Metaheuristics for Data Clustering and Image

Segmentation Meera Ramadas download

https://textbookfull.com/product/metaheuristics-for-data-
clustering-and-image-segmentation-meera-ramadas/

Download more ebook from https://textbookfull.com

We believe these products will be a great fit for you. Click
the link to download now, or visit textbookfull.com
to discover even more!

Recent Advances in Hybrid Metaheuristics for Data

Clustering Sourav De (Editor)

https://textbookfull.com/product/recent-advances-in-hybrid-
metaheuristics-for-data-clustering-sourav-de-editor/

Metaheuristic Algorithms for Image Segmentation Theory

and Applications Diego Oliva

https://textbookfull.com/product/metaheuristic-algorithms-for-
image-segmentation-theory-and-applications-diego-oliva/

Text Segmentation and Recognition for Enhanced Image

Spam Detection: An Integrated Approach Mallikka
Rajalingam

https://textbookfull.com/product/text-segmentation-and-
recognition-for-enhanced-image-spam-detection-an-integrated-
approach-mallikka-rajalingam/

Modern Technologies for Big Data Classification and

Clustering 1st Edition Hari Seetha

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

https://textbookfull.com/product/medical-image-recognition-
segmentation-and-parsing-machine-learning-and-multiple-object-
approaches-1st-edition-zhou/

Metaheuristics for Medicine and Biology 1st Edition

Amir Nakib

https://textbookfull.com/product/metaheuristics-for-medicine-and-
biology-1st-edition-amir-nakib/

Adaptive Resonance Theory in Social Media Data

Clustering Roles Methodologies and Applications Lei
Meng

https://textbookfull.com/product/adaptive-resonance-theory-in-
social-media-data-clustering-roles-methodologies-and-
applications-lei-meng/

Exam Preparatory Manual for Undergraduates Pathology

4th Edition Ramadas Nayak Rakshatha Nayak

https://textbookfull.com/product/exam-preparatory-manual-for-
undergraduates-pathology-4th-edition-ramadas-nayak-rakshatha-
nayak/

Multisensor Image Fusion and Data Mining for

Environmental Remote Sensing First Edition Bai

https://textbookfull.com/product/multisensor-image-fusion-and-
data-mining-for-environmental-remote-sensing-first-edition-bai/
Intelligent Systems Reference Library 152

Meera Ramadas
Ajith Abraham

Metaheuristics
for Data
Clustering
and Image
Segmentation
Intelligent Systems Reference Library

Volume 152

Series editors
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: kacprzyk@ibspan.waw.pl

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

Artificial Intelligence, University of Technology, Sydney, NSW, Australia;
Faculty of Science, Technology and Mathematics, University of Canberra,
Canberra, ACT, Australia;
Faculty of Science, Liverpool Hope University, Liverpool, UK
KES International, Shoreham-by-Sea, UK
e-mail: jainlakhmi@gmail.com; jainlc2002@yahoo.co.uk
The aim of this series is to publish a Reference Library, including novel advances
and developments in all aspects of Intelligent Systems in an easily accessible and
well structured form. The series includes reference works, handbooks, compendia,
textbooks, well-structured monographs, dictionaries, and encyclopedias. It contains
well integrated knowledge and current information in the field of Intelligent
Systems. The series covers the theory, applications, and design methods of
Intelligent Systems. Virtually all disciplines such as engineering, computer science,
avionics, business, e-commerce, environment, healthcare, physics and life science
are included. The list of topics spans all the areas of modern intelligent systems
such as: Ambient intelligence, Computational intelligence, Social intelligence,
Computational neuroscience, Artificial life, Virtual society, Cognitive systems,
DNA and immunity-based systems, e-Learning and teaching, Human-centred
computing and Machine ethics, Intelligent control, Intelligent data analysis,
Knowledge-based paradigms, Knowledge management, Intelligent agents,
Intelligent decision making, Intelligent network security, Interactive entertainment,
Learning paradigms, Recommender systems, Robotics and Mechatronics including
human-machine teaming, Self-organizing and adaptive systems, Soft computing
including Neural systems, Fuzzy systems, Evolutionary computing and the Fusion
of these paradigms, Perception and Vision, Web intelligence and Multimedia.

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

Meera Ramadas Ajith Abraham
•

Metaheuristics for Data

Clustering and Image
Segmentation

123
Meera Ramadas Ajith Abraham
Information Technology Scientific Network for Innovation and
University College of Bahrain Research Excellence
Manama, Bahrain Machine Intelligence Research Labs
(MIR Labs)
Auburn, WA, USA

ISSN 1868-4394 ISSN 1868-4408 (electronic)

Intelligent Systems Reference Library
ISBN 978-3-030-04096-3 ISBN 978-3-030-04097-0 (eBook)
https://doi.org/10.1007/978-3-030-04097-0

Library of Congress Control Number: 2018961197

© Springer Nature Switzerland AG 2019

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

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

artificial bee colony (ABC), and particle swarm optimization (PSO) have emerged
as potential algorithms for dealing with complex optimization problems, which are
otherwise difficult to solve by traditional methods. Researchers have continued with
their work to create different variants of the existing algorithms to find the most
optimal method. Differential evolution (DE) is considered to be a dominant tech-
nique for optimization and is being used to solve various real-time problems. DE is
a powerful, population-based, stochastic optimization algorithm. The mutation
strategy of DE algorithm is an important operator as it aids in generating a new
solution vector. Differential evolution algorithm has undergone various changes and
evolution by means of numerous researches. The performance of each of these
algorithms depends on the changes in mutation and crossover strategies. In this
research work, DE and their modified variants are applied to clustering of data and
images. Evolutionary algorithms work in a robust and efficient manner for clus-
tering. The ever-growing demand for acquiring, storing, and processing a huge
amount of data for analysis and the need for producing customized results have
given space for researchers to invent innovative ways for handling such large
volume of data. Clustering is an area where the data formed in clusters exhibit
efficient customer segmentation, organization of document, information retrieval,
extraction of topics, classification, collaborative filtering, visualization, and
indexing. Quick and accurate clustering of documents is instrumental in the area of
information retrieval system. In the case of images, image segmentation is per-
formed where grouping or clustering of bitmap images into different classes is done
such that each class gives one particular information about the bitmap image within
the entire image collection. In the past few years, image segmentation has emerged
as an active area of research in different domains including medical imaging,
satellite imaging. In computer vision, image segmentation is the process of
extracting the data from the image, which is very simple and easier to analyze.
Image segmentation is a complex task in computer science because it uses the
whole image as a data input, which needs to be processed further. Clustering of data
and image can be modeled as a nonlinear optimization problem for which suitable

v
vi Preface

techniques are needed to attain the global solution by overcoming the problem of
local maxima. The volume is divided into eight chapters with Introduction as
Chap. 1 and Conclusion as Chap. 8. In Chap. 2, detailed studies of the various
traditional algorithms are elaborated. Also, the detailed literature study on the
variants of the evolutionary algorithm and its application on clustering is given.
Chapter 3 deals with the newly created variant of differential evolution named
revised differential evolution (ReDE) and its comparison with the classical mutation
strategy of DE. Chapters 4 and 5 elaborate new variants of evolutionary algorithm
named as search strategy flower pollination algorithm with differential evolution
(ssFPA/DE) and forced strategy differential evolution (FSDE) and its usage on data
clustering. Chapter 6 deals with the usage of another variant of DE on the seg-
mentation of weather images. The algorithms are elaborately studied and are tested
on a large number of test images with their performance carefully analyzed in
reference to various performance measures including statistical analysis. It is
observed that although all the algorithms are suitable for dealing with clustering,
DE and its variants perform better.

Manama, Bahrain Meera Ramadas

Auburn, USA Ajith Abraham
Contents

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

vii
viii Contents

3 Revised Mutation Strategy for Differential Evolution

Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2 Revised Mutation Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3 Experimental Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4 Graphical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.5 Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4 Search Strategy Flower Pollination Algorithm with Differential
Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2 Proposed Hybrid Method ssFPA/DE . . . . . . . . . . . . . . . . . . . . . . 68
4.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.4 Graphical Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 ssFPA/DE in Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.6 Experimental Results on Clustering . . . . . . . . . . . . . . . . . . . . . . . 82
4.6.1 Cluster Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.6.2 Validation Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.6.3 Graphical Representation . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5 Forced Strategy Differential Evolution Used for Data
Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2 Proposed Variant of Differential Evolution . . . . . . . . . . . . . . . . . . 97
5.3 Test Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4 FSDE in Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.5 Experimental Results on Clustering . . . . . . . . . . . . . . . . . . . . . . . 106
5.5.1 Cluster Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.5.2 Validation Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.5.3 Graphical Representation . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6 Reconstructed Mutation Strategy for Differential Evolution
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.2 RDE Mutation Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.3 Experimental Results and Statistical Analysis . . . . . . . . . . . . . . . . 123
6.4 Multi-level Image Thresholding . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Contents ix

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

6.6 Test Results on Image Thresholding . . . . . . . . . . . . . . . . . . . . . . 129
6.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7 Enhanced Differential Evolution with Fuzzy c-Means
Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.2 Fuzzy c-Means Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.3 Enhanced Differential Evolution . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.4 Experimental Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.4.1 Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.5 eDE in FCM Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.6 Experimental Results on Clustering . . . . . . . . . . . . . . . . . . . . . . . 141
7.6.1 Validation Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.6.2 Graphical Representation . . . . . . . . . . . . . . . . . . . . . . . . . 152
7.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
8 Conclusion and Future Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.1 Future Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Chapter 1
Introduction

In the field of computer science and engineering, metaheuristic is a high level pro-
cedure, which finds, selects or generates a low level procedure that will give an
optimized solution to a complex problem. It is a high-level problem independent
framework that gives a set of strategies for developing heuristic algorithms. Meta-
heuristics are used for combinatorial optimization like genetic algorithm, evolution-
ary algorithms, particle swarm optimization, simulated annealing etc. It explores the
search space to find the near optimal solutions. Optimization is the selection of best
element from the set of alternative solutions.
Two components in metaheuristics are local exploitation and global exploration.
Exploitation consists of probing a limited region of search space in the hope of
improving the solution at hand. Exploration consists of probing a much larger por-
tion of the search space with the hope of finding other solutions that are yet to be
refined. Exploration looks for new solutions in new search space while exploita-
tion uses existing solutions and makes refinements to it to improve its fitness. Global
exploration explores the different feasible region in the entire search space for obtain-
ing global optimum. Local exploitation uses local information from local search to
ensure maximum convergence. These two concepts are conflicting and a balance need
to be maintained. There are different types of metaheuristics namely local search and
global search metaheuristics, single solution and population based metaheuristics,
hybridization and memetic algorithms, parallel metaheuristics and nature inspired
metaheuristics.
In past decades, several metaheuristic techniques have evolved by exploring vari-
ous concepts inspired from nature called nature inspired metaheuristics. These tech-
niques essentially mimic the characteristics that uniquely exist in the Mother Nature.
Nature has in itself evolved for many years continuously developing creative meth-
ods for problem solving and for adaption in this ever changing world environment.
During the Second World War, Alan Turing used heuristic algorithm for the first

© Springer Nature Switzerland AG 2019 1

M. Ramadas and A. Abraham, Metaheuristics for Data Clustering and Image
Segmentation, Intelligent Systems Reference Library 152,
https://doi.org/10.1007/978-3-030-04097-0_1
2 1 Introduction

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

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

ity, the information about the type of search landscape is focused. In algorithm-based
type, objective function is treated as a black box type and it uses the information dur-
ing iteration to find the best optimal solution. At times, combination of both the types
is also used.
Most of the metaheuristic approaches like DE, FPA has been applied over various
optimization problems like clustering. These approaches can find an optimal partition
of the dataset, when the number of clusters is given. In this research, variants of
evolutionary algorithm are introduced and these variants are applied in the field of
data clustering and image segmentation. Clustering is a process of breaking down
large population into smaller groups that are similar in character. It is mainly used for
data compression. This maximizes the inter class similarity. Good clustering has a
high within cluster similarity but low inter cluster similarity. It is used for improving
precision in information retrieval. Clustering is a type of unsupervised learning where
the data has no target attribute. It studies how system can learn to represent particular
input patterns in a way that reflects the statistical structure of the overall input pattern
assembly.
Similarly, clustering of text documents plays a vital role in efficient document
organization, summarization, topic extraction, information retrieval, customer seg-
mentation, classification, collaborative filtering, visualization and indexing. Fast and
accurate clustering of documents plays an important role in the field of automatic
information retrieval system. It is considered as a multi objective optimization prob-
lem. It involves an iterative task of trial and error. Clustering can be classified as hard
clustering and soft clustering. In hard clustering, each object belong to one cluster
or do not belong to any cluster. In soft clustering, each object may belong to more
than one cluster. Clustering algorithm can be classified into different types namely
exclusive clustering, overlapping clustering, hierarchical clustering and probabilistic
clustering. In exclusive clustering, each data is grouped into a particular cluster based
on its similarity to that cluster. K-means algorithm is an example for exclusive clus-
tering. Here, a given set of N data is partitioned into k different clusters. Grouping
of data is done by minimizing the Euclidian distance between the data and centroid.
It is one of the simplest unsupervised learning algorithms used for clustering. This
algorithm is significantly sensitive to initial randomly selected centroid.
Evolutionary algorithm can also be applied on to the area of unsupervised image
clustering where the implementation of cluster analysis can aid in easier understand-
ing of the images. Image clustering is often applied to segment and have easier
understanding of images based on the desired region of interest. The process of
segmentation is done using clustering techniques. Image segmentation refers to the
partitioning or dividing of a digital image into several smaller parts or segments in
order to study a given image in a detailed manner. Image segmentation is required to
study in detail some particular features (areas of interest) of a digital image. Thresh-
olding is one of the simplest techniques used for image segmentation. It is the method
of creating binary images from the given image. It forms an important and exigent
part of image processing and requires an exhaustive and robust search technique for
its implementation. Thresholding on colored images is done by manipulating the
4 1 Introduction

color components based on color spaces. Problems like thresholding are classified as
unsupervised learning and such issues can be solved using evolutionary algorithms.
Though evolutionary algorithms are being widely applied and accepted in diverse
areas like engineering, medicine, commerce, business etc., in reality they give only
marginal performance. Some of the major reasons attributed to the marginal perfor-
mance are inappropriate selection of parameters, inappropriate representation etc.
Hence, here there is tremendous scope and an urgent need to develop more hybrid
evolutionary algorithms that has the ability to optimize the performance of direct
evolutionary approach.
Hybridization of evolutionary algorithms are needed for the following:
i. improving standard of output
ii. improving the conduct
iii. to integrate evolutionary algorithm to a larger system.

1.1 Major Objectives

The core objective of this work is to create new variants of evolutionary algorithm
to solve specific optimization task. Subsequently, compare the newly created variant
with some of existing metaheuristic techniques that are commonly available to check
the efficiency of the new variants.
The basic objectives of this work are:
1. To propose the hybrids of Differential Evolution algorithm.
2. To perform the statistical analysis for each of the new variant to validate its
efficiency.
3. These proposed hybrids will be applied on to clustering of data and images to
demonstrate their application capability.

1.2 Organization

This volume is structured into several chapters with the first two chapters focusing
on the supporting information with respect to this research work and each of the
subsequent chapters explaining in detail the research work that was performed along
with the research outcome.
The below is the organization for the following sections:
Chapter 2 contains a detailed study of Differential Evolution and Flower Polli-
nation algorithm. This chapter not only explain in detail various steps involved in
these algorithms but also an in-depth study is done on the basic concepts of data
clustering and image segmentation. This chapter also discusses the literature review
about previous research works carried out on different variants created in evolution-
1.2 Organization 5

ary algorithms and their findings. This chapter also has references to the works in
the field of data clustering and image segmentation through clustering.
Hereafter, the following chapters form the core section of this work where each of
the variants created using Differential Evolution and Flower Pollination Algorithm
is explained in detail in each of Chaps. 3, 4, 5, 6 and 7 respectively. Each of these
chapters also comprise of the results obtained by applying these variants in the field
of data clustering and image segmentation.
Chapter 8 explains in brief the conclusion of the research work and future scope
of work that can follow up based on this research work.

Reference

Fogel, L.J., Owens, A.J., Walsh, M.J.: Intelligent decision making through a simulation of evolution.
Behav. Sci. 11(4), 253–272 (1966)
Chapter 2
Metaheuristics and Data Clustering

In this Chapter, the initial Sects. 2.1–2.4 give detailed analysis on types of metaheuris-
tics namely Genetic Algorithm, Particle Swarm Optimization, Differential Evolution
algorithm and Flower Pollination algorithm. Section 2.5 elaborates the basic data
clustering technique and Sect. 2.6 summarizes about the image segmentation. Sec-
tions 2.7–2.9 explains the works related to metaheuristics and their application on
data and image clustering respectively.

2.1 Genetic Algorithm

Holland [1] introduced genetic algorithms. Genetic algorithm (GA) is used to produce
high quality solutions to search and optimize the problems using the concepts of
mutation, crossover and selection. GA has been successfully applied in many areas
like clustering, image segmentation, image watermark etc.

2.1.1 Basics of Genetic Algorithms

Genetic algorithms is based on Darwin’s theory of survival of the fittest. The basic
idea behind GA is natural selection and natural genetic systems in which fittest
individuals will survive and the remaining individuals will be discarded. In GA, take
a set of candidate solution and iteratively refine them by alternating and selecting
the good solution for next generation. Select the candidate solutions on the basis of
fitness function to evaluate quality. The five phases of genetic algorithm are initial
population, fitness function, selection, crossover and mutation. Around the nature’s
scanty resources, competition within individuals lead to fittest dominating the weaker.
They are a computerized searching and optimization algorithm that uses natural
selection and natural genetics technique. They take large search spaces and look
© Springer Nature Switzerland AG 2019 7
M. Ramadas and A. Abraham, Metaheuristics for Data Clustering and Image
Segmentation, Intelligent Systems Reference Library 152,
https://doi.org/10.1007/978-3-030-04097-0_2
8 2 Metaheuristics and Data Clustering

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

defining optimization function, defining and implementing genetic representation,
and genetic operators.
GA begins with the chromosome representation of a parameter set{x 1 ,x 2 ,…x n };
generally, the chromosomes are designed as strings of 0’s and l’s. GAs employs three
basic operators over a limited number of strings for finding the global near-optimal
solution. These operators are:
(i) Reproduction/Selection
(ii) Cross-over
(iii) Mutation.

2.1.2 Reproduction

According to the objective function, copy the individual strings for reproduction
process. The strings created by reproduction depends upon the Darwinian survival
theory, where the fitter strings have more chance to get a higher number of offspring
in the succeeding generation. All the successful strings form a mating pool which
will behave like a new tentative population.

2.1.3 Crossover

Reproduction operation forms a mating pool of strings; after this phase, these strings
will form offspring for the new generation using crossover operation. Crossover
process consists of two steps: (1) all the members of the mating pool will mate
randomly and (2) each pair of strings undergo crossover as follows:
i. Select an integer position k at random between l and l − 1 where, l is the string
length greater than 1.
ii. Create two new strings by swapping all characters from position k + 1 to l.
Let a and b be two strings selected for the crossover operation from mating pool and
the random number generated be 10 (ten).

a  110001010101000 . . . 0111110001
b  100010111011101 . . . 0011010100

Then the newly produced offspring by swapping all characters after position 10
will be denoted as a and b :

a  110001010111101 . . . 0011010100
b  100010111001000 . . . 0111110001
2.1 Genetic Algorithm 9

2.1.4 Mutation

In the simple GA, mutation is the occasional random alteration of the value of a string
position. It helps to prevent the irrecoverable loss of potentially important genetic
material. Select a random bit position of a random string and replace it by another
character from the population.
For example, let the fifth bit of string ‘a’ given above, be selected for mutation.
Then the transformed string after mutation will be:

a  110011010101000 . . . 0111110001

The flow diagram for genetic algorithm is as given below in Fig. 2.1.
The general pseudo code for genetic algorithm is given below:

Initialize population with random candidate solution

Evaluate fitness of each candidate solution

While termination condition not true do

Select the best individuals to be used by genetic operators

Generate new individuals using crossover and mutation

Evaluate fitness of new individuals

Replace the worst individuals of population by best new individuals

End

Initialize Population

Selection Crossover Mutation

Termination
condition
No

Yes

End

Fig. 2.1 Flow diagram for genetic algorithm

10 2 Metaheuristics and Data Clustering

2.2 Particle Swarm Optimization Algorithm

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

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

+ c2 · rand (0, 1) · (gbesti − presenti ) (2.1)
presenti+1  presenti + vi,k+1 (2.2)

where, c1 and c2 are the weighted factor, presenti is the current position of the
candidate, vi,k+1 is the velocity of the candidate.
The general algorithm for PSO is given as follows:
2.2 Particle Swarm Optimization Algorithm 11

Initialize velocity and position of each particles

While stopping criteria not attained do
For each particle
Calculate fitness value
If fitness value is better than pbest value in history
Set current value as new pbest
End
Choose the particle with best fitness value of all particles as gbest
For each particle
Calculate particle velocity according to eq. 2.1
Calculate particle position according to eq. 2.2
End
End
End

PSO algorithm differs from other evolutionary algorithms as it doesn’t have the
operations of mutation and crossover. The particles revise itself by using the inter-
nal velocities. This algorithm also utilizes memory space unlike other evolutionary
algorithms. Flow diagram for PSO is given in Fig. 2.2.

2.3 Differential Evolution

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

and Price in [3]. It is used for minimizing non-differential, non-linear, continuous
space functions with real valued parameters. The most important characteristic of
DE is that it uses the differences between randomly sampled pairs of object vectors to
guide the mutation operation unlike the other EAs, which instead use the probability
distribution functions. DE is a technique that optimizes a problem by iteratively
trying to improve candidate solution with respect to given measure of quality.
The operations of crossover, mutation and selection make DE different from other
evolutionary algorithms. For genetic algorithm and evolutionary algorithm, muta-
tion is treated as a random change in some parameter. DE mutates the base vec-
tor with scaled population of derived difference vector. Evolutionary programming
requires adaption of absolute step size for each variable over generation, while DE
requires only specification of single relative scale factor for all variables. In crossover,
both DE and evolutionary computing uses crossover to create a single trial vector,
while genetic algorithm recombine two vectors to produce trial vectors by one point
12 2 Metaheuristics and Data Clustering

Fig. 2.2 Flow diagram for

particle swarm optimization Initialize population

Evaluate individual fitness and

rank the fitness

Yes
Stop
condition Stop

No

Generate new population

Selection Crossover
Mutation

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

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

2.3.1 Initialization

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

for each generation. At generation G, the population PG is composed of Xi,G where
i  1, 2, . . . N . For a given parameter Xi,G , three vectors Xr1,G , Xr2,G and Xr3,G
are randomly selected such that r1 , r2 , r3 are distinct. This operation extends the
workspace utilization as the individuals are chosen randomly by moving around in
the search space and giving an equal chance for all individuals to be selected for
mutation. The initial population can be chosen randomly under uniform probability
distribution if there is no prior information about the problem to be optimized. If
2.3 Differential Evolution 13

any prior knowledge about the problem is available, the preliminary solution can be
included to the initial population by adding normally distributed random deviations
to the nominal solution.

2.3.2 Mutation

The key characteristic of a DE is the way in which it generates trial parameter vectors
throughout the generation. This operator makes DE different from other Evolutionary
Algorithms. It computes the weighted difference between the vectors in population.
A weighted difference vector between two individuals is added to a third individual
to form a new parameter vector. The newly generated vector is evaluated by the
objective function. For each parent parameter vector, DE generates a candidate child
vector based on the distance of two other parameter vectors. For each dimension i,
the donor vector Vi,G is computed as:

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

Here F is the mutation factor which is a constant in (0,2). The above mutation
scheme is named as DE/rand/1. Mutation function demarcates one DE scheme from
another. The most often used DE codes are given below:

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

DE/best/1 Vi,G  Xbest,G + F · (Xr1,G − Xr2,G ) (2.5)
DE/best/2 Vi,G  Xbest,G + F · (Xr1,G − Xr2,G ) + F · (Xr3,G − Xr4,G ) (2.6)
DE/rand-to-best/1 Vi,G  Xr1,G + F · (Xbest,G − Xr2,G ) + F · (Xr3,G − Xr4,G )
(2.7)

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

r2  r3  i, F ∈ [0, 2], F is the control parameter proposed by Storn and Price [3].

2.3.3 Crossover

This process, also called recombination, incorporates successful solutions into the
population. The value of the corresponding objective function is compared with a pre-
determined individual. If the newly generated parameter vector has lower objective
function value, it will replace the predetermined parameter vector. The best param-
eter vector is evaluated for every generation in order to track the progress made
throughout the minimization process. The random deviations of DE are generated
by using both the search distance and the direction information from the popula-
tion. Correspondingly, this adaptive approach is associated with the normally fast
14 2 Metaheuristics and Data Clustering

convergence properties of a DE. The trial vector Ui,G is created for the target vector
Xi,G through binomial crossover. Elements of donor vector enter trial vector with
probability Cr ∈ [0, 1]. Cr is the crossover probability which is selected along with
population size NP ≥ 4.

Vj,i,G+1 if randi,j [0, 1] ≤ Cr or if j  Irand
Uj,i,G+1  (2.8)
Xj,i,G+1 if randi,j [0, 1] > Cr or if j  Irand

Here, randij ∼
 ∪[0, 1] and Irand is random integer from 1, 2,…NP.

2.3.4 Selection

This operation differs from the selection operation of other evolutionary algorithms.
Here, the population for next generation is selected from vectors in current population
and its corresponding trial vectors. The target vector Xi,G is compared with the trial
vector Vi,G and the lowest function value is taken into next generation.

Ui,G+1 if f (Ui,G+1 ) ≤ f (Xi,G ) where i  1, 2, . . . NP
Xi,G+1  (2.9)
Xi,G otherwise

Mutation, crossover and selection operations are continued until some stopping
criteria is reached.

2.3.5 Algorithm

The algorithm for Differential Evolution approach is given below:

2.3 Differential Evolution 15

The flow chart for Differential Evolution algorithm is given in Fig. 2.3.

2.4 Flower Pollination Algorithm

From the point of view of biological evolution, the purpose of pollination in flower is
to facilitate the optimal reproduction of plants taking into account their numbers and

Start

Input vectors of candidate solution

Calculate objective function

Perform mutation, crossover and selection

operation

No
Termination
condition

Yes

Print best vector Stop

Fig. 2.3 Differential Evolution algorithm

16 2 Metaheuristics and Data Clustering

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

the fittest. This is in effect an optimization process that take place within plant species.
The Flower Pollination Algorithm (FPA) uses the technique from the pollination of
flowering plants. Flowers reproduce using the concept of pollination. Yang et al.
[4] proposed a detailed work on FPA for multi-objective situations. Pollination (see
Fig. 2.4) is the method of relocating pollens from the male anther to the female stigma
of a flower. The goal of all living things is to produce offspring for the next genesis.
One of the methods that plants follow to produce offspring is by making seeds.
Pollination in plants is of two types:
• Cross Pollination
• Self-Pollination.
Cross Pollination occurs when the pollens from one plant is transferred to a
flower that belong to another plant. This process takes place with biotic and abiotic
creatures. Abiotic pollination occurs when external agents like wind, water etc. are
involved. Very few plants come under this category of pollination. Biotic pollination
is moving pollen from anther of one plant to stigma of another plant using external
pollinators like birds, insects, bees, etc. The pollinators are attracted by color, odor
and availability of nectar.
Self-Pollination on the other hand occurs when the flower pollinates within the
same plant. These pollination processes involves complex techniques been under-
taken for plant production strategies. Algorithm was developed based on these fun-
damental techniques. The global pollination occurs for cross pollination and biotic
pollen transfer whereas self-pollination and abiotic transfers are associated with local
2.4 Flower Pollination Algorithm 17

reproduction strategy. Plants can follow either method of pollination. The tendency
of individual pollinators to visit a plant depends on the attractiveness of that plant.
A solution of optimization problem is associated with a pollen gamete. Here, the
flower constancy is compared to solution fitness.
Biotic pollination takes place in 90% of flowering plants where pollen is trans-
ferred by a pollinator such as insects and animals. About 10% of pollination takes
in abiotic form, requiring no pollinators. Wind and water diffuses pollination in
these plants with flowers and grass being a good example. Pollinators that are some-
times called pollen vectors can be very diverse. They estimate that there are at least
200,000 varieties of pollinators, such as insects, bats and birds. Bees are considered
to be a good example of pollinators, which has also developed the so-called floral
constancy. These pollinators tend to visit certain unique species of flowers bypassing
other species of flowers. Such evidence may have evolutionary floral benefits, since
this process maximize the transfer of flower pollen within the same plant or of the
same species plants and thus maximizing the reproduction of the same species of
flowers. Instead of focusing on some unpredictable but potentially more rewarding
new flower species, flower constancy may require minimal investment cost and most
likely guaranteed intake nectar.
Using local and global pollination techniques, the pollens will be transferred to a
plant. Global method uses pollinators to transfer pollens to long distances towards
individual flowers characterized by higher fitness. Local pollination occurs in a lim-
ited range of individual flower using pollination mediators like air, water etc. Birds,
flies, bees and bats are the most common pollinators that fly for long distance. These
pollinators are considered as the carriers of global pollination.
FPA was formed by Yang [5]. This algorithm follows four rules.
• Self-pollination and abiotic method are considered as local pollination.
• Cross pollination and biotic method are considered as global pollination.
• Duplication probability is directly proportional to the likeness of two flowers
involved. So flower consistency is considered.
• Switch probability p controls the local and global pollination.
These rules can be converted to equations as given below:

xit+1  xit + ε(xit − xkt ) (2.10)

xit+1  xit + L(xit − g∗ ) (2.11)

where xit is the solution vector and g∗ is the current best during the iterations.
L is the step size from Levy distribution which denotes the strength of pollination.
Levy flight is used to imitate the movement of insects over a long distance.
The parameter L is computed as:
λ(λ) sin(π λ/2) 1
L∼ , (S S0 > 0) (2.12)
π S 1+λ
18 2 Metaheuristics and Data Clustering

where, λ is the probability density function, (λ) is the standard gamma function
and S is the step size of pollination. This is a continuous probability distribution for
positive random numbers. Flower Pollination algorithm has been used widely in the
field of fuzzy logic and engineering. It is flexible, simple and is exponentially better
to solve.
The algorithm for FPA is given below:

Objective function . Initialize n pollen seeds

Select optimal result in the original group

State as number of iterations and a switching probability as

While

for all n flowers in the group

If

For Levy distribution, select a step vector of

dimension

Do global pollination

Else
Select item from a uniform distribution within

Do local pollination

end if
Calculate new solution
Update population if the new solutions are improved
end for
Find current best solution

end while

Output solution
End

The flowchart of Flower Pollination Algorithm is given below in Fig. 2.5.

2.5 Data Clustering

With the advances in technology, the need for acquiring, storing and processing a
huge amount of data is ever increasing. Clustering is a process of partitioning a set
of data (or objects) into a set of meaningful sub-classes, called clusters. It helps the
user to understand the natural grouping or structure in a data set. It is breaking down
of large population into smaller groups that are similar in character. A cluster is
therefore a collection of objects which are similar internally, but clearly dissimilar to
2.5 Data Clustering 19

Start

Input population size

Load population of flowers with

random solution

Calculate current best

No
If rand> p

Local pollination
Yes

Global pollination using levy flight

Compute new solution

Update current global solution

No Yes
Termination Output best solution
Condition? Stop

Fig. 2.5 Flowchart for Flower Pollination algorithm

the objects belonging to other clusters. Dissimilarities and similarities are assessed
based on the attribute properties describing the objects. It is a connected region
of multidimensional space, containing a relatively high density. The quality of a
20 2 Metaheuristics and Data Clustering

clustering result also depends on both the similarity measure used by the method
and its implementation. The quality of a clustering method is also measured by
its ability to discover some or all of the hidden patterns. By organizing data into
clusters, there should be high intra cluster similarity and low inter cluster similarity.
In clustering, the objects of analysis could be persons, salaries, opinions, software
entities and many others. These objects must be carefully presented in terms of their
characteristics. These characteristics are the main variables of the problem and their
choice greatly influences the results of a clustering algorithm.
Clustering is based on three characteristics namely nesting, exclusiveness and
completeness. In the nested type, separation is built on characteristics of nesting
clusters. Hierarchical clustering is nested meaning it clusters to exist within the
bigger clusters. In hierarchical type, a set of nested clusters is organized into a
hierarchical tree. Partitioned clustering prohibits subsets of cluster. Here, a division
of data objects is grouped into non-overlapping subsets (clusters) such that each
data object is in exactly one subset. Exclusive separation is the characteristics in
which the data object is allowed to exist in one or more than one cluster. But, in
exclusive clustering, each of the data object shall exist in only one cluster. The data
objects can be grouped into two or more clusters through overlapping. For example,
in overlapping clustering, it can allow a student to be grouped as employee also
while in case of exclusive clustering, it demands that the person be grouped to the
once which are more important. In case of fuzzy clustering before also known as
probabilistic clustering, each of data object shall belong to every cluster. The data
objects here possess membership weightage between 0 and 1 where 0 shall point
that it does not belong to that cluster whereas 1 denote it belong to that particular
cluster. Completeness is a type of separation, which requires all the data objects to be
grouped. In complete clustering, every object is assigned to a cluster. Various steps are
involved in data clustering namely, data collection, initial screening, representation,
clustering tendency, clustering strategy, validation and interpretation. Data collection
includes extracting related data from the source. Initial screening is the messaging
of data after extraction from source. Representation means preparing data in order to
make it proper for algorithm. Clustering tendency verifies if the data in hand can be
considered in a cluster or not. Clustering strategy chooses the correct algorithm and
parameter. Validation is the manual examination of the data for validity of technique.
Interpretation includes combining clustering results with other studies and suggesting
further analysis.
In document clustering (see Fig. 2.6), it is possible to view the clustering prob-
lem as an optimization problem that locates the optimal centroids of the clusters
rather than to find an optimal partition. This view offers us a chance to apply evolu-
tionary algorithms for clustering problem. The appropriate clustering algorithm and
parameter settings (including values such as the distance function to use, a density
threshold or the number of expected clusters) depend on the individual data set and
their intended use of the results. Partitional algorithm permits to decompose data
into sets of disjoint clusters using similarity criterion. It allows updating of clus-
ter members if the clustering performance is improved. Partitional clustering is an
optimization problem as it minimizes the cluster assignment in probability density
Other documents randomly have
different content
 But the Lion of St Mark, though his claws were a little blunted and his
joints stiffened, had not lost his cunning. Moreover, he was forewarned. A
dramatic story of the premature disclosure of the plot is told in the Venetian
State papers. Spinola, an emissary of Gonsalvo of Cordova, came secretly to
Cornaro, the Venetian ambassador, at Valladolid, in February 1509, and
asked him to meet the great captain at mass in an unfrequented church at the
far end of the town. He went and the secret was revealed to him. He refused
to believe it, but later Spinola showed him a copy of a letter from Gonsalvo’s
wife at Genoa in which the details of the proposed partition were given, and
offered his master’s services to the Signory. Cornaro informed the Ten. They,
too, hesitated to believe in any cause for attack, advised caution, and asked
for further proof. Secret information from England soon brought
confirmation and the Ten sat day and night to prepare for the coming storm.
The weakness of the league lay in the fact that each spoiler was to seize his
own share of the prey. Self-interest was its motive power. Self-interest would
lead individual members to abandon the hunt if their portion were thrown to
them. This the Ten quickly saw and acted upon with consummate art and
patience while pushing on with all speed defensive military operations. The
aged and infirm Doge Loredano, so overwrought by emotion that it was
piteous to see, addressed the Great Council begging them to turn to
righteousness and offer their lives and substance in defence of the fatherland.
Himself would give an example by sending his silver plate to the mint. On
April 27, 1509, Julius flung a bull of excommunication couched in almost
savage terms against the Republic. The Ten forbade its publication and sent
officers to take down any copies posted on churches or on the walls. They
consulted learned canonists; drew up an appeal to a future council of the
Church and sent emissaries to Rome who nailed a copy on the doors of St
Peter’s. The secular arm swiftly followed. Sanudo tells us that while he and
two other Senators were examining a map of Italy painted on the walls of the
Senate hall, a courier arrived with the news that the French had crossed the
Adda, fallen on the fine army of the Republic at Agnadello and utterly routed
it with a loss of four thousand in killed alone. Faces gathered gloom and
despair. “Give me my cloak, wife,” said Paolo Barbo, one of the most
experienced of the fathers, “that I may go to the Senate, speak a couple of
words and die.” One disaster trod on the heels of another. Bergamo and
Brescia fell and before the month was ended nearly the whole of Lombardy
was lost. Preparations were even made to defend and victual Venice. Envoys
were sent to treat with the Kings of France and Spain. The Pope was tempted
by an offer of partial restitution and help towards a crusade against the
Turks. Meanwhile the Imperial Eagle swooped down from Trent. The
Signory, by ceding Verona and Vicenza, hoped to conciliate the Emperor and
save Padua. In vain were the civil commissioners with the army entreated to
make a stand, “lest the whole of our cities surrender in an hour.” Padua fell
and Treviso alone stood by the Republic. At bay she now turned to the
Sultan of Turkey and begged for money and men, especially men. If his
Highness would advance them one hundred thousand ducats and would
agree to buy no more cloth of the Genoese and Florentines, who only used
his money to help a League that sought his hurt, the Signory would send him
fifty thousand ducats’ worth of cloth, and jewels worth fifty thousand more,
as security. The Venetian consul at Alexandria was instructed to incite the
Sultan of Egypt to ruin Genoese and Florentine commerce in his dominions.
The good offices of the Kings of England and Scotland were sought.
But the gloom was wearing away. One day in July two tall, mysterious,
armed men were observed leaving Fusina in the gondola of the Ten. Arrived
at Venice they remained closeted with the Ten and the Doge far into the
night, then were rowed back whence they came. On the night of the 16th
there was a hurrying to and fro of transports and armed vessels between the
islands. The Doge’s two sons and two hundred noble youths, fully armed,
left for the mainland. The police boats of the Ten allowed no one to go out of
Venice without permission. Next day Padua, disgusted by the insolence and
exactions of the Imperialists, was won back for Venice before the laggard
Emperor could reach the city. Sanudo remembered the 17th of July, for did
he not buy a Hebrew Bible worth twenty ducats for a few pence as he was
going home? Two attacks by the Emperor were successfully resisted, and the
foiled Cæsar retired to Vicenza in October with anger in his heart against the
French. In February 1510, after long and tough negotiations, the Pope was
given his prey and detached from the league, but at the price of a bitter
abasement of Venice. Time had avenged the Empire. It was now the Queen
of the Adriatic who, in the person of her ambassadors, bowed the neck
before the enthroned Pope in the atrium of St Peter’s, surrendered her
ecclesiastical privileges, admitted the justice of the excommunication,
craved pardon for having provoked it, and was at length absolved and bidden
to do the penance of the seven churches. The Ten, however, entered in their
register a protest of nullity, declaring that the conditions had been extracted
from the Republic by violence. The Pope who, as he told Venice, had no
pleasure in seeing the ruin of her State to the aggrandisement of the
barbarians, now became her ally. Soon other cities, sickened by the atrocities
of the invaders, returned to their allegiance, and by skilful playing of King
against Emperor, and Pope against both, Venice was able to regain the bulk
of her territory.

CHAPTER XI
Loss of Cyprus—Lepanto—Paolo Sarpi—Attack on the Ten—Loss of Crete—Temporary
Reconquest of the Morea—Decadence—The End
“Alas, alas that great city, that was clothed in fine linen, and purple, and scarlet, and
decked with gold, and precious stones, and pearls!... Alas, alas that great city, wherein were
made rich all that had ships in the sea by reason of her costliness.”—The Revelation of St
John the Divine.

WE may not here attempt to tread the maze of chicanery and violence which
ended in the peace of Cambrai.[53] We are permitted to see a fighting Pope
exhorting his soldiers and directing siege operations against an Italian city,
and climbing by a scaling-ladder through the breach to take possession. In
1514 the Spaniards desolated the land up to the lagoons and levelled their
cannon at Venice. In 1515 the encampments of four armies were exhausting
and polluting Lombardy. King, Popes, and Emperor died and their
successors took up the unholy heritage of war and duplicity. Gaston de Foix,
Bayard and other renowned chevaliers perished. In 1521 the Emperor
Charles V. came upon the scene, and in alliance with the great Medicean
Pope, Leo X., swept the French and their Venetian allies out of Lombardy. In
July 1523, when the power of France was waning, the Venetians made terms
with the Emperor. They were suffered to retain their territory up to the Adda
in return for an annual tribute of 250,000 ducats. Venice excused herself to
Francis I. by professing solicitude for the peace of Christendom in view of
the threatening attitude of the Sultan. Before the year was ended King and
Emperor were competing for Venetian help in a renewed struggle for
mastery. While the Republic was temporising, the Imperialists had
descended on Lombardy, routed the French before Pavia and captured their
King. “Nothing is left to me,” wrote Francis, “but honour, and life which is
safe,”[54] and proceeded to send his ring secretly to the Sultan and to grovel
before Charles. The victorious Emperor brushed aside the subtleties of the
Venetian ambassadors. “If you were to send all your lawyers,” he cried, “you
would not convince me. You must pay 80,000 ducats for the troops you
failed to send to Pavia. You are rich: my expenses are heavy: you must help
me.”
After perjuring himself at the peace of Madrid, January 1526, the
Cristianissimo returned to France. In less than six weeks a “holy league” of
France, Venice and the Papacy had been signed at Cognac for the
“liberation” of Italy from the Imperialists. But Francis, whose moral fibre
had been rotted by lechery, was no match for the virile genius of Charles,
strong with the united resources of the Empire and of Spain in her greatness.
The Emperor was soon again master of Italy. Rome was captured and
sacked; Pope Clement VII. imprisoned. But the miserable condition of Italy
and the news that the Turks were threatening Vienna disposed Charles to
treat, and in July 1529 Margaret of Austria was once more at Cambrai
negotiating on behalf of the Emperor with Louise of Savoy, who represented
Francis.[55] Two adjacent houses were chosen and the party-wall pierced that
the ladies might confer with absolute secrecy. In two months, while the
Venetians were finessing, the “paix des dames” was concluded and Venice
left to make the best terms she might with the Emperor. Francis had given
way all along the line. “The peace of Cambrai,” says Michelet, “was the
moral annihilation of France in Europe.” During the coronation festivities at
Bologna the Emperor and the Pope found time to deal with the Venetians,
who agreed to pay the balance of the annual tribute of 250,000 ducats due on
the treaty of 1523; to restore the cities of Naples and Apulia to the Emperor;
and to the Papacy Ravenna and Cervia, which they had seized during the
Pope’s imprisonment at Rome. Thanks to the impassable lagoons Venice
preserved her capital inviolate, but her prestige and her military power were
gone.
After the League of Cambrai a change comes over the Venetian temper.
Patricians, instead of using their talent in commerce and discovery, chose to
live on their invested capital and on the revenues of their mainland estates.
The power of initiative was gone. In 1522, before Sebastian Cabot sailed for
the New World, he contrived to meet Contarini, an emissary of the Ten,
secretly at Valladolid, and told him he had no joy in selling his knowledge to
the foreigner; that he had refused tempting offers from Cardinal Wolsey and
was prepared to absolve himself from the King of Spain’s service and spend
his genius in the advancement of his fatherland. But Contarini talked of
things possible and impossible, and success is to those who will achieve the
impossible. The supreme opportunity of retrieving her mercantile position
was lost to Venice for ever. Sadder still, when Loredano had called on the
Senate for volunteers and patriotic gifts for Padua and Treviso, not a man
stirred. Venice had lost faith in herself.
In 1521 Leonardo Loredano died and was buried with more than usual
pomp at S. Zanipolo. Antonio Grimani, the disgraced of Sapienza, who had
redeemed himself by faithful service, reigned for two years and gave place
to Andrea Gritti, a distinguished civil commissioner with the army during the
wars. Between Gritti’s death in 1539 and the election of Sebastian Venier,
the hero of Lepanto, in 1577, there follows a line of Doges, Pietro Lando,
Francesco Donato, Marc’ Antonio Trevisano, Francesco Venier, Lorenzo and
Girolamo Priuli, Pietro Loredano, Luigi Mocenigo, worthy magistrates all,
but without distinction.
The wars had exhausted the State treasury. Her Indian trade was withered,
and the wealth of Venice was no more commensurate to the demands of a
long naval war. Her military pride had been chastened by the rod of the
Emperor, and a dread of Spanish arms and Spanish gold hung like a pall over
men’s minds. An era of subtle diplomacy begins, and the Council of the Ten,
with its new instrument of the Inquisitors of State, tightens its grip upon the
executive. Wave after wave of Ottoman fury surges against her Eastern
possessions; one by one they are engulfed. In 1535 she lost Egina, Paros and
Syra; in 1540, Malvasia and Nauplia. In 1570 Cyprus was marked out for
conquest and the usual appeal to the Christian Powers was made. Spain and
the Pope promised help. Zane, the Venetian commander, wasted his force
waiting at Zara, then learned that the allies were at Corfù. He reached the
island only to find the Spanish admiral without orders. Meanwhile the
season had worn along and operations were judged inopportune. The whole
island by this time was occupied by the Turks, Nicosia and Famagosta alone
holding out. While the futile admirals were squabbling about plans the
magnificent heroism of the garrisons and of the inhabitants was spent in
vain, and the cities fell to the horrors of a Turkish pillage, and Marc’ Antonio
Brigadin, the Venetian governor of Famagosta, was treacherously flayed
alive in the Piazza after having surrendered on terms to the enemy. Zane was
recalled to Venice, and Sebastian Venier given command. A new alliance of
Spain, the Papacy and Venice being concluded, at length on October 7, 1571,
the allied fleets came upon the Turkish armament off Lepanto in the Gulf of
Corinth. The Spanish admiral, Don John of Austria, was in supreme
command. Venier led the Venetians; Marc’ Antonio Colonna the Papalists.
It was a calm sunny morning. The line of the allied fleets was four miles
in extent, the two armaments were a mass of glittering steel as the rays of the
sun smote on the helmets, breastplates and shields, bright as polished
mirrors. The banners of gold and tall galley lamps were resplendent in many
colours. A beautiful, yet an awful spectacle. The Venetian flagship was
fiercely assailed. Venier, spite of his seventy-five years, was seen, sword in
hand, pressing to the thick of the fight, heartening his men and with
invincible courage striking down his enemies, so that he wrought deeds
beyond the belief of man. We cannot here linger on the vicissitudes of the
struggle. Scenes of comic relief were not absent from the tragedy. Some
Turks, their arms of offence failing, seized upon a quantity of oranges and
lemons and threw them at their enemies, who with mocking laughter cast
them back. At length after five hours of savage fighting, the Turks were
scoured off the seas. The allied and victorious admirals met, embraced each
other speechless from emotion; and as the venerable Venier and the youthful
Don John of Austria stood clasped in each other’s arms shedding tears of joy,
the eyes of even the most hardened of sea-dogs were moist with tears. Some
30,000 Turks are said to have perished; 3486 prisoners were divided among
the victors as slaves; 94 ships were burned; 130 ships and 356 guns
captured; 15,000 Christian slaves set at liberty. The allies lost heavily: 8000
men were slain including 25 Venetian nobles.[56] Among the Spanish was
Cervantes, who lost an arm in the engagement.
 A WINE SHOP.

As day broke on October 18th, a galley was seen sailing up to Lido

trailing the Turkish colours at her stern, a pile of turbans on her deck. Amid
the booming of the guns could be heard cries of “Victory! victory!” The
reaction from the gloom of Cyprian news was tremendous. A frenzy of joy
possessed the people. Shops were shut per la morte de’ Turchi. The streets
from the Rialto bridge to the Merceria were covered with a firmament of
blue cloth spangled with stars of gold; a pyramid of Turkish spoils stood on
the Piazza, which was gay with scarlet cloth, tapestry and pictures. Four
days’ rejoicings celebrated the triumph of the Cross. But to Venice the battle
of Lepanto (or alle Curzolari) was a sterile victory. Dynastic jealousy in
Spain and the old suspicion of Venice, which still clung to the allies,
permitted the Turks to recover from the blow, and in March 1573 Venice
agreed to purchase a separate peace at the cost of an indemnity of 300,000
ducats and a threefold increase of the tribute for Zante. The fair island of
Cyprus was lost for ever. “Was it the Turks who were the victors at
Lepanto?” asks Romanin.
On Doge Venier was conferred the consecrated golden rose, a supreme
token of papal favour, but during the seventy years’ peace from Lepanto to
the outbreak of the fifth Turkish war, the indictment at Rome against the
refractory children of the lagoons increased in gravity, and in the beginning
of the seventeenth century the Papacy determined to force them to yield to
discipline. The Venetians were a stubborn folk when their national dignity
was threatened by Rome. “These Signori of the Senate,” said Paul IV. to the
Venetian ambassador, “are tough fellows and take a lot of cooking” (non
sono molto buoni da cuocer). In 1527, when the Papacy was under the heel
of Charles V., the Republic had reasserted her rights to nominate to
ecclesiastical offices. Disputes as to the taxation of Church property, the
right of the Pope to inspect the monasteries, the right of the Republic to try
criminous clerics, exacerbated the situation until at length Gregory XIII.
declared in 1581 that he would no longer consent to be Pope everywhere but
in Venice, and sent his nuncio to make a visitation of the Venetian
monasteries. The Republic refused permission, and the conflict called of the
Interdetto began. But the fight between Italian Popes and Italian sovereigns,
even in our own times, has generally been a comedy played for the
mystification of Transalpine Powers and all ended in compromise. The
Signory appointed the Bishop of Verona as the nuncio’s colleague, and
persuaded itself that the nuncio went with the Bishop; the Pope satisfied his
dignity by claiming that the Bishop went with the nuncio. In 1605 the
Spanish party, ever the evil genius of the Sacred College, re-opened the
quarrel. The Republic had refused to send their nominee to the Patriarchate
of Venice for examination to Rome, and had tried and convicted two clerics
on the mainland for criminal offences. On Christmas Day a brief from Pope
Paul V. was delivered to the Signory threatening excommunication if they
did not submit in the matter of the taxation of ecclesiastical estates. The
Republic engaged the learned Augustinian friar Paolo Sarpi as their adviser
at a salary of two hundred ducats and prepared for the struggle. In February
1606, a second brief followed on the matter of the convicted clerics. The
Republic expressed her devotion to the Catholic Faith, but firmly though
respectfully declined to surrender her ancient rights and privileges. On April
16th, the Republic was given twenty-four days to submit under pain of
interdict. Venice calmly waited. In due time the bull of excommunication[57]
and interdict was delivered. The Signory forbade its publication and ordered
the clergy to continue their functions as usual. Some of the regular clergy
who disobeyed were expelled. Sarpi advised the Republic with excellent
prudence and wisdom, and became in the eyes of Europe one of the greatest
protagonists of national liberty against papal aggression. A Spanish army
having been mobilised on the Milanese frontier, Sir Henry Wotton, the
English ambassador, suggested an alliance of Venice with England, France,
the Grisons, and the German Protestant princes, but the Republic was deaf
on that side. She declared to the Pope that the Venetians were as good
Catholics as he himself, but that as Church property enjoyed the protection
of the State, it must share its burdens, and that criminals, whether lay or
cleric, must be equally subject to the laws of the land. Time wore on. Spain,
humbled by the defeat of her Armada and the revolt of the Netherlands, was
afraid to strike; the obsolete ghostly artillery of Rome failed to act; and the
secular clergy stood loyally by the Republic. The Papacy reverted to her
habitual policy of compromise. The services of Henry IV. of France as
mediator were accepted and a solemn comedy was played. The Republic
agreed to surrender the incriminated clerics, without prejudice, to the French
ambassador; the Pope agreed to withdraw informally the bull of
excommunication and interdict. The Republic continued to nominate for
Church offices and try clerical delinquents as before. But the Spanish
fanatics at the Vatican never forgave the Venetian friar for his share in their
discomfiture. On October 25, 1607, three ruffians fell upon him as he was
crossing the S. Fosca Bridge,[58] stabbed him, then left him for dead, and
escaped to Papal territory. Sarpi, however, recovered. The surgeon who was
dressing his wounds remarked on their jagged, inartistic nature. “Ah!”
replied the witty friar, “agnosco stylum curiæ romanæ.” On his recovery the
Republic gave him a pension of six hundred ducats and a house near the
Piazza, where the great patriot-scholar devoted the remainder of his life to
literature and science.[59] Two further attempts were made to assassinate
him, in 1609 and 1610. He spent his last breath in the service of the
Republic, advising the Senate in three important questions in 1623 as he lay
on his death-bed. His mind soon began to wander. “It is growing late,” he
murmured, “I must hasten to St Mark’s for I have much to do.” His last
words were a prayer for his country. “Esto perpetua.”
 The university of “Fair Padua, nursery of the Arts,” became under
Venetian auspices the most famous and most honoured centre of learning in
Europe. Liberal salaries and an atmosphere of intellectual freedom drew an
array of the most eminent teachers in Christendom. Fallopius, in physiology
and medicine; Galileo, in astronomy and mathematics, were names that
crowded its halls with eager students. As many as eighteen thousand of all
nations were gathered there daily in the sixteenth century. During his
professorship of twenty years, Galileo invented the thermometer and the
telescope. Tasso studied, and our own Harvey (Italians claim) learned the
secret of the circulation of the blood there. The Earl of Arundel sent his two
sons in 1622 to drink of its springs. The Admirable Crichton having called
on one of the Aldi in 1580, was introduced to the Signory, and improvised
before the Senate a Latin oration of “most rare and singular power.” The
Fathers voted the impecunious youth one hundred crowns as a courteous
recognition of his marvellous powers, and sent him to Padua with a warm
introduction.
To Sebastian Venier succeeded Doge Nicolo da Ponte in 1577, a worthy
scholar and student of theology, who had represented the Republic at the
Council of Trent. At his death, in 1585, Pasquale Cigogna, descended from
an apothecary ennobled after the war of Chioggia, was preferred to the ducal
office. Cigogna saw the erection of the new stone Rialto bridge, and, after
ten years’ peaceful reign, was followed by a popular and lavish prince,
Marino Grimani, whose consort was exceptionally honoured by a gorgeous
coronation ceremony.
Grimani died on Christmas Day 1605, the very evening of the delivery of
the first papal brief, and Leonardo Donato was chosen to open the document,
and to preside over the conflict with the Spanish papalists at Rome. When he
died in 1612, weird stories were whispered by fanatics of shrieks and cries
heard from his chamber as the Evil One bore him away.
During the short reigns of Marc’ Antonio Memmo, Giov. Bembo, Nicolo
Donato, and Antonio Priuli, the Ten had been accumulating evidence of a
vast conspiracy to seize the city, concerted by the Spanish Viceroy of Naples
and the Marquis of Bedmar, the Spanish ambassador at Venice. On May 12,
1618, three Frenchmen in Venetian pay were arrested, strangled, and hung
head downwards between the red columns, and orders were sent to the fleet
to despatch three others. The plot had been divulged by two of the
conspirators, and in all some three hundred persons of various nationalities,
including many poor Venetian patricians, were implicated, and paid the
penalty with their lives. The Spanish ambassador was for a time in danger,
and under guard. He protested his innocence of the plot, as did his colleague
of France. Both, however, soon sought a change of air. Two years were spent
in tracing the ramifications of the plot, and in 1620 a senator, Giambattista
Bragadin, was found to be in Spanish pay, and hanged between the columns.
In 1622 the atmosphere of dread and suspicion which encompassed the
State, so dulled the perceptions of the Ten that a grave miscarriage of justice
was laid to their charge. In 1618 Antonio Foscarini, a noble of high family
and Venetian ambassador at London, was accused by Mascorno, a
disaffected member of his staff, of licentiousness, blasphemy and treason.
Foscarini was recalled, arrested by the Ten, and, after a long trial, acquitted,
but kept under surveillance. In 1622, as he was leaving the Senate, a cloak
was flung over him, and he was hurried off to prison. His accuser, who had
been sentenced to two years’ detention in a fortress, had, on his release,
fabricated some documents which the tribunal deemed conclusive. Foscarini
was declared guilty of corresponding secretly with Spain and the Emperor,
strangled in prison, and his dead body hung by the leg between the red
columns. As he had been an occasional visitor at Casa Mocenigo, where
Lady Arundel resided, she was suspected also; but Sir Henry Wotton
prompted her to clear herself by asking an audience of the Doge. This she
did, and was allowed to make a statement in the Senate, the only woman
who ever addressed that Assembly. She was exonerated, and a present of
sweetmeats and wax offered to conciliate her. Four months later poor
Foscarini’s innocence was entirely proved, and two of his accusers were put
to death. The family was restored to honour, and his remains were dug up
and buried in the Frari with great splendour and pomp. His bust may still be
seen in the church of S. Eustacchio (S. Stae) near the old Foscarini Palace.
Doge Priuli died shortly after the Foscarini tragedy. The brief reign of
Francesco Contarini followed, and Giovanni Cornaro was chosen to fill the
ducal office in 1624. The shock of Foscarini’s judicial murder had given a
rallying cry to the poorer nobles in the Great Council, jealous of the power
of the Ten and the monopoly of office by the more influential patricians; and
Renier Zeno, a patrician, fearless and incorruptible—himself an ex-capo of
the Ten—led an attack on the tribunal. Banished for a year, he did but return
with added popularity, and forced himself again on the Ten as one of the
Capi. He used his power to accuse the Doge of nepotism, and his Serenity
was forced to cancel certain family appointments. Zeno, driving his
advantage further, came into conflict with his colleagues of the Ten, and,
leaning on the majority of the Great Council, emerged triumphant. Shortly
after, while standing at the Porta della Carta, he was attacked by five persons
and stabbed. The Doge’s son and certain alleged accomplices were
denounced to the Ten, whose laggard justice, however, made flight easy.
Again appointed one of the Ten, Zeno, on his recovery, renewed the struggle
with increased vigour; and, after a stormy scene in the Great Council, during
which he came to high words with the Doge, the stout reformer was ordered
to keep his house, and report himself to the Ten within three days. Ignoring
the summons, he was fined two thousand ducats and banished. The Great
Council quashed the sentence, and ordered it to be blotted out of the records
of the Ten. At length Zeno’s party succeeded in carrying a motion for a
committee of inquiry into the constitution of the Ten, but the four years’
bitter conflict ended in a virtual triumph for the tribunal, whose powers of
criminal jurisdiction over the nobles were reaffirmed, though it had to
submit to a modified capitulary.
During the latter half of the seventeenth century the power of Venice was
declining to its setting in an aureole of glory. In 1644, Crete, the oldest and
last remaining of her great possessions in the East, was marked for conquest,
and, like an old warrior who takes down his armour and girds himself to
make a last stand against his hereditary foes, Venice prepared to resist the
Turk to the uttermost. The old heroic times seem to return as we watch the
quarter of a century’s struggle, but our admiration is touched with pathos, for
we know that the dice are loaded against Venice. A Turkish pilgrim fleet for
Mecca had been pillaged by the Knights of Malta, and the pious buccaneers
had landed at Crete for provisions. This was pretext enough for hostilities. In
1645 a huge armament left the Bosphorus, ostensibly for Malta, actually for
the conquest of Crete. Canea quickly fell, and the Turks promised
themselves an easy occupation. But twenty-four years of fierce and
exhausting fighting ensued before the Crescent floated over the island.
Seven million ducats were quickly raised in Venice by the sale of patents
of nobility. By a marvellous re-birth of naval energy and capacity, her fleet
was reorganised and spread terror along the Dardanelles. The Venetian
Captain-General, Lazzaro Mocenigo, determined to force the passage and
attack Constantinople, but a well-aimed shell fired his ship, and he was
killed. Francesco Morosini, appointed his successor, won the admiration of
Europe by his twenty-two years’ defence of Candia. Inspired by his heroism,
companies of Flemish and French volunteers, eager and impetuous, joined
him, but their enthusiasm was soon spent, and, impatient of the long vigils
and toils of the war, they left the Venetians to fight alone. Morosini did not
save Crete, but he extorted an honourable peace. No indemnity was paid,
and the Venetian garrison marched proudly out of Candia unsubdued.[60]
Four thousand Candiots who opted for Venice were settled in Istria, where
traces of their language and customs are said still to survive. Suda and other
fortresses remained in the hands of the Venetians. To the Pope the result
seemed almost incredible.
In 1684 Venice was invited to join the Emperor and the King of Poland in
a league against the Turks. The Cretan war had cost her one hundred and
twenty-six million ducats, and she felt too exhausted to run with the
horsemen again. But bolder counsels prevailed. Morosini was despatched
with an army, and ably seconded by Koningsmark, the great Swedish
mercenary, overran the Morea, captured Coron, Sparta and Athens, which
last was won at the price of the ruin of the Parthenon, the Turkish powder
magazine there having been exploded by a Venetian shell. Morosini returned
in triumph, bringing the Greek lions, which still stand in front of the Arsenal.
He was made Doge in 1688, the coronation being deferred that he might
return to Greece. Vast designs of the recapture of Negropont, even of Crete,
lured him on, but ill-health soon necessitated his return, only, however, to be
again entreated to take up the command and retrieve the blunders of an
incompetent Captain-General. The veteran Doge and captain for the last time
sailed from Venice amid scenes that recalled the great crusading times of
old. After some successes at Corinth he went to winter at Nauplia, where he
died on January 9, 1694. He was the greatest of the modern Doges. A tomb
in S. Stefano and a triumphal arch in the Sala dello Scrutinio still witness to
his fame.
 PONTE DI RIALTO.

But Venice was too poor and too feeble to retain her conquests. During a
short campaign in 1715 she lost the whole of the Morea, and by the treaty of
Passarovitch in 1717 all that was left of her vast empire in the East were a
few fortresses in Dalmatia, Albania and Herzegovina. The treaty of 1717
bore her last signature as a European Power.
The procession of Doges that stretches from Giov. Cornaro, the opponent
of Renier Zeno, to the fall of the Republic contains but one name of historic
significance—Francesco Morosini. Marco Foscarini, elected in 1762, a
descendant of the ill-fated Antonio Foscarini, is known to students as the
author of a “History of Venetian Literature,” and Ludovico Manin has the
unhappy distinction of closing the line for ever. Through all the vicissitudes
of foreign affairs, the decadence of trade, the fear corroding at her
statesmen’s hearts, the social and ceremonial life of Venice waxed rather
than waned in pomp and splendour. The recurring ravages of plague
periodically purged her pride and luxury. Of all the great cities of Europe,
Venice bears the deepest traces of the passages of the destroying angel. In
her annals no less than seventy visitations are recorded. Two great churches,
the Redentore founded in 1575 and the Salute in 1630, are votive offerings
to Heaven for salvation from the scourge. Her greatest scuola is dedicated to
the chief plague saint, St Roch. Indeed in all her churches the figures of the
plague saints, St Roch, St Job, St Sebastian, have a sad pre-eminence. But
the danger past, the lesson faded from her memory, and the traditional
magnificence shone forth. She became again—

“The pleasant place of all festivity;

The Revel of the earth; the Masque of Italy.”

Sanudo gives a list of nineteen great annual pageants, and after his time
others were added. Besides these official festivals great patrician weddings
or the visits of foreign potentates were the occasions of stately pomp and
joyous revels. At the anniversaries of the greater scuole, each guild vied with
the other to excel in splendour. Never before nor since was such
magnificence. The greatest artists of the day were commissioned to execute
the decorations. The Bucintoro was carved by the best sculptors. Palladio,
Titian and Tintoretto designed and decorated triumphal arches.
The loan of the Bucintoro and a subsidy of five hundred ducats were
voted to the Calza to entertain the Duke of Milan in 1530. On this occasion a
bellissima colazion (luncheon) was prepared, says Sanudo, but so ill-
arranged that the Milanese nobles got nothing, while some Venetian Senators
filled the sleeves of their robes with sweetmeats to the shame of those who
saw it.
Venice surpassed herself in the reception given to Henry III. of France in
1574. The young king was met at Malghera—the modern traveller will pass
a fort erected there as he nears the railway bridge—by sixty Senators in
gondolas covered with velvet, oriental carpets and cloth of gold, and was
ferried to Murano, where he passed the night in one of the rich palaces with
delicious gardens for which the island was then noted. Sixty halberdiers
clothed in silk of azure and gold were his bodyguard: forty noble youths of
the Calza were his attendants. On the morrow amid salvos of artillery he
embarked for Venice in a great galley manned by four hundred Sclavonians
clothed in yellow and turquoise taffety, followed by an immense train of
galleys and gondolas decorated with carpets and tapestry, with banners and
flags waving in the breeze. The procession of the trade guilds, formed of a
hundred and seventy boats resplendent with crimson and silver and gold,
was a dazzling pageant. The glass-workers excelled in splendour and
invention. A marine monster, in whose body could be seen a furnace, and
craftsmen making most beautiful crystal vases, led their section, breathing
flames from his mouth. Then followed a boat in the shape of a great dolphin
bestridden by Neptune; on the poop stood two winged angels to waft it
along. Four river gods personifying the Brenta, the Adige, the Po and the
Piove plied the oars. At S. Nicolo del Lido, Palladio had constructed a
triumphal arch adorned with statues of Victory, Peace, Faith and Justice, and
with ten paintings by Titian and Tintoretto portraying events in the King’s
life. His Majesty lodged in the Palazzo Foscari from which an opening was
made into the Palazzo Giustiniani to accommodate his suite, the whole being
furnished with oriental magnificence. At a State ball given in the hall of the
Great Council, two hundred gorgeously attired ladies were present glittering
with jewels and precious stones. The Sala dello Scrutinio was made into a
supper-room where twelve hundred and sixty plates of sweetmeats in the
forms of griffins, ships, nymphs, deities, etc., tempted the palates of the
guests. Regattas, serenades and jousts made the whole visit seem a dream of
enchantment to the King. As trade languished and the population diminished,
public shows increased in splendour. The sum expended at the election and
coronation of the last Doge—forty-seven thousand, two hundred and ninety-
eight ducats, was beyond all precedent. Venice was still the temple of
pleasure. All the arts subservient to the luxury and vices of the rich
flourished in rankest exuberance despite the efforts of the Ten to cleanse
public morals and to enforce sumptuary laws. The excessive importance too
of the stage and of its tinselled heroes and tawdry queens, was an infallible
symptom of a decadent nation. The time came in the eighteenth century
when the State was torn by the petty jealousies and vanities of a playwright
and an actress, and when public appointments were controlled by the subtle
influence of the boudoir and the drawing-room, and an ambitious and
beautiful society lady was the central figure of Venetian life. It was the time
of the fatuous masquerades and futile pomposities portrayed for us by
Longhi, when the card table, the coffee house and the play were the
absorbing interest of Venetian minds. But before she sinks into the deep
night of subjection to Austria to rise again as a province of a free and united
Italy,[61] a faint hue of naval splendour lights up the horizon. Soon after
Goethe’s arrival in Venice in 1786 he ascended St Mark’s tower and under
the bright noon-day sun saw a fleet of galleys and frigates lying off Lido.
They were reinforcements for Tunis, where the last of the great Venetians,
Angelo Emo, was fighting the Algerian pirates. Emo humbled the Bey of
Tunis, cleared the seas, and died at Malta in 1792. Five years later Napoleon
marched his battalions towards the lagoons and before the mere breath of his
coming the Republic of Venice crumbled into dust. On May 16, 1797, for the
first time in a thousand years the Realtine islands were trodden by the foot of
a conqueror, and the hundred and twentieth Doge of Venice, handing his
biretta to an attendant, said: “Take it away, we shall not want it again.”

CHAPTER XII

THE FINE ARTS AT VENICE

Masons—Painters—Glass-workers—Printers

Some prefer the pure design:

Give me my gorge of colour, glut of gold
In a glory round the Virgin made for me!
Titian’s the man, not monk Angelico
Who traces you some timid chalky ghost
That turns the church into a charnel.
—Browning.

OWING to the absorption of her energies in commerce and the eastern trend
of her interests and activities Venice lagged behind the Tuscan masters in the
practice of the finer arts. Her earliest craftsmen were Byzantines, and St
Mark’s was modelled on the Church of the Holy Apostles at Constantinople
and adorned by mosaicists from the same city. They were artists, rich in
invention, and endowed with a perfect sense of beauty in design. The reliefs
imbedded in the façades of St Mark’s and in scores of houses about the city
of Venice bear ample testimony to their greatness. In the thirteenth and
fourteenth centuries Byzantine art had become degenerate, and traces of
native Venetian sculpture as early as the twelfth century have been argued
from the rude carvings on the pillars which support the tabernacle of the
high altar of St Mark’s. But it is not till we reach the masters of the
characteristic Gothic, Transitional, and early Renaissance styles, that the
important place due to Venice in the history of the mason’s craft is made
clear. It is doubtful whether any of the Pisani actually worked at Venice,
though their influence is beyond dispute. But Nicolo Lamberti, a Tuscan
sculptor, worked on the decorations of the main archivolt of St Mark’s,
Florentine artists carved some of the best figures and capitals in the façades
of the Ducal Palace, and wrought one of the finest tombs in S. Zanipolo;
Michelozzo is said to have built the Medici library at the monastery of St
Giorgio Maggiore; and the design so often met with in monumental
sculpture in Venice, two angels, one at either end of the tomb drawing aside
a curtain to display the recumbent effigy of the dead was invented by a
Florentine, Arnolfo di Cambio. Of the earliest Venetian masters it may be
truly said that their works live after them, for little beyond their works is
known of the Massegne and the Buoni who reached a comparatively higher
stage of excellence in sculpture than their contemporaries did in painting.
Jacobello and Piero delle Massegne (dei Macigni[62]), thus called because of
their craft, were working in Venice towards the close of the fourteenth
century. The statues of the apostles, the Virgin, and St Mark over the choir-
screen, and others in the choir chapels of St Mark’s, the main portal of S.
Stefano, the beautiful lunette over the Friar’s door at the Frari, and the tomb
of Simone Dandolo in the same church are excellent examples of the style of
these great artists. How much of the sculpture on the façades of the Ducal
Palace was due to the Buoni it is difficult to say. The Buoni seem to have
been Giovanni, the father, his son, Bartolomeo, and a certain Pantaleone
Buon, once believed to be another son, but actually of no kinship with him.
To Bartolomeo, sometimes known as Bartolomeo della Madonna dell’Orto,
is ascribed the Porta della Carta of the Ducal Palace (1439) on the strength
of an inscription opera Bartholomei. The reliefs of the Lion and of Doge
Foscari are modern reproductions, but the original head of Foscari, preserved
in the palace still, bears witness to the genius of this great craftsman.[63] He
must not be confounded with another Bartolomeo Buon, known as Master
Bartolomeo of Bergamo, also claimed as a native of Venice, who in 1493
superintended the painters in the Ducal Palace, and in 1500 presided over the
works designed by Pietro Lombardo for the Procuratie Vecchie.
Certain craftsmen dubbed Riccio or Rizzo (Curly pate) now claim
attention. Their identity is much canvassed by Italian authorities. Before the
use of surnames became common it was the custom to refer to
contemporaries by their Christian or nicknames to the confusion of
biographers and critics. We meet with three Ricci who are stated to have
worked at Venice, (1) Andrea Riccio of Padua living about 1400, who is said
by Vasari to have executed the statues of Adam and Eve to be referred to
presently. (2) Antonio Riccio or Rizzo, sometimes called Briosco, of Verona,
who, according to Zanotto[64] was employed by the Republic to assist
Antonio Loredan at the siege of Scutari.[65] He returned, after most effective
service, covered with wounds, and the grateful Senate voted him and his
sons in 1483 a pension for twenty years, and appointed him architect of the
Ducal Palace after the fire in the same year, authorising him to draw for
funds on the salt office. To him, and not to Andrea, are ascribed by Zanotto
the masterly statues of Adam and Eve in the niches opposite the Scala dei
Giganti, the Scala itself and the adjoining façade in the cortile and the rio
façade. These on Francesco Sansovino’s authority are more commonly
attributed to Antonio Bregno (also called Il Riccio or Rizzo). He is said to
have been a contemporary of Scarpagnino (Antonio Scarpagni), who, in
1514, submitted designs for the new stone Rialto Bridge, and succeeded
Sante Lombardo in the erection of the Scuola di S. Rocco. Bregno, however,
is a mysterious figure who, so Zanotto declares, either never existed or was
none other than Riccio the Veronese. It is clear, however, from the annals of
Malipiero that in 1498, one Antonio Riccio or Rizzo, architect of the Ducal
Palace, after spending 80,000 ducats left the work not half done, that he had
by forged vouchers defrauded the Salt Office to the extent of 12,000 ducats,
and bolted to Foligno, where he soon died.[66] To the family of Venetian
masters (or, according to some authorities, Lombard immigrants from
Carona on Lake Lugano), known as the Lombardi, are due the most beautiful
and original of the early renaissance architecture and sculpture in Venice.
Pietro Lombardo, said to have been the son of a mason named Martino, was
working in Venice in 1462. In 1481 his design for the Church of the Miracoli
was chosen, and the building was erected under his superintendence. When
Ant. Riccio fled from Venice, Pietro succeeded him at the Ducal Palace, with
a salary of 120 ducats, and for twelve years was the official architect of the
Republic. Among his works in Venice are the fine statues of St Anthony and
three other saints in S. Stefano. The altars of St James and St Paul in St
Mark’s are also attributed to him. He is probably best known as the sculptor
of the Dante Memorial at Ravenna. Antonio Lombardi, born before 1453,
assisted his father at the Miracoli and on the tomb of Doge Pietro Mocenigo
at S. Zanipolo. He collaborated with Aless. Leopardi on the bronze work in
the Cappella Zen at St Mark’s. The statue of St Thomas Aquinas in S.
Zanipolo is attributed to him. Martino Lombardo, whose relationship is
unknown, was architect of the Scuola di San Marco after 1485, and was
believed by Temanza to have built S. Zaccaria.[67] Moro Lombardo,
probably a son of Martino, assisted his father in the Scuola di S. Marco. In
1524-7, Giulio Lombardo, probably son of Pietro, was acting in an advisory
capacity to Sante or Zante Lombardo, a son of Pietro, born 1504, in the
works at the Scuola di S. Rocco. The Church of S. Giorgio dei Greci
ascribed to Sansovino is now attributed to Sante and one named Chiona.

PONTE DI RIALTO, FROM THE MARKET.

Tullio Lombardi (1453-1537) was the son of Pietro, and the best sculptor
of the family. Beside his work in the interior of the Miracoli, he executed the
reliefs on the façade of the Scuola di S. Marco and the monument to Giov.
Mocenigo in S. Zanipolo. He also collaborated with Leopardi on the
Vendramin tomb in the same church. To the Lombardi school we owe the
beautiful Cappella Giustiniano at S. Francesco della Vigna, and a fine relief
in S. Giov. Grisostomo (Coronation of the Virgin and the Twelve Apostles).
Aless. Leopardi (1450-1521) raised Venetian sculpture to its highest plane
of technical perfection. The Venetian artist was peculiarly privileged. Unlike
the Pisani and other Tuscans who drew their inspiration from Roman
antiques, he was able to draw from the fountain-head. The lands of Hellas
were subject to the Republic, and, doubtless, many a young apprentice spent
his Wanderjahre there. Enthusiasm gave insight, and both in technique and
design we seem to trace in Tullio Lombardo and Aless. Leopardi the
influence of Greek originals. To Leopardi are due the Vendramin tomb in S.
Zanipolo, the finest of renaissance sepulchral monuments, and the
completion of the Colleoni statue. He modelled the Six Virtues and the
Madonna Della Scarpa in the Cappella Zen. The three magnificent bronze
bases for the flagstaffs in front of St Mark’s were wrought by him. But soon
aversion from the study of nature, and the growing pomp of private and
public life reacted on the renaissance artists; their work became mannered
and feeble; they lost individuality and character. They found in Venice a rich
field for exploitation. She was not only the wealthiest, she was the most
tranquil of European states. Imperial in policy, oligarchical in government,
she sought by the splendour of the arts and by magnificent pageantry to feed
the pride of her nobles, and lay any spirit of political freedom that might
have survived in her people.
A giant among the sixteenth century masters who were attracted to Venice
was Jacopo Tatti (1477-1570) of Florence, the bosom friend and colleague of
Andrea del Sarto, known as Sansovino, from his intimate association with
his master, Andrea Contucci, of Monte Sansovino. Jacopo, while sketching
from the antique at Rome, attracted the notice of Bramante, who was
charmed by a wax model of the Laocoon executed by the young student, and
judged by Raphael to be the best of four others. It was cast in bronze, and
subsequently found its way to the Signory of Venice. In 1527, after the sack
of Rome, he came to Venice, and was employed by Doge Gritti to strengthen
the domes of St Mark’s. He did his work so amazingly well (fece stupire
Venezia, says Vasari) that he was appointed in 1529 chief architect, with a
house and a salary of 80 ducats, afterwards increased to 180. In 1536 the
Senate decreed the erection of a library to contain the books left to the
Republic by Petrarch and Cardinal Bassarione. Sansovino was charged with
the building, now known as the Libreria Vecchia, and esteemed by Palladio
to be probably the richest and most ornate edifice erected since the time of
the ancients. The Signory were royal pay-masters, but intolerant of bad
work; and when, on December 18, 1545, part of the vaulting fell, Sansovino
was imprisoned, fined a thousand crowns, and deprived of his office. He
succeeded, however, in proving his innocence, and was released and
compensated by a solatium of 900 crowns, and restored to his former
position. Sansovino’s work, however, ends at the sixteenth arch from the
Campanile corner. Twelve years after his death it was finished by Scamozzi.
He was a most lovable artist, ever ready with help and counsel to those who
entreated him; the friend of every great man of his time; in youth a most
winning personality; in age venerable and alert. At ninety-three, if we may
trust Vasari, his eyes were undimmed, and he bore himself erect as ever.
Among other works by him at Venice may be specified the beautiful loggia
destroyed by the collapse of the Campanile, in July 1902, and the bronze
doors leading to the sacristy, St Mark’s, on which he is said to have worked
during a period of twenty years; the six bronze reliefs in the choir of the
same church; the colossal statues of Mars and Neptune at the top of the
giants’ staircase, and the Scala d’Oro in the Ducal Palace, and many
mansions and churches, the choicest of which, S. Geminiano, no longer
exists.
Of all his followers, Girolamo Campagna is the most talented. Good
examples of his works are the bronze statues of St Mark and St Francis in
the Redentore, the small statues of St Francis and St Clare in the Miracoli,
and the reclining figure of Doge Cicogna (1595) in the Gesuiti. Aless.
Vittoria of Trent (1525-1603) was a facile artist. Among his works are the
statue of St Sebastian in S. Salvatore, the fine bust of Cardinal Gasparo in
the Madonna dell’ Orto, the ruined chapel of the Rosary in S. Zanipolo, and
his own tomb in S. Zaccaria. Michele Sammichele (1484-1559), the great
Veronese master and famous military engineer, was employed by the
Republic between 1530 and 1550, and designed the great fortifications in the
mainland provinces, on the Dalmatian coast, at Corfù, Cyprus and Romania,
many of which remain to this day. On his return to Venice, he constructed
the magnificent fortress of S. Andrea del Lido, a stupendous work, now
threatened with ruin, owing to erosion by currents set up by the new dykes
near the Lido. The Palazzo Grimani on the Grand Canal, the Ponte del
Bucintoro at the Arsenal are by this master, whose architecture so dominates
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com