Metaheuristics in Machine Learning Theory and Applications Studies in Computational Intelligence 967 Diego Oliva (Editor) Download PDF

Full download test bank at ebookmeta.
com
Metaheuristics in Machine Learning Theory and

Applications Studies in Computational Intelligence
967 Diego Oliva (Editor)
For dowload this book click LINK or Button below
https://ebookmeta.com/product/metaheuristics-in-
machine-learning-theory-and-applications-studies-
in-computational-intelligence-967-diego-oliva-
editor/
OR CLICK BUTTON
DOWLOAD EBOOK
Download More ebooks from https://ebookmeta.com

More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Artificial Intelligence Theory and Applications 973

Studies in Computational Intelligence 973 Endre Pap
(Editor)
https://ebookmeta.com/product/artificial-intelligence-theory-and-
applications-973-studies-in-computational-intelligence-973-endre-
pap-editor/
Modern Metaheuristics in Image Processing 1st Edition

Diego Oliva Salvador Hinojosa Noe Ortega Sánchez Marco
Pérez Cisneros
https://ebookmeta.com/product/modern-metaheuristics-in-image-
processing-1st-edition-diego-oliva-salvador-hinojosa-noe-ortega-
sanchez-marco-perez-cisneros/
Data-Driven Evolutionary Optimization: Integrating

Evolutionary Computation, Machine Learning and Data
Science (Studies in Computational Intelligence, 975)
Yaochu Jin
https://ebookmeta.com/product/data-driven-evolutionary-
optimization-integrating-evolutionary-computation-machine-
learning-and-data-science-studies-in-computational-
intelligence-975-yaochu-jin/
Machine learning and artificial intelligence with

industrial applications : from big data to small data
1st Edition Diego Carou
https://ebookmeta.com/product/machine-learning-and-artificial-
intelligence-with-industrial-applications-from-big-data-to-small-
data-1st-edition-diego-carou/
Innovations in Machine and Deep Learning Case Studies
and Applications Gilberto Rivera
https://ebookmeta.com/product/innovations-in-machine-and-deep-
learning-case-studies-and-applications-gilberto-rivera/
Reliability Engineering and Computational Intelligence

(Studies in Computational Intelligence, 976) Coen Van
Gulijk (Editor)
https://ebookmeta.com/product/reliability-engineering-and-
computational-intelligence-studies-in-computational-
intelligence-976-coen-van-gulijk-editor/
Computational Intelligence in Machine Learning Select

Proceedings of ICCIML 2021 1st Edition Amit Kumar
https://ebookmeta.com/product/computational-intelligence-in-
machine-learning-select-proceedings-of-icciml-2021-1st-edition-
amit-kumar/
Swarm Intelligence and Machine Learning: Applications

in Healthcare 1st Edition Manish Gupta (Editor)
https://ebookmeta.com/product/swarm-intelligence-and-machine-
learning-applications-in-healthcare-1st-edition-manish-gupta-
editor/
Combating Fake News with Computational Intelligence

Techniques Studies in Computational Intelligence 1001
Mohamed Lahby
https://ebookmeta.com/product/combating-fake-news-with-
computational-intelligence-techniques-studies-in-computational-
intelligence-1001-mohamed-lahby/
Studies in Computational Intelligence 967
Diego Oliva
Essam H. Houssein
Salvador Hinojosa Editors
Metaheuristics
in Machine
Learning: Theory
and Applications
Studies in Computational Intelligence
Volume 967
Series Editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
The series “Studies in Computational Intelligence” (SCI) publishes new develop-
ments and advances in the various areas of computational intelligence—quickly and
with a high quality. The intent is to cover the theory, applications, and design methods
of computational intelligence, as embedded in the fields of engineering, computer
science, physics and life sciences, as well as the methodologies behind them. The
series contains monographs, lecture notes and edited volumes in computational
intelligence spanning the areas of neural networks, connectionist systems, genetic
algorithms, evolutionary computation, artificial intelligence, cellular automata, self-
organizing systems, soft computing, fuzzy systems, and hybrid intelligent systems.
Of particular value to both the contributors and the readership are the short publica-
tion timeframe and the world-wide distribution, which enable both wide and rapid
dissemination of research output.
Indexed by SCOPUS, DBLP, WTI Frankfurt eG, zbMATH, SCImago.
All books published in the series are submitted for consideration in Web of Science.
More information about this series at http://www.springer.com/series/7092

Diego Oliva · Essam H. Houssein ·
Salvador Hinojosa
Editors
Metaheuristics in Machine
Learning: Theory
and Applications
Editors
Diego Oliva Essam H. Houssein
Computer Sciences Department Department of Computer Science
CUCEI Faculty of Computers and Information
University of Guadalajara, Guadajalara, Minia University
Jalisco, Mexico Minia, Egypt
Salvador Hinojosa
Computer Sciences Department
CUCEI
University of Guadalajara, Guadajalara,
Jalisco, Mexico
ISSN 1860-949X ISSN 1860-9503 (electronic)

Studies in Computational Intelligence
ISBN 978-3-030-70541-1 ISBN 978-3-030-70542-8 (eBook)
https://doi.org/10.1007/978-3-030-70542-8
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2021
This work is subject to copyright. All rights are solely and exclusively licensed 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, expressed 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
In recent years, metaheuristics (MHs) have become important tools for solving
hard optimization problems encountered in industry, engineering, biomedical, image
processing, as well as in the theoretical field. Several different metaheuristics exist,
and new ones are under constant development. One of the most fundamental prin-
ciples in our world is the search for an optimal state. Therefore, choosing the right
solution method for an optimization problem can be crucially important in finding the
right solutions for a given optimization problem (unconstrained and constrained opti-
mization problems). There exist a diverse range of MHs for optimization. Optimiza-
tion is an important and decisive activity in science and engineering. Engineers will
be able to produce better designs when they can save time and decrease the problem
complexity with optimization methods. Many engineering optimization problems are
naturally more complex and difficult to solve by conventional optimization methods
such as dynamic programming. In recent years, more attention has been paid to inno-
vative methods derived from the nature that is inspired by the social or the natural
systems, which have yielded outstanding results in solving complex optimization
problems. Metaheuristic algorithms are a type of random algorithm which is used
to find the optimal solutions. Metaheuristics are approximate types of optimization
algorithms that can better escape from the local optimum points and can be used in
a wide range of engineering problems.
Recently, metaheuristics (MHs) and Machine learning (ML) became a very impor-
tant and hot topic to solve real-world applications in the industrial world, science,
engineering, etc. Among the subjects to be considered are theoretical developments in
MHs; performance comparisons of MHs; cooperative methods combining different
types of approaches such as constraint programming and mathematical programming
techniques; parallel and distributed MHs for multi-objective optimization; adapta-
tion of discrete MHs to continuous optimization; dynamic optimization; software
implementations; and real-life applications. Besides, Machine learning (ML) is a
data analytics technique to use computational methods. Therefore, recently, MHs
have been combined with several ML techniques to deal with different global and
engineering optimization problems, also real-world applications. Chapters published
in the “Metaheuristics in machine learning: theory and applications (MAML2020)”
book describe original works in different topics in both science and engineering,
v
vi Preface
such as: Metaheuristics, Machine learning, Soft Computing, Neural Networks,

Multi-criteria decision making, energy efficiency, sustainable development, etc.
In short, it can be said that metaheuristic algorithms and machine learning are
advanced and general search strategies. Therefore, the main contribution of this
book is to indicate the advantages and importance of metaheuristics with machine
learning in various real-world applications.
Guadajalara, Mexico Diego Oliva

Minia, Egypt Essam H. Houssein
Guadajalara, Mexico Salvador Hinojosa
Introduction
This book “MAML2020” collects several hybridized metaheuristics (MHs) with

machine learning (ML) methods for various real-world applications. Hence, the MHs
have become essential tools for solving hard optimization problems encountered in
industry, engineering, biomedical, image processing, as well as in the theoretical field.
Besides, machine learning (ML) is a data analytics technique to use computational
methods. Therefore, recently, MHs have been combined with several ML techniques
to deal with different global and engineering optimization problems, also real-world
applications. However, this book addresses the issues of two important computer
sciences strategies: MHs and ML. The idea of combining the techniques is to improve
the performance of the original methods in different applications.
The book guides the reader along with different and exciting implementations,
but it also includes the theoretical support that permits understanding of all the ideas
presented in the chapter. Moreover, each chapter that offers applications includes
comparisons and updated references that support the results obtained by the proposed
approaches. At the same time, every chapter provides the reader with a practical guide
to go to the reference sources. The book was designed for graduate and postgrad-
uate education, where students can find support for reinforcing or as the basis for
their consolidation; researchers can polish their knowledge. Also, professors can
find support for the teaching process in areas involving machine vision or as exam-
ples related to main techniques addressed. Additionally, professionals who want to
learn and explore the advances in concepts and implementation of optimization and
machine learning-based algorithms applied to several real-world applications can
find in this book an excellent guide for such purpose.
This exciting book has 30 chapters organized considering an overview of meta-
heuristics (MHs) and machine learning (ML) methods applied to solve various real-
world applications. In this sense, Chapters 1 and 2 provide the cross-entropy-based
thresholding segmentation of magnetic resonance prostatic images and hyperpa-
rameter optimization in a convolutional neural network using metaheuristic algo-
rithms, respectively. Chapter 3 presents a diagnosis of collateral effects in climate
change through the identification of leaf damage using a novel heuristics and machine
learning framework. Chapter 4 explains the feature engineering for machine learning
and deep learning-assisted wireless communication. Chapter 5 introduces the genetic
vii
viii Introduction
operators and their impact on the training of deep neural networks. In Chapter 6,
the implementation of metaheuristics with extreme learning machines is described.
Chapter 7 presents the architecture optimization of convolutional neural networks
by microgenetic algorithms. In Chapter 8, optimizing connection weights in neural
networks using a memetic algorithm incorporating chaos theory is introduced.
Further, Chapter 9 provides a review of metaheuristic optimization algorithms
for wireless sensor networks. In Chapter 10, a metaheuristic algorithm for white
blood cell classification in healthcare informatics is presented. In Chapter 11, a
review of multi-level thresholding image segmentation using nature-inspired opti-
mization algorithms is introduced. Chapter 12 explains the hybrid Harris hawks
optimization with differential evolution for data clustering. Chapter 13 introduces
the variable mesh optimization for continuous optimization and multimodal prob-
lems. Chapter 14 provides traffic control using image processing and deep learning
techniques. Chapter 15 introduces the drug design and discovery: theory, applica-
tions, open issues, and challenges. The thresholding algorithm applied to chest X-ray
images with pneumonia is presented in Chapter 16.
Moreover, Chapter 17 presents a comprehensive review of artificial neural
networks on stock market prediction. Chapter 18 introduces the image classifica-
tion with convolutional neural networks. Chapter 19 provides the applied machine
learning techniques to find patterns and trends in bicycle-sharing systems influenced
by traffic accidents and violent events in Guadalajara, Mexico. In Chapter 20, a review
on machine reading comprehension (LSTM) is presented. Chapter 21 introduces a
survey of metaheuristic algorithms for solving optimization problems. Chapter 22
integrates metaheuristic algorithms and minimum cross-entropy for image segmen-
tation in mist conditions. Chapter 23 provides a machine learning application for
particle physics: Mexico’s involvement in the Hyper-Kamiokande observatory.
Besides, Chapter 24 provides a novel metaheuristic approach for image contrast
enhancement based on grayscale mapping. Chapter 25 presents the geospatial data
mining technique survey. In Chapter 26, an integration of Internet of things and
cloud computing for cardiac health recognition is discussed. Chapter 27 introduces
the combinatorial optimization for artificial intelligence-enabled mobile network
automation. Chapter 28 presents the performance optimization of a PID controller
based on parameter estimation using metaheuristic techniques. Chapter 29 provides
the solar irradiation change detection for photovoltaic systems through ANN trained
with a metaheuristic algorithm. Finally, in Chapter 30, the genetic algorithm-based
global and local feature selection approach for handwritten numeral recognition is
presented.
It is important to mention that an advantage of this structure is that each chapter
could be read separately. This book is an important reference for hybridized meta-
heuristics (MHs) with machine learning (ML) methods for various real-world appli-
cations. These areas are relevant and are in constant evolution. For that reason, it
Introduction ix
is hard to collect all the information in a single book. I congratulate the authors for
their effort and dedication to assembling the topics addressed in the book.
Diego Oliva
Essam H. Houssein
Salvador Hinojosa
Contents
Cross Entropy Based Thresholding Segmentation of Magnetic

Resonance Prostatic Images Using Metaheuristic Algorithms . . . . . . . . . . 1
Omar Zárate and Daniel Záldivar
Hyperparameter Optimization in a Convolutional Neural Network
Using Metaheuristic Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Angel Gaspar, Diego Oliva, Erik Cuevas, Daniel Zaldívar,
Marco Pérez, and Gonzalo Pajares
Diagnosis of Collateral Effects in Climate Change Through
the Identification of Leaf Damage Using a Novel Heuristics
and Machine Learning Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Juan Salazar, Eddy Sánchez-De La Cruz, Alberto Ochoa-Zezzatti,
Martin Montes, Roberto Contreras-Masse, and José Mejia
Feature Engineering for Machine Learning and Deep Learning
Assisted Wireless Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Vijay Kumar and Sarat Kumar Patra
Genetic Operators and Their Impact on the Training of Deep
Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
David Eliel Bocanegra Michel and Daniel Zaldivar Navarro
Implementation of Metaheuristics with Extreme Learning
Machines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Hector Escobar and Erik Cuevas
Architecture Optimization of Convolutional Neural Networks
by Micro Genetic Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Edgar Saul Marquez Casillas and Valentín Osuna-Enciso
Optimising Connection Weights in Neural Networks Using
a Memetic Algorithm Incorporating Chaos Theory . . . . . . . . . . . . . . . . . . . 169
Seyed Jalaleddin Mousavirad, Gerald Schaefer,
and Hossein Ebrahimpour-Komleh
xi
xii Contents
A Review of Metaheuristic Optimization Algorithms in Wireless

Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Essam H. Houssein, Mohammed R. Saad, Kashif Hussain,
Hassan Shaban, and M. Hassaballah
A Metaheuristic Algorithm for Classification of White Blood Cells
in Healthcare Informatics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Ana Carolina Borges Monteiro, Yuzo Iano, Reinaldo Padilha França,
and Rangel Arthur
Multi-level Thresholding Image Segmentation Based
on Nature-Inspired Optimization Algorithms: A Comprehensive
Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Essam H. Houssein, Bahaa El-din Helmy, Diego Oliva,
Ahmed A. Elngar, and Hassan Shaban
Hybrid Harris Hawks Optimization with Differential Evolution
for Data Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Laith Abualigah, Mohamed Abd Elaziz, Mohammad Shehab,
Osama Ahmad Alomari, Mohammad Alshinwan, Hamzeh Alabool,
and Deemah A. Al-Arabiat
Variable Mesh Optimization for Continuous Optimization
and Multimodal Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Jarvin A. Antón-Vargas, Luis A. Quintero-Domínguez,
Guillermo Sosa-Gómez, and Omar Rojas
Traffic Control Using Image Processing and Deep Learning
Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Rafael Baroni, Sthefanie Premebida, Marcella Martins, Diego Oliva,
Erikson Freitas de Morais, and Max Santos
Drug Design and Discovery: Theory, Applications, Open Issues
and Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
Essam H. Houssein, Mosa E. Hosney, Diego Oliva,
No Ortega-Sánchez, Waleed M. Mohamed, and M. Hassaballah
Thresholding Algorithm Applied to Chest X-Ray Images
with Pneumonia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
Jesus Murillo-Olmos, Erick Rodríguez-Esparza,
Marco Pérez-Cisneros, Daniel Zaldivar, Erik Cuevas,
Gerardo Trejo-Caballero, and Angel A. Juan
Artificial Neural Networks for Stock Market Prediction:
A Comprehensive Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
Essam H. Houssein, Mahmoud Dirar, Kashif Hussain,
and Waleed M. Mohamed
Contents xiii
Image Classification with Convolutional Neural Networks . . . . . . . . . . . . . 445

Alfonso Ramos-Michel, Marco Pérez-Cisneros, Erik Cuevas,
and Daniel Zaldivar
Applied Machine Learning Techniques to Find Patterns
and Trends in the Use of Bicycle Sharing Systems Influenced
by Traffic Accidents and Violent Events in Guadalajara, Mexico . . . . . . . 475
Adrian Barradas, Andrea Gomez-Alfaro, and Rosa-María Cantón-Croda
Machine Reading Comprehension (LSTM) Review (State of Art) . . . . . . 491
Marcos Pedroza, Alberto Ramírez-Bello, Adrián González Becerra,
and Fernando Abraham Fausto Martínez
A Survey of Metaheuristic Algorithms for Solving Optimization
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515
Essam H. Houssein, Mohamed A. Mahdy, Doaa Shebl,
and Waleed M. Mohamed
Integrating Metaheuristic Algorithms and Minimum Cross
Entropy for Image Segmentation in Mist Conditions . . . . . . . . . . . . . . . . . . 545
Mario A. Navarro, Diego Oliva, Daniel Zaldívar, and Gonzalo Pajares
Machine Learning Application for Particle Physics: Mexico’s
Involvement in the Hyper-Kamiokande Observatory . . . . . . . . . . . . . . . . . . 583
S. Cuen-Rochin, E. de la Fuente, L. Falcon-Morales,
R. Gamboa Goni, A. K. Tomatani-Sanchez, F. Orozco-Luna,
H. Torres, J. Lozoya, J. A. Baeza, J. L. Flores, B. Navarro-Garcia,
B. Veliz, A. Lopez, and B. Gonzalez-Alvarez
A Novel Metaheuristic Approach for Image Contrast Enhancement
Based on Gray-Scale Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609
Alberto Luque-Chang, Itzel Aranguren, Marco Pérez-Cisneros,
and Arturo Valdivia
Geospatial Data Mining Techniques Survey . . . . . . . . . . . . . . . . . . . . . . . . . . 635
Jorge Antonio Robles Cárdenas and Griselda Pérez Torres
Integration of Internet of Things and Cloud Computing
for Cardiac Health Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
Essam H. Houssein, Ibrahim E. Ibrahim, M. Hassaballah,
and Yaser M. Wazery
Combinatorial Optimization for Artificial Intelligence Enabled
Mobile Network Automation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663
Furqan Ahmed, Muhammad Zeeshan Asghar, and Ali Imran
Performance Optimization of PID Controller Based on Parameters
Estimation Using Meta-Heuristic Techniques: A Comparative
Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 691
Mohamed Issa
xiv Contents
Solar Irradiation Changes Detection for Photovoltaic Systems

Through ANN Trained with a Metaheuristic Algorithm . . . . . . . . . . . . . . . 711
Efrain Mendez-Flores, Israel Macias-Hidalgo, and Arturo Molina
Genetic Algorithm Based Global and Local Feature Selection
Approach for Handwritten Numeral Recognition . . . . . . . . . . . . . . . . . . . . . 745
Sagnik Pal Chowdhury, Ritwika Majumdar, Sandeep Kumar,
Pawan Kumar Singh, and Ram Sarkar
Cross Entropy Based Thresholding
Segmentation of Magnetic Resonance
Prostatic Images Using Metaheuristic
Algorithms
Omar Zárate and Daniel Záldivar
1 Introduction
Medical images (MI) are important sources of information to help to detect and
diagnose illnesses and abnormalities in the human body. Magnetic Resonance Images
(MRIs) systems measure the spatial distribution of several distinct tissue-related
parameters, such as relaxation times and proton density [1]. MRI measurements are
collections of features (that is, numerical characteristics) from a spatial array that are
aggregated into multidimensional data (from a single anatomical slice).
Prostate MRIs are primarily used for the medical diagnosis of prostate diseases.
The most common varieties are:
Prostatitis: The prostate can become inflamed secondary to an infectious process,
this is known as Prostatitis, it has a cure, but it deserves long-term treatment. Its major
complication is the development of an abscess (collection of pus) in the prostate.
Benign Prostatic Hyperplasia (BPH): It is the most frequent benign tumor in the
male sex and the cause in most cases of the annoying voiding symptoms (prostatism)
that appear after the 40 years in which it develops. It can be treated with medications
or through surgical procedures (surgeries).
Prostate cancer (malignant growth): It is one of the three most frequent cancers
and a cause of death in men. It usually occurs after age 50, and its symptoms are NOT
O. Zárate (B) · D. Záldivar

División de Electrónica y Computación, CUCEI, Universidad, de Guadalajara, Av. Revolución
1500, Guadalajara, Jalisco, México
e-mail: [email protected]; [email protected]
D. Záldivar
e-mail: [email protected]
O. Zárate
Departamento de Tecnologías de La Información, Universidad, Tecnologíca de Jalisco, Luis J.
Jiménez 577, Guadalajara, Jalisco, México
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 1

D. Oliva et al. (eds.), Metaheuristics in Machine Learning: Theory and Applications,
Studies in Computational Intelligence 967,
https://doi.org/10.1007/978-3-030-70542-8_1
2 O. Zárate and D. Záldivar
different from those of benign prostate growth (BPH). Its timely detection allows its
healing to be achieved either through surgery or some other type of procedure [2].
It is essential to make a correct differential diagnosis to indicate the appropriate
treatment.
Prostate MRIs analysis is performed by experts using visual evaluations based
on their professional experience and skills. However, this type of inspection is
limited and time-consuming. Due to these limitations, computer-assisted techniques
have been developed with the objective of extracting the anatomical structures with
segmentation being the principal methodology [3].
Image segmentation consists of obtaining underlying structures to facilitate their
interpretation, obtaining borders or groups of pixels that form regions with some
property such as intensity or texture, but this approach is computationally expensive
[4].
To reduce the computational time required, metaheuristic algorithms (MA) have
been proposed to solve complex engineering problems. MA are stochastic search
algorithms that use rules or heuristics applicable to any problem to accelerate their
convergence to near-optimal solutions.
The vast majority of metaheuristic algorithms (MA) have been derived from
biological behavior systems or physical systems in nature. Each of these algorithms
has advantages and disadvantages [5].
Within the family of metaheuristic algorithms, the methods that are of interest for
research are genetic algorithms that emulate evolution by passing genes [6].
Genetic Algorithm is the most popular stochastic optimization algorithm, was
proposed to alleviate the drawbacks of the deterministic algorithms. Taking the
evolution of species as inspiration, they set out to try to solve problems, particu-
larly optimization problems, opening a wide field of research that continues to be in
force as evolutionary algorithms (Mirjalili, Moth-Flame Optimization Algorithm: A
Novel Nature-inspired Heuristic Paradigm [7].
Holland and his colleagues proposed one of the first genetic algorithms; this
algorithm is based on the adaptation theory of artificial populations of individuals,
which is inspired by the mixture of Darwin’s theories of the evolution of species
with the genetics of Mendel [8].
Optimization is the process of finding the best possible solution for a particular
problem, finding optimal values for the parameters from all the possible values to
maximize or minimize its output. The vast majority of optimization problems with
practical implications in science, engineering, economics, and business are very
complex and challenging to solve. Such problems cannot be solved in an exact
way using classical optimization methods. Under these circumstances, metaheuristic
algorithm methods have established themselves as an alternative solution (Mirjalili,
SCA: A Sine Cosine Algorithm for Solving Optimization Problems [9].
The feasibility of performing different image processing tasks, oriented towards
segmentation, has been demonstrated through the use of MA, such as image thresh-
olding. Image thresholding performs the separation of pixels in an image by consid-
ering certain intensity values (thresholds) that partition the image histogram into a
finite number of classes based on pixel values relative to thresholds [10].
Cross Entropy Based Thresholding Segmentation of Magnetic … 3
This chapter evaluates three recently published evolutionary algorithms in

multilevel thresholding with the classic thresholding criteria.
The MA applied for segmentation of prostate magnetic resonance images (MRIs)
were Moth-Flame Optimization Algorithm (MFO) that is a metaheuristic algorithm
recently published (2015) with a considerable number of citations by the scientific
community, Sine Cosine Algorithm for Solving Optimization Problems (SCA) is
another metaheuristic algorithm published in 2016. Sunflower Optimization (SFO)
Algorithm is an algorithm published in 2018, these three metaheuristic algorithms
have been used in several investigations of digital image processing, and these three
have a large number of citations in this field.
1.1 Applied Metaheuristic Algorithms
These three metaheuristic algorithms were selected to perform the experiments and
apply it to prostate magnetic resonance imaging.
• Moth-Flame Optimization Algorithm (MFO)
This algorithm is inspirited in the navigation method of moths in nature called
transverse orientation (Mirjalili, Moth-Flame Optimization Algorithm: A Novel
Nature-inspired Heuristic Paradigm [7].
• Sine Cosine Optimization Algorithm (SCA)
This algorithm fluctuates outwards or towards the best solution using a mathe-
matical model based on sine and cosine functions (Mirjalili, SCA: A Sine Cosine
Algorithm for Solving Optimization Problems [9].
• Sunflower Optimization Algorithm (SFO)
This algorithm is inspirited on sunflowers’ motion; every day, they awaken and
accompany the sun like the needles of a clock. At night, they travel the opposite
direction to wait again for their departure the next morning (Ferreira Gomes, Simões
da Cunha and Ancelotti [11].
The three metaheuristics mentioned above were applied, which seek the best
results for thresholding with 2, 3, 4, 5, 8, 16, and 32 thresholds, applied to five
prostate magnetic resonance images, the three algorithms use Minimum Cross
Entropy Thresholding as the objective function.
The proposed algorithms are evaluated in terms of quality, and statistical analysis
is presented to compare the results of these algorithms against traditional approaches
Table 1.
It was observed in the results that the MFO algorithm generated the best results,
in fitness, evaluated by the objective function and in the statistical tests. The results
were competitive between the SCA and MFO algorithms, where SCA had shorter
processing time, but which is still not the decisive factor.
This work is divided into seven sections: In the first section are explained the three
used algorithms; Moth-flame optimizer algorithm, Sine Cosine Optimization. Algo-
rithm and Sunflower Optimization Algorithm. In Sects. 2, 3, 4 and 5 are described the
Table 1 Parameters for MFO, SCA and SFO algorithms

MFO SCA SFO
Population: 60 Search agents number: 60 Number of sunflowers: 60
Number of experiments: 30 Number of experiments: 30 Number of experiments: 30
Iterations: 1000 Iterations: 1000 Iterations/generations: 1,000
Lower bound: 1 Lower bound: 1 Lower bound: 1
Upper bound: 255 Upper bound: 255 Upper bound: 255
Number of thresholds Number of thresholds Number of thresholds
nt: 2, 3, 4, 5, 8, 16, 32 nt: 2, 3, 4, 5, 8, 16, 32 nt: 2, 3, 4, 5, 8, 16, 32
Objective function: minimum Objective function: minimum Objective function: minimum
cross entropy thresholding cross entropy thresholding cross entropy thresholding
Pollination rate best
Values p: 0.05
Mortality rate, best values m:
0.1
Survival rate, best values: 1-(p
+ m)
Problem dimension: 2
Image Segmentation Using Minimum Cross Entropy. In Sect. 6 and 7 are analyzed
results obtained and the argue of final conclusions.
2 Moth-Flame Optimizer Algorithm
Moths are insects, which are similar to the family of butterflies. Basically, there are
over 160,000 various species of this insect in nature. The most interesting fact about
moths is their special navigation methods at night.
They use a mechanism called transverse orientation. In this method, a moth flies
by maintaining a fixed angle with respect to the moon, a very effective mechanism
for traveling long distances in a straight path.
The inspiration of this optimizer is the navigation method of moths in nature called
transverse orientation. The MFO algorithm mathematically models this behavior to
perform optimization (Fig. 1).
The MFO considers a spiral as the main update mechanism of moths. However,
any types of spiral can be utilized here subject to the following conditions:

S Mi , F j = Di · ebt · cos(2π t) + F j (1)
where Di indicates the distance of the ith moth for the jth flame, b is a constant for
defining the shape of the logarithmic spiral, and t is a random number in [−1, 1].
D is calculated as follows:

Di = F j − Mi (2)
Fig. 1 Moth flame

transverse orientation for
navigation
Fig. 2 Fly spirally around

the lights
where Mi indicates the ith moth, Fj indicates the jth flame, and Di indicates the
distance of the ith moth for the jth flame (Fig. 2).
The moth eventually converges towards the light. Modeled this behavior and
proposed an optimizer called Moth-Flame Optimization (MFO) algorithm [7].

N −1
moth f lameno = r ound N − l ∗ (3)
T
where l is the current number of iteration, N is the maximum number of moth flames,
and T indicates the maximum number of iterations.
Begin
Input parameters of the algorithm and the initial data
Initialize the positions of moths and evaluate their fitness values
While (the stop criterion is not satisfied or )
Update moth flame no.
OM = Fitness Function( M )
If iteration = 1
F = sort( M )
OF = sort( OM )
Else
F = sort( Mt – 1, Mt )
OF = sort ( Mt – 1, Mt )
End if
For i = 1:N
For j = 1:D
Update r and t
Calculate D with respect to the corresponding moth
End for j
End for i
End While
End
Computational complexity of the MFO algorithm depends on the number of moths

flames, number of variables, maximum number of iterations, and sorting mechanism
of flames in each iteration [7].
3 Sine Cosine Optimization Algorithm
Population-based optimization techniques create multiple initial random candidate

solutions. They are evaluated repeatedly by an objective function and improved by a
set of rules that is the kernel of an optimization technique. An optimization algorithm
combines the random solutions in the set of solutions abruptly with a high rate of
randomness to find the promising regions of the search space.
SCA creates multiple initial random candidate solutions and requires them to
fluctuate outwards or towards the best solution using a mathematical model based on
sine and cosine functions. Several random and adaptive variables also are integrated
into this algorithm to emphasize exploration and exploitation [9].
SCA uses the following position updating:

X it+1 = X it + r1 × sin(r 2) × r3 Pit − X it

X t+1 = X t + r1 × cos(r 2) × r3 P t − X t
i i i ii (4)
Fig. 3 SCA with the range in [−2, 2]
where X it is the position of the current solution in ith dimension at tth iteration,
r1 /r2 /r3 are random numbers, Pi is the position of the destination point in ith
dimension.
These two equations are combined to be used as follows:

X it+1 = X it + r1 × sin(r 2) × r3 Pit − X it , r4 < 0.5
X it+1 = (5)
X it+1 = X it + r1 × cos(r 2) × r3 Pit − X iti , r4 ≥ 0.5
where r4 is a random number in [12] (Fig. 3).
begin
InitializeSearchAgents[X];
while t<maximum number of iterations do
EvaluateSearchAgents[X];
UpdateBestSolution(P=X*);
Update ;
UpdateSearchAgentPosition;
end while
return BestSolution
end
4 Sunflower Optimization Algorithm
The cycle of a sunflower is always the same: every day, they awaken and accompany
the sun like the needles of a clock. At night, they travel in the opposite direction to
wait again for their departure the next morning.
An important nature-based metaphor for optimization can be found here; the

inverse square law radiation. The law says that the intensity of the radiation is
inversely proportional to the square of the distance; the strength of radiation reduces
in proportion to the square of the increment in the distance [13].
Then, the amount of heat Q received by each plant is given by:
P
Qi = (6)
4πri2
where P is the power of the source and ri the distance between the current best and
the plant i.
The direction of the sunflowers to the sun is:
X ∗ − Xi
→ = , i = 1, 2, . . . , n p (7)
Si ||X ∗ − X i ||
The step of the sunflowers on the direction s is calculated by:
di = γ × Pi (||X i + X i−1 ||) × ||X i + X i−1 || (8)
where γ is the constant value that defines an “inertial” displacement of the plants,
Pi (||X i + X i−1 ||) is the probability of pollination. The sunflower i pollinates with its
nearest neighbor i − 1 generating a new individual in a random position that varies
according to each distance between the flowers. That is, individuals closer to the sun
will take smaller steps in search of a local refinement while more distant individuals
will move normally. It is also necessary to restrict the maximum step given by each
individual, in order not to skip regions prone to be global minimum candidates. Here
we define the maximum step as:
||X max − X min ||

dmax = (9)
2 × N pop
where X max and X min are the upper and lower bound values, and N pop the number
of plants of the total population.
The new plantation will be:
→ X i+1 =→ X i +di × → Si (10)
Sunflower Optimization Algorithm initially generates a uniform/random popula-

tion of n flowers find the sun (best solution s ∗ ). In the initial population orient all
plants toward the sun:
Begin
while (k < MaxDays)
Calculate Calculate the orientation vector for each plant
Remove m (%) plants further away from the sun
Calculate the step for each plant
Best b plants will pollinate around the sun
Evaluate the new individuals
If a new individual is a global best, update the sun
end while
Best solution found
End
Ferreira Gomes, Simões da Cunha and Ancelotti [14].
5 Image Segmentation Using Minimum Cross Entropy
The main purpose of thresholding is to determine the best thresholds that divide
pixel intensity values. One method of measuring the best thresholds that have shown
excellent results is the minimum cross entropy.
The cross entropy, also known as divergence, is an information-theoretic metric
used to verify the amount of information in a random process [15].
The cross entropy was proposed by Kullback under the name of directed diver-
gence [16]. The cross entropy measures the information-theoretic distance between
two distributions P = { p1 , p2 , . . . , pn } and Q = {q1 , q2 , . . . , qn } by

N
qk
D(Q, P) = qk log2 (11)
k=1
pk
where Q = {Q 1 , Q 2 , Q 3 , . . . , Q N } and P = {P1 , P2 , P3 , . . . , PN } are two proba-

bilistic distributions from the same set and D is the cross entropy, it can be interpreted
as the hope of change in the content of the information when Q is used instead of P.
The minimum cross entropy method can be seen as an extension of the maximum
entropy method by setting equal initial estimates for all Pi when no prior information
is available [15].
In the image segmentation, it is possible to select a set of threshold values th =
[th 1 , th 2 , th 3 , . . . , th nt ] from the histogram h of the image.
Histograms are statistical tools that allow to visualize the distribution of values
described by the frequency with which a set of events occurs. From the perspective
of image processing, histograms consider all possible levels of intensity in the image
as events, and each of them is associated with the number of occurrences of that
intensity value throughout the image. In the case of grayscale images, one image
I (u, v) containing intensity values in the interval [0, L − 1] will generate an H

histogram with exactly L different values. Since it is common to encode intensities
in grayscale images using 8 bits, a total of L = 28 = 256 levels intensity. Therefore,
the frequency of occurrence of intensity level i is defined as h(i) and is equal to the
number of pixels with the intensity i existing in image I for all values 0 ≤ i ≥ L.
h(i) = car d{(u, v)|I (u, v) = i } (12)
where car d{. . .} represent the number of elements. So, h(0) is the number of pixels
that have an intensity value of 0. Similarly h(1) expresses the number of pixels with
a value of 1 and thus continues until h(255). The resulting histogram is presented in
the form of a one-dimensional vector of length L [17].
The simplest example of image segmentation is using a single threshold th = [th 1 ]
which, once selected, allows separating the pixels of the image according to the
following rule:

μ(1, th 1 ) si Ior (i, j) < th 1 ,
Is (i, j) = (13)
μ(th 1 , L − 1) si Ior (i, j) ≥ th 1
where L is the maximum intensity value of the image histogram (L = 256 in a

grayscale image with 8 rendering bits).
Considering the histogram h of the original image, the normalized value for a
specific range restricted by a and b it is calculated as:
b−1
i−a i h(i)
μ(a, b) = b−1
, i = 0, 1, . . . , L − 1 (14)
i−a h(i)
The Minimum cross entropy is defined for a single threshold (th = [th 1 ]):

th 1 −1
L
i i
D(th) = i h(i) log + i h(i) log (15)
i=1
μ(1, th 1 i=th 1
μ(th 1 , L + 1)
The proposed segmentation method has been implemented considering the

optimization algorithms and using as objective function:
th opt = arg min th (D(th)) (16)
Diego et al. [18].

6 Results of the Experiments
In this work, three MAs are adapted to determine the optimal number of thresholds
for Prostatic MRIs segmentation using the cross entropy minimal technique. In Table
2 is observed the best thresholds results from each algorithms applied 30 times each
evaluation and that the thresholds proposed by the 3 optimization algorithms are
homogeneous.
According to Table 3, the MFO algorithm presents the best results in fitness and
standard deviation. After 30 individual executions to each Prostatic MRIs the PSNR,
SSIM and FSIM results without significant differences as shows in the Table 4.
In Table 5 is observed the results of computational average time calculated in the
complete process for each images. The SCA algorithm generate the best processing
times, so we can conclude that it is the most efficient in this sense.
The experimental data obtained suggests that MFO has interesting search prop-
erties when applied to the ECM problem. One of the interesting characteristics of
ECM-MFO is that requires few dataset and even though it did not have the best
execution times, it exceeded the fitness results of the other metaheuristic algorithms.
Table 6 presents the thresholding images with the results generated by ECM
applying the MFO algorithm which presented the best fitness. The improvement in
the visual definition of the MRIs can be observed, which can help experts in the
diagnosis of prostate diseases.
Table 2 Best thresholds found by MFO, SCA y SFO

MRI nt MFO SCA SFO
MRI 01 2 32 112 32 112 32 112
3 28 84 142 28 84 142 28 84 143
4 11 38 90 147 11 38 90 150 11 38 93 150
5 10 33 70 109 158 10 33 71 110 163 9 31 69 108 160
8 6 14 31 56 80 109 143 177 3 13 27 49 77 114 143 175 5 15 35 59 82 109 146 183
16 3 7 12 18 29 44 58 70 82 95 2 3 5 10 13 14 22 38 49 61 75 4 10 17 27 41 55 70 84 98
111 129 147 166 184 203 90 104 127 164 176 115 134 146 157 171 184 209
32 2 4 6 8 12 16 20 25 30 36 41 2 6 10 12 17 21 22 25 26 35 3 6 9 11 13 18 27 33 41 51 54
46 55 62 71 77 83 89 98 106 43 57 63 66 74 87 109 119 62 73 76 83 90 101 113 124
114 122 131 139 148 159 168 122 132 140 155 159 170 179 132 134 143 151 163 174 180
177 187 197 208 220 182 197 209 215 221 228 246 193 201 208 224 230 240
MRI 02 2 30 110 30 110 30 110
3 26 82 139 26 83 140 26 82 140
4 9 35 85 141 10 37 85 143 10 37 86 144
5 9 31 69 106 153 9 31 66 106 155 9 33 72 109 155
8 5 13 30 55 80 109 141 176 5 14 28 54 76 109 130 170 4 13 29 56 79 105 136 171
16 3 6 10 15 22 32 45 58 70 84 2 2 6 11 19 30 49 58 63 82 88 3 8 13 23 34 42 60 73 84 100
100 118 136 155 174 194 101 125 150 168 194 119 146 159 180 188 205
32 2 4 6 8 12 15 19 23 30 38 45 1 2 3 3 5 7 9 10 13 18 19 30 3 6 8 11 15 21 31 39 47 54 63
53 60 67 75 82 89 97 104 112 32 42 55 55 74 75 84 104 106 66 71 79 86 97 103 109 117
120 129 137 147 156 166 173 111 114 130 142 152 162 171 125 135 140 148 157 166 175
180 187 195 204 216 174 190 214 255 183 197 201 218 225 232
(continued)
Table 2 (continued)
MRI nt MFO SCA SFO
MRI 03 2 32 112 32 112 32 112
3 27 83 141 27 82 140 28 84 142
4 10 37 88 144 10 37 88 143 10 36 90 146
5 9 31 68 105 153 8 29 64 102 154 10 33 71 108 155
8 5 13 29 53 77 103 134 170 5 13 24 41 73 97 134 172 6 15 32 56 77 108 142 179
16 3 7 11 16 24 37 51 64 77 90 1 3 8 15 18 38 54 74 88 101 4 9 16 27 40 52 65 82 89 102
105 121 138 157 176 197 121 151 169 204 241 255 116 127 140 164 187 206
32 2 4 6 9 12 16 21 27 35 42 48 2 3 4 4 5 6 9 12 15 25 27 31 2 4 9 12 15 21 29 34 41 50 58
54 62 66 71 74 78 84 90 97 46 60 74 82 96 96 114 120 69 75 81 89 96 103 113 121
103 111 119 128 137 148 160 133 135 142 148 166 186 191 127 131 137 145 155 161 170
171 182 194 209 255 197 227 228 232 234 182 198 199 203 216 232
MRI 04 2 29 110 29 110 29 110
3 26 86 144 26 86 145 26 86 144
4 9 35 89 146 10 37 89 146 10 38 92 148
5 8 30 68 105 154 9 29 67 105 153 8 29 68 106 154
8 4 11 27 53 78 106 138 173 3 12 25 46 67 106 132 167 5 12 33 60 90 117 154 182
16 3 6 9 14 21 32 47 60 71 86 2 3 8 8 15 20 28 47 51 60 74 4 9 15 26 45 61 75 87 103
101 118 137 156 175 196 84 100 117 147 186 121 137 157 171 185 199 215
32 2 4 6 9 12 16 21 27 32 41 49 2 2 2 3 6 9 13 20 30 33 44 46 2 5 8 13 17 24 31 38 47 53 58
57 63 71 80 87 95 103 111 54 56 64 80 82 92 103 103 63 68 76 86 98 107 114 118
118 126 132 139 147 161 172 108 119 122 129 138 139 148 125 129 135 146 154 163 174
180 188 197 205 215 226 176 185 198 206 255 183 188 197 202 218 243
MRI 05 2 31 111 31 111 31 111
3 27 88 147 28 88 147 27 88 147
4 10 38 92 149 10 39 94 149 10 37 91 149
5 8 30 67 105 156 8 30 66 106 154 8 30 67 105 155
8 5 13 29 54 78 105 138 174 3 12 22 45 70 91 132 176 5 13 29 55 79 108 138 173
16 3 6 11 16 24 39 54 66 78 93 2 2 4 7 14 24 33 50 64 75 103 3 8 15 29 44 57 69 87 105
110 128 148 166 184 204 107 121 153 168 196 122 140 156 169 186 202 225
32 2 4 6 9 12 15 20 26 35 45 53 2 2 2 3 5 8 9 9 10 15 19 25 28 2 4 7 9 14 18 25 32 38 41 47
61 68 75 84 92 101 111 121 30 35 39 49 54 69 78 82 89 54 65 76 87 93 96 102 108
129 136 142 149 156 163 170 90 92 105 120 136 149 157 113 119 125 139 155 160 176
179 189 197 206 216 229 180 213 255 187 200 213 225 231 233
The proposed approach is an interesting alternative to traditional prostate magnetic

resonance image analysis techniques. Furthermore, the ECM-MFO is suitable for
improving prostate MRIs. Since timely diagnosis of medical conditions is crucial,
these issues have gained the attention of the scientific community in recent years. The
ECM-MFO results, could incentivate medical professionals to use these techniques
to support the improvement the diagnosis of prostate diseases through MRIs.
Table 3 Objective function values of the various ECM implementations

MRI nt MFO SCA SFO
Mean σ Mean σ Mean σ
MRI 01 2 2.74380 1.355E−15 2.74378 0 2.744206 7.56E−04
3 1.81172 6.477E−05 1.81348 1.60E−03 1.834469 2.75E−02
4 1.20439 1.057E−03 1.21701 6.62E−03 1.310557 8.41E−02
5 0.82768 6.334E−04 0.86014 1.93E−02 0.919299 8.80E−02
8 0.39427 7.347E−03 0.52981 5.36E−02 0.456838 2.87E−02
16 0.12714 1.343E−02 0.24938 2.22E−02 0.169912 2.22E−02
32 0.04127 4.817E−03 0.10075 1.32E−02 0.063071 1.05E−02
MRI 02 2 2.70180 4.517E−16 2.701812 0 2.702699 1.338E−03
3 1.69380 9.034E−16 1.696641 2.12E−03 1.717922 2.989E−02
4 1.13993 9.223E−05 1.150822 5.31E−03 1.269306 7.673E−02
5 0.82356 6.418E−04 0.845354 9.37E−03 0.865446 5.184E−02
8 0.38813 2.107E−02 0.500945 5.63E−02 0.455621 4.153E−02
16 0.12339 1.240E−02 0.243122 3.35E−02 0.174285 2.208E−02
32 0.03978 4.287E−03 0.099806 1.31E−02 0.064814 1.100E−02
MRI 03 2 2.87490 1.807E−15 2.874859 2.90E−05 2.876068 1.46E−03
3 1.80901 5.477E−05 1.81121 1.45E−03 1.828635 2.38E−02
4 1.17258 8.491E−04 1.183529 6.69E−03 1.288846 1.16E−01
5 0.83651 6.284E−04 0.864371 1.19E−02 0.901905 5.85E−02
8 0.39987 2.043E−02 0.528934 7.21E−02 0.467832 4.78E−02
16 0.12559 1.299E−02 0.240691 2.75E−02 0.165565 1.85E−02
32 0.04108 5.551E−03 0.098469 1.12E−02 0.065349 1.06E−02
MRI 04 2 2.706900 0 2.706862 0 2.708078 1.72E−03
3 1.683900 9.034E−16 1.686771 1.83E−03 1.705813 2.64E−02
4 1.126953 8.398E−04 1.139438 6.88E−03 1.213077 7.55E−02
5 0.802267 1.352E−03 0.831658 1.29E−02 0.871645 7.91E−02
8 0.383007 8.243E−03 0.529497 5.25E−02 0.458168 3.87E−02
16 0.118650 6.324E−03 0.243228 2.38E−02 0.176595 2.43E−02
32 0.040608 5.305E−03 0.096557 1.24E−02 0.063178 1.18E−02
MRI 05 2 2.809200 4.517E−16 2.809254 2.60E−05 2.810057 1.218E−03
3 1.814400 6.775E−16 1.816462 1.58E−03 1.841243 3.235E−02
4 1.169577 2.417E−04 1.18423 7.92E−03 1.253173 1.073E−01
5 0.836586 2.433E−03 0.866872 1.24E−02 0.902196 5.720E−02
8 0.399083 6.802E−03 0.522246 5.49E−02 0.468348 3.913E−02
16 0.124061 1.016E−02 0.256123 3.59E−02 0.181385 2.899E−02
32 0.040649 5.104E−03 0.095375 1.12E−02 0.065289 1.212E−02
Table 4 Comparisons of the values of PSNR, SSIM and FSIM generated by algorithms MFO,
SCA and SFO
Nt MFO SCA SFO
PSNRV SSIMV FSIMV PSNRV SSIMV FSIMV PSNRV SSIMV FSIMV
MRI 2 15.5319 0.6620 0.5842 15.5310 0.6620 0.5842 15.5589 0.6635 0.5842
01 3 18.3902 0.7307 0.6987 18.4196 0.7318 0.7021 18.5022 0.7284 0.6896
4 19.2696 0.8044 0.7307 19.2306 0.8039 0.7300 20.0791 0.7929 0.7227
5 21.2501 0.8458 0.7894 21.0732 0.8421 0.7949 21.5196 0.8410 0.7920
8 24.7215 0.9124 0.8815 22.9643 0.8877 0.8289 24.9516 0.9058 0.8782
16 30.4152 0.9693 0.9651 26.5277 0.9362 0.9331 30.1566 0.9630 0.9653
32 36.1296 0.9904 0.9908 30.5782 0.9691 0.9641 34.8870 0.9850 0.9870
MRI 2 15.6062 0.6795 0.5962 15.6062 0.6795 0.5962 15.6384 0.6803 0.5980
02 3 18.5669 0.7588 0.7102 18.6115 0.7595 0.7111 18.7499 0.7567 0.7133
4 19.3497 0.8202 0.7392 19.3076 0.8185 0.7309 20.3910 0.8095 0.7297
5 21.3259 0.8544 0.7984 21.1932 0.8508 0.7846 21.5822 0.8506 0.7999
8 24.8353 0.9162 0.8826 23.4018 0.8953 0.8189 25.1689 0.9105 0.8941
16 30.8091 0.9717 0.9665 26.3253 0.9360 0.9185 30.3199 0.9644 0.9534
32 36.3488 0.9910 0.9910 30.5565 0.9695 0.9664 34.8035 0.9848 0.9777
MRI 2 15.6000 0.6641 0.5972 15.5993 0.6641 0.5972 15.6286 0.6647 0.5960
03 3 18.4860 0.7450 0.7033 18.5022 0.7460 0.7059 18.6366 0.7441 0.7125
4 19.3222 0.8145 0.7349 19.3091 0.8127 0.7263 20.1043 0.8082 0.7354
5 21.2873 0.8551 0.7988 21.0815 0.8506 0.7912 21.5958 0.8499 0.8019
8 24.7123 0.9158 0.8792 22.9894 0.8921 0.8560 25.0835 0.9111 0.8968
16 30.4639 0.9706 0.9641 26.7542 0.9399 0.9294 30.2575 0.9649 0.9630
32 36.1901 0.9908 0.9910 30.5924 0.9701 0.9539 34.5439 0.9846 0.9674
MRI 2 15.5282 0.6876 0.6025 15.5282 0.6876 0.6025 15.5560 0.6885 0.6025
04 3 18.4324 0.7645 0.7162 18.4568 0.7649 0.7143 18.6152 0.7656 0.7207
4 19.2854 0.8230 0.7428 19.2569 0.8216 0.7425 20.0710 0.8203 0.7414
5 21.2413 0.8624 0.8028 21.0151 0.8599 0.8116 21.6469 0.8552 0.8159
8 24.8419 0.9202 0.8825 23.0459 0.8982 0.8319 25.2778 0.9154 0.8934
16 30.7924 0.9730 0.9664 26.6280 0.9404 0.8932 30.1195 0.9645 0.9653
32 36.5294 0.9915 0.9915 30.6340 0.9707 0.9544 34.8836 0.9858 0.9617
MRI 2 15.5052 0.6785 0.5950 15.5060 0.6785 0.5950 15.5183 0.6772 0.5964
05 3 18.2529 0.7483 0.7073 18.2676 0.7487 0.7050 18.3836 0.7469 0.7041
4 19.2291 0.8158 0.7362 19.0962 0.8120 0.7268 19.7729 0.8124 0.7623
5 21.0730 0.8554 0.7943 21.0336 0.8508 0.7845 21.2859 0.8498 0.7930
8 24.7252 0.9187 0.8780 22.9904 0.8944 0.8494 25.1746 0.9142 0.9004
16 30.4716 0.9717 0.9650 26.1674 0.9369 0.9214 29.8469 0.9625 0.9415
32 36.2035 0.9912 0.9912 30.6326 0.9716 0.9658 34.6977 0.9856 0.9887
Table 5 Average computational time of algorithms MFO, SCA and SFO for image segmentation
using the ECM
Computational time in seconds
nt MFO SCA SFO
MRI 01 2 2.5938 1.4677 5.5310
3 2.6676 1.5076 5.6581
4 2.6709 1.5581 5.7025
5 2.5827 1.5919 5.7904
8 2.7198 1.6831 6.0866
16 2.9796 1.9097 6.4186
32 3.4662 2.3905 7.2317
MRI 02 2 2.8429 1.4649 5.3162
3 2.6343 1.5145 5.5558
4 2.7083 1.5518 5.7175
5 2.6584 1.5927 5.8099
8 2.7406 1.6855 5.9610
16 2.9451 1.9060 6.5197
32 3.4673 2.3855 7.2022
MRI 03 2 2.8769 1.4620 5.3361
3 2.5274 1.5068 5.6983
4 2.7187 1.5670 5.7049
5 2.5627 1.5959 5.8611
8 2.6861 1.6867 5.9571
16 2.8826 1.9070 6.2847
32 3.5169 2.3920 7.1995
MRI 04 2 2.8023 1.4678 5.5429
3 2.5618 1.5071 5.6096
4 2.4792 1.5557 5.8121
5 2.5381 1.5865 5.8561
8 2.7391 1.6821 5.8789
16 2.8766 1.9167 6.3065
32 3.4146 2.4022 7.2223
MRI 05 2 2.6412 1.4631 5.5622
3 2.5150 1.5081 5.6057
4 2.5056 1.5621 5.5439
5 2.6239 1.6021 5.8496
8 2.6670 1.6794 5.8986
16 2.8790 1.9114 6.3024
32 3.4602 2.3909 7.1916
Table 6 Results obtained applying the MFO algorithm results for ECM on the selected images
MRI 01, nt = 2 MRI 01, nt = 3
MRI 01, nt = 4 MRI 01, nt = 5
(continued)
Table 6 (continued)
MRI 01, nt = 8 MRI 01, nt = 16
(continued)
Table 6 (continued)
MRI 01, nt = 32
MRI 02, nt = 2 MRI 02, nt = 3
(continued)
Table 6 (continued)
MRI 02, nt = 4 MRI 02, nt = 5
(continued)
Table 6 (continued)
MRI 02, nt = 8 MRI 02, nt = 16
MRI 02, nt = 32
(continued)
Table 6 (continued)
MRI 03, nt = 2 MRI 03, nt = 3
(continued)
Table 6 (continued)
MRI 03, nt = 4 MRI 03, nt = 5
MRI 03, nt = 8 MRI 03, nt = 16
(continued)
Table 6 (continued)
MRI 03, nt = 32
(continued)
Table 6 (continued)
MRI 04, nt = 2 MRI 04, nt = 3
MRI 04, nt = 4 MRI 04, nt = 5
(continued)
Table 6 (continued)
MRI 04, nt = 8 MRI 04, nt = 16
(continued)
Table 6 (continued)
MRI 04, nt = 32
MRI 05, nt = 2 MRI 05, nt = 3
(continued)
Another random document with
no related content on Scribd:
burrow, she lays some eggs, and when these hatch, feeds and tends the larvae and pupae;
the first specimens of these latter that become perfect Insects are workers of all sizes, and at
once undertake the duties of tending the young and feeding the mother, who, being thus freed
from the duties of nursing and of providing food while she is herself tended and fed, becomes
a true queen-ant. Thus it seems established that in the case of this species the division of
labour found in the complex community, does not at first exist, but is correlative with increasing
numbers of the society. Further observations as to the growth of one of these nascent
communities, and the times and conditions under which the various forms of individuals
composing a complete society first appear, would be of considerable interest.
An American species of the same genus, C. pennsylvanicus, the carpenter-ant, establishes its
nests in the stumps of trees. Leidy observed that solitary females constructed for themselves
cells in the wood and closed the entrances, and that each one in its solitary confinement
reared a small brood of larvae. The first young produced in this case are said to be of the
dwarf caste, and it was thought by the observer that the ant remained not only without
assistance but also without food during a period of some weeks, and this although she was
herself giving food to the larvae she was rearing.
Adlerz states that the females or young queens take no food while engaged in doing their
early work, and that the large quantity of fat-body they possess enables them to undergo
several months of hunger. In order to feed the young larvae they use their own eggs or even
the younger larvae. It is to the small quantity of food rather than to its nature that he attributes
the small size of the first brood of perfect workers. M. Janet[63] has recently designed an
ingenious and simple apparatus for keeping ants in captivity. In one of these he placed a
solitary female of Lasius alienus, unaccompanied by any workers or other assistants, and he
found at the end of 98 days that she was taking care of a progeny consisting of 50 eggs, 2
larvae, 5 pupae in cocoons, 5 without cocoons. On the 102nd day workers began to emerge
from the cocoons.[64] From these observations it is evident that the queen-ant, when she
begins her nest, lives under conditions extremely different from those of the royal state she
afterwards reaches.
In many kinds of ants the full-grown individuals are known to feed not only the larvae by
disgorging food from their own mouths into those of the little grubs, but also to feed one
another. This has been repeatedly observed, and Forel made the fact the subject of
experiment in the case of Camponotus ligniperdus. He took some specimens and shut them
up without food for several days, and thereafter supplied some of them with honey, stained
with Prussian blue; being very hungry, they fed so greedily on this that in a few hours their
hind bodies were distended to three times their previous size. He then took one of these
gorged individuals and placed it amongst those that had not been fed. The replete ant was at
once explored by the touches of the other ants and surrounded, and food was begged from it.
It responded to the demands by feeding copiously a small specimen from its mouth: when this
little one had received a good supply, it in turn communicated some thereof to other
specimens, while the original well-fed one also supplied others, and thus the food was
speedily distributed. This habit of receiving and giving food is of the greatest importance in the
life-history of ants, and appears to be the basis of some of the associations that, as we shall
subsequently see, are formed with ants by numerous other Insects.
Fig. 60—Oecophylla smaragdina. Worker using a larva for spinning.
Oecophylla smaragdina, a common ant in Eastern Asia, forms shelters on the leaves of trees
by curling the edges of leaves and joining them together. In doing this it makes use of an
expedient that would not be believed had it not been testified by several competent and
independent witnesses. The perfect ant has no material with which to fasten together the
edges it curls; its larva, however, possesses glands that secrete a supply of material for it to
form a cocoon with, and the ants utilise the larvae to effect their purpose. Several of them
combine to hold the foliage in the desired position, and while they do so, other ants come up,
each one of which carries a larva in its jaws, applies the mouth of the larva to the parts where
the cement is required, and makes it disgorge the sticky material. Our figure is taken from a
specimen (for which we are indebted to Mr. E. E. Green) that was captured in the act of
bearing a larva.
Formica rufa, the Red-ant, Wood-ant, or Hill-ant, is in this country one of the best-known
members of the Formicidae. It frequents woods, especially such as are composed, in whole or
part, of conifers, where it forms large mounds of small sticks, straws, portions of leaves, and
similar material. Although at first sight such a nest may appear to be a chaotic agglomeration,
yet examination reveals that it is arranged so as to leave many spaces, and is penetrated by
galleries ramifying throughout its structure. These mound-nests attain a considerable size
when the operations of the industrious creatures are not interfered with, or their work
destroyed, as it too often is, by ignorant or mischievous persons. They may reach a height of
three feet or near it, and a diameter of twice that extent. The galleries by which the heaps are
penetrated lead down to the earth below. From the mounds extend in various directions paths
constantly traversed by the indefatigable ants. M‘Cook observed such paths in the Trossachs;
they proceed towards the objects aimed at in lines so straight that he considers they must be
the result of some sense of direction possessed by the ants; as it is impossible to suppose
they could perceive by the sense of sight the distant objects towards which the paths were
directed: these objects in the case M‘Cook describes were oak-trees up which the ants
ascended in search of Aphides.
M‘Cook further observed that one of the oak-trees was reached by individuals from another
nest, and that each of the two parties was limited to its own side of the tree, sentinels being
placed on the limits to prevent the trespassing of an intruder; he also noticed that the ants saw
an object when the distance became reduced to about an inch and a half from them. This
species is considered to be wanting in individual courage; but when acting in combination of
vast numbers it does so with intelligence and success. It does not make slaves, but it has
been observed by Bignell and others that it sometimes recruits its numbers by kidnapping
individuals from other colonies of its own species. Its nests are inhabited by forty or fifty
species of guests of various kinds, but chiefly Insects. Another ant, Myrmica laevinodis,
sometimes lives with it in perfect harmony, and Formicoxenus nitidulus lives only with F. rufa.
Amongst the most peculiar of its dependants we may mention large beetles of the genera
Cetonia and Clythra, which in their larval state live in the hills of the wood-ant. It is probable
that they subsist on some of the vegetable matter of which the mounds are formed. Adlerz has
given some attention to the division of labour amongst the different forms of the workers of
ants, and says that in F. rufa it is only the bigger workers that carry building and other
materials, the smaller individuals being specially occupied in the discovery of honey-dew and
other Aphid products. In Camponotus it would appear, on the other hand, that the big
individuals leave the heavy work to be performed by their smaller fellows.
The wood-ant and its near allies have been, and indeed still are, a source of great difficulty to
systematists on account of the variation that occurs in the same species, and because this
differs according to locality. Our European F. rufa has been supposed to inhabit North
America, and the interesting accounts published by M‘Cook of the mound-making ant of the
Alleghanies were considered to refer to it. This Insect, however, is not F. rufa, as was
supposed by M‘Cook, but F. exsectoides, Forel. It forms colonies of enormous extent, and
including an almost incredible number of individuals. In one district of about fifty acres there
was an Ant City containing no less than 1700 of these large ant-hills, each one teeming with
life. It was found by transferring ants from one hill to another that no hostility whatever existed
between the denizens of different hills; the specimens placed on a strange hill entered it
without the least hesitation. Its habits differ in some particulars from those of its European
congener; the North American Insect does not close the formicary at night, and the inquilines
found in its nest are very different from those that live with F. rufa in Europe. Whether the
typical wood-ant occurs in North America is doubtful, but there are races there that doubtless
belong to the species.
F. sanguinea is very similar in appearance to its commoner congener F. rufa, and is the only
slave-making ant we possess in Britain. This species constructs its galleries in banks, and is
of very courageous character, carrying out its military operations with much tactical ability. It is
perfectly able to live without the assistance of slaves, and very frequently does so; indeed it
has been asserted that it is in our own islands (where, however, it is comparatively rare) less
of a slave-owner than it is in Southern Europe, but this conclusion is very doubtful. It appears
when fighting to be rather desirous of conquering its opponents by inspiring terror and making
them aware of its superiority than by killing them; having gained a victory it will carry off the
pupae from the nest it has conquered to its own abode, and the ants of the stranger-species
that develop from these pupae serve the conquerors faithfully, and relieve them of much of
their domestic duties. The species that F. sanguinea utilises in this way in England are F.
fusca, F. cunicularia, and possibly Lasius flavus. Huber and Forel have given graphic accounts
of the expeditions of this soldier-ant. In the mixed colonies of F. sanguinea and F. fusca the
slaves do most of the house-work, and are more skilful at it than their masters. Adlerz says
that one of the slaves will accomplish twice as much work of excavation in the same time as
the slave-owner; these latter being lazy and fond of enjoyment, while the slaves are very
industrious.
Fig. 61—Head of Polyergus rufescens. (After André.)
Polyergus rufescens, an European ant allied to Formica, is renowned since the time of Huber
(1810) as the slave-making or Amazon ant. This creature is absolutely dependent on its
auxiliaries for its existence, and will starve, it is said, in the midst of food unless its servitors
are there to feed it. Wasmann, however, states that Polyergus does possess the power of
feeding itself to a certain extent. Be this as it may, the qualities of this ant as warrior are
superb. When an individual is fighting alone its audacity is splendid, and it will yield to no
superiority of numbers; when the creatures are acting as part of an army the individual
boldness gives place to courage of a more suitable sort, the ants then exhibiting the act of
retreating or making flank movements when necessary. If a Polyergus that is acting as a
member of a troop finds itself isolated, and in danger of being overpowered, it has then no
hesitation in seeking safety even by flight. This species is provided with mandibles of a
peculiar nature; they are not armed with teeth, but are pointed and curved; they are therefore
used after the manner of poignards, and when the ant attacks a foe it seizes the head
between the points of these curved mandibles, and driving them with great force into the brain
instantly paralyses the victim.
Mandibles of this shape are evidently unfitted for the purposes of general work, they can
neither cut, crush, nor saw, and it is not impossible that in their peculiar shape is to be found
the origin of the peculiar life of Polyergus: we find similar mandibles reappearing amongst the
aberrant Dorylides, and attaining a maximum of development in the ferocious Eciton; they also
occur, or something like them, in a few aberrant Myrmicides; and in the male sex of many
other ants, which sex exercises no industrial arts, this sort of mandible is present.
The ants that Polyergus usually attacks in order to procure slaves are Formica fusca and F.
fusca, race auricularia; after it has routed a colony of one of these species, P. rufescens
pillages the nest and carries off pupae and some of the larger larvae to its own abode. When
the captives thus deported assume the imago state, they are said to commence working just
as if they were in their own houses among their brothers and sisters, and they tend their
captors as faithfully as if these were their own relatives: possibly they do not recognise that
they are in unnatural conditions, and may be quite as happy as if they had never been
enslaved. The servitors are by no means deficient in courage, and if the place of their
enforced abode should be attacked by other ant-enemies they defend it bravely. The fact that
P. rufescens does not feed its larvae has been considered evidence of moral degeneration,
but it is quite possible that the Insect may be unable to do so on account of some deficiency in
the mouth-parts, or other similar cause. The larvae of ants are fed by nutriment regurgitated
from the crop of a worker (or female), and applied to the excessively minute mouth of the
helpless grub: for so delicate a process to be successfully accomplished, it is evident that a
highly elaborated and specialised arrangement of the mouth-parts must exist, and it is by no
means improbable that the capacity of feeding its young in true ant-fashion is absent in
Polyergus for purely mechanical reasons.
M‘Cook states that the North American ant, Polyergus lucidus, which some entomologists
consider to be merely a variety of the European species, makes slaves of Formica schaufussi,
itself does no work, and partakes of food only when fed by its servitors. He did not, however,
actually witness the process of feeding. When a migration takes place the servitors deport
both the males and females of P. lucidus. M‘Cook adds that the servitors appear to be really
mistresses of the situation, though they avail themselves of their power only by working for the
advantage of the other species.
Fig. 62—Myrmecocystus mexicanus. Honey-pot ant, dorsal view.
Fig. 63—Myrmecocystus mexicanus. Lateral view.
The honey-ant of the United States and Mexico has been investigated by M‘Cook and others;
the chief peculiarity of the species is that certain individuals are charged with a sort of honey
till they become enormously distended, and in fact serve as leather bottles for the storage of
the fluid. The species Myrmecocystus hortideorum and M. melliger, are moderate-sized
Insects of subterranean habits, the entrance to the nest of M. hortideorum being placed in a
small raised mound. The honey is the product of a small gall found on oak leaves, and is
obtained by the worker-ants during nocturnal expeditions, from which they return much
distended; they feed such workers left at home as may be hungry, and then apparently
communicate the remainder of the sweet stuff they have brought back to already partly
charged "honey-bearers" left in the nest. The details of the process have not been observed,
but the result is that the abdomens of the bearers become distended to an enormous extent
(Figs. 62, 63), and the creatures move but little, and remain suspended to the roof of a special
chamber. It is considered by M‘Cook that these living honey-tubs preserve the food till a time
when it is required for the purposes of feeding the community. The distension is produced
entirely by the overcharging of the honey-crop, the other contents of the abdomen being
forced by the distention to the posterior part of the body. Lubbock has since described an
Australian ant, Melophorus inflatus, having a similar peculiar habit, but belonging to the allied
tribe Plagiolepisii. Quite recently a South African honey-tub ant belonging to the distinct genus
Plagiolepis (Ptrimeni For.) has been discovered, affording a proof that an extremely
specialised habit may arise independently of relation between the Insects, and in very different
parts of the world.
Species of the genus Lasius are amongst the most abundant of the ant-tribe in Britain. They
are remarkable for their constructive powers. L. niger, the common little black garden-ant,
forms extensive subterranean galleries, and is extremely successful in the cultivation of
various forms of Aphidae, from the products of which the species derives a large part of its
subsistence. The ants even transport the Aphidae to suitable situations, and thus increase
their stock of this sugary kind of cattle, and are said to take the eggs into their own dwellings
in the autumn so that these minute and fragile objects may be kept safe from the storms and
rigours of winter. These little creatures are brave, but when attacked by other ants they defend
themselves chiefly by staying in their extensive subterranean galleries, and blocking up and
securing these against their assailants.
L. fuliginosus, another of our British species, has very different habits, preferring old trees and
stumps for its habitation; in the hollows of these it forms dwellings of a sort of card; this it
makes from the mixture of the secretions of its salivary glands with comminuted fragments of
wood, after the fashion of wasps. It is a moderate-sized ant, much larger than the little L.
niger, and is of a black colour and remarkably shining; it gives off a very strong but by no
means disagreeable odour, and may be seen on the hollow trees it frequents, stalking about in
large numbers in a slow and aimless manner that contrasts strikingly with the active, bustling
movements of so many of its congeners. When this species finds suitable trees near one
another, a colony is established in each; the number of individuals thus associated becomes
very large, and as the different colonies keep up intercommunication, this habit is very useful
for purposes of defence. Forel relates that he once brought a very large number of Formica
pratensis and liberated them at the base of a tree in which was a nest of L. fuliginosus; these
latter, finding themselves thus assaulted and besieged, communicated in some way,
information of the fact to the neighbouring colonies, and Forel soon saw large columns of the
black creatures issuing from the trees near by and coming with their measured paces to the
assistance of their confrères, so that the invaders were soon discomfited and destroyed.
Although the European and North American representatives of the sub-family Camponotides
live together in assemblies comprising as a rule a great number of individuals, and although
the separate nests or formicaries which have their origin from the natural increase of a single
original nest keep up by some means a connection between the members, and so form a
colony of nests whose inhabitants live together on amicable terms, yet there is no definite
information as to how long such association lasts, as to what is the nature of the ties that
connect the members of the different nests, nor as to the means by which the colonies
become dissociated. It is known that individual nests last a long time. Charles Darwin has
mentioned in a letter to Forel that an old man of eighty told him he had noticed one very large
nest of Formica rufa in the same place ever since he was a boy. But what period they usually
endure is not known, and all these points probably vary greatly according to the species
concerned. It has been well ascertained that when some ants find their nests, for some
unknown reason, to be unsuitable the inhabitants leave their abodes, carrying with them their
young and immature forms, and being accompanied or followed by the various parasites or
commensals that are living with them. Wasmann and other entomologists have observed that
the ants carry bodily some of their favourite beetle-companions, as well as members of their
own species. Forel observed that after a nest of Formica pratensis had been separated into
two nests placed at a considerable distance from one another so as to have no
intercommunication, the members yet recognised one another as parts of the same family
after the lapse of more than a month; but another observation showed that after some years of
separation they were no longer so recognised.
Although it is now well ascertained that ants are able to distinguish the individuals belonging to
their own nests and colonies from those that, though of their own species, are not so related to
them, yet it is not known by what means the recognition is effected; there is, however, some
reason to suppose that it is by something of the nature of odour. One observer has noticed
that if an ant fall into water it is on emerging at first treated as if it were a stranger by its own
friends; but other naturalists have found this not to be the case in other species. Contact with
corrosive sublimate deprives ants for a time of this power of recognising friends, and under its
influence they attack one another in the most ferocious mariner.
The nests and colonies of the species of Camponotides we have considered are all
constructed by societies comprising a great number of individuals; there are, however, in the
tropics numerous species that form their nests on foliage, and some of these contain only a
few individuals. The leaf-nests (Fig. 64) of certain species of Polyrhachis are said to be formed
of a paper-like material, and to contain each a female and about 8 or 10 worker ants. Forel[65]
has examined nests of several Indian species, and finds they differ from those of other ants in
consisting of a single cavity, lined with silk like that of a spider. These nests are further said to
be constructed so as to render them either inconspicuous or like other objects on the leaves;
P. argentea covers its small dwelling with little bits of vegetable matter, and a nest of P. rastella
was placed between two leaves in such a manner as to be entirely hidden. All the species of
the genus do not, however, share these habits, P. mayri making a card-nest, like Dolichoderus
and some other ants. The species of the genus Polyrhachis are numerous in the tropics of the
Old World.
Fig. 64—Nest of Polyrhachis sp. (After Smith.)
Forbes noticed that a species of this genus, that makes its paper-like nest on the underside of
bamboo-leaves produces a noise by striking the leaf with its head in a series of spasmodic
taps. The same observer has recorded a still more interesting fact in the case of another
species of this genus—a large brown ant—found in Sumatra. The individuals were "spread
over a space, perhaps a couple of yards in diameter, on the stem, leaves, and branches of a
great tree which had fallen, and not within sight of each other; yet the tapping was set up at
the same moment, continued exactly the same space of time, and stopped at the same
instant; after the lapse of a few seconds all recommenced at the same instant. The interval
was always of about the same duration, though I did not time it; each ant did not, however,
beat synchronously with every other in the congeries nearest to me; there were independent
tappings, so that a sort of tune was played, each congeries dotting out its own music, yet the
beginnings and endings of the musical parties were strictly synchronous."
Fig. 65—Polyrhachis pandurus, worker. Singapore.
Mr. Peal has also recorded that an ant—the name is not mentioned, but it may be presumed
to be an Assamese species—makes a concerted noise loud enough to be heard by a human
being at twenty or thirty feet distance, the sound being produced by each ant scraping the
horny apex of the abdomen three times in rapid succession on the dry, crisp leaves of which
the nest is usually composed. These records suggest that these foliage-ants keep up a
connection between the members of different nests somewhat after the same fashion as do so
many of the terrestrial Camponotides. Although the species of Camponotides have no special
organ for the production of sound in the position in which one is found in Myrmicides and
Ponerides, yet it is probable that they are able to produce a sound by rubbing together other
parts of the abdomen.
Sub-Fam. 2. Dolichoderides.—Hind body furnished with but one constriction so that

only a single scale or node is formed; Sting rudimentary; the poison-sac without cushion.
Fig. 66—Tapinoma erraticum, worker. Britain. Upper side and profile.
The Dolichoderides are similar to Camponotides in appearance, and are distinguished chiefly
by the structure of the sting and the poison apparatus. To this we may add that Forel also
considers the gizzard to be different in the two sub-families, there being no visible calyx in the
Dolichoderides, while this part is largely developed in the Camponotides. This is one of the
least extensive of the sub-families of ants, not more than 150 species being yet discovered.
Comparatively little is known of the natural history of its members, only a very small number of
species of Dolichoderides being found in Europe. The best known of these (and the only
British Dolichoderid) is Tapinoma erraticum, a little ant of about the size of Lasius niger, and
somewhat similar in appearance, but very different in its habits. T. erraticum does not cultivate
or appreciate Aphides, but is chiefly carnivorous in its tastes. Our knowledge of it is due to
Forel, who has noticed that it is very fond of attending the fights between other ants. Here it
plays the part of an interested spectator, and watching its opportunity drags off the dead body
of one of the combatants in order to use it as food. Although destitute of all power of stinging,
this Insect has a very useful means of defence in the anal glands with which it is provided;
these secrete a fluid having a strong characteristic odour, and possessing apparently very
noxious qualities when applied to other ants. The Tapinoma has no power of ejecting the fluid
to a distance, but is very skilful in placing this odorous matter on the body of an opponent by
touching the latter with the tip of the abdomen; on this being done its adversary is usually
discomfited. This Insect is subterranean in its habits, and is said to change its abode very
frequently. T. erraticum occurs somewhat rarely in Britain. Forel has also noted the habits of
Liometopum microcephalum, another small European species of Dolichoderides. It is a tree-
ant, and by preference adopts, and adapts for its use, the burrows made by wood-boring
beetles. It forms extremely populous colonies which may extend over several large trees, the
inhabitants keeping up intercommunication by means of numerous workers. No less than
twelve mighty oaks were found to be thus united into a colony of this ant in one of the
Bulgarian forests. The species is very warlike, and compensates for the extreme minuteness
of its individuals by the skilful and rapid rushes made by combined numbers on their ant-foes
of larger size.
Fritz Müller has given a brief account, under the name of the Imbauba ant, of a Brazilian
arboreal ant, that forms small nests in the interior of plants. The species referred to is no doubt
an Azteca, and either A. instabilis, or A. mulleri. The nests are founded by fertilised females
which may frequently be found in the cells on young Cecropia plants. Each internode, he says,
has on the outside, near its upper part, a small pit where the wall is much thinner, and in this
the female makes a hole by which she enters. Soon afterwards the hole is completely closed
by a luxuriant excrescence from its margins, and it remains thus closed until about a dozen
workers have developed from the eggs of the female, when the hole is opened anew from
within by the workers. It is said that many of the larvae of these ants are devoured by the
grubs of a parasite of the family Chalcididae. This Insect is thought to protect the plant from
the attacks of leaf-cutting ants of the genus Atta.
We may here briefly remark that much has been written about the benefits conferred on plants
by the protection given to them in various ways by ants: but there is reason to suppose that a
critical view of the subject will not support the idea of the association being of supreme
importance to the trees.[66]
Sub-Fam. 3. Myrmicides.—Pedicel of abdomen formed of two well-marked nodes

(knot-like segments). Sting present (absent in the Cryptocerini and Attini). (It should be
noted that the workers of the genera Eciton and Aenictus of the subfamily Dorylides
have, like the Myrmicides, two nodes in the pedicel.)
This sub-family consists of about 1000 species, and includes a great variety of forms, but, as
they are most of them of small size, they are less known than the Camponotides, and much
less attention has been paid to their habits and intelligence. Forel, until recently, adopted four
groups: Myrmicini, Attini, Pseudomyrmini and Cryptocerini; but he is now disposed to increase
this number to eight.[67] They are distinguished by differences in the clypeus, and in the form
of the head; but it must be noted that the characters by which the groups are defined are not
in all cases fully applicable to the males. The Cryptocerini are in external structure the most
highly modified of Hymenoptera, if not of all the tribes of Insecta.
Fig. 67—Pheidologeton laboriosus, large and small workers. East India.
i. The Myrmicini proper are defined by Forel as having the antennae inserted near the middle,
a little behind the front, of the head, which has carinae on the inner sides, but none on the
outer sides, of the insertions of the antennae; the clypeus extends between the antennae.
Fig. 68—Formicoxenus nitidulus, male. (After Adlerz.)
Certain genera of small European ants of the group Myrmicini display some most anomalous
phenomena. This is especially the case in Formicoxenus, Anergates and Tomognathus. The
facts known have, however, been most of them only recently discovered, and some obscurity
still exists as to many of even the more important points in these extraordinary life-histories.
Fig. 69.—Anergates atratulus. Europe. A, male, with part of hind leg broken off; B, female, with
wings: C, female, after casting the wings and becoming a queen.
It has long been known that the little Formicoxenus nitidulus lives as a guest in the nests of
Formica rufa, the wood-ant; and another similar ant, Stenamma westwoodi, which shares the
same life, was declared by Nylander and Smith to be its male; it was however shown some
years ago by André that this is a mistake, and that S. westwoodi is really the male of another
ant that had till then been called Asemorhoptrum lippulum. This correction left the workers and
females of Formicoxenus nitidulus destitute of a male, but Adlerz has recently discovered that
the male of this species is wingless and similar to the worker, the female being a winged
Insect as usual. It is very curious that the characters by which the male is distinguished from
the worker should vary in this species; but according to Adlerz this is the case, individuals
intermediate in several points between the males and workers having been discovered. This
phenomenon of quite wingless males in species where the female is winged is most
exceptional, and is extremely rare in Insects; but it occurs, as we shall see, in one or two other
Myrmicides. Charles Darwin made the very reasonable suggestion that winged males may be
developed occasionally as an exceptional phenomenon, and it is very probable that this may
be the case, though it has not yet been demonstrated. Formicoxenus nitidulus occurs in
England in the nests of Formica rufa and of F. congerens, but we are not aware that the male
has ever been found in this country. The genus Anergates is allied to Formicoxenus, and
occurs in Central Europe, but has not been found in Britain; the female, as in Formicoxenus, is
winged and the male wingless, but there is no worker-caste; the male is a rather helpless
creature, and incapable of leaving the nest. The species lives in company with Tetramorium
caespitum a little ant very like Myrmica, and not uncommon in South-East England. The
female Anergates is at first an active little creature with wings, but after these are lost the body
of the Insect becomes extremely distended as shown in Fig. 69, C; the creature is in this state
entirely helpless, and as there are no workers, the Anergates is completely dependent, for the
existence of itself and its larvae, on the friendly offices of the Tetramorium that lives with it.
The mode of the association of these two Insects is at present both unparalleled and
inexplicable, for only workers of the Tetramorium are found in company with the ♂ and ♀
Anergates; the community, in fact, consisting of males and females of one species and
workers of another. The nests of Anergates are so rare that only a few naturalists have been
able to observe them (Schenk, von Hagens, and Forel may be specially mentioned), but in the
spots where they occur, nests of the Tetramorium, containing all the forms of that species, are
numerous, and it therefore seems probable that a young fertile female of the Anergates may
leave a nest in which it was born, enter a nest of the Tetramorium, and, destroying the queen
thereof, substitute herself in the place of the victim; but if this be really the case, the larvae
and pupae of the Tetramorium must also be destroyed, for no young of the Tetramorium are
ever found in these strange associations. It is very difficult to believe that the Tetramorium
workers should be willing to accept as their queen a creature that commenced her
acquaintance with them by destroying their own queen or queens and a number of their young
sisters; especially as the Tetramorium is a more powerful ant than the Anergates, and could
readily dispose of the murderous intruder if it were disposed to do so. It is known, however,
that colonies of Tetramorium completely destitute of queens sometimes occur, and Wasmann
has suggested that the female Anergates may seek out one of these, and installing herself
therein as a substitute, may be accepted by the orphaned colony. This plausible hypothesis
has still to be verified.
The genus Cardiocondyla also exhibits the phenomenon of apterous, worker-like males, while
in one species, C. emeryi, a winged male is also known to exist.
Tomognathus sublaevis is a little Myrmicid ant, found rarely in Denmark and Sweden, where
its habits have recently been studied by Adlerz. A band of the Tomognathus attack the nest of
another little Myrmicid, Leptothorax acervorum, and succeed by their own pertinacity and the
fears of the Leptothorax in obtaining possession of it; the legitimate owners disappear, leaving
the Tomognathus in possession of their larvae and pupae; these complete their development
only to find themselves the slaves of Tomognathus. The subsequent relations of the two ants
are friendly, the slaves even preventing their masters from wandering from the nest when they
wish to do so. If an established mixed community of this nature is in want of additional
servitors, the Tomognathus secure a supply by raids after the fashion of the Amazon-ant,
bringing back to their abode larvae and pupae of Leptothorax to be developed as slaves. It
was formerly supposed that the Tomognathus continued its species by perpetual
parthenogenesis of the workers, for neither males nor females could be found. Adlerz[68] has,
however, now discovered the sexual individuals. The male is an ordinary winged ant, and is so
like that sex of the Leptothorax, that Adlerz had failed to distinguish the two before he reared
them. The females are apterous, and in fact like the workers. It would perhaps be more correct
to say that the workers of this species vary greatly but never become winged; some of them
have ocelli and a structure of the thorax more or less similar to that of winged females, though
none have been found with wings. Certain of these females possess a receptaculum seminis,
and Adlerz treats this as the true distinction between female and worker. In accordance with
this view the female of Tomognathus may be described as a worker-like individual possessing
a receptaculum seminis, and having more or less of the external structures of winged females,
though never being actually winged. It is probable that other workers reproduce
parthenogenetically. The males of this species will not unite with females from the same nest,
thus differing from many other ants, in which union between individuals of the same nest is the
rule. Finally, to complete this curious history, we should remark that the larvae of the
Tomognathus are so similar to those of the Leptothorax that Adlerz is quite unable to
distinguish the two.
Strongylognathus testaceus and S. huberi live in association with Tetramorium caespitum, and
are cared for by these latter ants; it is notable that as in the case of the slave-making
Polyergus rufescens the mandibles of the Strongylognathus are cylindrical and pointed, and
therefore unsuitable for industrial occupations. S. testaceus is a weak little ant, and lives in
small numbers in the nests with T. caespitum, which it is said to greatly resemble in
appearance. The proportions of the forms of the two species usually associated is peculiar,
there being a great many workers of T. caespitum both in the perfect and pupal states, and
also all the sexes of the Strongylognathus, of which, however, only a few are workers. This
would seem to suggest that S. testaceus attacks and pillages the nests of T. caespitum in
order to carry off worker pupae, just as Polyergus rufescens does. But the facts that S.
testaceus is a weaker Insect than the Tetramorium, and that only a few of its worker-caste are
present in a community where there are many workers of the Tetramorium, seem to negative
the view that the latter were captured by the former; and the mode in which the associated
communities of these two species are started and kept up is still therefore in need of
explanation.
Strongylognathus huberi is a much stronger Insect than its congener, S. testaceus, and Forel
has witnessed its attack on Tetramorium caespitum. Here the raid is made in a similar manner
to that of Polyergus rufescens on Formica; the Tetramorium is attacked, and its pupae carried
off to the abode of the Strongylognathus to serve in due time as its slaves. The mandibles of
S. huberi, being similar in form to those of Polyergus rufescens are used in a similar manner.
Although T. caespitum is common enough in South-East England, it is to be regretted that

none of the guests or associates we have mentioned in connection with it occur in this country.
It is a most variable species, and is distributed over a large part of the globe.
Our British species of Myrmicides, about ten in number, all belong to the group Myrmicini;
none of them are generally common except Myrmica rubra, which is a most abundant Insect,
and forms numerous races that have been considered by some entomologists to be distinct
species; the two most abundant of these races are M. ruginodis and M. scabrinodis, which
sometimes, at the time of the appearance of the winged individuals, form vast swarms.
The tiny Monomorium pharaonis is a species that has been introduced into Britain, but now
occurs in houses in certain towns; it sometimes accumulates on provisions in such numbers
as to be a serious nuisance. Seventeen thousand individuals weigh 1 gramme, and it is
probable that a nest may include millions of specimens.
The genus Aphaenogaster[69] and its immediate allies include the harvesting ants of Europe
and North America: they form subterranean nests consisting of isolated chambers connected
by galleries; some of the chambers are used as store-houses or granaries, considerable
quantities of corn, grass, and other seeds being placed in them. A. structor and A. barbarus
have been observed to do this in Southern Europe, by Lespès, Moggridge, and others.
Fig. 70—Aphaenogaster (Messor) barbarus. Algeria. A, male; B, winged female; C, large
worker or soldier; D, small worker. × 3⁄2.
In the deserts about Algeria and Tunis a harvesting ant, Aphaenogaster (Messor) arenarius, is
an important creature: its subterranean dwellings are very extensive, and are placed at a
depth of several feet from the surface. Entrance to these dwellings is obtained by small holes,
which are the orifices of galleries many feet in length: the holes are surrounded by pellets of
sand projecting somewhat above the general surface, and consequently making the places
conspicuous. The subterranean works occupy an area of fifty or a hundred square yards
excavated at a depth of three to six feet. In these immense nests there exists a form of worker,
of very small size, that never comes to the surface.[70]
Pogonomyrmex barbatus and other species have been observed to do harvesting in North
America. After the workers of P. barbatus have taken the seeds into the nest they separate the
husks and carry them out, depositing them on a heap or kitchen-midden, formed near by.
M‘Cook has witnessed and described the process of stripping the seeds.
Certain genera—e.g. Aphaenogaster, Pheidole—exhibit great disparity in the forms of the

workers, some of which are of size much superior to the others, and possess
disproportionately large heads; these large individuals are found in the same nest as the
smaller forms. All the intermediate forms may frequently be found, and at the same time, in
the genus Aphaenogaster; but in Pheidole intermediates are of the utmost rarity.
The genus Cremastogaster is remarkable on account of the shape of the hind body and its
articulation, which give the abdomen the appearance of being put on upside down. This mode
of articulation may allow the Insect to threaten its enemies when they are in front of it; but it is
doubtful whether the Cremastogaster possesses an effective sting.
Fig. 71—Cremastogaster tricolor, worker. A, with abdomen extended; B, uplifted.
ii. The group Attini is distinguished by the presence of a carina near the eye, by the antennae
being inserted at a moderate distance from one another, by the clypeus being prolonged
backwards between them; and by the absence of a sting. The group is not represented in
Europe, but in Tropical America the ants belonging to it are amongst the most important of
natural objects. The species of the genus Atta (usually styled Oecodoma) are the formidable
leaf-cutting ants of America. They occur in enormous colonies in certain places, and will in a
short time completely strip a tree of its leaves. As they appear to prefer trees of a useful kind,
especially those planted by man, their ravages are often of the most serious nature. The
natives, feeling it hopeless to contend with these Insect hordes, only too frequently abandon
all attempts to cultivate the trees and vegetables the Insects are fond of. Both Bates and Belt
have given accounts of some points in the economy of these ants. They are amongst the
largest of the Formicidae, the females in some cases measuring about two and a half inches
across their expanded wings; the males are much smaller, but are less dissimilar to their
partners than is usual among ants. The workers, on the other hand, are so extremely different,
that no one would suppose them to be at all related to the males and females (see vol. v. Fig.
339).
The mode of operation of these ants is to form paths from their formicary extending for a
considerable distance in various directions, so that they have a ready access to any spot in a
district of considerable extent; when a tree or bush is found bearing leaves suitable for their
purposes, the worker ants ascend it in large numbers and cut up the leaves by biting out of
them pieces similar in size and shape to a small coin; these pieces are then carried back in
the jaws of the ants to their nests; the ant-paths are therefore constantly traversed by bands of
little creatures carrying burdens homewards, or hurrying outwards in search of suitable trees.
The formicaries are of considerable size, and are described as consisting of low mounds of
bare earth of considerable extent. Bates speaks of as great a circumference as forty yards;
these accumulations of earth have frequently an appearance different from the adjoining soil,
owing to their being formed of subsoil brought up from below; they are kept bare by the ants
constantly bringing to and depositing on the surface fresh material resulting from their
subterranean excavations. The true abodes, beneath the earth, are of greater extent than the
mounds themselves, and extend to a considerable depth; they consist of chambers connected
by galleries.
The leaf-cutting ants extend their range to North America, and M‘Cook has recently called
attention to a case there in which A. fervens made an underground route at an average depth
of 18 inches, and at an occasional depth of 6 feet, extending 448 feet entirely beneath the
earth, after which it was continued for 185 feet to reach a tree which the ants were engaged in
defoliating. This route, extending altogether to a length of more than 600 feet, presented only
a very slight deviation from a straight line drawn between the point of departure and the object
to be attained. By what sense this ant was enabled to make a subterranean tunnel in a
straight line to a desired object situated at so great a distance, we know not.
The use the leaf-cutting ants make of the enormous amount of material they gather was for
long a subject of debate, and has only recently been ascertained by the observations of
Möller. After being carried to the nest the pieces of leaves are cut into small fragments by
another set of workers and formed into balls, which are packed in various parts of the nest,
and amongst which the mycelium of a fungus—Rozites gongylophora—ramifies. This fungus
the ants cultivate in the most skilful manner: they manage to keep it clear from mouldiness
and bacterial agents, and to make it produce a modified form of growth in the shape of little
white masses, each one formed by an agglomeration of swellings of the mycelium. These
form the chief food of the colony. Möller ascertained by experiment that the results were due to
a true cultivation on the part of the ants: when they were taken away from the nests, the
mycelium produced two kinds of conidia instead of the ant-food.
Many details of the economy of these leaf-cutting ants are still very imperfectly known. The
large-headed forms, called soldiers, have been the subject of contradictory statements; Bates
having concluded from his own observations that they are harmless, while Mr. J. H. Hart
assures us that they are very fierce and vindictive, and inflict very serious wounds by biting
(the Attini do not sting). We anticipate that the observations of both these naturalists will prove
to be substantially correct, and that the differences in habits will be found to be owing to
distinctions in the conditions of the community. In connection with this point we may remark
that the function of the excessively large heads of certain kinds of soldier-ants is still obscure.
In the East Indian Pheidologeton diversus the big soldiers are quite one hundred times as
large as the smaller workers. As these latter bite viciously it would naturally be supposed that
their gigantic confrères with enormous heads would be warriors of a most formidable nature;
but, as a matter of fact, the giants are unable to bite even when they try to do so. Aitken has
somewhere suggested that these enormous individuals play the part of state elephants; and
we have been informed by Colonel Bingham that the small ants may frequently be seen riding
in numbers on their unwieldy fellows. We anticipate however, that some other function will be
found to exist for these forms with enormous heads. An examination of their organs of sense
and of voice is very desirable.
Details of the modes in which the great communities of the leaf-cutting Attidae are maintained,
are still wanting. The females do not, we have been informed by Mr. Hart, possess any
considerable powers of aftergrowth, so that there is no reason to suppose them to be
unusually prolific. At certain seasons great swarms of winged individuals are produced, and
after leaving the nests pair in the manner of our European Myrmica. Possibly the females may,
after losing their wings, again enter the large communities. Von Ihering states that the workers
of Atta lundi are fertile.
iii. The group Pseudomyrmini includes the genera Pseudomyrma and Sima, which are by
some entomologists treated as but a single genus. The antennae are inserted near together
on the front of the head; there is no carina on the head external to their insertion, and the
clypeus does not extend forwards between them. The Insects are usually of elongate form,
possess a sting, and have a naked pupa. The group occurs in both hemispheres, but is
exclusively exotic, and but little is known of the habits of its members. Forel has recently
observed that numerous species live inside dried stems of grass or in hollow twigs, and are
beautifully adapted for this mode of life by their elongate form, some of them being as slender
as needles. Some interesting observations have been made in Nicaragua by Belt on
Pseudomyrma bicolor and its relations with an acacia-tree, in the thorns of which it lives. The
acacia in question is called the bull's-horn thorn, because the branches and trunk are armed
with strong curved spines, set in pairs, and much resembling the horns of the quadruped
whose name they bear. The ant takes possession of a thorn by boring a small hole near the
distal extremity, and forms its nest inside. The leaves of this plant are provided with glands
that secrete a honey-like fluid, which it appears forms the chief, if not the sole, subsistence of
the ant. Belt considers that the presence of the ant is beneficial to the acacia; he supposes
that the ants assume the rights of proprietors, and will not allow caterpillars or leaf-cutting ants
to meddle with their property; the leaves are, he thinks, so preserved to the benefit of the tree.
Fig. 72—Sima rufo-nigra and its associates. A, winged female; B,
worker, of the ant; C, Rhinopsis ruficornis; a fossorial wasp of the
sub-family Ampulicides; D, a spider, Salticus sp. The coloration is
extremely similar in all these creatures.
Rothney has given some particulars of the habits of Sima rufo-nigra,

an ant of this group that appears to be not uncommon near Calcutta,
where it lives on the trunks of trees in company with a spider and a
wasp that greatly resemble it in form and in colour. The three
creatures seem to associate together on amicable terms; indeed the
wasp and the ant occasionally indulge in wrestling matches without
doing one another any serious harm. In connection with this fact we
may observe that other species of ants have been observed to
indulge in sports and feats of agility.
S. leviceps, an Australian species of this genus, is furnished with a

stridulating file that has the appearance of being constructed so as to
produce two very different kinds of sounds.
Fig. 73—Stridulating file of Sima leviceps.
iv. The Cryptocerini are distinguished from other ants by their

antennae being inserted at the sides of the head, where they are
placed between ridges or in a groove into which they can be
withdrawn; when in some cases they are entirely concealed. These
ants assume a great variety of shapes and forms, some of which
look almost as if they were the results of an extravagant imagination.
The skeleton is usually much harder than it is in other ants; the
abdomen consists almost entirely of one very large segment, there
being, however, three others visible at its extremity; these segments
can be only slightly protruded, and the ants have no power of
stinging. They are probably most of them arboreal in their habits.
Nearly all of the known forms are exotic. According to the
observations of Bates the species of the genus Cryptocerus in the
Amazons Valley may frequently be observed in dry open places on
low trees and bushes, or running on branches of newly felled trees;
they also visit flowers abundantly. The species generally are wood-
borers, usually perforating the dead branches of trees. C. atratus has
been observed to construct its nests in the dead, suspended
branches of woody climbers; a number of neatly drilled holes are all
that can be seen externally; but, inside, the wood is freely perforated
with intercommunicating galleries. Each community appears to
consist of a single female and two kinds of workers; the latter in
some species are quite unlike each other, differing in the form of the
head, and in the armature of the thorax and nodes of the peduncle.
The species of Cryptocerus appear to be omnivorous, and are
frequently attracted by the excrement of birds. The pupae are not
enclosed in a cocoon. In the South of Europe two very minute ants,
of the genera Strumigenys and Epitritus, belonging to this family, are
met with under very large stones partly embedded in the earth. They
are of the greatest rarity.
Fig. 74—Cyrptocerus atratus, worker. Amazons. The compressed first

joint of the hind foot is shown at a and b in different positions.
Sub-fam. 4. Ponerides.—Hind body elongate, furnished with
one node at the base, and having also great capacity of
movement between the first and second segments, between
which there is usually a slight constriction. Sting well developed.
This sub-family includes numerous genera and about 400 species.

The Ponerides have an elongate hind-body; the second segment
behind the node is capable of great movement in and out of the
preceding segment, and for this purpose is furnished with a basal
portion slightly more slender than the apical part; this basal part is
usually concealed within the more anterior segment, the hind margin
of which embraces it very closely. On the middle of the dorsal aspect
of this articulation there is usually placed a stridulating organ,
consisting of an elongate band or patch of very fine lines; this gives
out a sound when the second segment is moved in and out of the
first at a time when the posterior edge of the latter is slightly
depressed.
We follow Forel in including the Australian bull-dog ants—Myrmecia

—in Ponerides, as well as the Odontomachi. The former have,
however, a definite pedicel, consisting of two nodes (Fig. 76). In the
Odontomachi the mandibles are approximate at their bases, being
inserted on the middle of the front of the head (Fig. 77).
This sub-family includes a considerable number of species, and is

found in all parts of the world. Extremely little is known as to the
habits, but the true Ponerides do not, so far as is known, occur in
large communities, and it seems probable that they are destitute of
the powers of combined action that are so remarkable in the
Camponotides, and in some of the Myrmicides and Dorylides. Most
of the species that have been described are known by only one sex,
so that very little knowledge exists as to the sexual distinctions; but
from the little that is known it would appear that the three sexual
forms are not so differentiated as they are in most of the
Camponotides and Myrmicides.
Fig. 75—Dinoponera grandis, worker. Amazons.
The species of the genus Leptogenys are believed by Emery and

Forel to possess an apterous female. Mr. Perkins has observed that
the Hawaiian L. falcigera has workers with different kinds of sting,
but no true female. Males of this species are, however, abundant.
Wroughton has recently discovered that one member of this genus is
of Termitophagous habits, but this is not the case with L. falcigera.
Dinoponera grandis (Fig. 75) is the largest of the Ponerides, its
workers attaining an inch and a quarter in length. This Insect,
according to Bates, marches in single file in the thickets at Pará; its
colonies consist of a small number of individuals, and are
established at the roots of slender trees. The effects of its powerful
sting are not so serious as is the case with some of the smaller ants.
In Britain we have only two representatives of the sub-family, viz.

Ponera contracta, a small ant of dirty-yellow colour, found rarely in
the Southern counties, living in moss or under stones. Its colonies
consist of only a few individuals; Forel giving fifty as the highest
number he has observed. The second species, P. punctatissima,
presents the almost unique peculiarity of possessing two forms of
the male sex, one of them resembling the worker in most of its
peculiarities, and in being destitute of wings, while the other is
winged, as is usual in male ants. In the island of St. Vincent another
species of Ponera has been discovered having an apterous and
worker-like male, and was named by Forel P. ergatandria.[71] The
discovery of this form has led him to express some doubt as to
whether Ponera punctatissima has two forms of males; but it seems
probable that it really is so, the ergatoid males being produced under
somewhat different circumstances from the normal males. We shall
subsequently see that Cardiocondyla and a few other Myrmicides
exhibit an analogous peculiarity.
The genus Myrmecia is confined to the Australian continent and

Tasmania, and includes a considerable number of species of large
and moderate-sized ants, the classification of which has been a
subject of difference of opinion. This has arisen from the fact that the
nodes of the abdominal pedicel are more similar to those existing in
the Myrmicides than to those of the typical Ponerides. There are,
however, some American members of the latter sub-family
(Paraponera clavata, e.g.) that differ but little in this point from
Myrmecia, and, moreover, the pupae of Myrmecia are enclosed in a
cocoon, while in the Myrmicides they are usually naked. On the other
hand the nests are, it appears, very large and populous, more like
what exists in the Myrmicides; there is no true stridulating organ on
the first abdominal segment. The genus is therefore one of those
interesting anomalies that form so large a proportion of the
Australian fauna, and will probably be ultimately treated as a distinct
sub-family. There are about thirty species.
The ants of this genus are well known to the residents in Australia,
where they are called "bull-dog ants." They form large mounds of
earth for their nests. The workers, and females (Fig. 76) are much
alike except during the period when the latter are still carrying their
wings. The males, however, differ considerably, being of more
slender form, and possessing only insignificant mandibles, and
straight antennae with a quite short basal joint.
Fig. 76—Myrmecia pyriformis. Australia. Female after casting wings.

Metaheuristics in Machine Learning Theory and Applications Studies in Computational Intelligence 967 Diego Oliva (Editor) Download PDF

Uploaded by

Copyright:

Available Formats

Metaheuristics in Machine Learning Theory and Applications Studies in Computational Intelligence 967 Diego Oliva (Editor) Download 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

Metaheuristics in Machine Learning Theory and Applications Studies in Computational Intelligence 967 Diego Oliva (Editor) Download PDF

Uploaded by

Copyright:

Available Formats

Full download test bank at ebookmeta.

Metaheuristics in Machine Learning Theory and

Download More ebooks from https://ebookmeta.com

Artificial Intelligence Theory and Applications 973

Modern Metaheuristics in Image Processing 1st Edition

Data-Driven Evolutionary Optimization: Integrating

Machine learning and artificial intelligence with

Reliability Engineering and Computational Intelligence

Computational Intelligence in Machine Learning Select

Swarm Intelligence and Machine Learning: Applications

Combating Fake News with Computational Intelligence

More information about this series at http://www.springer.com/series/7092

ISSN 1860-949X ISSN 1860-9503 (electronic)

such as: Metaheuristics, Machine learning, Soft Computing, Neural Networks,

Guadajalara, Mexico Diego Oliva

This book “MAML2020” collects several hybridized metaheuristics (MHs) with

Cross Entropy Based Thresholding Segmentation of Magnetic

A Review of Metaheuristic Optimization Algorithms in Wireless

Image Classification with Convolutional Neural Networks . . . . . . . . . . . . . 445

Solar Irradiation Changes Detection for Photovoltaic Systems

Omar Zárate and Daniel Záldivar

O. Zárate (B) · D. Záldivar

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 1

This chapter evaluates three recently published evolutionary algorithms in

1.1 Applied Metaheuristic Algorithms

Table 1 Parameters for MFO, SCA and SFO algorithms

2 Moth-Flame Optimizer Algorithm

Fig. 1 Moth flame

Fig. 2 Fly spirally around

Computational complexity of the MFO algorithm depends on the number of moths

3 Sine Cosine Optimization Algorithm

Population-based optimization techniques create multiple initial random candidate

Fig. 3 SCA with the range in [−2, 2]

where r4 is a random number in [12] (Fig. 3).

4 Sunflower Optimization Algorithm

An important nature-based metaphor for optimization can be found here; the

The step of the sunflowers on the direction s is calculated by:

di = γ × Pi (||X i + X i−1 ||) × ||X i + X i−1 || (8)

||X max − X min ||

→ X i+1 =→ X i +di × → Si (10)

Sunflower Optimization Algorithm initially generates a uniform/random popula-

Ferreira Gomes, Simões da Cunha and Ancelotti [14].

5 Image Segmentation Using Minimum Cross Entropy

where Q = {Q 1 , Q 2 , Q 3 , . . . , Q N } and P = {P1 , P2 , P3 , . . . , PN } are two proba-

I (u, v) containing intensity values in the interval [0, L − 1] will generate an H

h(i) = car d{(u, v)|I (u, v) = i } (12)

where L is the maximum intensity value of the image histogram (L = 256 in a

The proposed segmentation method has been implemented considering the

th opt = arg min th (D(th)) (16)

Diego et al. [18].

6 Results of the Experiments

Table 2 Best thresholds found by MFO, SCA y SFO

The proposed approach is an interesting alternative to traditional prostate magnetic

Table 3 Objective function values of the various ECM implementations

MRI 01, nt = 4 MRI 01, nt = 5

MRI 01, nt = 8 MRI 01, nt = 16

MRI 02, nt = 2 MRI 02, nt = 3

MRI 02, nt = 4 MRI 02, nt = 5

MRI 03, nt = 2 MRI 03, nt = 3