Textbook Metaheuristics For Data Clustering and Image Segmentation Meera Ramadas Ebook All Chapter PDF

Metaheuristics for Data Clustering and
Image Segmentation Meera Ramadas

Visit to download the full and correct content document:
https://textbookfull.com/product/metaheuristics-for-data-clustering-and-image-segmen
tation-meera-ramadas/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Recent Advances in Hybrid Metaheuristics for Data

Clustering Sourav De (Editor)
https://textbookfull.com/product/recent-advances-in-hybrid-
metaheuristics-for-data-clustering-sourav-de-editor/
Metaheuristic Algorithms for Image Segmentation Theory

and Applications Diego Oliva
https://textbookfull.com/product/metaheuristic-algorithms-for-
image-segmentation-theory-and-applications-diego-oliva/
Text Segmentation and Recognition for Enhanced Image

Spam Detection: An Integrated Approach Mallikka
Rajalingam
https://textbookfull.com/product/text-segmentation-and-
recognition-for-enhanced-image-spam-detection-an-integrated-
approach-mallikka-rajalingam/
Modern Technologies for Big Data Classification and

Clustering 1st Edition Hari Seetha
https://textbookfull.com/product/modern-technologies-for-big-
data-classification-and-clustering-1st-edition-hari-seetha/
Medical image recognition, segmentation and parsing :
machine learning and multiple object approaches 1st
Edition Zhou
https://textbookfull.com/product/medical-image-recognition-
segmentation-and-parsing-machine-learning-and-multiple-object-
approaches-1st-edition-zhou/
Metaheuristics for Medicine and Biology 1st Edition

Amir Nakib
https://textbookfull.com/product/metaheuristics-for-medicine-and-
biology-1st-edition-amir-nakib/
Adaptive Resonance Theory in Social Media Data

Clustering Roles Methodologies and Applications Lei
Meng
https://textbookfull.com/product/adaptive-resonance-theory-in-
social-media-data-clustering-roles-methodologies-and-
applications-lei-meng/
Exam Preparatory Manual for Undergraduates Pathology

4th Edition Ramadas Nayak Rakshatha Nayak
https://textbookfull.com/product/exam-preparatory-manual-for-
undergraduates-pathology-4th-edition-ramadas-nayak-rakshatha-
nayak/
Multisensor Image Fusion and Data Mining for

Environmental Remote Sensing First Edition Bai
https://textbookfull.com/product/multisensor-image-fusion-and-
data-mining-for-environmental-remote-sensing-first-edition-bai/
Intelligent Systems Reference Library 152
Meera Ramadas
Ajith Abraham
Metaheuristics
for Data
Clustering
and Image
Segmentation
Intelligent Systems Reference Library
Volume 152
Series editors
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: [email protected]
Lakhmi C. Jain, Faculty of Engineering and Information Technology, Centre for

Artificial Intelligence, University of Technology, Sydney, NSW, Australia;
Faculty of Science, Technology and Mathematics, University of Canberra,
Canberra, ACT, Australia;
Faculty of Science, Liverpool Hope University, Liverpool, UK
KES International, Shoreham-by-Sea, UK
e-mail: [email protected]; [email protected]
The aim of this series is to publish a Reference Library, including novel advances
and developments in all aspects of Intelligent Systems in an easily accessible and
well structured form. The series includes reference works, handbooks, compendia,
textbooks, well-structured monographs, dictionaries, and encyclopedias. It contains
well integrated knowledge and current information in the field of Intelligent
Systems. The series covers the theory, applications, and design methods of
Intelligent Systems. Virtually all disciplines such as engineering, computer science,
avionics, business, e-commerce, environment, healthcare, physics and life science
are included. The list of topics spans all the areas of modern intelligent systems
such as: Ambient intelligence, Computational intelligence, Social intelligence,
Computational neuroscience, Artificial life, Virtual society, Cognitive systems,
DNA and immunity-based systems, e-Learning and teaching, Human-centred
computing and Machine ethics, Intelligent control, Intelligent data analysis,
Knowledge-based paradigms, Knowledge management, Intelligent agents,
Intelligent decision making, Intelligent network security, Interactive entertainment,
Learning paradigms, Recommender systems, Robotics and Mechatronics including
human-machine teaming, Self-organizing and adaptive systems, Soft computing
including Neural systems, Fuzzy systems, Evolutionary computing and the Fusion
of these paradigms, Perception and Vision, Web intelligence and Multimedia.
More information about this series at http://www.springer.com/series/8578

Meera Ramadas Ajith Abraham
•
Metaheuristics for Data

Clustering and Image
Segmentation
123
Meera Ramadas Ajith Abraham
Information Technology Scientific Network for Innovation and
University College of Bahrain Research Excellence
Manama, Bahrain Machine Intelligence Research Labs
(MIR Labs)
Auburn, WA, USA
ISSN 1868-4394 ISSN 1868-4408 (electronic)

Intelligent Systems Reference Library
ISBN 978-3-030-04096-3 ISBN 978-3-030-04097-0 (eBook)
https://doi.org/10.1007/978-3-030-04097-0
Library of Congress Control Number: 2018961197
© Springer Nature Switzerland AG 2019

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, 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.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Metaheuristic algorithms like genetic algorithms (GAs), differential evolution (DE),

artificial bee colony (ABC), and particle swarm optimization (PSO) have emerged
as potential algorithms for dealing with complex optimization problems, which are
otherwise difficult to solve by traditional methods. Researchers have continued with
their work to create different variants of the existing algorithms to find the most
optimal method. Differential evolution (DE) is considered to be a dominant tech-
nique for optimization and is being used to solve various real-time problems. DE is
a powerful, population-based, stochastic optimization algorithm. The mutation
strategy of DE algorithm is an important operator as it aids in generating a new
solution vector. Differential evolution algorithm has undergone various changes and
evolution by means of numerous researches. The performance of each of these
algorithms depends on the changes in mutation and crossover strategies. In this
research work, DE and their modified variants are applied to clustering of data and
images. Evolutionary algorithms work in a robust and efficient manner for clus-
tering. The ever-growing demand for acquiring, storing, and processing a huge
amount of data for analysis and the need for producing customized results have
given space for researchers to invent innovative ways for handling such large
volume of data. Clustering is an area where the data formed in clusters exhibit
efficient customer segmentation, organization of document, information retrieval,
extraction of topics, classification, collaborative filtering, visualization, and
indexing. Quick and accurate clustering of documents is instrumental in the area of
information retrieval system. In the case of images, image segmentation is per-
formed where grouping or clustering of bitmap images into different classes is done
such that each class gives one particular information about the bitmap image within
the entire image collection. In the past few years, image segmentation has emerged
as an active area of research in different domains including medical imaging,
satellite imaging. In computer vision, image segmentation is the process of
extracting the data from the image, which is very simple and easier to analyze.
Image segmentation is a complex task in computer science because it uses the
whole image as a data input, which needs to be processed further. Clustering of data
and image can be modeled as a nonlinear optimization problem for which suitable
v
vi Preface
techniques are needed to attain the global solution by overcoming the problem of
local maxima. The volume is divided into eight chapters with Introduction as
Chap. 1 and Conclusion as Chap. 8. In Chap. 2, detailed studies of the various
traditional algorithms are elaborated. Also, the detailed literature study on the
variants of the evolutionary algorithm and its application on clustering is given.
Chapter 3 deals with the newly created variant of differential evolution named
revised differential evolution (ReDE) and its comparison with the classical mutation
strategy of DE. Chapters 4 and 5 elaborate new variants of evolutionary algorithm
named as search strategy flower pollination algorithm with differential evolution
(ssFPA/DE) and forced strategy differential evolution (FSDE) and its usage on data
clustering. Chapter 6 deals with the usage of another variant of DE on the seg-
mentation of weather images. The algorithms are elaborately studied and are tested
on a large number of test images with their performance carefully analyzed in
reference to various performance measures including statistical analysis. It is
observed that although all the algorithms are suitable for dealing with clustering,
DE and its variants perform better.
Manama, Bahrain Meera Ramadas

Auburn, USA Ajith Abraham
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Major Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Metaheuristics and Data Clustering . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Genetic Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Basics of Genetic Algorithms . . . . . . . . . . . . . . . . . . . . . . 7
2.1.2 Reproduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.3 Crossover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.4 Mutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Particle Swarm Optimization Algorithm . . . . . . . . . . . . . . . . . . . . 10
2.3 Differential Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.1 Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.2 Mutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.3 Crossover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.4 Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.5 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Flower Pollination Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 Data Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.5.1 Various Clustering Techniques . . . . . . . . . . . . . . . . . . . . . 21
2.5.2 Distance Measures Used in Clustering . . . . . . . . . . . . . . . 22
2.5.3 K-Means Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6 Image Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7 Work Done on Metaheuristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.8 Application of Metaheuristics on Data Clustering . . . . . . . . . . . . . 35
2.9 Application of Metaheuristics on Image Segmentation . . . . . . . . . 41
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
vii
viii Contents
3 Revised Mutation Strategy for Differential Evolution

Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2 Revised Mutation Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3 Experimental Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4 Graphical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.5 Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4 Search Strategy Flower Pollination Algorithm with Differential
Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2 Proposed Hybrid Method ssFPA/DE . . . . . . . . . . . . . . . . . . . . . . 68
4.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.4 Graphical Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 ssFPA/DE in Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.6 Experimental Results on Clustering . . . . . . . . . . . . . . . . . . . . . . . 82
4.6.1 Cluster Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.6.2 Validation Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.6.3 Graphical Representation . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5 Forced Strategy Differential Evolution Used for Data
Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2 Proposed Variant of Differential Evolution . . . . . . . . . . . . . . . . . . 97
5.3 Test Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4 FSDE in Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.5.1 Cluster Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6 Reconstructed Mutation Strategy for Differential Evolution
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.2 RDE Mutation Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.3 Experimental Results and Statistical Analysis . . . . . . . . . . . . . . . . 123
6.4 Multi-level Image Thresholding . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Contents ix
6.5 Multi-level Thresholding Using RDE Strategy . . . . . . . . . . . . . . . 129

6.6 Test Results on Image Thresholding . . . . . . . . . . . . . . . . . . . . . . 129
6.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7 Enhanced Differential Evolution with Fuzzy c-Means
Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.2 Fuzzy c-Means Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.3 Enhanced Differential Evolution . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.4 Experimental Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.4.1 Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.5 eDE in FCM Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
8 Conclusion and Future Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.1 Future Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Chapter 1
Introduction
In the field of computer science and engineering, metaheuristic is a high level pro-
cedure, which finds, selects or generates a low level procedure that will give an
optimized solution to a complex problem. It is a high-level problem independent
framework that gives a set of strategies for developing heuristic algorithms. Meta-
heuristics are used for combinatorial optimization like genetic algorithm, evolution-
ary algorithms, particle swarm optimization, simulated annealing etc. It explores the
search space to find the near optimal solutions. Optimization is the selection of best
element from the set of alternative solutions.
Two components in metaheuristics are local exploitation and global exploration.
Exploitation consists of probing a limited region of search space in the hope of
improving the solution at hand. Exploration consists of probing a much larger por-
tion of the search space with the hope of finding other solutions that are yet to be
refined. Exploration looks for new solutions in new search space while exploita-
tion uses existing solutions and makes refinements to it to improve its fitness. Global
exploration explores the different feasible region in the entire search space for obtain-
ing global optimum. Local exploitation uses local information from local search to
ensure maximum convergence. These two concepts are conflicting and a balance need
to be maintained. There are different types of metaheuristics namely local search and
global search metaheuristics, single solution and population based metaheuristics,
hybridization and memetic algorithms, parallel metaheuristics and nature inspired
metaheuristics.
In past decades, several metaheuristic techniques have evolved by exploring vari-
ous concepts inspired from nature called nature inspired metaheuristics. These tech-
niques essentially mimic the characteristics that uniquely exist in the Mother Nature.
Nature has in itself evolved for many years continuously developing creative meth-
ods for problem solving and for adaption in this ever changing world environment.
During the Second World War, Alan Turing used heuristic algorithm for the first
© Springer Nature Switzerland AG 2019 1

M. Ramadas and A. Abraham, Metaheuristics for Data Clustering and Image
Segmentation, Intelligent Systems Reference Library 152,
https://doi.org/10.1007/978-3-030-04097-0_1
2 1 Introduction
time. The search method that he used then was named as heuristic search. In 1948, a
report was released by Turing in which he elaborated his innovative ideas on machine
intelligence and learning, evolutionary algorithm and neural networks. Since 1950,
the concept coined by Darwin started to make inroads into the world of computation
world. The great scientist Darwin had coined the concept of survival of the fittest. It
dealt with the origin, adaption and survival of the species in nature. The 1960s and
1970s saw a great development in evolutionary algorithms.
There is a trend set among the scientific community towards the use of nature
inspired metaheuristics to effectively solve and model complex optimization prob-
lems. Nature inspired metaheuristics algorithm uses all evolutionary algorithm tech-
niques. Evolutionary algorithm perform parallel search in complex search space.
The main disadvantage of classical optimization problem is its inability to adapt
its solution algorithm to a given problem. Owing to this need, more adaptable and
flexible general purpose algorithms are in demand. Based on this spur, many nature
inspired algorithms were developed like Genetic Algorithm (GA), Particle Swarm
Optimization (PSO), Ant Bee Colony (ABC), Flower Pollination Algorithm (FPA),
etc. These algorithms provide better solutions in comparison to traditional algo-
rithms. Evolutionary algorithm has become dominant problem solving methodology
among researchers. Robustness, population based collective learning process and
self-adaption are some of the advanced features of evolutionary algorithms when
compared with other global optimization techniques.
Fogel [1] introduced the concept of evolutionary computing. Evolutionary Com-
puting is a subfield of Artificial Intelligence that uses continuous combinatorial opti-
mization problems. It uses iterative process of growth and development. Evolution-
ary algorithm is a subset of evolutionary computing. Evolutionary algorithm uses
the basic steps of selection, crossover, mutation and replacement.
The basic steps for evolutionary algorithm are:
1. Initialize the population.
2. Calculate the fitness of each element in the population.
3. Repeat until termination.
4. Select the best fit individuals for reproduction.
5. Create new individuals through crossover and mutation.
6. Calculate the fitness of the new product.
7. Compare the new product with the elements in the population and replace the
least fit element with the new product.
Exploration is obtained by randomization that enables the algorithm to reach the
capacity to come out of the local optimum and thereby perform the search globally.
Randomization is used in local searches near the current best if limited to local region.
If steps are large, then search is done globally. Balancing global and local search and
correct randomness are important for efficiency of metaheuristic algorithm. Explo-
ration uses local information like history of search and gradients. Knowledge from
simulation and observations on convergence behavior of algorithms implies that
exploration increases speed of convergence and decreases rate of convergence of
1 Introduction 3
algorithm. There are two types of optimality approach. In landscape-based optimal-

ity, the information about the type of search landscape is focused. In algorithm-based
type, objective function is treated as a black box type and it uses the information dur-
ing iteration to find the best optimal solution. At times, combination of both the types
is also used.
Most of the metaheuristic approaches like DE, FPA has been applied over various
optimization problems like clustering. These approaches can find an optimal partition
of the dataset, when the number of clusters is given. In this research, variants of
evolutionary algorithm are introduced and these variants are applied in the field of
data clustering and image segmentation. Clustering is a process of breaking down
large population into smaller groups that are similar in character. It is mainly used for
data compression. This maximizes the inter class similarity. Good clustering has a
high within cluster similarity but low inter cluster similarity. It is used for improving
precision in information retrieval. Clustering is a type of unsupervised learning where
the data has no target attribute. It studies how system can learn to represent particular
input patterns in a way that reflects the statistical structure of the overall input pattern
assembly.
Similarly, clustering of text documents plays a vital role in efficient document
organization, summarization, topic extraction, information retrieval, customer seg-
mentation, classification, collaborative filtering, visualization and indexing. Fast and
accurate clustering of documents plays an important role in the field of automatic
information retrieval system. It is considered as a multi objective optimization prob-
lem. It involves an iterative task of trial and error. Clustering can be classified as hard
clustering and soft clustering. In hard clustering, each object belong to one cluster
or do not belong to any cluster. In soft clustering, each object may belong to more
than one cluster. Clustering algorithm can be classified into different types namely
exclusive clustering, overlapping clustering, hierarchical clustering and probabilistic
clustering. In exclusive clustering, each data is grouped into a particular cluster based
on its similarity to that cluster. K-means algorithm is an example for exclusive clus-
tering. Here, a given set of N data is partitioned into k different clusters. Grouping
of data is done by minimizing the Euclidian distance between the data and centroid.
It is one of the simplest unsupervised learning algorithms used for clustering. This
algorithm is significantly sensitive to initial randomly selected centroid.
Evolutionary algorithm can also be applied on to the area of unsupervised image
clustering where the implementation of cluster analysis can aid in easier understand-
ing of the images. Image clustering is often applied to segment and have easier
understanding of images based on the desired region of interest. The process of
segmentation is done using clustering techniques. Image segmentation refers to the
partitioning or dividing of a digital image into several smaller parts or segments in
order to study a given image in a detailed manner. Image segmentation is required to
study in detail some particular features (areas of interest) of a digital image. Thresh-
olding is one of the simplest techniques used for image segmentation. It is the method
of creating binary images from the given image. It forms an important and exigent
part of image processing and requires an exhaustive and robust search technique for
its implementation. Thresholding on colored images is done by manipulating the
4 1 Introduction
color components based on color spaces. Problems like thresholding are classified as
unsupervised learning and such issues can be solved using evolutionary algorithms.
Though evolutionary algorithms are being widely applied and accepted in diverse
areas like engineering, medicine, commerce, business etc., in reality they give only
marginal performance. Some of the major reasons attributed to the marginal perfor-
mance are inappropriate selection of parameters, inappropriate representation etc.
Hence, here there is tremendous scope and an urgent need to develop more hybrid
evolutionary algorithms that has the ability to optimize the performance of direct
evolutionary approach.
Hybridization of evolutionary algorithms are needed for the following:
i. improving standard of output
ii. improving the conduct
iii. to integrate evolutionary algorithm to a larger system.
1.1 Major Objectives
The core objective of this work is to create new variants of evolutionary algorithm
to solve specific optimization task. Subsequently, compare the newly created variant
with some of existing metaheuristic techniques that are commonly available to check
the efficiency of the new variants.
The basic objectives of this work are:
1. To propose the hybrids of Differential Evolution algorithm.
2. To perform the statistical analysis for each of the new variant to validate its
efficiency.
3. These proposed hybrids will be applied on to clustering of data and images to
demonstrate their application capability.
1.2 Organization
This volume is structured into several chapters with the first two chapters focusing
on the supporting information with respect to this research work and each of the
subsequent chapters explaining in detail the research work that was performed along
with the research outcome.
The below is the organization for the following sections:
Chapter 2 contains a detailed study of Differential Evolution and Flower Polli-
nation algorithm. This chapter not only explain in detail various steps involved in
these algorithms but also an in-depth study is done on the basic concepts of data
clustering and image segmentation. This chapter also discusses the literature review
about previous research works carried out on different variants created in evolution-
1.2 Organization 5
ary algorithms and their findings. This chapter also has references to the works in
the field of data clustering and image segmentation through clustering.
Hereafter, the following chapters form the core section of this work where each of
the variants created using Differential Evolution and Flower Pollination Algorithm
is explained in detail in each of Chaps. 3, 4, 5, 6 and 7 respectively. Each of these
chapters also comprise of the results obtained by applying these variants in the field
of data clustering and image segmentation.
Chapter 8 explains in brief the conclusion of the research work and future scope
of work that can follow up based on this research work.
Reference
Fogel, L.J., Owens, A.J., Walsh, M.J.: Intelligent decision making through a simulation of evolution.
Behav. Sci. 11(4), 253–272 (1966)
Chapter 2
Metaheuristics and Data Clustering
In this Chapter, the initial Sects. 2.1–2.4 give detailed analysis on types of metaheuris-
tics namely Genetic Algorithm, Particle Swarm Optimization, Differential Evolution
algorithm and Flower Pollination algorithm. Section 2.5 elaborates the basic data
clustering technique and Sect. 2.6 summarizes about the image segmentation. Sec-
tions 2.7–2.9 explains the works related to metaheuristics and their application on
data and image clustering respectively.
2.1 Genetic Algorithm
Holland [1] introduced genetic algorithms. Genetic algorithm (GA) is used to produce
high quality solutions to search and optimize the problems using the concepts of
mutation, crossover and selection. GA has been successfully applied in many areas
like clustering, image segmentation, image watermark etc.
2.1.1 Basics of Genetic Algorithms
Genetic algorithms is based on Darwin’s theory of survival of the fittest. The basic
idea behind GA is natural selection and natural genetic systems in which fittest
individuals will survive and the remaining individuals will be discarded. In GA, take
a set of candidate solution and iteratively refine them by alternating and selecting
the good solution for next generation. Select the candidate solutions on the basis of
fitness function to evaluate quality. The five phases of genetic algorithm are initial
population, fitness function, selection, crossover and mutation. Around the nature’s
scanty resources, competition within individuals lead to fittest dominating the weaker.
They are a computerized searching and optimization algorithm that uses natural
selection and natural genetics technique. They take large search spaces and look
© Springer Nature Switzerland AG 2019 7
M. Ramadas and A. Abraham, Metaheuristics for Data Clustering and Image
Segmentation, Intelligent Systems Reference Library 152,
https://doi.org/10.1007/978-3-030-04097-0_2
8 2 Metaheuristics and Data Clustering
for optimal combination of solutions. The three important aspects in GA include

defining optimization function, defining and implementing genetic representation,
and genetic operators.
GA begins with the chromosome representation of a parameter set{x 1 ,x 2 ,…x n };
generally, the chromosomes are designed as strings of 0’s and l’s. GAs employs three
basic operators over a limited number of strings for finding the global near-optimal
solution. These operators are:
(i) Reproduction/Selection
(ii) Cross-over
(iii) Mutation.
2.1.2 Reproduction
According to the objective function, copy the individual strings for reproduction
process. The strings created by reproduction depends upon the Darwinian survival
theory, where the fitter strings have more chance to get a higher number of offspring
in the succeeding generation. All the successful strings form a mating pool which
will behave like a new tentative population.
2.1.3 Crossover
Reproduction operation forms a mating pool of strings; after this phase, these strings
will form offspring for the new generation using crossover operation. Crossover
process consists of two steps: (1) all the members of the mating pool will mate
randomly and (2) each pair of strings undergo crossover as follows:
i. Select an integer position k at random between l and l − 1 where, l is the string
length greater than 1.
ii. Create two new strings by swapping all characters from position k + 1 to l.
Let a and b be two strings selected for the crossover operation from mating pool and
the random number generated be 10 (ten).
a 110001010101000 . . . 0111110001
b 100010111011101 . . . 0011010100
Then the newly produced offspring by swapping all characters after position 10
will be denoted as a and b :
a 110001010111101 . . . 0011010100
b 100010111001000 . . . 0111110001
2.1 Genetic Algorithm 9
2.1.4 Mutation
In the simple GA, mutation is the occasional random alteration of the value of a string
position. It helps to prevent the irrecoverable loss of potentially important genetic
material. Select a random bit position of a random string and replace it by another
character from the population.
For example, let the fifth bit of string ‘a’ given above, be selected for mutation.
Then the transformed string after mutation will be:
a 110011010101000 . . . 0111110001
The flow diagram for genetic algorithm is as given below in Fig. 2.1.
The general pseudo code for genetic algorithm is given below:
Initialize population with random candidate solution
Evaluate fitness of each candidate solution
While termination condition not true do
Select the best individuals to be used by genetic operators
Generate new individuals using crossover and mutation
Evaluate fitness of new individuals
Replace the worst individuals of population by best new individuals
End
Initialize Population
Selection Crossover Mutation
Termination
condition
No
Yes
End
Fig. 2.1 Flow diagram for genetic algorithm

2.2 Particle Swarm Optimization Algorithm
Eberhart and Kennedy [2] introduced a method for solving continuous nonlinear
function. It is the simulation of social behavior of birds and was termed as Particle
Swarm Optimization (PSO). It moves solutions through search space. Swarm is an
identified collection of active individuals, which tend to group together while each
individual seems to be moving in arbitrary direction. It is very simple in concept,
easy to implement and computationally efficient. This concept has its origin from
evolutionary computing and artificial life, which generate the population by assign-
ing random position and velocities. Unlike GA, PSO does not have operators like
mutation and crossover.
Each particle keeps track of best position in the hyperspace. It also keeps track of
the previous best position for individual particle, global best position in population
and local best position in the defined neighborhood. Particles take benefit from the
previous particle’s experience and makes search for higher objective function values.
Treat each particle as a point in N-dimensional space which adjusts its “flying”
according to its own flying experience as well as the flying experience of other
particles. The particles revise itself by using the internal velocities. This algorithm
also utilizes memory space unlike other evolutionary algorithms.
In PSO algorithm, each particle consist of a data representing a solution, personnel
best position pbest and velocity value. pbest is the personnel best position of a particle
having its co-ordinate in the solution space with the best solution achieved by the
particle. Velocity value indicates how much the value has changed. PSO keeps track of
three global variables namely target value, global best position gbest and the stopping
value indicating when the algorithm should terminate. gbest is the global best value
achieved by a particle. Basic idea behind PSO is to accelerate each particle towards
its gbest and pbest positions using a random weighted acceleration. Position of each
particle will change on the basis of the (i) current positions (ii) current velocities (iii)
distance between pbest and current positions and (iv) the distance between gbest and
current positions. Therefore, the modifications will be done on the basis of given
equation:
vi,k+1 vi,k + c1 · rand (0, 1) · (pbesti − presenti )

+ c2 · rand (0, 1) · (gbesti − presenti ) (2.1)
presenti+1 presenti + vi,k+1 (2.2)
where, c1 and c2 are the weighted factor, presenti is the current position of the
candidate, vi,k+1 is the velocity of the candidate.
The general algorithm for PSO is given as follows:
2.2 Particle Swarm Optimization Algorithm 11
Initialize velocity and position of each particles

While stopping criteria not attained do
For each particle
Calculate fitness value
If fitness value is better than pbest value in history
Set current value as new pbest
End
Choose the particle with best fitness value of all particles as gbest
For each particle
Calculate particle velocity according to eq. 2.1
Calculate particle position according to eq. 2.2
End
End
End
PSO algorithm differs from other evolutionary algorithms as it doesn’t have the
operations of mutation and crossover. The particles revise itself by using the inter-
nal velocities. This algorithm also utilizes memory space unlike other evolutionary
algorithms. Flow diagram for PSO is given in Fig. 2.2.
2.3 Differential Evolution
Differential Evolution (DE) is a heuristic optimization technique proposed by Storn

and Price in [3]. It is used for minimizing non-differential, non-linear, continuous
space functions with real valued parameters. The most important characteristic of
DE is that it uses the differences between randomly sampled pairs of object vectors to
guide the mutation operation unlike the other EAs, which instead use the probability
distribution functions. DE is a technique that optimizes a problem by iteratively
trying to improve candidate solution with respect to given measure of quality.
The operations of crossover, mutation and selection make DE different from other
evolutionary algorithms. For genetic algorithm and evolutionary algorithm, muta-
tion is treated as a random change in some parameter. DE mutates the base vec-
tor with scaled population of derived difference vector. Evolutionary programming
requires adaption of absolute step size for each variable over generation, while DE
requires only specification of single relative scale factor for all variables. In crossover,
both DE and evolutionary computing uses crossover to create a single trial vector,
while genetic algorithm recombine two vectors to produce trial vectors by one point
Fig. 2.2 Flow diagram for

particle swarm optimization Initialize population
Evaluate individual fitness and

rank the fitness
Yes
Stop
condition Stop
No
Generate new population

Selection Crossover
Mutation
crossover. In selection, genetic algorithm selects parent based on fitness while DE

give all individuals an equal chance to be selected for mutation.
In a population of potential solution within an n-dimensional search space, a fixed
number of vectors are randomly initialized which is evolved over time to explore
search space and to locate the minima of objective function. At each iteration, new
vectors are generated by combination of vectors randomly chosen from current pop-
ulation. The outcome vectors are then combined with a predefined target vector. This
operation is called recombination. Thereafter, the trial vector is accepted for next
generation if and only if it yields a reduction in value of objective function. This is
called selection. The various steps involved in Differential Evolution approach are
explained in detail in the below section.
2.3.1 Initialization
DE is a parallel direct search method using a population of N parameter vectors

for each generation. At generation G, the population PG is composed of Xi,G where
i 1, 2, . . . N . For a given parameter Xi,G , three vectors Xr1,G , Xr2,G and Xr3,G
are randomly selected such that r1 , r2 , r3 are distinct. This operation extends the
workspace utilization as the individuals are chosen randomly by moving around in
the search space and giving an equal chance for all individuals to be selected for
mutation. The initial population can be chosen randomly under uniform probability
distribution if there is no prior information about the problem to be optimized. If
2.3 Differential Evolution 13
any prior knowledge about the problem is available, the preliminary solution can be
included to the initial population by adding normally distributed random deviations
to the nominal solution.
2.3.2 Mutation
The key characteristic of a DE is the way in which it generates trial parameter vectors
throughout the generation. This operator makes DE different from other Evolutionary
Algorithms. It computes the weighted difference between the vectors in population.
A weighted difference vector between two individuals is added to a third individual
to form a new parameter vector. The newly generated vector is evaluated by the
objective function. For each parent parameter vector, DE generates a candidate child
vector based on the distance of two other parameter vectors. For each dimension i,
the donor vector Vi,G is computed as:
Vi,G Xr1,G + F · (Xr2,G − Xr3,G ) (2.3)
Here F is the mutation factor which is a constant in (0,2). The above mutation
scheme is named as DE/rand/1. Mutation function demarcates one DE scheme from
another. The most often used DE codes are given below:
DE/rand/2 Vi,G Xr1,G + F · (Xr2,G − Xr3,G ) + F · (Xr4,G − Xr5,G ) (2.4)

DE/best/1 Vi,G Xbest,G + F · (Xr1,G − Xr2,G ) (2.5)
DE/best/2 Vi,G Xbest,G + F · (Xr1,G − Xr2,G ) + F · (Xr3,G − Xr4,G ) (2.6)
DE/rand-to-best/1 Vi,G Xr1,G + F · (Xbest,G − Xr2,G ) + F · (Xr3,G − Xr4,G )
(2.7)
where, i 1, 2 . . . NP, r1 , r2 , r3 ∈ {1, 2, . . . , NP} are randomly selected and r1

r2 r3 i, F ∈ [0, 2], F is the control parameter proposed by Storn and Price [3].
2.3.3 Crossover
This process, also called recombination, incorporates successful solutions into the
population. The value of the corresponding objective function is compared with a pre-
determined individual. If the newly generated parameter vector has lower objective
function value, it will replace the predetermined parameter vector. The best param-
eter vector is evaluated for every generation in order to track the progress made
throughout the minimization process. The random deviations of DE are generated
by using both the search distance and the direction information from the popula-
tion. Correspondingly, this adaptive approach is associated with the normally fast
convergence properties of a DE. The trial vector Ui,G is created for the target vector
Xi,G through binomial crossover. Elements of donor vector enter trial vector with
probability Cr ∈ [0, 1]. Cr is the crossover probability which is selected along with
population size NP ≥ 4.

Vj,i,G+1 if randi,j [0, 1] ≤ Cr or if j Irand
Uj,i,G+1 (2.8)
Xj,i,G+1 if randi,j [0, 1] > Cr or if j Irand
Here, randij ∼
∪[0, 1] and Irand is random integer from 1, 2,…NP.
2.3.4 Selection
This operation differs from the selection operation of other evolutionary algorithms.
Here, the population for next generation is selected from vectors in current population
and its corresponding trial vectors. The target vector Xi,G is compared with the trial
vector Vi,G and the lowest function value is taken into next generation.

Ui,G+1 if f (Ui,G+1 ) ≤ f (Xi,G ) where i 1, 2, . . . NP
Xi,G+1 (2.9)
Xi,G otherwise
Mutation, crossover and selection operations are continued until some stopping
criteria is reached.
2.3.5 Algorithm
The algorithm for Differential Evolution approach is given below:

2.3 Differential Evolution 15
The flow chart for Differential Evolution algorithm is given in Fig. 2.3.
2.4 Flower Pollination Algorithm
From the point of view of biological evolution, the purpose of pollination in flower is
to facilitate the optimal reproduction of plants taking into account their numbers and
Start
Input vectors of candidate solution
Calculate objective function
Perform mutation, crossover and selection

operation
No
Termination
condition
Yes
Print best vector Stop
Fig. 2.3 Differential Evolution algorithm

Fig. 2.4 Pollination in flowers (Source https://biology.tutorvista.com/)
the fittest. This is in effect an optimization process that take place within plant species.
The Flower Pollination Algorithm (FPA) uses the technique from the pollination of
flowering plants. Flowers reproduce using the concept of pollination. Yang et al.
[4] proposed a detailed work on FPA for multi-objective situations. Pollination (see
Fig. 2.4) is the method of relocating pollens from the male anther to the female stigma
of a flower. The goal of all living things is to produce offspring for the next genesis.
One of the methods that plants follow to produce offspring is by making seeds.
Pollination in plants is of two types:
• Cross Pollination
• Self-Pollination.
Cross Pollination occurs when the pollens from one plant is transferred to a
flower that belong to another plant. This process takes place with biotic and abiotic
creatures. Abiotic pollination occurs when external agents like wind, water etc. are
involved. Very few plants come under this category of pollination. Biotic pollination
is moving pollen from anther of one plant to stigma of another plant using external
pollinators like birds, insects, bees, etc. The pollinators are attracted by color, odor
and availability of nectar.
Self-Pollination on the other hand occurs when the flower pollinates within the
same plant. These pollination processes involves complex techniques been under-
taken for plant production strategies. Algorithm was developed based on these fun-
damental techniques. The global pollination occurs for cross pollination and biotic
pollen transfer whereas self-pollination and abiotic transfers are associated with local
2.4 Flower Pollination Algorithm 17
reproduction strategy. Plants can follow either method of pollination. The tendency
of individual pollinators to visit a plant depends on the attractiveness of that plant.
A solution of optimization problem is associated with a pollen gamete. Here, the
flower constancy is compared to solution fitness.
Biotic pollination takes place in 90% of flowering plants where pollen is trans-
ferred by a pollinator such as insects and animals. About 10% of pollination takes
in abiotic form, requiring no pollinators. Wind and water diffuses pollination in
these plants with flowers and grass being a good example. Pollinators that are some-
times called pollen vectors can be very diverse. They estimate that there are at least
200,000 varieties of pollinators, such as insects, bats and birds. Bees are considered
to be a good example of pollinators, which has also developed the so-called floral
constancy. These pollinators tend to visit certain unique species of flowers bypassing
other species of flowers. Such evidence may have evolutionary floral benefits, since
this process maximize the transfer of flower pollen within the same plant or of the
same species plants and thus maximizing the reproduction of the same species of
flowers. Instead of focusing on some unpredictable but potentially more rewarding
new flower species, flower constancy may require minimal investment cost and most
likely guaranteed intake nectar.
Using local and global pollination techniques, the pollens will be transferred to a
plant. Global method uses pollinators to transfer pollens to long distances towards
individual flowers characterized by higher fitness. Local pollination occurs in a lim-
ited range of individual flower using pollination mediators like air, water etc. Birds,
flies, bees and bats are the most common pollinators that fly for long distance. These
pollinators are considered as the carriers of global pollination.
FPA was formed by Yang [5]. This algorithm follows four rules.
• Self-pollination and abiotic method are considered as local pollination.
• Cross pollination and biotic method are considered as global pollination.
• Duplication probability is directly proportional to the likeness of two flowers
involved. So flower consistency is considered.
• Switch probability p controls the local and global pollination.
These rules can be converted to equations as given below:
xit+1 xit + ε(xit − xkt ) (2.10)

xit+1 xit + L(xit − g∗ ) (2.11)
where xit is the solution vector and g∗ is the current best during the iterations.
L is the step size from Levy distribution which denotes the strength of pollination.
Levy flight is used to imitate the movement of insects over a long distance.
The parameter L is computed as:
λ(λ) sin(π λ/2) 1
L∼ , (S S0 > 0) (2.12)
π S 1+λ
where, λ is the probability density function, (λ) is the standard gamma function
and S is the step size of pollination. This is a continuous probability distribution for
positive random numbers. Flower Pollination algorithm has been used widely in the
field of fuzzy logic and engineering. It is flexible, simple and is exponentially better
to solve.
The algorithm for FPA is given below:
Objective function . Initialize n pollen seeds

Select optimal result in the original group
State as number of iterations and a switching probability as
While
for all n flowers in the group
If
For Levy distribution, select a step vector of

dimension
Do global pollination
Else
Select item from a uniform distribution within
Do local pollination
end if
Calculate new solution
Update population if the new solutions are improved
end for
Find current best solution
end while
Output solution
End
The flowchart of Flower Pollination Algorithm is given below in Fig. 2.5.
2.5 Data Clustering
With the advances in technology, the need for acquiring, storing and processing a
huge amount of data is ever increasing. Clustering is a process of partitioning a set
of data (or objects) into a set of meaningful sub-classes, called clusters. It helps the
user to understand the natural grouping or structure in a data set. It is breaking down
of large population into smaller groups that are similar in character. A cluster is
therefore a collection of objects which are similar internally, but clearly dissimilar to
2.5 Data Clustering 19
Start
Input population size
Load population of flowers with

random solution
Calculate current best
No
If rand> p
Local pollination
Yes
Global pollination using levy flight
Compute new solution
Update current global solution
No Yes
Termination Output best solution
Condition? Stop
Fig. 2.5 Flowchart for Flower Pollination algorithm
the objects belonging to other clusters. Dissimilarities and similarities are assessed
based on the attribute properties describing the objects. It is a connected region
of multidimensional space, containing a relatively high density. The quality of a
clustering result also depends on both the similarity measure used by the method
and its implementation. The quality of a clustering method is also measured by
its ability to discover some or all of the hidden patterns. By organizing data into
clusters, there should be high intra cluster similarity and low inter cluster similarity.
In clustering, the objects of analysis could be persons, salaries, opinions, software
entities and many others. These objects must be carefully presented in terms of their
characteristics. These characteristics are the main variables of the problem and their
choice greatly influences the results of a clustering algorithm.
Clustering is based on three characteristics namely nesting, exclusiveness and
completeness. In the nested type, separation is built on characteristics of nesting
clusters. Hierarchical clustering is nested meaning it clusters to exist within the
bigger clusters. In hierarchical type, a set of nested clusters is organized into a
hierarchical tree. Partitioned clustering prohibits subsets of cluster. Here, a division
of data objects is grouped into non-overlapping subsets (clusters) such that each
data object is in exactly one subset. Exclusive separation is the characteristics in
which the data object is allowed to exist in one or more than one cluster. But, in
exclusive clustering, each of the data object shall exist in only one cluster. The data
objects can be grouped into two or more clusters through overlapping. For example,
in overlapping clustering, it can allow a student to be grouped as employee also
while in case of exclusive clustering, it demands that the person be grouped to the
once which are more important. In case of fuzzy clustering before also known as
probabilistic clustering, each of data object shall belong to every cluster. The data
objects here possess membership weightage between 0 and 1 where 0 shall point
that it does not belong to that cluster whereas 1 denote it belong to that particular
cluster. Completeness is a type of separation, which requires all the data objects to be
grouped. In complete clustering, every object is assigned to a cluster. Various steps are
involved in data clustering namely, data collection, initial screening, representation,
clustering tendency, clustering strategy, validation and interpretation. Data collection
includes extracting related data from the source. Initial screening is the messaging
of data after extraction from source. Representation means preparing data in order to
make it proper for algorithm. Clustering tendency verifies if the data in hand can be
considered in a cluster or not. Clustering strategy chooses the correct algorithm and
parameter. Validation is the manual examination of the data for validity of technique.
Interpretation includes combining clustering results with other studies and suggesting
further analysis.
In document clustering (see Fig. 2.6), it is possible to view the clustering prob-
lem as an optimization problem that locates the optimal centroids of the clusters
rather than to find an optimal partition. This view offers us a chance to apply evolu-
tionary algorithms for clustering problem. The appropriate clustering algorithm and
parameter settings (including values such as the distance function to use, a density
threshold or the number of expected clusters) depend on the individual data set and
their intended use of the results. Partitional algorithm permits to decompose data
into sets of disjoint clusters using similarity criterion. It allows updating of clus-
ter members if the clustering performance is improved. Partitional clustering is an
optimization problem as it minimizes the cluster assignment in probability density
Another random document with
no related content on Scribd:
He reached the motor cruiser just as the rowboat drifted back empty
on a tow-rope at the vessel's stern. The tide was running toward the
stern and he let himself drift with it, keeping close to the dark green
hull of the craft until he had rounded the stern and was swimming
just underneath the tightly stretched rope.
He was swimming close enough to the cruiser to be invisible from
the deck, unless the boarders had remained by the rail and were
staring straight down at him. He treaded water for an instant,
listening for sounds on deck, his ears alert for a startled grunt or
shout of anger.
When he was convinced that he had not been spotted he grasped
the rope firmly and ascended hand over hand to the rail. The aftdeck
was deserted. Its polished surface glimmered in the sunlight and was
encumbered only by a brass stanchion, and a waist-high coil of rope.
Although the cruiser was quite large there was no stern or forward
deckhouse, just the curving back of what appeared to be a
companionway entrance shaped like a gigantic scallop shell.
A moment later Fenton was crouching just inside the shell, above a
descending flight of stairs. A faint light was visible from the top of the
stairs, but it wasn't the light that interested him. It was the hum of
angry voices.
He started to move cautiously downward in order to hear better, but
stopped when the voices rose sharply, becoming so heated and
enraged that he could catch every word.
"We've kept our hands off you so far, but it wouldn't be a sharp idea
for you to keep shaking your head like that and pretending you don't
know what the score is. Sit down, Hansen. Sit down. We told you to
relax, didn't we?"
"He's stalling," a deeper voice cut in. "Darby would never have gone
this far if he didn't think Gerstle told him more than it's safe for him to
know. Why don't you come clean, kid? What have you to gain by
stalling?"
There was a slight pause and then Hansen's voice rose as high in
fright as the other two voices had in anger. "He didn't tell me a thing.
Only that he was collecting information for a series of articles that he
was hoping he could persuade Miss Lathrup to let him bring out
under his own byline. Sensational material which would name names
and be backed up with affidavits. Would I tell you even that much if
he'd turned any of that material over to me, as you seem to think he
did? I'd just pretend to know nothing at all about it."
"No—you're too smart for that, kid. You want us to think you're
leveling with us, and if you denied you'd seen those names—"
"But I didn't. Not one name. He took me into his confidence most of
the time, but this was too big, I guess."
"You'd make a good salesman, kid. The way you tell it ... I can
almost see myself buying it. But not quite. And that's going to make
a big difference to you, kid ... you're not quite selling us."
"What do you want me to do? Lie about it?"
"We're wasting time," the one with the lighter voice said. "We should
at least try to persuade him."
"We should try, by all means. You want to begin, Foldes?"
There was a meaty thud, followed by a groan and a low, barely
audible sobbing sound.
Fenton stiffened in instant concern, his lips tightening. He forced
himself to remain where he was for a moment longer, however. It
was safe to assume the two below would work up to what they were
doing gradually and there were things he desperately wanted to
know. Under stress of rage they might let something drop—some
clue, some pointer, which would enable him to save both Gerstle and
Hansen. They must know what had happened to Gerstle.
There was another thud, and Hansen's cry of pain was too loud, this
time, to permit of further delay. Fenton unholstered his revolver,
snapped off the safety catch and was down the companionway and
in the cabin so fast the two kidnappers were taken completely by
surprise.
He gave them no time to recover. The one who had struck Hansen
was just raising a reversed automatic for another blow and was
facing away from Fenton near the base of the stairs. A sudden
tensing of his muscles failed to save him. Just as a glint of
awareness flickered at the perimeter of his vision Fenton's fist
caught him flush on the jaw, and sent him crashing backwards.
He hit the opposite bulkhead, rebounded and sank with a groan to
his knees. Fenton moved in close again, and chopped downward on
his wrist, sending the gun clattering. It was a needless precaution,
for the man was already going limp, and had held fast to the gun in
his backward lurch by convulsively contracting his fingers.
He collapsed forward on his face and Fenton did not wait to see if he
would try to rise again. The danger that his companion would get to
the dropped gun first was too urgent. He could have prevented that
by putting a bullet in him with his own gun, but he did not want to kill
a key accessory in a murder case and he was too excited to be sure
of merely splintering the man's kneecap.
The weapon had skidded half-way across the cabin, but Fenton
raised his right foot and kicked it two yards further a split second
before it could pass into dangerous hands again.
The second man had dropped on all fours, and his hand hit the deck
with a thud when the gun was kicked beyond reach of his fingers.
Fenton reversed his own gun and clobbered the unsuccessful
weapon snatcher across the back of his skull with a blow that
flattened him out almost at his companion's side.
The big detective was breathing harshly when he straightened, his
face very white and looked with concern at Hansen, who was
moaning and slumping a little in a straight-backed chair, with a
swelling ugly-looking bruise on his right forehead. One eye was half-
closed, and his breathing was harsher than Fenton's.
Fenton bent and gripped him firmly by the shoulders, easing him into
a less strained position. "Take it easy," he cautioned. "Just lean back
and don't try to talk for a minute. You're going to be all right."
"Thanks," Hansen muttered, disregarding the advice. "They ...
slugged me twice. Felt like the whole top of my head was coming off
the second time."
Fenton nodded. "They were just being gentle," he said. "I know the
breed. Each time they hit you a little harder and they don't stop until
you black out. But they do it the slow way, even when it's the butt of
a gun they slug you with. They keep hoping you'll talk...."
"They were going to kill me," Hansen said.
"I know. I heard them talking from just up above."
"Thank God for that," the young man breathed.
"Yes. I'm glad I could get here in time."
He turned and swept the cabin with his eyes. The two kidnappers
were still out.
Fenton hesitated for an instant, then tapped the handle of his gun,
letting it rest on Hansen's right knee, and tightening his grip on
Hansen's shoulder. "I don't suppose you've ever shot a man, in self-
defense or otherwise. But do you think you could handle a gun if you
had to ... handle it well? There's a coil of rope on deck, and I've got
to tie these two up. The quicker it's done the better."
Hansen nodded, an angry glint coming into his eyes. "You can trust
me," he said. "Just let one of them make a move—"
"All right," Fenton said. "But be careful—the safety catch is off. If one
of them comes to, and tries to take the gun away from you—shoot to
kill."
Fifteen minutes later Fenton stood by the rail of the cruiser, staring
down into the clear, blue-green water, a deeply worried, almost
tormented look in his eyes. He had no right, he told himself, to feel
the way he did, for Hansen was alive and safe and the two
kidnappers securely bound.
If he'd stayed on shore until help arrived and Hansen had finished
dead, a dark cloud would have hung over him for the rest of his life,
even if he turned in his badge. He had a lot to be thankful for, for
self-reproach to a man like himself could be harder to live with than
the sternest kind of official censure.
But it took more than what had happened to drive away all of the
clouds—far more. He still didn't know what had become of Gerstle,
and although it wasn't too hard to picture what might have happened
to the elderly exposé editor it was bad ... very bad ... for a cop to
allow his imagination to paint a picture so ugly that he'd stop thinking
seriously about how to rescue a living man and concentrate solely on
capturing a remorseless killer who had included that man in his list of
victims.
He was still confronted with the same problem which had prevented
him from shooting it out with the kidnappers in front of the Eaton-
Lathrup building. The pair might be persuaded to talk, since they'd
have more time to reflect now, and would realize they could only
hope to escape the chair by turning State's evidence. It would be a
slim reed for them even then, but they might seize upon it. They
might ... but it couldn't be counted on.
He was rather glad that the motor cruiser had a tiny kitchen, and that
he'd persuaded young Hansen to spend a few minutes there
percolating some coffee before they both went ashore in the rowboat
with the securely bound pair. It gave him a chance to straighten his
shoulders, collect his thoughts and breathe in the brine-scented air. It
wasn't the open sea, only the fingertip of a bay, with the shoreline
close on both sides. But there was something about any part of the
ocean that could give a harassed man perspective, make him realize
how small and quick-passing all human tragedies were, when you
contrasted them with eternally breaking waves, and the vast shining
permanence of the sea.
He had paused for only a moment by the rail, to stare down into the
clear water, seeking perhaps to make that realization even stronger,
to keep it more forcefully in mind. Or perhaps only because he was
so inwardly preoccupied. He could not have said exactly why.
He could see every rock and crevice, every waving seaweed, every
darting silvery fish between the cruiser's keel and the sandy bottom,
for the inlet was now as still as a sheet of glass.
He was just turning from the rail when a faint gust of wind ruffled the
water, spoiling its crystal-clear transparency. The ripples attracted his
attention and he did not turn, simply continued on for a few paces
parallel with the rail.
He was staring down idly when the ripples vanished and he could
see all the way to the bottom again.
A look of horror came into his eyes and he gripped the rail with both
hands, cold sweat oozing from the pores of his skin, bringing a
glistening to his bare back, drenching him from waist to armpits.
The corpse was wedged in a narrow rock crevice, in a rigidly
contorted attitude, the face white and staring and turned upward, the
legs grotesquely bent. It was clothed only in shorts and the blonde
hair on the naked chest was matted with seaweed, which swayed
back and forth in the underwater current.
Curiously enough, the hair on the dead man's head did not move
with the current, but the slack jaw seemed to move slightly, as if
protesting against the indignity which had been thrust upon him.
There was neither strength nor weakness in Gerstle's lifeless
features now, but there was something about the configuration of the
face which suggested that great energy and firmness of purpose had
once been dominant characteristics of the man. The cheeks were
faintly blue with a two-days' growth of beard, the eyes wide and
staring, the lips purplish.
Both the wrists and ankles of the slain cafe society exposé editor
had been bound with heavy wire which glistened in the
downstreaming sunlight, and had cut cruelly into the flesh, whether
before or after death Fenton had no way of knowing.
How long the corpse had remained at the bottom of the inlet was
another thing which the detective had no way of knowing. But he
was almost sure that it could not have been longer than two or three
days, for no trace of decomposition was visible on either the face or
the body.
It could have been dropped overboard from any part of the inlet and
been carried by the tides to where it now was, but somehow he
doubted that it had been carried far. It did not have a sea-battered
look.
Fenton did not remain for more than a minute or two by the rail
speculating about it. A grapple might have drawn it to the surface,
but he had no stomach for such a procedure at that particular
moment, even if he could have found a grapple somewhere on the
cruiser.
He was content to let the body remain where it was, securely
wedged in a rock crevice, until the police could follow their usual
procedure, and examine it in situ before removing it.
A sardonic thought flashed for an instant across his mind, but he put
it from him as unworthy. No dusting for fingerprints here, or
surrounding the corpse with chalk marks. The bottom of a Flushing
Bay inlet was quite different from a magazine office.
Chapter IX
It might have been a repeat of a conference that had taken place in
the Eaton-Lathrup offices several days earlier, if twice the original
number of people hadn't been present. Like the earlier conference, it
was held in Macklin's office and in addition to Macklin, Eaton, young
Hansen and Ellers, there were two women and two police officers
present.
It was a conference ... in a strictly official sense. Lieutenant Fenton
had made it plain that there were a number of weighty matters to be
discussed and that he wished precisely eight people to be present.
The eight, of course, included First Grade Detective Gallison and
himself. One of the women was Lynn Prentiss, the other Susan Weil,
who presided over the seldom-idle switchboard in the outer office.
As before, it was Macklin who seemed the most intent on asking
blunt questions, and challenging the opinion of the majority. Even
Fenton came under challenge, and the big detective seemed content
to let Macklin talk on for several minutes in almost uninterrupted
fashion, for many of the points which the boyish-looking editor
brought up—he was almost phenomenally youthful-looking for a man
in his forties, Fenton told himself—seemed both discerning and well-
taken.
"I don't understand," Macklin was saying, "why there should be any
doubt left in your mind, Lieutenant, as to the guilt of the man you've
just arrested and charged with Gerstle's murder."
"Not as to Gerstle's murder," Fenton said. "I thought I made that very
plain. He'll go before a jury for killing Gerstle. But the other two
slayings—"
"For Pete's sake, Lieutenant," Macklin said, leaning a little forward
and giving him no time to finish. "You've arrested the owner of that
motor cruiser, and he turns out to be someone Helen Lathrup had
known for five years. Not to mince words, John Darby had been her
lover for that length of time, if not longer. They quarreled and she
broke off with him a half-dozen times. But just recently the fire
started burning again, for both of them. He practically admits all that,
because I guess he knows he'd gain nothing by denying it. What he
doesn't admit is that they quarreled again even more recently."
Macklin paused an instant, as if to emphasize the importance of
what he was about to say. Then he went on in a tone of absolute
conviction: "Isn't it all pretty obvious? When the fires started burning
again she sat on Gerstle's story, wouldn't let him run it. It was big,
but that didn't matter. She was determined to protect Darby. Then it
stands to reason they must have quarreled again. You'll never get
him to admit that, because it supplies the strongest kind of motive for
murder and would be the equivalent of a complete confession. They
quarreled again and she threatened to give Gerstle the green light,
and that's why he killed her."
"A pretty drastic way of making sure," Fenton said. "Why didn't he try
making up again with her?"
"He probably did and got nowhere. When once her mind was made
up, it usually stayed that way. Really made up, I mean. She might
have quarreled with Darby off and on for years, enjoyed letting him
dangle, but this time she probably turned absolutely venomous.
"Consider what kind of man he is. That's important, too. A cafe
society procurer—a flesh-for-sale racketeer with a capital-gains
league clientele—big names in Hollywood and the TV industry, not to
mention the magazine field. Consider how far that kind of upper-
echelon pimp would go if she thought she was about to blow his five-
million-a-year racket sky high, and himself along with it? Of course
Darby killed her."
"And Ruth Porges too?" Fenton asked.
Macklin nodded. "She worked here, didn't she? I knew that Gerstle
was working on something big which she didn't want him to publish,
and she may have found out more than either Hansen or I knew or
suspected. She may have found out too much for him to let her go
on living. It all hangs together, doesn't it? It would be the wildest kind
of coincidence if Darby just murdered Gerstle and someone else
killed Helen Lathrup and Ruth Porges."
"I'm afraid I can't agree," Fenton said quietly. "It would be the wildest
kind of coincidence if the three crimes were not closely related. But
they were, very closely, even if a different person committed the first
two and for an entirely different reason."
"And I'm afraid I can't follow your line of reasoning," Macklin said. He
grinned suddenly. "I don't know why I should be raising problems for
the police to worry about, when you're so convinced that you know
just who did murder Helen Lathrup, and why!"
"We know," Fenton said.
"Then why don't you arrest him then?"
"We intend to," Fenton said. "But first I'd like to point out the flaw in
your line of reasoning. You've just said the flame started up again
between Helen Lathrup and Darby, and she sat on Gerstle's exposé.
That we know—Darby, as you say, has practically admitted it,
because he knows it would remove any motive he might have had
for killing Helen Lathrup. He'd have a very special reason for wanting
her to stay alive. And he did want her to stay alive, you can be sure
of that."
"Not if she quarreled with him again still later," Macklin said.
"Just have patience," Fenton said. "I'm coming to that. What makes
you so sure she quarreled with him again, in a deadly serious way
this time, and was going to expose him? We haven't uncovered a
shred of evidence pointing in that direction. It just possibly might be
true, because she was a quarrelsome woman, but even if it were
true, he didn't kill her for that, or any other reason. He didn't kill her,
period. If they had another quarrel, I rather suspect he'd have known
how to talk himself back into her good graces again. So the whole
quarrel motive is pure assumption on your part."
"A very plausible assumption," Macklin pointed out.
"Under ordinary circumstances it would be," Fenton conceded. "But it
carries no weight at all now, because we know who murdered Helen
Lathrup and—well, we can make a pretty good guess as to why he
did it, even if we're not absolutely sure about his motive."
Fenton nodded, his lips tightening a little. "By the same token, we
know that Darby couldn't have murdered her. He happens to have an
unbreakable alibi for that particular morning, and the two hoodlums
who kidnapped Hansen have just as good an alibi, and they were the
only professional, gun-carrying characters in his employ, as far as
we've been able to determine. But Darby couldn't have murdered her
anyway—because someone else did. Someone else went into her
office and shot her dead and it was that which started the fire under
Darby. With Lathrup dead, Darby had no longer a beautiful,
protective, guardian angel—or crime-blinking witch, if you'd prefer
that term—to keep the exposé under wraps, and Gerstle would have
had a field day. And Gerstle was going ahead with it, not telling
anyone, not even Mr. Eaton or Hansen here ... although Darby made
the mistake of thinking Hansen did know and had him kidnapped
also, to silence him."
Macklin shook his head. "I'm sorry, Lieutenant. I just can't buy it. I'm
willing to be convinced, of course, if you've strong evidence to
support it. But otherwise—"
"We've the strongest kind of evidence," Fenton said, "short of what
an eye-witness would be able to tell us. No jury could listen to expert
testimony regarding it, and stay locked up for more than ten or fifteen
minutes."
He paused for a moment to stare around the office. His gaze
lingered for an instant on Ellers and it seemed to Lynn Prentiss—up
to that moment her own gaze had rested on Macklin and the
detective—that the elderly editor paled slightly. But Hansen and
Eaton appeared ill at ease too, the publisher extremely so, and even
Susan Weil grew a little restive under the big detective's prolonged
scrutiny.
It was to Susan that he spoke. "The switchboard keeps you pretty
busy, doesn't it, Miss Weil," he said. "If someone came out of one of
the offices on this side of the reception desk and darted past you—
I'm using that word deliberately because it describes the way he
probably moved—it would be quite possible for him to reach Miss
Lathrup's office on the other side of the desk-division unobserved, I
should imagine. Your back would be turned, because the
switchboard faces toward the outer door. Isn't that so?"
Susan wet her lips before replying. But her answer was decisive
enough. "Yes, sir. I'm quite sure I wouldn't have seen him, if I was
plugging in a call."
"Then, if he'd darted back again a few moments later, it stands to
reason you wouldn't have seen him then either ... if you happened to
be answering another call or even—if you just happened to have
your back turned. It would have been very easy for him to pause an
instant before darting back to make sure that your back would be
turned."
Susan nodded. "That's right, sir. When someone comes into the
outer office I usually see him right away, because I just have to
glance sideways. But I seldom turn around and look behind me,
unless an editor speaks to me. Editors pass back and forth all the
time and even when I hear them I seldom turn around."
"I see," Fenton said. "That would have cut down the risks for him,
made it even easier. But I imagine he did his best to dart past as
quickly and silently as possible. Did you hear the sound of the
silenced gun, Miss Weil?"
"Yes, I'm sure I did. But it didn't make much of an impression on me,
until Miss Prentiss came out a minute or two later and told me that
she'd heard it also. It seemed to trouble her a great deal."
"Yes, well—that's all I wanted to know."
He looked directly at Lynn. "How long was it, Miss Prentiss, before
you got up, after hearing the sound, and went to investigate? I
questioned you about that on the day of the murder, but it wasn't of
such vital importance then. I mean—it didn't seem so to us at the
time. It should have, and I blame myself for it. Please try to think
back again, and narrow it down as much as you can. A half-minute,
a minute and a half?"
"It's hard to be sure," Lynn said, wetting her lips as Susan had done.
"I should say—about one minute. I remember that I just sat staring at
the manuscript I'd been working on, blue-penciled mentally a third of
a page, in fact. Then my curiosity got the better of me—"
She sighed helplessly. "I really don't know. It could have been as
long as two or three full minutes."
"Time enough for the murderer to get back into his office on this side
of the switchboard, if he moved quickly."
"Yes, I should think so."
"That was not a question, Miss Lynn. It was a statement. For my
part, I'm sure of it ... in view of what you've just said. It could easily
have been all of three minutes. Time enough—and to spare."
"Yes...." Lynn murmured, moistening her lips again.
"Just one more question, Miss Prentiss. The typewriters you said you
heard clattering away. If one of them had stopped for as long as ten
minutes, would you have realized it had stopped—if it started up
again the instant you stepped out of your office? Think now. Several
typewriters, a great deal of sound, and you were engrossed in your
editing. Even if it had been the typewriter in the office next to yours
—"
That too, Lynn was to realize later, had been more of a statement
than a question, for Fenton did not even wait for her to reply.
He looked directly at Macklin and said: "It's your typewriter I'm talking
about, I'm afraid. You didn't know she'd heard the sound of your
silenced gun, but the instant she stepped out into the hall some
instinct warned you that you'd have nothing to lose by battering away
on your machine again. It would certainly help to make her believe
you hadn't left this office at all—not even long enough to dart past
the reception desk, shoot Helen Lathrup through the head and dart
straight back again."
Macklin paled visibly, but not a muscle of his face moved. He sat
very quietly returning Fenton's accusing stare, a strangely withdrawn
look in his eyes, as if he had half-anticipated exposure and had
steeled himself to endure the agony of it, if it came, by erecting a
kind of mental block within himself.
"We found the gun you killed her with in Ruth Porges' apartment,"
Fenton said, not unkindly—he could never bring himself to speak
without compassion to a man who was certain to die. "We'll never
know where she found it, unless you tell us, but we don't have to
know. Ballistics has identified it as the murder gun. You searched her
apartment after you strangled her, even tore apart two mattresses in
your search. But you didn't look inside the flushing compartment of
the toilet. It would have been so easy for you to go into the
bathroom, lift the lid and look. But I guess you just didn't think of it.
"It's your gun, Macklin. A war souvenir gun with a long black barrel,
the kind of gun some men, with your kind of war record, like to show
to friends. Possibly you showed it to her once, but that's also
something we'll never know unless you tell us. But she must have
known it was your gun or she would not have attempted to hide it. If
she'd found a stranger's gun—the gun of someone she had no
reason to respect or like or want to protect, she'd have gone straight
to the police with it. She must have felt you were justified in killing
Helen Lathrup.
"If you did show the gun to her once—that wouldn't have prevented
you from using it. You had no way of knowing she'd find it and that it
would be traced to you. War souvenir guns are often very hard to
trace to their owners, but we had very little trouble tracing this one....
You're tagged with it, Macklin. You're also tagged with a fingerprint
you left in Ruth Porges' apartment. I imagine you wore gloves and
were very careful, but not careful enough. Remember taking one
glove off for a moment? Well ... it's not too important, so long as we
have that one very fine print."
Fenton sighed and his voice hardened a little. "Would you like to tell
us why you killed her, Macklin? I must warn you, though, that
anything you say now—"
Macklin seemed not to hear him. He spoke softly, almost gently, as if
the violence which had taken two human lives had been long since
spent.
"There are two kinds of men in this world—leaving abnormality out of
it. One kind, I think, is very rare. The old saying: 'Love is a woman's
whole life—to men a thing apart' isn't always true. There are men to
whom love is everything. I have always been ... that kind of man.
And when she betrayed the great love I had for her, as she betrayed
the others, she—"
A look of torment came into his eyes. "It would have been better if
she had been the one to slay, to kill me then without compassion and
without remorse. But that was one cruelty she was incapable of, and
so I had no choice...."
"Every man has a choice, Macklin," Fenton said. "There was no
need—"
Fenton was later to regret that he had not been more careful, not
stayed more alert and on guard. But when a man does not in the
least resemble a killer in his outer aspect, when he can grin boyishly,
and disarm everyone with his blunt forthrightness, his wry humor, his
complete absence of even the everyday, garden-variety kind of
neuroticism which afflicts nine men and women out of ten—when, in
short, he seems more robustly wholesome, normal than a football
player with a well-rounded love life, it is very easy for a man to go a
little astray emotionally and assume that he can't be too dangerous
in an immediate way.
Fenton had not realized that Gallison was standing so close to
Macklin's desk, facing away from Macklin and that the police positive
on Gallison's hip could be a very formidable weapon in the hands of
killer still bent on saving his skin.
The realization came a split second too late. Macklin had reached for
the gun, whipped it from its holster and was gripping it firmly before
Gallison could swing about. And swinging about did Gallison no
good, for by that time he was weaponless.
Macklin snapped off the safety catch and fired twice. The first bullet
struck Gallison in the right thigh, wrenching a groan from him, and
dropping him to his knees. Blood spurted, spraying out over the floor.
The second shot, also aimed at Gallison, missed its mark. It went
wild, causing Lynn Prentiss to cry out and clutch at her side. Beneath
her fingers a dull stain grew. She swayed a little, staggered toward
the desk and clung to it, supporting herself with one hand, staring at
Fenton in mute appeal.
Fenton had his own gun out now and was taking careful aim at
Macklin, who was heading for the door. But before he could fire
Macklin was out of the office and was racing down the hall,
Gallison's gun still in his clasp.
Macklin reached the reception desk, and started toward the door of
the outer office. But he never reached the door. A tall, very thin
young man with a sheaf of drawings under his arm had just entered
the outer office and he had heard Lynn's stricken cry.
He was all very confused about everything. But the cry did
something to him, because he recognized Lynn's voice and realized
instantly that something quite terrible had happened.
And when he saw the distraught-looking man come rushing out of
the corridor, with a gun in his hand, he acted on impulse, threw out
one leg and tripped the man up, sending him sprawling. A moment
later, while the man with the gun was still sprawling, another very big
strong-looking man with another gun emerged from the corridor,
piled on top of the man he'd tripped and clobbered him over the
head with the butt of the gun until he gave up trying to rise.
It was all very confusing and hard to understand.
Chapter X
He had entered the hospital room so quietly that Lynn Prentiss was
unaware that she was not alone—the nurse had left fifteen minutes
before—until he was standing by the bed with a sheaf of drawings
under arm and the strangest, oddest assortment of yellow flowers in
the other she had ever seen.
"You'll have to turn on the light to look at these, I guess," he said and
she didn't know at first whether he was referring to the flowers or the
drawings. But when she switched on the light directly over the bed
she saw that it was one of the drawings which he was extending
toward her. The flowers he was holding a little awkwardly, not even
venturing to offer them to her, as if he wasn't quite sure that she
would approve of his taste in flowers or would not think him over-
presumptuous.
Since he seemed to want her to look at the drawings first, she did so,
studying them carefully as he passed them to her one by one.
"Well," he said, when she had remained for a long while silent. "What
do you think."
She sat up very straight, took the flowers from him and pressed them
to her nostrils, looking at him very steadily and with a strange
warmth in her eyes—a warmth that made him return her gaze
incredulously.
"Yes," she said.
"The flowers aren't too bad, is that what you're trying to tell me? But
my drawings—"
"No," she said. "That isn't what I'm trying to tell you. I like both the
flowers and the drawings. But these new drawings—well, they're a
little on the terrific side, if you don't mind my saying so. And they are
saleable. I can guarantee it. I can't do too much to help, because we
only need six more drawings this month, and six sales to the Eaton-
Lathrup publications will do no more than start you off. But when
we've published six, I'm sure you won't have any difficulty in selling
the rest to other publications."
"It went against the grain," he said. "I don't quite know why I did it."
She continued to look at him, and the warmth in her eyes told him
why, but it took him quite a long while to grasp it.
It was over and they'd released him. Ralph Gilmore still had to
appear in court on an illegal firearms possession charge, but that big
detective he'd disliked so much at first had assured him that the
worst he'd get would be a suspended sentence.
The law was designed, apparently, to discourage gangsters from
carrying weapons—although it didn't always work that way—and a
young writer, without a criminal record, would be shown a great deal
of leniency.
Everyone, even judges, expected writers to behave a little strangely,
a little differently from other people and that would count in his favor.
He hoped the big detective was right about that. It would worry him
and keep him awake nights until the ordeal was over, because just
the thought of appearing in court to answer a quite serious charge
terrified him.
No reason why it should now, he told himself. He'd had the book
thrown at him, hadn't he? He'd been booked at a police station,
taken into court, fingerprinted and confined in a cell for more than a
week. If he could survive that, he could survive anything.
And the way they'd questioned him, in a room without windows and
a bright light flooding down—Not the third degree really, nothing as
bad as that. But it had been bad enough.
There was a tap on the door and he looked up quickly.
"Who is it?" he demanded.
"It's me—Nora. I heard about all what happened to you. I wanted to
die myself, Ralph—I swear it."
He arose slowly, went to the door and opened it. "I told your mother
about it," he said. "I went to that office with the intention of killing her.
But at the last moment—I couldn't do it. I went there on the very
morning of the murder. And I bought a gun—"
"I know, Ralph ... I know, darling. It's painful to talk about and there's
no reason why you should, now. You know I love you."
"Yes ... and I love you, Nora. I must have been crazy not to realize it
sooner."
"Don't blame yourself too much, Ralph. She must have been a very
beautiful woman."
"Well—"
"Please don't worry about it, Ralph. It's all over now, done with. Don't
even think about it."
"I'm trying very hard not to. But it isn't so easy—"
"Kiss me, Ralph. Kiss me and take me into your arms and make
passionate love to me, like you did once."
"We're going to get married, Nora. You know that, don't you? We'll
go right down to City Hall tomorrow, and apply for a marriage
license."
"You don't have to marry me, Ralph. You don't—"
"But I want to. Don't you understand, you little fool? I want to and I'll
do it ... if I have to drag you all the way downtown by the hair!"
"Ralph...."
"Yes?"
"Oh, Ralph, darling—"
Fenton had slept so soundly for ten hours his wife had to tug at his
arm three times to awaken him.
"Joseph," she said. "The hospital just phoned again. Gallison is
completely out of danger. You said he was last night, but I could see
you were still a little worried."
Fenton came wide awake in an instant. "They're absolutely sure
about it, eh?"
"Yes, the bullet's out and he'll be reporting back for duty in two
weeks. He told them to tell you that."
"Is he crazy?" Fenton grumbled. "He rates a month's sick leave, at
least."
"I'm just repeating the message he asked them to give you."
Fenton sighed and rolled over on his side. "I always thought he was
a little crazy. Well ... it's good news, anyway. Will you please go
away now."
"But why? Breakfast is—"
"Never mind about my breakfast," Fenton said, drawing the sheets
up over his head. "I'm not anything like as crazy as Gallison is. I'm
going to sleep for a month."
*** END OF THE PROJECT GUTENBERG EBOOK THE HORROR
EXPERT ***
Updated editions will replace the previous one—the old editions will
be renamed.
Creating the works from print editions not protected by U.S.

copyright law means that no one owns a United States copyright in
these works, so the Foundation (and you!) can copy and distribute it
in the United States without permission and without paying copyright
royalties. Special rules, set forth in the General Terms of Use part of
this license, apply to copying and distributing Project Gutenberg™
electronic works to protect the PROJECT GUTENBERG™ concept
and trademark. Project Gutenberg is a registered trademark, and
may not be used if you charge for an eBook, except by following the
terms of the trademark license, including paying royalties for use of
the Project Gutenberg trademark. If you do not charge anything for
copies of this eBook, complying with the trademark license is very
easy. You may use this eBook for nearly any purpose such as
creation of derivative works, reports, performances and research.
Project Gutenberg eBooks may be modified and printed and given
away—you may do practically ANYTHING in the United States with
eBooks not protected by U.S. copyright law. Redistribution is subject
to the trademark license, especially commercial redistribution.
START: FULL LICENSE

THE FULL PROJECT GUTENBERG LICENSE
PLEASE READ THIS BEFORE YOU DISTRIBUTE OR USE THIS WORK
To protect the Project Gutenberg™ mission of promoting the free

distribution of electronic works, by using or distributing this work (or
any other work associated in any way with the phrase “Project
Gutenberg”), you agree to comply with all the terms of the Full
Project Gutenberg™ License available with this file or online at
www.gutenberg.org/license.
Section 1. General Terms of Use and

Redistributing Project Gutenberg™
electronic works
1.A. By reading or using any part of this Project Gutenberg™
electronic work, you indicate that you have read, understand, agree
to and accept all the terms of this license and intellectual property
(trademark/copyright) agreement. If you do not agree to abide by all
the terms of this agreement, you must cease using and return or
destroy all copies of Project Gutenberg™ electronic works in your
possession. If you paid a fee for obtaining a copy of or access to a
Project Gutenberg™ electronic work and you do not agree to be
bound by the terms of this agreement, you may obtain a refund from
the person or entity to whom you paid the fee as set forth in
paragraph 1.E.8.
1.B. “Project Gutenberg” is a registered trademark. It may only be

used on or associated in any way with an electronic work by people
who agree to be bound by the terms of this agreement. There are a
few things that you can do with most Project Gutenberg™ electronic
works even without complying with the full terms of this agreement.
See paragraph 1.C below. There are a lot of things you can do with
Project Gutenberg™ electronic works if you follow the terms of this
agreement and help preserve free future access to Project
Gutenberg™ electronic works. See paragraph 1.E below.
1.C. The Project Gutenberg Literary Archive Foundation (“the
Foundation” or PGLAF), owns a compilation copyright in the
collection of Project Gutenberg™ electronic works. Nearly all the
individual works in the collection are in the public domain in the
United States. If an individual work is unprotected by copyright law in
the United States and you are located in the United States, we do
not claim a right to prevent you from copying, distributing,
performing, displaying or creating derivative works based on the
work as long as all references to Project Gutenberg are removed. Of
course, we hope that you will support the Project Gutenberg™
mission of promoting free access to electronic works by freely
sharing Project Gutenberg™ works in compliance with the terms of
this agreement for keeping the Project Gutenberg™ name
associated with the work. You can easily comply with the terms of
this agreement by keeping this work in the same format with its
attached full Project Gutenberg™ License when you share it without
charge with others.
1.D. The copyright laws of the place where you are located also
govern what you can do with this work. Copyright laws in most
countries are in a constant state of change. If you are outside the
United States, check the laws of your country in addition to the terms
of this agreement before downloading, copying, displaying,
performing, distributing or creating derivative works based on this
work or any other Project Gutenberg™ work. The Foundation makes
no representations concerning the copyright status of any work in
any country other than the United States.
1.E. Unless you have removed all references to Project Gutenberg:
1.E.1. The following sentence, with active links to, or other

immediate access to, the full Project Gutenberg™ License must
appear prominently whenever any copy of a Project Gutenberg™
work (any work on which the phrase “Project Gutenberg” appears, or
with which the phrase “Project Gutenberg” is associated) is
accessed, displayed, performed, viewed, copied or distributed:
This eBook is for the use of anyone anywhere in the United
States and most other parts of the world at no cost and with
almost no restrictions whatsoever. You may copy it, give it away
or re-use it under the terms of the Project Gutenberg License
included with this eBook or online at www.gutenberg.org. If you
are not located in the United States, you will have to check the
laws of the country where you are located before using this
eBook.
1.E.2. If an individual Project Gutenberg™ electronic work is derived

from texts not protected by U.S. copyright law (does not contain a
notice indicating that it is posted with permission of the copyright
holder), the work can be copied and distributed to anyone in the
United States without paying any fees or charges. If you are
redistributing or providing access to a work with the phrase “Project
Gutenberg” associated with or appearing on the work, you must
comply either with the requirements of paragraphs 1.E.1 through
1.E.7 or obtain permission for the use of the work and the Project
Gutenberg™ trademark as set forth in paragraphs 1.E.8 or 1.E.9.
1.E.3. If an individual Project Gutenberg™ electronic work is posted

with the permission of the copyright holder, your use and distribution
must comply with both paragraphs 1.E.1 through 1.E.7 and any
additional terms imposed by the copyright holder. Additional terms
will be linked to the Project Gutenberg™ License for all works posted
with the permission of the copyright holder found at the beginning of
this work.
1.E.4. Do not unlink or detach or remove the full Project

Gutenberg™ License terms from this work, or any files containing a
part of this work or any other work associated with Project
Gutenberg™.
1.E.5. Do not copy, display, perform, distribute or redistribute this

electronic work, or any part of this electronic work, without
prominently displaying the sentence set forth in paragraph 1.E.1 with
active links or immediate access to the full terms of the Project
Gutenberg™ License.
1.E.6. You may convert to and distribute this work in any binary,
compressed, marked up, nonproprietary or proprietary form,
including any word processing or hypertext form. However, if you
provide access to or distribute copies of a Project Gutenberg™ work
in a format other than “Plain Vanilla ASCII” or other format used in
the official version posted on the official Project Gutenberg™ website
(www.gutenberg.org), you must, at no additional cost, fee or expense
to the user, provide a copy, a means of exporting a copy, or a means
of obtaining a copy upon request, of the work in its original “Plain
Vanilla ASCII” or other form. Any alternate format must include the
full Project Gutenberg™ License as specified in paragraph 1.E.1.
1.E.7. Do not charge a fee for access to, viewing, displaying,

performing, copying or distributing any Project Gutenberg™ works
unless you comply with paragraph 1.E.8 or 1.E.9.
1.E.8. You may charge a reasonable fee for copies of or providing

access to or distributing Project Gutenberg™ electronic works
provided that:
• You pay a royalty fee of 20% of the gross profits you derive from
the use of Project Gutenberg™ works calculated using the
method you already use to calculate your applicable taxes. The
fee is owed to the owner of the Project Gutenberg™ trademark,
but he has agreed to donate royalties under this paragraph to
the Project Gutenberg Literary Archive Foundation. Royalty
payments must be paid within 60 days following each date on
which you prepare (or are legally required to prepare) your
periodic tax returns. Royalty payments should be clearly marked
as such and sent to the Project Gutenberg Literary Archive
Foundation at the address specified in Section 4, “Information
about donations to the Project Gutenberg Literary Archive
Foundation.”
• You provide a full refund of any money paid by a user who

notifies you in writing (or by e-mail) within 30 days of receipt that
s/he does not agree to the terms of the full Project Gutenberg™
License. You must require such a user to return or destroy all
copies of the works possessed in a physical medium and
discontinue all use of and all access to other copies of Project
Gutenberg™ works.
• You provide, in accordance with paragraph 1.F.3, a full refund of

any money paid for a work or a replacement copy, if a defect in
the electronic work is discovered and reported to you within 90
days of receipt of the work.
• You comply with all other terms of this agreement for free
distribution of Project Gutenberg™ works.
1.E.9. If you wish to charge a fee or distribute a Project Gutenberg™

electronic work or group of works on different terms than are set
forth in this agreement, you must obtain permission in writing from
the Project Gutenberg Literary Archive Foundation, the manager of
the Project Gutenberg™ trademark. Contact the Foundation as set
forth in Section 3 below.
1.F.
1.F.1. Project Gutenberg volunteers and employees expend

considerable effort to identify, do copyright research on, transcribe
and proofread works not protected by U.S. copyright law in creating
the Project Gutenberg™ collection. Despite these efforts, Project
Gutenberg™ electronic works, and the medium on which they may
be stored, may contain “Defects,” such as, but not limited to,
incomplete, inaccurate or corrupt data, transcription errors, a
copyright or other intellectual property infringement, a defective or
damaged disk or other medium, a computer virus, or computer
codes that damage or cannot be read by your equipment.
1.F.2. LIMITED WARRANTY, DISCLAIMER OF DAMAGES - Except

for the “Right of Replacement or Refund” described in paragraph
1.F.3, the Project Gutenberg Literary Archive Foundation, the owner
of the Project Gutenberg™ trademark, and any other party
distributing a Project Gutenberg™ electronic work under this
agreement, disclaim all liability to you for damages, costs and
expenses, including legal fees. YOU AGREE THAT YOU HAVE NO
REMEDIES FOR NEGLIGENCE, STRICT LIABILITY, BREACH OF
WARRANTY OR BREACH OF CONTRACT EXCEPT THOSE
PROVIDED IN PARAGRAPH 1.F.3. YOU AGREE THAT THE
FOUNDATION, THE TRADEMARK OWNER, AND ANY
DISTRIBUTOR UNDER THIS AGREEMENT WILL NOT BE LIABLE
TO YOU FOR ACTUAL, DIRECT, INDIRECT, CONSEQUENTIAL,
PUNITIVE OR INCIDENTAL DAMAGES EVEN IF YOU GIVE
NOTICE OF THE POSSIBILITY OF SUCH DAMAGE.
1.F.3. LIMITED RIGHT OF REPLACEMENT OR REFUND - If you

discover a defect in this electronic work within 90 days of receiving it,
you can receive a refund of the money (if any) you paid for it by
sending a written explanation to the person you received the work
from. If you received the work on a physical medium, you must
return the medium with your written explanation. The person or entity
that provided you with the defective work may elect to provide a
replacement copy in lieu of a refund. If you received the work
electronically, the person or entity providing it to you may choose to
give you a second opportunity to receive the work electronically in
lieu of a refund. If the second copy is also defective, you may
demand a refund in writing without further opportunities to fix the
problem.
1.F.4. Except for the limited right of replacement or refund set forth in
paragraph 1.F.3, this work is provided to you ‘AS-IS’, WITH NO
OTHER WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR ANY PURPOSE.
1.F.5. Some states do not allow disclaimers of certain implied

warranties or the exclusion or limitation of certain types of damages.
If any disclaimer or limitation set forth in this agreement violates the
law of the state applicable to this agreement, the agreement shall be
interpreted to make the maximum disclaimer or limitation permitted
by the applicable state law. The invalidity or unenforceability of any
provision of this agreement shall not void the remaining provisions.
1.F.6. INDEMNITY - You agree to indemnify and hold the
Foundation, the trademark owner, any agent or employee of the
Foundation, anyone providing copies of Project Gutenberg™
electronic works in accordance with this agreement, and any
volunteers associated with the production, promotion and distribution
of Project Gutenberg™ electronic works, harmless from all liability,
costs and expenses, including legal fees, that arise directly or
indirectly from any of the following which you do or cause to occur:
(a) distribution of this or any Project Gutenberg™ work, (b)
alteration, modification, or additions or deletions to any Project
Gutenberg™ work, and (c) any Defect you cause.
Section 2. Information about the Mission of

Project Gutenberg™
Project Gutenberg™ is synonymous with the free distribution of
electronic works in formats readable by the widest variety of
computers including obsolete, old, middle-aged and new computers.
It exists because of the efforts of hundreds of volunteers and
donations from people in all walks of life.
Volunteers and financial support to provide volunteers with the

assistance they need are critical to reaching Project Gutenberg™’s
goals and ensuring that the Project Gutenberg™ collection will
remain freely available for generations to come. In 2001, the Project
Gutenberg Literary Archive Foundation was created to provide a
secure and permanent future for Project Gutenberg™ and future
generations. To learn more about the Project Gutenberg Literary
Archive Foundation and how your efforts and donations can help,
see Sections 3 and 4 and the Foundation information page at
www.gutenberg.org.
Section 3. Information about the Project

Gutenberg Literary Archive Foundation
The Project Gutenberg Literary Archive Foundation is a non-profit
501(c)(3) educational corporation organized under the laws of the
state of Mississippi and granted tax exempt status by the Internal
Revenue Service. The Foundation’s EIN or federal tax identification
number is 64-6221541. Contributions to the Project Gutenberg
Literary Archive Foundation are tax deductible to the full extent
permitted by U.S. federal laws and your state’s laws.
The Foundation’s business office is located at 809 North 1500 West,

Salt Lake City, UT 84116, (801) 596-1887. Email contact links and up
to date contact information can be found at the Foundation’s website
and official page at www.gutenberg.org/contact
Section 4. Information about Donations to

the Project Gutenberg Literary Archive
Foundation
Project Gutenberg™ depends upon and cannot survive without
widespread public support and donations to carry out its mission of
increasing the number of public domain and licensed works that can
be freely distributed in machine-readable form accessible by the
widest array of equipment including outdated equipment. Many small
donations ($1 to $5,000) are particularly important to maintaining tax
exempt status with the IRS.
The Foundation is committed to complying with the laws regulating

charities and charitable donations in all 50 states of the United
States. Compliance requirements are not uniform and it takes a
considerable effort, much paperwork and many fees to meet and
keep up with these requirements. We do not solicit donations in
locations where we have not received written confirmation of
compliance. To SEND DONATIONS or determine the status of
compliance for any particular state visit www.gutenberg.org/donate.
While we cannot and do not solicit contributions from states where

we have not met the solicitation requirements, we know of no

Textbook Metaheuristics For Data Clustering and Image Segmentation Meera Ramadas Ebook All Chapter PDF

Uploaded by

Copyright:

Available Formats

Textbook Metaheuristics For Data Clustering and Image Segmentation Meera Ramadas Ebook All Chapter PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Textbook Metaheuristics For Data Clustering and Image Segmentation Meera Ramadas Ebook All Chapter PDF

Uploaded by

Copyright:

Available Formats

Metaheuristics for Data Clustering and

Image Segmentation Meera Ramadas

Recent Advances in Hybrid Metaheuristics for Data

Metaheuristic Algorithms for Image Segmentation Theory

Text Segmentation and Recognition for Enhanced Image

Modern Technologies for Big Data Classification and

Metaheuristics for Medicine and Biology 1st Edition

Adaptive Resonance Theory in Social Media Data

Exam Preparatory Manual for Undergraduates Pathology

Multisensor Image Fusion and Data Mining for

Lakhmi C. Jain, Faculty of Engineering and Information Technology, Centre for

More information about this series at http://www.springer.com/series/8578

Metaheuristics for Data

ISSN 1868-4394 ISSN 1868-4408 (electronic)

Library of Congress Control Number: 2018961197

© Springer Nature Switzerland AG 2019

Metaheuristic algorithms like genetic algorithms (GAs), differential evolution (DE),

Manama, Bahrain Meera Ramadas

3 Revised Mutation Strategy for Differential Evolution

6.5 Multi-level Thresholding Using RDE Strategy . . . . . . . . . . . . . . . 129

© Springer Nature Switzerland AG 2019 1

algorithm. There are two types of optimality approach. In landscape-based optimal-

1.1 Major Objectives

2.1 Genetic Algorithm

2.1.1 Basics of Genetic Algorithms

for optimal combination of solutions. The three important aspects in GA include

Initialize population with random candidate solution

Evaluate fitness of each candidate solution

While termination condition not true do

Select the best individuals to be used by genetic operators

Generate new individuals using crossover and mutation

Evaluate fitness of new individuals

Replace the worst individuals of population by best new individuals

Selection Crossover Mutation

Fig. 2.1 Flow diagram for genetic algorithm

2.2 Particle Swarm Optimization Algorithm

vi,k+1 vi,k + c1 · rand (0, 1) · (pbesti − presenti )

Initialize velocity and position of each particles

2.3 Differential Evolution

Differential Evolution (DE) is a heuristic optimization technique proposed by Storn

Fig. 2.2 Flow diagram for

Evaluate individual fitness and

Generate new population

crossover. In selection, genetic algorithm selects parent based on fitness while DE

DE is a parallel direct search method using a population of N parameter vectors

Vi,G Xr1,G + F · (Xr2,G − Xr3,G ) (2.3)

DE/rand/2 Vi,G Xr1,G + F · (Xr2,G − Xr3,G ) + F · (Xr4,G − Xr5,G ) (2.4)

where, i 1, 2 . . . NP, r1 , r2 , r3 ∈ {1, 2, . . . , NP} are randomly selected and r1 

The algorithm for Differential Evolution approach is given below:

2.4 Flower Pollination Algorithm

Input vectors of candidate solution

Calculate objective function

Perform mutation, crossover and selection

Print best vector Stop

Fig. 2.3 Differential Evolution algorithm

Fig. 2.4 Pollination in flowers (Source https://biology.tutorvista.com/)

xit+1 xit + ε(xit − xkt ) (2.10)

Objective function . Initialize n pollen seeds

State as number of iterations and a switching probability as

where, i 1, 2 . . . NP, r1 , r2 , r3 ∈ {1, 2, . . . , NP} are randomly selected and r1