
°: ; u ::; t ::; 1, then

Proposition 2.2 Let H = (13, T) be a partial fuzzy subgraph of G =

(A,R). For any thresholdt,O::; t::; 1, (Et,Tt) is a subgraph of (At, Rt) . •

We say that the partial fuzzy subgraph (13, T) spans the fuzzy graph
(A, R) if A = E. In this case, the two graphs have the same fuzzy vertex set;
they differ only in the edge weights. For any fuzzy subset 13 of S such that
13 ~ A, the partial fuzzy subgraph of (A, R) induced by 13 is the maximal
partial fuzzy subgraph of (A, R) that has fuzzy vertex set E. Evidently, this
is just the partial fuzzy graph (13, T), where T(x, y) = E(x) I\E(y) I\R(x, y)
for all x, YES.

2.1 Paths and Connectedness

Let G = (V, X) be a graph. A walk of G is an alternating sequence of
vertices and edges vo, xl, VI, ... , Vn-l, x n , Vn , where vo, Vi E V, Xi E X, Xi =
(Vi-I, Vi), i = 1, ... , n. A walk is sometimes denoted by VOVI ... Vn , where the
edges are evident by context. A path is a walk with all vertices distinct. Let
VOVI." Vn be a path. If n 2: 2 and Vo = Vn , then the path is called a cycle.
G = (V, X) is said to be complete if (u, v) EX \fu, V E V, u -# v. A clique
of a graph is a maximal complete subgraph.
A path p in a fuzzy graph (A, R) is a sequence of distinct vertices
XO,XI, ... ,Xn such that R(Xi-I,Xi) > 0,1::; i::; n. Here n 2: 1 is called the
 2.1 Paths and Connectedness 23

length of the path p. The consecutive pairs (:ri-1,:ri) are called the edges
ofthe path. The strength of p is defined as /'<'=1 R(.ri-1, Xi), In other words,
the strength of a path is defined to be the weight of the weakest edge of the
path. A single vertex x may also be considered as a path. In this case, the
path is of length O. If the path has length 0, it is convenient to define its
strength to be A(xo). It may be noted that any path of length n > 0 can as
well be defined as a sequence of edges (Xi-I, Xi), 1 ~ i ~ n, satisfying the
condition R(Xi-1,Xi) > 0 for 1 ~ i ~ n. A partial fuzzy suhgraph (A,R)
is said to be connected if \fx, y E supp(A), RDO(X, y) > O.
\Ve call p a cycle if Xo = Xn and n 2 2. Two vertices that are joined
by a path arc said to be connected. It is evident that "connected" is an
equivalence relation. In fact, X and yare connected if and only if RDO(X, y) >
O. The equivalence classes of vertices under this relation are called connected
components of the given fuzzy graph. They arc just its maximal connected
partial fuzzy subgraphs. A strongest path joining any two vertices .c, y
has strength ROO (:r, y). We shall sometimes refer to this as the strength of
connectedness between the vertices.

Proposition 2.3 fI(B, T) is a paTtialjuzzy subgraph of (A, R), then Too ~

ROO .•
Let (A, R) be a fuzzy graph. We now provide two popular ways of defining
the distance between a pair of vertices. One way is to define the "distance"
dis(:r, y) between x and y as the length of the shortest strongest path
between them. This "distance" is symmetric and is such that dis(x, x) = 0
since by our definit.ion of a fuzzy graph, no path from x to x can have
strength greater than A( x), which is the strength of the path of length
O. Hovvever, it docs not satisfy the triangle property, as we see from the
following example.

1.0
0.5

Y 1.0

V
0.5
1.0

Here any path from x to y or from y to z has strength ~ 1/2 since it

must involve either edge (x,y) or edge (y,z). Thus the shortest strongest
paths between them have length 1. On the other hand, there is a path
from .1' to z. through u and v, that has length 3 and strength 1. Thus
dis(x, z) = 3 > 1 + 1 = dis(x, y) + dis(y, z) in this case.
24 2. FUZZY GRAPHS

A better notion of distance in a fuzzy graph can be defined as follows:

For any path p = Xo, ... ,xn , define the R-length of p as the sum of the
reciprocals of p's edge weights, that is,

If n = 0, we define l(p) = O. Clearly, for n 2: 1, we have l(p) 2: 1. For

any two vertices x, y, we can now define their R-distance 8(x, y) as the
smallest R-Iength of any path from x to y. Thus 8(x,y) = !\{l(p)lp is a
path between x and y} if x and y are connected. We define 8(x, y) = 00 if
x and yare not connected.

Proposition 2.4 8 is a metric on S. That is, 't/x, y, z E S,

(i) 8(x, y) = 0 if and only ~f x = y,

(ii) 8(x, y) = 8(y, x),

(iii) 8(x, z) ::; 8(x, y) + 8(y, z) .•

In the non-fuzzy case, l (p) is just the length n of p since all the ks are 1.
Hence, 8(x, y) becomes the usual definition of distance, that is, the length
of the shortest path between x and y.

2.2 Clusters
In graph theory, there are several ways of defining "clusters" of vertices.
One approach is to call a subset C of S a cluster of order k if the following
two conditions hold:

(a) for all vertices X,y in C, d(x,y)::; k;

(b) for all vertices z ¢: C, d(z,w) > k for some w E C;

where d( u, v) is the length of a shortest path between any two vertices u

and v.
Thus in a k-cluster C, every pair of vertices are within distance k of each
other, and C is maximal with respect to this property. That is, no vertex
outside C is within distance k of every vertex in C.
A I-cluster is called a clique; it is a maximal complete subgraph. That
is, a maximal subgraph in which each pair of vertices is joined by an edge.
At the other extreme, if we let k ---t 00, a k-cluster becomes a connected
component, that is, a maximal subgraph in which each pair of vertices is
joined by a path (of any length).
 2.2 Clusters 25

These ideas can be generalized to fuzzy graphs as follows: In G = (A, R),

we can call C <;;; S a fuzzy cluster of order k if

/\{Rk(X,y) I x,y E C} > V{/\{Rk(w,z) I w E C} I z tt C}.

Note that C is an ordinary subset of S, not a fuzzy subset. If G is an
ordinary graph, we have Rk(a, b) = 0 or 1 for all a and b. Hence this
definition reduces to

(a) Rk(x, y) = 1 for all x, yin C,

(b) Rk(w, z) = 0 for all z tt C and some w E C.

Property (a) implies that for all x, yin C, there exists a path of length
:S k between x and y and property (b) implies that for all z tt C and some
w E C, there does not exist a path of length :S k. This is the same as the
definition of a cluster of order k.
In fact, the k-clusters obtained using this definition are just ordinary
cliques in graphs obtained by thresholding the kth power of the given fuzzy
graph. Indeed, let C be a fuzzy k-cluster, and let /\ {Rk (x, y) I x, Y E C} = t.
If we threshold Rk (and A) at t, we obtain an ordinary graph in which C
is now an ordinary clique.
Example 2.1 Let
v = {x,y,z,u,v}
and
x = {(x,y),(x,z),(y,z),(z,u,),(u,v)}.
I:,et A(x) = A(Y2 = A(z) =_ A(u) = A(v) = 1 and R(x, y) = R(x, z)
R(y, z) = 1/2, R(z, u,) = R(u, v) = 1/4. Let C = {x, y, z}. Then /\
c,dEC
R k (c,d)=1/2fork=1,2, ... , V (/\ R 1 (c,e)) = (1/4/\0/\0)V(0/\0/\
e~C cEC
0) = 0, V (/\ R2(c, e)) = (1/4/\ 1/4/\ 1/4) V (1/4/\ 0 /\ 0) = 1/4, and
e~C cEC
V (/\ Rk(c,e)) = (1/4/\1/4/\1/4) V (1/4/\ 1/4/\1/4) =1/4fork2:3.
e~C cEC
Hence C = {x, y, z} is a fuzzy cluster of order k for all k 2: 1.
Now let R(x, y) = R(x, z) = R(y, z) = 1/8, R(z, u) = R(u, v) = 1/4.
Then /\ R k (c,d)=1/8fork=1,2, .... V (/\ Rl(c,e)) =(1/4/\0/\
c,dEC e~C cEC
O)V(O/\O/\O) = 0, V (/\
e~C cEC
R 2 (c,e)) = (1/4/\1/8/\1/8)V(1/4/\0/\0) = 1/8,
and V (/\ Rk(c,e)) = (1/4/\ 1/8/\ 1/8) V (1/4/\ 1/8/\ 1/8) = 1/8 for
e~C cEC
k 2: 3. Hence C is a fuzzy cluster of order 1, but not of order k for k 2: 2.

Bridges and Cut Vertices

Let G = (A, R) be a fuzzy graph, let x, y be any two distinct vertices,
and let G' be the partial fuzzy subgraph of G obtained by deleting the
26 2. FUZZY GRAPHS

edge (x, y). That is, G' = (A, k), where k(x, y) = 0 and k = R for all
other pairs. We say that (x,y) is a bridge in G if kCXJ(u,v) < RCXJ(u,v) for
some u, v. In other words, if deleting the edge (x, y) reduces the strength of
connectedness between some pair of vertices. Thus, (x, y) is a bridge if and
only if there exist vertices u, v such that (x, y) is an edge of every strongest
path from u to v.

Theorem 2.5 The following statements are equivalent:

(i) (x, y) is a bridge;

(ii) kCXJ(x,y) < R(x,y);

(iii) (x,y) is not the weakest edge of any cycle.

Proof. (ii) =} (i): If (x,y) is not a bridge, we must have kCXJ(x,y)

RCXJ(x, y) ~ R(x, y).
(i)=}(iii): If (x, y) is a weakest edge of a cycle, then any path involving
edge (x, y) can be converted into a path not involving (x, y) but at least
as strong, by using the rest of the cycle as a path from x to y. Thus (x, y)
cannot be a bridge.
(iii)=}(ii): If kCXJ(x, y) ~ R(x, y), there is a path from x to y, not involv-
ing (x, y), that has strength ~ R( x, y), and this path together with (x, y)
forms a cycle of which (x, y) is a weakest edge . •

Let w be any vertex and let G* be the partial fuzzy subgraph of G

obtained by deleting the vertex w. That is, G* is the partial fuzzy subgraph
induced by A *, where A * (w) = 0; A * = A for all other vertices.
Note that in G* = (A*, R*), we must have R*(w, z) = 0 for all z. We
say that w is a cutvertex in G if R*CXJ (u, v) < RCXJ (u, v) for some u, v (other
than w). In other words, if deleting the vertex w reduces the strength
of connectedness between some other pair of vertices. Evidently, w is a
cutvertex if and only if there exist u, v, distinct from w such that w is on
every strongest path from u to v. G* is called nonseparable (or sometimes:
a block) if it has no cut vertices. It should be pointed out that a block may
have bridges (this cannot happen for non-fuzzy graphs). For example, in
Figure 2.1 edge (x, y) is a bridge since its deletion reduces the strength of
connectedness between x and y from 1 to 0.5. However, it is easily verified
that no vertex of this fuzzy graph is a cutvertex.
If between every two vertices x, y of G there exist two strongest paths
that are disjoint (except for x, y themselves), G is evidently a block. This
is analogous to the "if" of the non-fuzzy graph theorem that G is a block
(with at least three vertices) if and only if every two vertices of G lie on
a common cycle. The "only if", on the other hand, does not hold in the
fuzzy case, as the example just given shows.
 2.2 Clusters 27

FIGURE 2.1 Fuzzy graph with a bridge; but no cut vertices.

1.0

0.5

0.5

Forests and Trees

We recall that a graph that has no cycles is called acyclic, or a forest; and
a connected forest is called a tree. We shall call a fuzzy graph a forest if the
graph consisting of its nonzero edges is a forest, and a tree if this graph is
also connected. More generally, we call the fuzzy graph G ~ (A, R) a fuzzy
forest if it has a partial fuzzy spanning subgraph F = (A, T) which is a
forest, where for all edges (x, y) not in F (i.e., such that T(x, y) = 0), we
have R(x,y) < T=(x,y). In other words, if (x,y) is in G but (x,y) is not
in F, there is a path in F between x and y whose strength is greater than
R(x, y). It is clear that a forest is a fuzzy forest.
The fuzzy graphs in Figure 2.2 are fuzzy forests and the fuzzy graphs in
Figure 2.3 are not fuzzy forests.

FIGURE 2.2 Fuzzy forests.

x 1.0 Y
X,~l
V :~25
0.5
~--+-----~~----~~w

0.5
1.0
1.0

1.0

If G is connected, then so is F since any edge of a path in G is either in

F, or can be diverted through F. In this case, we call G a fuzzy tree. The
examples offuzzy forests given above are all fuzzy trees. Note that if we re-
placed < by :S in the definition, then even the fuzzy graph (8, A, R), where
8 = {x,y,z}, A(x) = A(Y) = A(z) = 1,R(x,y) = R(x,z) = R(y,z) = 1,
would be a fuzzy forest since it has partial fuzzy spanning subgraphs such
as (8,A,k), where k(x,y) = k(x,z) = 1 and R(y,z) = o.
28 2. FUZZY GRAPHS

FIGURE 2.3 Fuzzy graphs; but not fuzzy forests.

0.50 510 [~105

x 1.0 Y X 1.0 Y
1.0
0.5

w 1.0 w 0.5 z 1.0

0.5 1.0
w

Theorem 2.6 G is a fuzzy forest if and only if in any cycle of G, there

is an edge (x, y) such that R(x, y) < k= (x, y), where G' = (ii, k) is the
partial fuzzy subgraph obtained by the deletion of the edge (x, y) from G.

Proof. Suppose (x, y) is an edge, belonging to a cycle, which has the

property of the theorem and for which R(x, y) is smallest. (If there are no
cycles, G is a forest and we are done.) If we delete (x, y), the resulting partial
fuzzy subgraph satisfies the path property of a fuzzy forest. If there are still
cycles in this graph, we can repeat the process. Note that at each stage, no
previously deleted edge is stronger than the edge being currently deleted;
hence the path guaranteed by the property of the theorem involves only
edges that have not yet been deleted. When no cycles remain, the resulting
partial fuzzy sub graph is a forest F. Let (x, y) not be an edge of F; thus
(x, y) is one of the edges that we deleted in the process of constructing
F, and there is a path from x to y that is stronger than R(x, y) and that
does not involve (x, y) nor any of the edges deleted prior to it. If this path
involves edges that were deleted later, it can be diverted around them using
a path of still stronger edges; if any of these were deleted later, the path
can be further diverted; and so on. This process eventually stabilizes with
a path consisting entirely of edges of F. Thus G is a fuzzy forest.
Conversely, if G is a fuzzy forest and p is any cycle, then some edge (x, y)
of p is not in F. Thus by definition of a fuzzy forest we have R(x, y) <
T= (x, y) :s: k= (x, y) . •
Note that if G is connected, so is the constructed F in the first part of
the proof since no step of the construction disconnects.

Proposition 2.7 If there is at most one strongest path between any two
vertices of G, then G must be a fuzzy forest.

Proof. Suppose G is not a fuzzy forest. Then by Theorem 2.6, there is a

cycle pin G such that R(x,y) 2: k(x,y) for all edges (x,y) of p. Thus
(x, y) is a strongest path from x to y. If we choose (x, y) to be a weakest
edge of p, it follows that the rest of the p is also a strongest path from x
to y, a contradiction . •
 2.3 Cluster Analysis and Modeling of Information Networks 29

The converse of Proposition 2.7 is false; G can be a fuzzy forest and

still have multiple strongest paths between vertices. This is because the
strength of a path is that of its weakest edge, and as long as this edge lies
in F, there is little constraint on the other edges. For example, the fuzzy
graph in Figure 2.4 is a fuzzy forest. Here F consists of all edges except
(u, y). The strongest paths between x and y have strength 1/4, due to the
edge (x, u); both x, u, v, y and x, u, yare such paths, where the former lies
in F but the latter does not.
Proposition 2.8 If G is a fuzzy forest, then the edges of F are just the
bridges of G.

Proof. An edge (x, y) not in F cannot be a bridge since R( x, y) < Too (x, y)
::; R'oo (x, y). Suppose that (x, y) is an edge in F. If it were not a bridge, we
would have a path P from x to y, not involving (x, y), of strength 2: R( x, y).
This path must involve edges not in F since F is a forest and has no cycles.
However, by definition, any such edge (Ui' Vi) can be replaced by a path Pi
in F of strength> R( u, v). Now Pi cannot involve (x, y) since all its edges
are strictly stronger than R( u, v) 2: R( x, y). Thus by replacing each (Ui' Vi)
by Pi' we can construct a path in F from x to y that does not involve (x, y),
giving us a cycle in F, contradiction . •

FIGURE 2.4 A fuzzy forest with no multiple strongest paths between

vertices.

0.25 0.5

1.0

1.0

2.3 Cluster Analysis and Modeling of Information

Networks
In this section, fuzzy graphs will be analyzed from the connectedness view-
point. The results will be applied to cluster analysis and modeling of in-
formation networks. We do not assume our (fuzzy) graphs are necessarily
undirected in this section.
Let G = (A, R) be a fuzzy graph. We denote by 1\11R the corresponding
fuzzy matrix of a fuzzy graph G. In other words, (MR)ij = R(Vi,Vj).
30 2. FUZZY GRAPHS

Theorem 2.9 Let G = (B, A, R.) be a fuzzy graph such that cardinality of
B is n. Then

(i) ~f R is refiexive, there exists k < n such that IV!R < I'vI'A < ... <
M~ =I'vI~+l.
R3 R'"

(ii) if R is irrefiexive, the sequence M A, AIk, ... is eventually periodic . •

Definition 2.1 Let G be a fuzzy graph. Let 0 <::: f <::: 1. A vertex v is said
to be E-reachable fTOm another vertex u if there exists a positive integer
k such that Rk(u,v) 2' E. The reachability matrix ofG. denoted by Mk=,
-is the matrix of the fuzzy graph (A, R=). The E-reachability matrix of G,
k k
denoted by I'vI =, is de.fi>ned as follows: M =(u, v) = lif R( u, v) 2' E and
A
M = (u, v) = 0, otherwise.

The following algorithm can determine the reachability between any pair
of vertices in a fuzzy graph G.
Algorithm 2.1. Determination of M ROO
1. Let Ri = (ail, ... , ain) denote the ith row.
2. Obtain the new Hi by the following procedure:

3. Repeat Step 2 until no further changes occur.

4. I'vIA=(i,j) = aij(new).
Note that a similar algorithm can be constructed for the determination
k
of M =, 0 <::: E <::: 1.

Definition 2.2 Let G be a fuzzy graph. The connectivity of a pair of ver-

tices 11 and v, denoted by C( 11, v) is defined tQ be R=( 11, v) 1\ R=( v, 11). The
connectivity matrix ofG, denoted by CG. is defined s11ch that CG(11,V) =
CCLL, v). For 0 <::: E <::: 1, the E-connectivity matrix of G, denoted by Co, is
defined as follows: Co (11, v) = 1 if C( u, v) 2' E and CO( 11. v) = 0 otherwise.

Algorithm 2.2. Determination of C G .

1. Construct tvIk= .
2. Cc;(i,j) = Cc;(j,i) = JlIR=(i,j) 1\ MR=(j.i).
An algorithm for determining C(] is similar to Algorithm 2.2.

Definition 2.3 Let G be a f11zzy graph. G is called strongly E-connected

if every paiT of vertices are mutually E-reachable. G is said to be initial E-
connected if there exists v E V such that every vertex u in G is E-reachable
from v. A maximal strongly E-connected fuzzy subgraph (AI BEGB) of G is
a strongly E-connected fuzzy subqraph not properly contained in any other
MBECB.
 2.3 Cluster Analysis and Modeling of Information Networks 31

TABLE 2.1 Fuzzy matrix and connectivity matrix of a fuzzy graph.

1.0 0.6 0.4 0.0 0.0 1 0 0 0 0
0.0 1.0 0.2 0.6 0.3 0 1 0 1 0
MR = 0.0 0.8 1.0 0.0 0.9 CO.5 -
G - 0 0 1 0 1
0.2 0.7 0.3 1.0 0.2 0 1 0 1 0
0.4 0.0 0.5 0.3 1.0 0 0 1 0 1

Clearly strongly f-connectedness implies initial f-connectedness. Also,

the following result is straightforward.

Theorem 2.10 A fuzzy graph G is strongly f-connected if and only if there

exists a vertex u such that for any other vertex v in G, R= (u, v) 2: f and
R=(v, u) 2: f .•

Algorithm 2.3. Determination of all MSfCS in G.

1. Construct C'G.
2. The number of MSfCS in G is given by the number of distinct row
vectors in C'G. For each row vector a in C'G, the vertices contained in
the corresponding !VIS fC S are the nonzero elements of the corresponding
columns of a.

Example 2.2 Let G be a fuzzy graph whose corresponding fuzzy matrices

MR and cg 5 is given in Table 2.1. We see that the MSO.5CS's of G
contain the following vertex sets, {1}, {2, 4}, {3, 5}, respectively.

The previous result is now applied to clustering analysis. We assume that

a data fuzzy graph G = (V, R) is given, where V is a set of data and R( u, v)
is a quantitative measure of the similarity of the two data items u and v.
For 0 < f ~ 1, an f-cluster in V is a maximal subset W of V such that each
pair of elements in W is mutually f-reachable. Therefore, the construction
of f-clusters of V is equivalent to finding all maximal strongly f-connected
fuzzy subgraphs of G.
Algorithm 2.4. Construction of f-clusters
1. Compute R, R2, ... , Rk. where k is the smallest integer such that Rk =
Rk+1.,
_ k_.
2. Let S = U R'.
i=l
3. Construct F,B.
Then, each element in F,B is an f-cluster.
We may also define an f-cluster in V as a maximal subset W of V such
that every element of W is f-reachable from a special element v in W. In
this case, the construction of f-clusters is equivalent to finding all maximal
initial f-connected fuzzy subgraphs of G. However, the relation induced by
initial f-connected fuzzy subgraphs is not, in general, a similarity relation.
32 2. FUZZY GRAPHS

Another application is the use of fuzzy graphs to model information

networks. Such a model was proposed in [34] utilizing the concepts of a
directed graph. In [34] a measure of flexibility of a network was introduced.
More specifically, let N be a network with m edges and n vertices. Then
the measure of flexibility on N, denoted by Z(N), is defined as follows:

Z(N) _ m-n
- n(n - 2)

The equation above is quite useful in classifying certain graph structures

related to information networks. However, it is insensitive to certain classes
of graphs. It seems that the use of fuzzy graphs is a more desirable model for
information networks. The weights in each edge could be used as parameters
such as number of channels between stations, costs for sending messages,
etc. Thus, we propose here the use of a fuzzy graph to model an information
network. Let N have n vertices. Define two measures of N : flexibility and
balancedness, denoted by Z(N) and B(N) respectively, as follows:

These two measures are much more sensitive to the structure of graphs
as the one given in [34].

2.4 Connectivity in Fuzzy Graphs

In this section, connectivity of fuzzy graphs will be investigated. In this
section and in fact for the remainder of the chapter, we assume all graphs
are undirected. Let G = (V, R) be a fuzzy graph. Define the degree of
a vertex v to be d(v) = L R(v,u). The minimum degree of G is 8(G) =
uf.v
/\{d(v) I v E V}, and the maximum degree of G is ~(G) = V{d(v) I v E V}.

Definition 2.4 Let Gi = (Vi, ~), i = 1,2 be two fuzzy subgraphs of G =

(V, R2. The union of G l and G2 , denoted by G l U G2 , is the fuzzy graph
(V', R' ), where V' = Vl U V2 and

k(u, v) = { R(~, v) if {u, v} s;:: Vl U V2

if {u, v} r.J;. Vl U V2

Lemma 2.11 Let G = (V, R) be a fuzzy graph and Gi (Vi, ~), i

1, ... , n, be fuzzy subgraphs of G such that Vi n Vi = 0 for i i= j, 1 :S i, j :S n,
n
and U Gi is connected. Then
i=l
 2.4 Connectivity in Fuzzy Graphs 33

n
(i) 8( U Gi ) 2 1\{8(Gi ) Ii = 1,,,. ,n},
i=I

n
(ii) ~(U Gi ) 2 V{8(G i ) Ii = 1, ... ,n} .•
i=I

Recall that G is said to be connected if for each pair of vertices u and v

in V, there exists a k > 0 such that Rk (u, v) > O.

Definition 2.5 Let G = (V, R) be a fuzzy graph. G is called T-degree

connected, for some T 2 0, ~f 8( G) 2 T and G is connected. AT-degree
component of G is a maximal T-degree connected fuzzy subgraph of G.

Theorem 2.12 For any T > 0, the T-degree components of a fuzzy graph
are disjoint.

Proof. Let G I and G 2 be two T-degree components of G such that their

vertex sets have at least one common element. Since 8( G I U G2 ) 2 8( Gd 1\
8(G 2 ) by Lemma 2.11, G I uG 2 is T-degree connected. Since G I and G 2 are
maximal with respect to T-degree connectedness, we have that G I = G 2 . •

Algorithm 2.5. Determination of T-degree components of a finite fuzzy

graph G.
1. Calculate the row sums of NIH-
2. If there are rows whose sums are less than T, then obtain a new reduced
matrix by eliminating those vertices, and go to 1.
3. If there is no such row, then stop.
4. Each disjoint fuzzy subgraph of the graph induced by the vertices in
the last matrix as well as each eliminated vertex is a maximal T-degree
connected fuzzy subgraph.

Definition 2.6 Let G be a fuzzy graph, and {VI, V2 } be a partition of its

vertex set V. The set of edges joining vertices of VI and vertices of V2 is
called a cut-set of G, denoted by (VI, V2), relative to the partition {VI, V2}'
The weight of the cut-set (VI, V2 ) is defined to be

L R(u,v).
uEV" vEV2

Definition 2.7 Let G be a fuzzy graph. The edge connectivity of G, de-

noted by '>-(G), is defined to be the minimum weight of cut-sets of G. G is
called T-edge connected if G is connected and '>-(G) 2 T. A T-edge compo-
nent of G is a maximal T-edge connected subgraph of G.

Example 2.3 Consider the fuzzy graph G given below.

Other documents randomly have
different content
regard, une fin de regard : immobilisé vers le plafond, attentif à des
choses, il méditait infatigablement, et plusieurs fois, dans mon
délire, j'eus l'impression que cet œil fixe travaillait à recueillir dans
l'espace toutes les pensées de mon mutisme : Berthe écoutait par lui
les paroles que ma voix était incapable de proférer, et que mon âme
jetait à la sienne.
— N'est-ce pas, chérie, tu m'entends?
J'ai demandé cela, à un moment ; je me rappelle très bien avoir
demandé cela. Mais l'œil ne m'a pas répondu, et j'ai compris qu'il
m'entendait, mais qu'il ne daignait pas répondre.
D'abord, je me suis résigné, comme un enfant ; puis, j'ai
recommencé et j'ai supplié. L'œil immuable déclarait : « Il a tué une
créature vivante, et, maintenant, il l'implore. »
— Berthe…
— Je ne veux pas répondre.
— Berthe! Berthe!
— Je ne peux pas répondre. Je ne bouge plus. C'est ton œuvre.
Je me suis mis à regarder le plafond, moi aussi, cherchant
l'endroit que Berthe fixait si âprement, et je le cherchais avec
obstination, convaincu d'y lire sa pensée, comme si l'œil de la morte
eût écrit au plafond les choses qu'elle avait à me dire. Et je les ai
lues, les réponses de Berthe : c'était des paroles tranquilles et
nettes. Elle disait : « Tu m'as tuée. C'est fait. Laisse-moi. »
J'ai voulu crier : « Pardon! »
Mais elle déclara : « Tu as fait la chose irréparable. Il ne sert à
rien de demander pardon. Ton remords ne me ressuscitera jamais.
— Je t'aimais tant!
— L'amour n'est pas une excuse au crime de tuer.
— J'étais jaloux!
— Une vie n'appartient qu'à elle-même! L'épouse n'est pas le
meuble de l'époux, un bibelot qu'il peut casser à sa guise. Je vivais :
chacun est le seul maître de sa vie.
 — J'ai tué parce que tu m'as trompé.
— Chacun est le seul maître de son corps. J'avais le droit de
préférer un autre amour ; et tu n'avais pas le droit de me tuer.
— Oui, Berthe, ta faute fut légère, si elle existe ; la mienne fut
atroce. Je le sais maintenant.
— Trop tard.
— Pardonne-moi!
— Laisse-moi. »
A partir de cet instant, le regard de Berthe n'a plus voulu
répondre. J'ai cru voir qu'il s'endormait. Je fus horriblement seul.
Je dois supposer que mon délire prit alors un caractère plus
proche encore de la folie, car, désormais, tout se brouille dans mon
souvenir. J'y retrouve pourtant un îlot de clarté, et je me souviens de
ceci : par intermittence, je poussais mon cri maniaque, dans l'air
fétide. L'odeur de la chambre avait empiré. Une espèce de buée
opalisait les vitres, et le soleil la diaprait en passant au travers. Les
feuillages du jardin, remués par le vent, secouaient leur ombre sur la
vitre et sur le tapis ; j'observais cette fluctuation de lumières et
d'ombres ; ma tête tournait à les voir ; tout à coup, ce grouillement
prit corps et fut le corps de Berthe, qui houlait, qui m'attirait ; et,
tout à coup, le corps de Berthe fut le mien, étalé sous mes propres
yeux, et je me voyais pourrir.
D'effroi, je poussai un cri strident. La peur de mourir me dressa
sur mon séant. Mais, trop faible, je perdis aussitôt l'équilibre et je
roulai à bas du lit…
Après cela, c'est une nouvelle lacune dans ma mémoire : je ne
sais pas comment s'acheva la journée. J'ai la vague réminiscence
d'être revenu à moi, vers le soir, et je grelottais nu, sur le sol. Je
m'entends geindre. Ensuite, j'ai dû dormir.
Ce sommeil m'a sauvé. Probablement il fut long, car il faisait
grand jour, lorsque je m'éveillai, voyant tout, jugeant tout, épuisé,
mais redevenu un homme.
 Le premier mouvement que mes bras purent exécuter fut de se
tendre vers l'ancienne adorée. A genoux, au bord de sa couche, je
levais vers elle mes mains ressuscitées, mes regards de prière, mon
remords inutile. Ah! comme j'ai pleuré sur le bord de ce lit, et
comme elle est entrée dans moi, à travers mes larmes, l'image de ce
corps qu'il ne faut plus décrire, la vénérable horreur de cette morte
que j'avais faite!
Ah! oui, ce matin-là, je l'ai aimée saintement, l'impassible
victime, et religieusement, d'un bel amour que je n'ai jamais connu
durant ma vie, d'un grand amour expiatoire. Éclairé par la mort et
dégagé de moi, je l'ai chérie pour elle et non plus pour moi-même,
et je l'adorais de tout mon respect, de toute ma douleur, mille fois
mieux qu'au temps de sa beauté!
En cet état d'esprit, une idée fixe s'intronisa en moi : « Je ne
veux pas que nul la voie ainsi ».
Dans le vœu de l'ensevelir, je me traînai sur les genoux. Je
traversai la chambre. Je gagnai la fenêtre ; je pus me hausser, et
l'ouvrir. Tout le printemps entra chez nous, et l'infection s'évada dans
le bleu.
*
* *

Vous savez le reste : un passant qui m'aperçut, debout à la

fenêtre, complètement nu et m'écroulant sur le parquet ; les gens
qui sont venus, et ma convalescence, l'enquête, le jugement…
On a eu tort de m'acquitter. On déclame des inepties! Qu'on
hésite à guillotiner un homme, je le conçois, moi qui ai tué! Je le sais
mieux que personne : nul n'a le droit de punir ; ni le mari, ni le juge,
nul n'a le droit de tuer. Mais, si ce droit-là n'existe pour aucun, quelle
aberration peut inspirer les êtres qui osent, sous couleur de justice,
trouver à l'assassin des circonstances atténuantes? Il n'y a pas
d'excuses au meurtre, quel qu'il soit! Afin de m'épargner, on a
stupidement invoqué la passion, les lois du mariage, l'adultère de
Berthe! Le rouge de la honte m'en montait au visage pendant que
j'écoutais ces bavardages monstrueux! Avocats et jurés, on voit bien
que ceux-là n'ont pas vécu, comme moi, face à face avec un cadavre
qu'ils venaient de faire! Mais, voilà! ces messieurs ont une loi qui
tue : ils n'osent plus l'appliquer, et ils n'osent pas l'abolir. Alors,
lâchement, ils me cherchent des excuses, ce qui fait leur ignominie ;
ils les trouvent, ce qui fait leur crime, et ils se détournent de moi en
se lavant les mains.
Soyez tranquilles, Pilates! Ce que vous n'avez pas le courage de
prescrire, pour l'exemple, je m'en charge, et moi, j'en ai le droit,
n'est-ce pas? Je suis le maître de ma vie. De nulle autre, entendez-
vous? mais je suis le maître de celle-là, et je la jette. Je n'en veux
plus. Bonsoir.
 LE PRIE-DIEU

Le procès-verbal relate :
« Ce mardi 3 février, à quatre heures dix du soir, etc… Nous nous
sommes transporté au cimetière Montmartre, etc… Le préposé nous
ayant conduit à la sépulture Derouville, sise, etc…, monument
formant chapelle muni d'une porte de bronze, pleine dans sa partie
inférieure, ajourée dans le haut par une grille à décor de feuillages
et de rinceaux ; avons fait ouvrir ladite porte et trouvé, à l'intérieur
du monument, le corps d'une jeune femme qui gisait sur le dallage,
parmi des flaques de sang coagulé ; laquelle donnant encore signe
de vie, mais sans connaissance, serrait dans sa main droite une
petite clef d'acier, reconnue pour être celle du caveau, et qui était
maintenue par une chaîne d'or attachée à la ceinture ; les meubles
meublants de la chapelle, éparpillés dans le plus grand désordre,
présentaient les traces d'une lutte : un prie-dieu était renversé, des
vases en porcelaine, ayant contenu des fleurs et décoré l'autel,
étaient en éclats sur le sol ; une marche en marbre blanc qui règne
en avant de l'autel était couverte de boue et brisée en un endroit par
une balle de revolver petit calibre que nous avons retrouvée et
recueillie (pièce jointe)… La blessée, formellement reconnue par le
préposé pour être la dame veuve Léon Derouville, âgée de vingt-
trois ans, a été par nos soins transportée à l'hôpital, etc. Cinq
blessures ont été relevées sur son corps, toutes les cinq provenant
d'une arme à feu, et toutes intéressant le côté droit : deux au
sommet de l'épaule droite, une à la cuisse droite, une à la cheville
du pied droit, une au crâne, formant séton, en arrière de l'oreille
droite. Après un premier examen, les médecins ont déclaré que l'état
de la victime était des plus graves, les blessures remontant à vingt-
quatre heures au moins, et des complications restant à redouter,
autant en raison du retard apporté aux premiers pansements qu'en
raison d'un séjour prolongé dans la température humide et froide du
monument funéraire ; que la dame Derouville, même si l'on parvient
à la sauver, ne sera pas de longtemps en état de subir un
interrogatoire, etc. »
Le sieur N…, gardien attaché au cimetière, dépose :
« … La dame veuve Léon Derouville lui est parfaitement connue ;
depuis le décès de son mari (octobre dernier), elle vient
régulièrement au cimetière, deux fois par semaine, le lundi et le
vendredi, apportant chaque fois des fleurs et procédant elle-même,
avec les plus grands soins, au nettoyage et rangement de la
chapelle ; le lundi 2 février, elle se présenta à son heure ordinaire,
soit une heure et demie après-midi ; la neige tombait en abondance
depuis le matin, et les allées du cimetière étaient absolument
désertes… Se souvient d'avoir salué la visiteuse, et de l'avoir suivie
des yeux, tandis qu'elle s'éloignait dans la neige, où ses pas
laissaient une trace qui fut promptement recouverte ; mais il ne l'a
pas vue sortir, et d'ailleurs n'y a point pris garde, supposant qu'elle
avait repassé à son insu. A constaté, le soir, devant la sépulture
Derouville, des inégalités de la neige, attestant que le sol avait été
piétiné, mais n'en a tiré aucune conclusion inquiétante, puisqu'il
avait noté lui-même la visite de la veuve. Les rondes réglementaires
n'ont amené aucune constatation anormale, pas plus dans la soirée
du lundi que dans la matinée du mardi. La neige ayant continué à
tomber, le mardi matin, les visiteurs furent très rares : cependant,
vers midi environ, un jeune garçon, âgé d'une douzaine d'années,
inconnu, interpella le préposé, dénonçant la présence de
« revenants » dans la région Nord-Ouest, où il disait avoir entendu
des soupirs et des bruits de chaînes ; à quoi le préposé n'a pas pris
garde, croyant à une plaisanterie macabre. Dans l'après-midi du
même jour, une dame âgée, qui sortait du cimetière, entra dans le
bureau, et, très émue, déclara que, dans la même région Nord-
Ouest, « un mort avait été enterré vivant, et qu'il appelait ».
Rapprochant alors cette seconde déclaration de celle qui l'avait
précédée, s'est rendu dans la région indiquée, et là, au cours d'une
ronde, mais seulement après de longues recherches, a effectivement
entendu des soupirs ou des râles ; a découvert enfin que ces bruits
provenaient de la sépulture Derouville, où il a reconnu la présence
d'un être vivant, enfermé dans la chapelle ; est aussitôt revenu au
bureau, pour y prendre ses clefs, et a fait informer le
commissariat… »
La dame veuve Alexis Derouville, mère de feu Léon Derouville et
belle-mère de la victime, habitant avec cette dernière, boulevard
Malesherbes, no…, dépose :
« Sa belle-fille, bien qu'elle soit jeune et extrêmement jolie,
mène l'existence la plus solitaire, ne sortant que pour se rendre sur
la tombe de son mari ; elle a toujours été d'un caractère aimable et
doux, fort timide, qui ne permet pas de supposer qu'elle ait donné
lieu à l'exercice d'une vengeance ; après trois ans d'une union
parfaitement heureuse, la jeune femme, devenue veuve, en ressentit
un chagrin si profond que la vie lui sembla désormais à charge ; très
calme autrefois, elle est à présent fort impressionnable et passe ses
nuits à pleurer ; la famille et le médecin, inquiets de sa santé, ont dû
intervenir pour restreindre à deux par semaine ses visites au
cimetière, visites qui, dans les premiers temps du deuil, avaient été
quotidiennes ; la veuve s'est résignée à obéir, mais son occupation
perpétuelle consiste à décompter les heures qui la séparent du
moment où il lui sera permis de retourner vers les restes de son
époux. Le jour du crime, sa belle-mère s'attacha vainement à la
dissuader, en raison du mauvais temps, d'effectuer le pèlerinage
ordinaire ; la jeune femme répondit :
« — Il doit avoir si froid, là-bas, dans la neige ; il aurait encore
plus froid, s'il voyait que je l'abandonne… etc. »
Le 9 février, la dame Léon Derouville, dont l'état s'améliore, peut
enfin être interrogée ; elle dépose :
 « Je suis arrivée au cimetière vers une heure et demie ; il était
tout blanc de neige et absolument désert ; cette grande solitude m'a
serré le cœur, à cause du pauvre ami qui me paraissait plus
abandonné que jamais ; je croyais l'entendre pleurer et j'ai hâté le
pas, pour le rejoindre plus vite ; le chemin m'a paru bien long et
j'avais une espèce de peur. Je n'ai rencontré personne dans les
allées, mais, tout d'un coup, au tournant d'un sentier, je me suis
trouvée en présence de deux hommes qui s'abritaient de la neige,
appuyés contre la porte d'un monument ; ils fumaient des cigarettes,
ce que j'ai remarqué, parce que cela m'a choquée, mais je ne leur
en ai rien laissé voir ; ils étaient jeunes et ils avaient mauvaise
mine ; cette rencontre imprévue m'a fait une impression pénible, car
je suis maintenant très nerveuse, et j'ai marché plus vite ; j'ai
entendu derrière moi ces hommes, qui me criaient des choses que je
n'ai pas comprises ; je ne me suis rassurée qu'en arrivant au caveau,
où je me sentais protégée par Lui, et j'ai bientôt oublié cette
rencontre. J'ai rangé mes fleurs et je me suis agenouillée sur le prie-
dieu, pour causer avec Lui. J'étais là depuis un certain temps,
lorsque tout d'un coup j'ai éprouvé du malaise, et je ne pouvais plus
penser, et Il ne me répondait plus, et c'était comme si quelqu'un
nous avait écoutés. Je sentais un poids sur mon cou ; alors j'ai
instinctivement tourné les yeux vers la porte, qui était restée
ouverte, et j'ai vu la figure des deux hommes, qui étaient cachés et
qui avançaient la tête, chacun d'un côté ; ils m'ont paru encore plus
méchants, et, de saisissement, j'ai poussé un cri. Alors ils sont
entrés en même temps, et je les ai suppliés de sortir, parce qu'ils
profanaient le repos ; je n'avais plus aussi peur, mais je pensais
qu'ils marchaient sur Lui, et ça me faisait mal. Je leur ai parlé bien
poliment, mais ils riaient, et ils me disaient encore des choses, des
compliments ; j'ai recommencé à prendre peur, et un des deux m'a
touchée : alors, j'ai crié de nouveau ; alors, ils m'ont enfoncé un
foulard dans la bouche, et j'étouffais ; ils riaient toujours, et ils me
serraient, avec leurs mains, par tout le corps ; ensuite, pendant que
l'un me tenait, l'autre a jeté un sou en l'air, comme pour jouer à pile
ou face. J'ai entendu le sou tomber sur la pierre de mon pauvre
mari, et l'un des deux hommes a poussé un juron, puis il s'est retiré
dans l'allée, en me laissant seule avec l'autre. A ce moment-là, j'ai
compris que j'étais l'enjeu, et je me suis débattue, mais le perdant
est venu au secours de son camarade : à eux deux, ils m'ont tordu
les reins pour me jeter sur le prie-dieu, et l'un m'a prise par les
poignets, l'autre par les chevilles, et je ne sais plus, je ne veux plus
savoir. J'ai entendu mon pauvre ami qui pleurait, dessous, et je me
suis évanouie.
« Lorsque j'ai repris connaissance, un des deux hommes était
debout devant moi, et l'autre, du dehors, appelait son camarade, en
criant qu'il venait du monde : alors, le vilain homme est sorti en
courant, et je me croyais sauvée, et je me soulevais ; mais j'ai
compris bien vite, à leurs rires, que l'un avait fait une farce à l'autre,
et que personne ne venait…
« Alors, le second a dit que c'était son tour, à présent, et je le
regardais avec frayeur ; alors, il s'est mis à danser devant moi, dans
le sentier, en faisant des gestes affreux. J'avais retrouvé ma raison,
et, d'un coup, sans avoir l'air, j'ai poussé la porte, si vite et si fort,
qu'elle s'est refermée, au nez de l'homme. Alors, il a été furieux, et
son camarade riait en se tapant les genoux : je les voyais à travers
la grille, et, à mesure que l'un riait, l'autre devenait plus en colère. Il
avait la figure collée contre le bronze du grillage, et il me criait des
injures, des menaces, il m'ordonnait de rouvrir la porte, et, aussi,
par instants, il prenait une toute petite voix pour me dire des choses
que je n'ai pas comprises, et me faire des promesses qui n'avaient
pas de sens, pendant que son camarade riait de plus en plus.
« Je m'étais blottie dans le coin, contre l'autel, pour être aussi
loin que possible. Mais il a sorti son revolver et me l'a montré à
travers la grille, jurant qu'il allait tirer et me tuer, si je n'ouvrais pas.
Je me suis jetée à terre et traînée vers la porte, pour m'appuyer tout
contre en me faisant petite, afin qu'on ne pût pas me viser ; mais, si
j'ai eu cette idée-là, bien sûr mon pauvre mari me l'envoyait, par
pitié pour moi, car je n'ai réfléchi à rien, et l'idée m'est venue toute
seule.
« Alors, le brigand a tiré un coup de feu qui a sonné fort dans le
caveau, et j'ai senti comme si on me frappait l'épaule avec un
bâton ; il a tiré, sans arrêter, plusieurs coups de revolver ; tout s'est
mis à tourner sur ma tête, l'autel, les murs, et je n'ai plus rien
entendu.
« Quand je suis revenue à moi, la nuit tombait ; j'ai essayé de
me relever et je n'ai pas pu : je souffrais partout. J'ai essayé de crier,
et je n'ai pas pu. La nuit est venue tout à fait. J'avais mal dans la
poitrine et à l'épaule, chaque fois que je respirais. Toute la nuit, la
douleur m'a empêchée de dormir ; j'étais glacée, et je devenais folle,
tant j'avais peur de tous ces morts, autour de moi. Je ne veux plus
m'en souvenir! Le matin, j'ai entendu des pas ; j'ai appelé. On ne
m'a pas répondu. Je tremblais de fièvre : dans un vase de fleurs, le
seul qui ne fût pas brisé, j'ai bu de l'eau. La neige est tombée
encore : le vent la soufflait sous la porte. Deux fois, j'ai entendu des
pas, mais personne n'a fait attention à moi. Si quelqu'un était venu,
j'aurais jeté la clef, pour qu'on m'ouvrît! J'ai fait une prière et j'ai
compris qu'il faudrait mourir là ; au moins je mourais près de Lui. Je
ne sais pas comment on m'a retirée… »
Note pour le Parquet de la Seine : « 17 février. La dame veuve
Léon Derouville est décédée à la date de ce jour, atteinte de
pleurésie ; les auteurs de l'attentat sont activement recherchés. »
 LA BARATTE

— Oh! ma foi, Dieu oui, monsieur le juge, c'est bien vrai que j'ai
tué, on peut le dire, et même je n'y ai pas regret, vous savez bien.
Je suis une pauvre malheureuse femme, et si vous croyez qu'il vaut
mieux qu'on me coupe le cou, ce n'est pas moi, bien sûr, qui vous
dédirai ; je n'en aurai pas de la peine, monsieur le juge, bien sûr! Il
faut le faire, si vous voyez que c'est mieux, et vous ne devez pas
vous déranger à cause de moi, quand c'est votre idée, parce que
moi, ça ne me fait rien, vrai comme je vous parle.
On peut bien dire que j'ai fait ça, de tuer, et toute seule, car ma
fille n'y est pour rien, je vous le promets : elle m'a regardé faire, oui,
mais pas plus, monsieur le juge, et vous pouvez me croire, car je ne
voudrais pas vous faire tort avec un mensonge, quand vous avez été
toujours bien honnête avec moi, et pas méchant, comme on raconte
qu'il y en a chez vous autres. Ma fille a su que j'allais tuer son petit,
ça, ça est ; mais pour m'avoir donné un coup de main, ça, non, elle
n'a pas fait. Tout de même, vous pouvez bien lui couper la tête à elle
aussi, comme à moi, monsieur le juge, et vous lui rendrez service :
car elle n'a rien de bon à attendre sur la terre, et elle sera mieux
dessous, comme de raison. Elle se reposera, et il n'est que temps :
elle l'a bien gagné, et son paradis avec, car nous n'avons jamais fait
tort à personne, ni l'une ni l'autre, et le bon Dieu le sait bien.
Mais je vais vous dire le tout, et vous m'excuserez si je vous
retiens un peu de temps à m'écouter : il faut que je remonte en
arrière pour que vous compreniez le fin de la chose, n'est-ce pas?
Je n'ai pas eu la vie heureuse, moi non plus : c'est la boisson qui
a fait le mal, toujours la boisson! Le cidre, et l'eau-de-vie, surtout!
L'eau-de-vie fait tout le mal, chez nous! Pas à moi, mon bon
monsieur, car je n'en ai jamais touché une goutte, et ça me fait peur,
tenez, comme le feu! On a eu trop de misère, rapport à la boisson!
Mon père était bon marin, et il gagnait, à Islande, des écus et de
l'or, tant et plus ; mais, une fois à terre, il buvait tout, et toujours en
bordée! Pour lors, on ne mangeait pas, l'hiver, et nous étions sept
enfants, sans compter mon frère Yves-Marie, qui faisait huit, et qui
était drôle, comme on appelle : je veux dire qu'il n'avait pas toute sa
tête ; mais il était fort, dame! et solide, et il lui fallait des patates à
son souper, plus qu'à un autre, encore. Mais on n'en avait pas à lui
donner tous les jours, ni de la soupe, bien sûr, et personne ne
mangeait à sa faim.
C'est dans ce temps-là que je me suis mariée avec mon mari ; au
commencement, ça marchait : il était bon marin, lui aussi ; mais il
n'allait pas à Islande ; il ne buvait que le dimanche et le lundi. C'était
un brave garçon, je dois le dire, pas mauvais et courageux à
l'ouvrage, qui savait la mer ; mais, quand il était en boisson, il ne
connaissait plus rien et il cassait tout. Mon meilleur temps, c'est
quand il rentrait tout mort à rouler : alors, ça allait ; je n'avais qu'à
le ramasser pour le mettre dans le lit, et, comme ça, il ne faisait pas
de dégât dans la maison. Ça coûte, quand on casse! Et même sans
casser, on avait de la peine à vivre, tenez! Nous avions cinq enfants,
en comptant Toussaint, qui était drôle, comme son oncle, et puis
Honorine, la cadette, qui ne savait pas parler, à cause d'une maladie,
et qui était muette, sauf votre respect. On dit que toutes ces
maladies-là, c'est la boisson qui les fait, la boisson des parents, vous
comprenez : moi, je ne peux pas croire, parce que ça ne serait pas
juste, et le bon Dieu est juste. Mais on dit que c'est vrai tout de
même. Pour lors, quoique ça, j'avais du mal. Mon mari, à la fin, se
soûlait trois et quatre fois la semaine. C'était trop, mon cher
monsieur, vous ne trouvez pas? Un jour qu'il était bu, mais pas
assez, il a voulu aller sur son bateau, malgré le temps : il a attrapé
un coup de gui à la tête, et il est tombé dans l'eau ; on l'a trouvé,
après trois jours, sous le courant, parce qu'il faut vous dire que le
courant, par chez nous, est fort comme un diable, et il vous
emporte : jamais on ne reste en place, avec lui.
Quand on a retrouvé mon homme, on est venu me chercher : je
l'ai vu, couché tout en grand sur la grève, même que les crabes
l'avaient haché, et qu'il avait encore deux bigorneaux, un sur chaque
œil, à le manger. Voilà l'eau-de-vie, monsieur le juge, et ce qu'elle
fait! Ça n'est pas une pitié?
Aussi, quand il a fallu marier mon aînée, Céline, j'ai bien gardé,
allez, pour voir si son prétendu n'allait pas aussi à la boisson, comme
le mien et celui de ma mère. Dans le pays, ils buvaient tous, ou
autant dire ; alors, je l'ai pris ailleurs, pas bien loin, à dix lieues. Et il
avait l'air doux, je vous assure, ce garçon, et gentil, et il jurait sa foi
que jamais il n'avait touché une bolée, et qu'il prenait seulement un
rien de piquette, à son souper, comme de juste. Un homme, non
plus, ne peut pas se priver de tout. Oui, mais, mon bon monsieur le
juge, il mentait, celui-là, et j'ai bien su, quand il a été marié avec
Céline, qu'il était tout pareil aux autres, devers la boisson. Le cidre
et tout, ça marchait! Chaque matin un verre d'eau-de-vie, avec son
café, et un grand verre, tenez! Il était bon maçon, et il se faisait des
journées de trois et quatre francs, quand il voulait, et on le
demandait, car on fait assez bien de bâtisses, dans tout le pays,
autour de Brest. Mais il ne cherchait pas souvent le travail, et il
refusait d'aller au chantier, s'il avait dix sous dans sa poche, pour se
soûler. Quand il n'avait plus rien, il travaillait un jour, deux jours,
quelquefois trois, mais pas plus, car le samedi arrivait tout de suite,
et, le soir, vite au cabaret, pour le dimanche, le lundi, le mardi ; soûl
mercredi, il dormait avec sa boisson.
Mais vous, mon bon monsieur, qui êtes capable, et qui êtes
quelque chose dans la justice, est-ce que vous ne pourriez pas faire
une loi pour empêcher qu'on vende de l'alcool dans tous ces
cabarets de misère? Ça serait charité pour les pauvres femmes, et
pour les hommes, aussi, puisqu'ils se tuent la santé avec ça! Si c'est
poison, comme on dit, il faut vendre ça chez le pharmacien, avec
tant d'autres poisons qu'il a! Je dis peut-être bien une bêtise, et vous
m'excuserez, peut-être ; mais si j'étais juge dans le pays, moi, ou le
président, ou quelque chef, comme vous, j'opposerais de vendre la
mort, tant que ça.
La vérité, c'est que la misère avait commencé chez Céline, dès au
bout d'un an qu'elle était mariée, autant dire tout de suite : chez le
meunier ou chez l'épicier, on ne voulait plus lui faire crédit, vous
pensez bien, puisqu'elle ne payait pas.
Mais ça a bien été une autre histoire, un jour. Voilà-t-il pas que le
mari de Céline s'en est retourné dans son pays, tranquille comme
Baptiste?
— On ne peut plus aller, qu'il dit, je m'en vais.
Et il a fait. Le vrai, voyez-vous, c'est qu'il a mieux aimé garder
tout son argent pour la boisson, et être tranquille avec ses amis,
comme avant le mariage, quoi! On l'a plus revu. Céline est venue
demeurer avec moi. Mais c'était de la misère, tenez! Car, moi,
comment voulez-vous que je gagne? Ah, oui, c'était de la misère, et
vous pouvez me croire.
Pour lors, je vous dirai que ma fille, dans ce temps-là, nourrissait
son petit, qu'elle avait eu. Mais elle ne faisait guère de lait, vous
entendez bien, parce qu'elle ne mangeait pas, et il faut savoir,
monsieur le juge, qu'une femme a besoin de manger, quand elle
nourrit. Le petit prenait la bouillie. Il venait bel enfant, tenez,
magnifique! Il forçait à vue d'œil. Tout de même, il ne marchait pas,
et quand il a eu ses quatorze mois, impossible qu'il se tienne
debout ; alors, on a bien vu qu'il avait une jambe un peu courte, ou
la hanche, tenez, là, qui était faible, et qui pliait. Je l'aimais bien. Il
me riait. Céline me le laissait, quand elle allait laver au puits.
J'essayais de le mettre droit, et je halais sur sa jambe, pas trop fort,
pour qu'elle allonge.
Un jour, un beau monsieur de Paris, qui se promenait sur nos
grèves, était là à me regarder faire, et il dit comme ça : « Qu'est-ce
qu'il a, ce petit? » Et il le touche avec un air de s'y connaître.
— Vous êtes un médecin? que je lui dis.
 — Oui, dit-il.
Il remuait la tête et il n'était pas content. Je demande :
— Qu'est-ce que c'est avec le petit, s'il vous plaît?
— L'hérédité, qu'il me répond.
— C'est mauvais, cette maladie-là? Et d'où qu'il l'a prise?
Il a ri un peu, pas beaucoup, et il m'a demandé :
— Le père boit?
— Vous êtes sorcier? que je dis.
— Non, qu'il dit.
— Ah! que je dis, vous savez tout de même le vrai.
Alors, il me demande si le gamin a des frères, des sœurs, si je
suis la grand'mère, si j'ai eu plusieurs enfants, s'ils étaient bien
allants, si mon mari buvait, tout, quoi, il me demande tout. Je lui
raconte Yves-Marie et puis Toussaint, qui étaient drôles, que je vous
ai dit, et Honorine qui est muette. Toujours le monsieur remuait la
tête comme s'il avait eu de la peine, ou comme s'il s'attendait par
avance à ce que j'allais lui raconter.
— Mais il guérira, que je dis, n'est-ce pas?
Cette fois-là, il ne m'a pas répondu, et il remuait encore la tête ;
il m'a mis dans la main une pièce blanche, et puis il s'en est allé, le
monsieur de Paris, et j'ai bien senti qu'il ne voulait pas me dire que
le petit ne guérirait pas, jamais, et que ma fille l'avait sur les bras,
pour toujours, à le regarder souffrir, sans rien pouvoir contre.
Alors j'y ai pensé toute la nuit, et je me disais : « Vaudrait mieux
qu'il soit mort. Ça n'est qu'un moment à passer. »
Parce que, il faut bien vous dire ça, sur nos grèves, on ne meurt
pas comme dans les villes : on en a l'habitude, voyez-vous, et ça ne
nous dérange guère, vu qu'à tout moment il y en a qui s'en vont
dans la mer, et c'est chacun son tour. Il faut ce qu'il faut, et on ne
change pas sa destinée, vous pouvez me croire.
Pour lors, le lendemain, j'ai dit à Céline :
 — Ma fille, c'est ça et ça ; ton petit ne guérira jamais : il est
empoisonné par la boisson. C'est pas ta faute ; mais, plutôt que de
le laisser souffrir, il vaudrait mieux lui faire délivrance, n'est-ce pas?
— Sûr, qu'elle dit, puisqu'on ne peut pas le nourrir, et qu'il a du
mal.
— Si tu veux, que je dis, moi je ferai.
Elle m'a répondu :
— Bien sûr que moi je ne ferai pas, parce que je ne pourrais
pas ; mais tout de même je vois bien que c'est le mieux et, si tu
crois, tu peux faire.
Vous pensez bien, mon cher monsieur, qu'elle en avait, du
chagrin, en disant ça, et des larmes tout plein les yeux, malgré
qu'elle se tenait, pour être forte et ne pas pleurer.
— Pour ça, si ce n'est que ça, je ferai, moi, que je dis.
— Alors, qu'elle dit, fais.
Il était, tenez, à ce moment-là, environ quatre heures, puisque
nous avions deux bonnes heures avant le bas de l'eau, et que j'ai eu
le temps de faire la toilette au petit. Et il riait, mon bon monsieur!
— Peut-être bien qu'il comprend, qu'elle dit, Céline : il sait que
nous lui faisons de ne pas souffrir, puisqu'il rit tant!
Elle l'embrassait à tous les coups qu'elle passait devant, et elle lui
riait en se tenant de pleurer.
— Mais, qu'elle dit, comment tu vas faire ça?
— Oh! que je dis, on le mettra dans la baratte, avec le flot.
Vous pensez bien, monsieur le juge, qu'on ne regarde pas à une
baratte, pour un enfant, et j'ai même pris la plus propre, celle où je
fais ma buée, et qui est comme neuve.
On avait bien arrangé le petit, avec du linge frais, et on n'y a pas
regardé, vu que Céline n'aura jamais un autre enfant, comme vous
pouvez croire. Elle lui a mis au cou sa petite croix d'argent qu'elle
avait de sa première communion, avec une médaille de la Vierge.
Elle ne les aurait données à personne, non, dame! pour or ni pour
argent ; mais, pour son petit, elle n'y a pas eu regret.
On faisait tout ça sans rien dire, à cause de la gorge, qui nous
serrait, et on ne voulait pas avoir l'air. Quand le petit a été bien gréé,
Céline l'a encore embrassé, et puis elle a dit : « C'est toi qui le
porteras ; moi je n'y ai pas le cœur. » Elle a pris la baratte qui était
assez joliment grande.
— C'est celle, que je dis, où j'ai lavé tes affaires de noces.
— Du propre, dit-elle.
Je portais le petit, et elle marchait derrière, avec la baratte, et
une couette qu'elle avait prise pour coucher l'enfant.
Il n'y avait personne par les grèves, rapport au temps, qui était
vilain, ce jour-là. C'était grande marée. Nous avons mis un grand
quart d'heure, dans les roches, pour attraper le bas de l'eau. Parce
que, vous comprenez bien, si vous connaissez la mer, qu'il fallait
joindre le bas de l'eau, pour que la mer, quand il y aurait le flot,
prenne notre baratte et l'emporte dans le courant, qui est fort,
comme je vous ai dit…
Mais c'était mal aisé d'aller, comme vous pensez, rapport à la
pluie, car il n'y a rien de plus glissant que les roches de mer, quand il
pleut dessus : c'est tout ciré. J'avais peur de tomber avec le petit,
rapport à mes sabots, et je lui aurais fait du mal : je les ai tirés, et
Céline a aussi tiré les siens, qui faisaient du bruit sur les cailloux, car
on ne voulait pas être dérangées, vous jugez.
Quand on est arrivé, il y avait flot, déjà. Céline a posé la baratte,
en la calant dessous avec du galet, pour qu'elle soit bien d'équerre,
et le petit à son aise. On lui a fait un lit, nous deux, avec la couette,
et Céline l'a embrassé pendant que je le tenais ; mais elle ne l'a pas
pris, monsieur le juge, ça, je vous jure, vrai comme je suis là! Elle ne
l'a pas pris pour me laisser tout faire, et je n'aurais pas voulu le lui
laisser prendre.
On l'a couché sur la couette, si mignon qu'il était! Il avait l'air
d'un Enfant Jésus dans sa crèche.
 Ah! dame! vous pensez bien, quand ça été fini d'arranger, et qu'il
a fallu partir, c'était des cris et des cris! La pauvre Céline
l'embrassait, son petit, fallait voir ça et ça fendait le cœur! Même
qu'un moment j'ai cru qu'elle n'allait plus vouloir, la malheureuse, j'ai
dû la remonter. Je lui disais : « Ça vaut-il pas mieux, tout de même,
que de le laisser mourir de faim, quand tu n'as plus de pain à lui
donner, et quand il va souffrir de son mal de boisson, sans pouvoir
guérir?
— Oui, qu'elle dit.
— Mourir pour mourir, que je dis, et il ne souffrira plus.
— Non, dit-elle.
Elle pleurait, comme vous pensez, et je l'ai emmenée…
Souvent, elle se retournait, pour voir si le flot arrivait déjà. En
route, nous avons retrouvé nos sabots. Il tombait une pluie fine.
— Pauvre petit, qu'elle dit Céline, il sera mouillé.
Vous me croirez si vous voulez, monsieur le juge, il n'avait pas
pleuré en nous voyant partir. On aurait dit qu'il comprenait que
c'était pour son bien, ce qu'on lui faisait là.
Quand nous avons été de retour sur la lande, Céline s'est
retournée et moi aussi, comme de juste. Le flot avait monté, et nous
reconnaissions notre baratte, quoique, vous savez bien, la baratte et
les roches, c'est de la même couleur…
La mer venait tout autour.
Céline a levé les deux bras, un peu, et les a tendus vers. Et puis
elle a été prise d'un hoquet.
Alors, je l'ai emmenée à la maison, pour qu'elle ne voie pas
monter le flot…
Welcome to Our Bookstore - The Ultimate Destination for Book Lovers
Are you passionate about books and eager to explore new worlds of
knowledge? At our website, we offer a vast collection of books that
cater to every interest and age group. From classic literature to
specialized publications, self-help books, and children’s stories, we
have it all! Each book is a gateway to new adventures, helping you
expand your knowledge and nourish your soul
Experience Convenient and Enjoyable Book Shopping Our website is more
than just an online bookstore—it’s a bridge connecting readers to the
timeless values of culture and wisdom. With a sleek and user-friendly
interface and a smart search system, you can find your favorite books
quickly and easily. Enjoy special promotions, fast home delivery, and
a seamless shopping experience that saves you time and enhances your
love for reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!

ebookgate.com