100% found this document useful (2 votes)

904 views

Kansei Engineering and Soft Computing Theory and Practice Premier Reference Source 1st Edition Ying Dai 2024 scribd download

Engineering

Uploaded by

jehehematiwe

Available Formats

Download as PDF, TXT or read online on Scribd

100% found this document useful (2 votes)

904 views

Kansei Engineering and Soft Computing Theory and Practice Premier Reference Source 1st Edition Ying Dai 2024 scribd download

Engineering

Uploaded by

jehehematiwe

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 72

Visit https://ebookultra.

com to download the full version and

explore more ebooks

Kansei Engineering and Soft Computing Theory and

Practice Premier Reference Source 1st Edition Ying
Dai

_ Click the link below to download _

https://ebookultra.com/download/kansei-engineering-and-
soft-computing-theory-and-practice-premier-reference-
source-1st-edition-ying-dai/

Explore and download more ebooks at ebookultra.com

Here are some recommended products that might interest you.
You can download now and explore!

Gender and Social Computing Interactions Differences and

Relationships Premier Reference Source 1st Edition Celia
Romm Livermore
https://ebookultra.com/download/gender-and-social-computing-
interactions-differences-and-relationships-premier-reference-
source-1st-edition-celia-romm-livermore/
ebookultra.com

Electronic Resource Management in Libraries Research and

Practice Premier Reference Source 1st Edition Holly Yu

https://ebookultra.com/download/electronic-resource-management-in-
libraries-research-and-practice-premier-reference-source-1st-edition-
holly-yu/
ebookultra.com

Successes and New Directions in Data Mining Premier

Reference Source 1st Edition Pascal Poncelet

https://ebookultra.com/download/successes-and-new-directions-in-data-
mining-premier-reference-source-1st-edition-pascal-poncelet/

ebookultra.com

Managing Web Service Quality Measuring Outcomes and

Effectiveness Premier Reference Source 1st Edition Khaled

https://ebookultra.com/download/managing-web-service-quality-
measuring-outcomes-and-effectiveness-premier-reference-source-1st-
edition-khaled/
ebookultra.com
Open Information Management Applications of
Interconnectivity and Collaboration Premier Reference
Source 1st Edition Samuli Niiranen
https://ebookultra.com/download/open-information-management-
applications-of-interconnectivity-and-collaboration-premier-reference-
source-1st-edition-samuli-niiranen/
ebookultra.com

Ethics and Game Design Teaching Values Through Play

Premier Reference Source 1st Edition Karen Schrier

https://ebookultra.com/download/ethics-and-game-design-teaching-
values-through-play-premier-reference-source-1st-edition-karen-
schrier/
ebookultra.com

Agent and Web Service Technologies in Virtual Enterprises

Premier Reference Source 1st Edition Nicolaos Protogeros

https://ebookultra.com/download/agent-and-web-service-technologies-in-
virtual-enterprises-premier-reference-source-1st-edition-nicolaos-
protogeros/
ebookultra.com

Information Technology Governance and Service Management

Frameworks and Adaptations Premier Reference Source 1st
Edition Aileen Cater-Steel
https://ebookultra.com/download/information-technology-governance-and-
service-management-frameworks-and-adaptations-premier-reference-
source-1st-edition-aileen-cater-steel/
ebookultra.com

Social Software and Web 2 0 Technology Trends Premier

Reference Source 1st Edition P. Candace Deans

https://ebookultra.com/download/social-software-and-
web-2-0-technology-trends-premier-reference-source-1st-edition-p-
candace-deans/
ebookultra.com
Kansei Engineering and Soft Computing Theory and
Practice Premier Reference Source 1st Edition Ying Dai
Digital Instant Download
Author(s): Ying Dai
ISBN(s): 9781616927998, 1616927992
Edition: 1
File Details: PDF, 23.98 MB
Year: 2010
Language: english
Kansei Engineering and
Soft Computing:
Theory and Practice

Ying Dai
Iwate Prefectural University, Japan

Basabi Chakraborty
Iwate Prefectural University, Japan

Minghui Shi
Xiamen University, China

EnginEEring sciEncE rEfErEncE

Hershey • New York
Director of Editorial Content: Kristin Klinger
Director of Book Publications: Julia Mosemann
Acquisitions Editor: Lindsay Johnston
Development Editor: Christine Bufton
Publishing Assistant: Keith Glazewski
Typesetter: Keith Glazewski
Production Editor: Jamie Snavely
Cover Design: Lisa Tosheff

Published in the United States of America by

Engineering Science Reference (an imprint of IGI Global)
701 E. Chocolate Avenue
Hershey PA 17033
Tel: 717-533-8845
Fax: 717-533-8661
E-mail: cust@igi-global.com
Web site: http://www.igi-global.com

Copyright © 2011 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in
any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or com-
panies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.

Library of Congress Cataloging-in-Publication Data

Kansei engineering and soft computing : theory and practice / Ying Dai, Basabi Chakraborty, and Minghui Shi, editors.
p. cm.
Includes bibliographical references and index.
Summary: "This book focuses on the qualitative and quantitative evaluation of kansei, including the measurement and analysis
of the impressions, emotions, and sense perceptions of groups and individuals, as well as the development of procedures for
the representation, design and creation of products, machines and systems that embody the kansei of groups and individuals"-
-Provided by publisher.
ISBN 978-1-61692-797-4 (hardcover) -- ISBN 978-1-61692-799-8 (ebook) 1. Human engineering. 2. Soft computing. 3.
Senses and sensation. 4. System design--Psychological aspects. 5. Industrial design--Psychological aspects. I. Dai, Ying,
1963- II. Chakraborty, Basabi, 1956- III. Shi, Minghui, 1973-
TA166.K363 2011
006.3--dc22
2009053466

British Cataloguing in Publication Data

A Cataloguing in Publication record for this book is available from the British Library.

All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the
authors, but not necessarily of the publisher.
Editorial Advisory Board
Alan Chin-Chen Chang, Feng Chia University Taiwan, China
Goutam Chakraborty, Iwate Prefectural University, Japan
Guozheng Li, Tongji University, China
Hideyuki Takagi, Kyushu University, Japan
Hisao Shiizuka, Kogakuin University, Japan
Hugh de Garis, Xiamen University, China
Jayanta Basak IBM, India Ltd., India
Kaori Yoshida, Kyushu Institute of Technology, Japan
Kenzoh Itoh, Iwate Prefectural University, Japan
S. Bandopadhyay, Indian Statistical Institute, Calcutta, India
Shaozi Li, Xiamen University, China
Sushmita Mitra, Indian Statistical Institute, Calcutta, India
Toshi Kato, Chuo University, Japan
Wei Dai, Victoria University, Australia
Wenhua Zeng, Xiamen University, China
Yaokai Feng, Kyushu University, Japan
Yin Feng, Xiamen University, China
Zhibin Pan, Xian Jiaotong University, China
Zhongpan Qiu, Xiamen University, China

List of Reviewers
C. A. Murthy, University of Calcutta, India
Elizabeth Damour, Act&be, Suivi d’Effets / Creative Management, France
Goutam Chakraborty, Iwate Prefectural University, Japan
Guozheng Li, Tongji University, China
Hugo de Garis, Xiamen University, China
Jayanta Basak, IBM, India Ltd., India
Kamyar Mehran, Newcastle University, UK
Kaori Yoshida, Kyushu Institute of Technology, Japan
Masayoshi Kanoh, Chukyo University, Japan
Mingyu You, Tongji University, China
Naotoshi Sugano, Tamagawa University, Japan
Pierre Lévy, Eindhoven University of Technology, The Netherlands
S. Bandopadhyay, Indian Statistical Institute, India
S.V. Ulyanov, International University of Nature, Society, and Man, Russia
Sanghamitra Bandyopadhyay, University of Burdwan, India
Shang-fei Wang, University of Science and Technology of China, China
Shaozi Li, Xiamen University, China
Soma Mitra, Consultant, India
Sushmita Mitra, Indian Statistical Institute, India
Tadashi Hasebe, Tohoku University, Japan
Tao Song, Consultant, China
Wei Dai, Victoria University, Australia
Weijie Wang, Southeast University, China
Yanping Lu, Xiamen University, China
Yaokai Feng, Kyushu University, Japan
Yasuhiro Yamada, Shimane University, Japan
Yin Feng, Xiamen University, China
Zhibin Pan, Xian Jiaotong University, China
Zhongpan Qiu, Xiamen University, China
Table of Contents

Foreword ............................................................................................................................................. xv

Preface ............................................................................................................................................... xvii

Acknowledgment ............................................................................................................................... xxii

Section 1
Basic Concepts, Frameworks and Techniques

Chapter 1
Soft Computing and its Applications ...................................................................................................... 1
Siddhartha Bhattacharyya, The University of Burdwan, India
Ujjwal Maulik, Jadavpur University, India
Sanghamitra Bandyopadhyay, Indian Statistical Institute, India

Chapter 2
Particle Swarm Optimizer for High-Dimensional Data Clustering ...................................................... 31
Yanping Lu, Xiamen University, China
Shaozi Li, Xiamen University, China

Chapter 3
Text Mining for Analysis of Interviews and Questionnaires ................................................................ 52
Yasuhiro Yamada, Shimane University, Japan
Kanji Kato, GK Sekkei Incorporated, Japan
Sachio Hirokawa, Kyushu University, Japan

Chapter 4
Intra-Class Threshold Selection in Face Space Using Set Estimation Technique ................................ 69
Madhura Datta, University of Calcutta, India
C.A. Murthy, Indian Statistical Institute, India
Chapter 5
DNA Computing ................................................................................................................................... 85
Tao Song, Huazhong University of Science and Technology, China
Xun Wang, University of Tsukuba, Japan
Shudong Wang, Shandong University of Science and Technology, China
Yun Jiang, Huazhong University of Science and Technology, China

Chapter 6
Kansei Database and AR*-Tree for Speeding up the Retrieval .......................................................... 111
Yaokai Feng, Kyushu University, Japan

Chapter 7
Emotional Semantic Detection from Multimedia: A Brief Overview ................................................ 126
Shang-fei Wang, University of Science and Technology of China, China
Xu-fa Wang, University of Science and Technology of China, China

Chapter 8
Fuzzy Logic for Non-Smooth Dynamical Systems ............................................................................ 147
Kamyar Mehran, Newcastle University, UK
Bashar Zahawi, Newcastle University, UK
Damian Giaouris, Newcastle University, UK

Chapter 9
An Intimate Relation: Human Beings with Humanoids ..................................................................... 169
Elisabeth Damour, Act & Be - Creative Management Consultancy, France

Section 2
Measurement, Analysis, and Representation of Kansei

Chapter 10
Music and Kansei: Relations between Modes, Melodic Ranges, Rhythms, and Kansei .................... 180
Shigekazu Ishihara, Hiroshima International University, Japan
Mitsuo Nagamachi, Hiroshima International University, Japan
Jun Masaki, Nagasaki Prefectural Government, Japan

Chapter 11
Analyses of People’s Perceptions on Sidewalk Environments Combining Factor Analysis
and Rough Sets Approach ................................................................................................................... 199
Weijie Wang, Southeast University, China
Wei Wang, Southeast University, China
Moon Namgung, Wonkwang University, South Korea
Chapter 12
Affective Facial Expressions Using Auto-Associative Neural Network
in Kansei Robot ‘‘Ifbot’’ ..................................................................................................................... 215
Masayoshi Kanoh, Chukyo University, Japan
Tsuyoshi Nakamura, Nagoya Institute of Technology, Japan
Shohei Kato, Nagoya Institute of Technology, Japan
Hidenori Itoh, Nagoya Institute of Technology, Japan

Chapter 13
Natural Effect of Spatial and Temporal Color Sequence on Human Color Impression ..................... 237
Naotoshi Sugano, Tamagawa University, Japan

Chapter 14
Rural Scenery Narrative and Field Experiences: From an Aspect of Kansei ..................................... 255
Tadashi Hasebe, Tohoku University, Japan
Michiaki Omura, Tohoku University, Japan
Hisashi Bannai, Rural Finance Research Institute, Japan

Chapter 15
Psychophysiological Applications in Kansei Design.......................................................................... 266
Pierre Lévy, Eindhoven University of Technology, The Netherlands & Chiba University, Japan
Toshimasa Yamanaka, University of Tsukuba, Japan
Oscar Tomico, Eindhoven University of Technology, The Netherlands

Chapter 16
Analyzing Coordinate Relations in Handwriting Activity: Tacit Skill and Individuality ................... 287
Yusuke Manabe, Chiba Institute of Technology, Japan
Kenji Sugawara, Chiba Institute of Technology, Japan

Chapter 17
Kansei’s Physiological Measurement and its Application (1): Salivary Biomarkers
as a New Metric for Human Mental Stress ......................................................................................... 303
Shusaku Nomura, Nagaoka University of Technology, Japan

Chapter 18
Kansei’s Physiological Measurement and its Application (2): Estimation of Human
States Using PCA and HMM .............................................................................................................. 319
Santoso Handri, Nagaoka University of Technology, Japan
Shusaku Nomura, Nagaoka University of Technology, Japan
Chapter 19
The China Brain Project: An Evolutionary Engineering Approach to Building China’s First
Artificial Brain Consisting of 10,000s of Evolved Neural Net Minsky-Like Agents ......................... 330
Hugo de Garis, Xiamen University, China
Xiaoxi Chen, Xiamen University, China
Ben Goertzel, Novamente LLC, USA & Singularity Institute, USA & Xiamen University, China

Compilation of References .............................................................................................................. 360

About the Contributors ................................................................................................................... 395

Index ................................................................................................................................................... 406

Detailed Table of Contents

Foreword ............................................................................................................................................. xv

Preface ............................................................................................................................................... xvii

Acknowledgment ............................................................................................................................... xxii

Section 1
Basic Concepts, Frameworks and Techniques

The architecture of neural networks, the underlying concepts of fuzzy set theory and fuzzy logic, the
heuristic search techniques [including genetic algorithm (GA), simulated annealing (SA), and ant col-
ony optimization (ACO)], and an outline of the rough set theory are briefly discussed. The application
of soft computing techniques in kansei-related issues is also discussed.

New Particle Swarm Optimization (PSO) algorithms for clustering high-dimensional data and auto-
matically determining the number of clusters for data mining applications are developed. As these
algorithms are global search techniques, they can also be used to determine optimal feature subspace
selection or to solve subspace clustering problems.
Chapter 3
Text Mining for Analysis of Interviews and Questionnaires ................................................................ 52
Yasuhiro Yamada, Shimane University, Japan
Kanji Kato, GK Sekkei Incorporated, Japan
Sachio Hirokawa, Kyushu University, Japan

A brief survey of basic text mining techniques used to analyze interviews and questionnaires, including
keyword extraction, word graphs, clustering of text and association rule mining, is introduced. Then,
an example of text mining applied to interview and questionnaire analysis is provided. Finally, the ad-
vantages and disadvantages of text mining techniques are discussed.

The development of a face recognition system for person authentication with a novel threshold selec-
tion technique for correct decision-making incorporating human cognitive behavior is presented. The
effectiveness of the approach is demonstrated by simulation experiments with real data in a real-world
environment.

Chapter 5
DNA Computing ................................................................................................................................... 85
Tao Song, Huazhong University of Science and Technology, China
Xun Wang, University of Tsukuba, Japan
Shudong Wang, Shandong University of Science and Technology, China
Yun Jiang, Huazhong University of Science and Technology, China

The development of DNA computing, a form of biomolecular computing, is discussed, along with an
introduction to the theoretical framework and formal models and a description of several models used
to solve real-world problems. Practical problems regarding DNA encoding and the bottlenecks for this
new tool are also discussed.

Chapter 6
Kansei Database and AR*-Tree for Speeding up the Retrieval .......................................................... 111
Yaokai Feng, Kyushu University, Japan

After reviewing existing kansei retrieval systems, the general flow of kansei retrieval systems and the
structure of a typical kansei database are presented. Furthermore, multidimensional index structures
used in kansei retrieval systems, including the AR*-tree used to speed up the retrieval process, are
described and discussed.
Chapter 7
Emotional Semantic Detection from Multimedia: A Brief Overview ................................................ 126
Shang-fei Wang, University of Science and Technology of China, China
Xu-fa Wang, University of Science and Technology of China, China

Advances in research on emotional semantic detection from multimedia are surveyed. A general over-
view of research is provided; research on emotional semantic detection from images, videos, and mu-
sic, as well as their typical prototype systems, is introduced; and, finally, future research directions are
identified.

The fundamental concepts of non-smooth dynamical systems theory, together with case studies, are
described, and a successful Takagi-Sugeno (TS) fuzzy modeling approach in modeling the dynamics
of complex nonlinear systems is introduced. Then, a new type of TS-based fuzzy model is proposed to
represent a non-smooth dynamical system, and the method of predicting the onset of instability leading
to chaos for the non-smooth dynamical system is depicted.

The relationship between human beings and humanoid robots, including some characteristic with their
influence of future humanoids, such as immortality, cyber-body, human identity, and consciousness, are
discussed. Further, strategies for making robots benevolent and grateful towards humans are presented.

Section 2
Measurement, Analysis, and Representation of Kansei

Music-related research, classified into 6 categories, is described. Kansei evaluation experiments on

modes, melodic ranges, and rhythms are introduced, and a real-time kansei music recognition system
based on the results of analysis is developed. Further, internet-based music services incorporating kan-
sei concepts are reviewed.
Chapter 11
Analyses of People’s Perceptions on Sidewalk Environments Combining Factor Analysis
and Rough Sets Approach ................................................................................................................... 199
Weijie Wang, Southeast University, China
Wei Wang, Southeast University, China
Moon Namgung, Wonkwang University, South Korea

The usefulness of factor analysis, combined with the rough sets theory as an operational tool for linking
participants’ perceptions with the physical components of sidewalk environments, is illustrated. The
methods of sidewalk photograph collection, field survey, and psychological survey are introduced. Fur-
thermore, methods for extracting decision rules from data generated by factor analysis using the rough
sets theory are presented in detail.

Chapter 12
Affective Facial Expressions Using Auto-Associative Neural Network
in Kansei Robot ‘‘Ifbot’’ ..................................................................................................................... 215
Masayoshi Kanoh, Chukyo University, Japan
Tsuyoshi Nakamura, Nagoya Institute of Technology, Japan
Shohei Kato, Nagoya Institute of Technology, Japan
Hidenori Itoh, Nagoya Institute of Technology, Japan

The development of Ifbot, a kansei robot capable of communicating inner emotions through facial
expressions, is discussed. The mapping of human emotions and facial expressions has been achieved
using an auto-associative neural network, a soft computing tool. A method for creating personality us-
ing facial expressions is also proposed.

Chapter 13
Natural Effect of Spatial and Temporal Color Sequence on Human Color Impression ..................... 237
Naotoshi Sugano, Tamagawa University, Japan

The different effects of spatial (or temporal) ordered color sequences on the naturalness are examined.
These effects were investigated by subjects who assessed the level of “naturalness” of linear, toroidal,
or circular sequences of six colors. The relationship between sequence patterns and the impression of
“naturalness” is explored, and some simple fuzzy rules are extracted by calculating the projected route
area and the route complexity of a hexagonal diagram of six color sequences.

Kitarou Nishida’s theory of environment is adapted for the purpose of explaining the theory of scenery.
In order to present the concept of a scenery narrative, this theory of scenery and Keiichi Noe’s narrative
theory are integrated. In the subsequent section, the theoretical relationship between scenery narratives
and previous or current work experiences in the field are discussed. Finally, a hypothesis from the nar-
rative theory of scenery is developed and tested in a case study.

An overview of the emerging approaches for kansei design based on its relation to psychophysiology
has been presented. Tools and techniques are developed in order to incorporate human inspiration and
mental images into the design process and are supported by real-world examples and applications.

A framework for analyzing handwriting as an example of a human skill with hidden elements of indi-
viduality is described. An approach for identity detection using handwriting captured by a computer
pen tablet with the help of soft computing tools is presented and is supported by simulation experiments
with benchmark data.

Biomarkers and biomarker studies, namely saliva sampling and quantitative determination of salivary
biomarkers, are introduced. After describing the experimental procedure of biomarkers against short-
term stressors, the results of subjects’ salivary concentration changes and accumulative changes in task/
break experiments are discussed. Then, a mathematical model reflecting biomarker response against
intermittent short-term stressors is proposed.

Modes for acquiring several types of biosignals, including electrocardiogram (ECG) and ECG pro-
cessing, are introduced. After an overview of the use of ECG as an indicator of fatigue, an experiment
evaluating mental health problems is introduced. Approaches integrating PCA, HMM and NN are used
to estimate human states of fatigue and vigor based on ECG, electroencephalogram (EEG), and ther-
mograph data.

Chapter 19
The China Brain Project: An Evolutionary Engineering Approach to Building China’s First
Artificial Brain Consisting of 10,000s of Evolved Neural Net Minsky-Like Agents ......................... 330
Hugo de Garis, Xiamen University, China
Xiaoxi Chen, Xiamen University, China
Ben Goertzel, Novamente LLC, USA & Singularity Institute, USA & Xiamen University, China

The “Parcone” (Partially Connected Neural Evolutionary) model, which is used as the basis for all
artificial brain projects, and the operating system software “IMSI” (Inter Module Signaling Interface)
are described. The capabilities of the ABL’s NAO robots and the interface between the IMSI operating
system and NAO are introduced. A series of demos and discussions about the main aims and major
challenges of this research project are presented. Finally, an improved model called CUDA-based par-
tially connected neural evolutionary model, which is developed to speed up the evolution process, is
presented in the appendix.

Compilation of References .............................................................................................................. 360

About the Contributors ................................................................................................................... 395

Index ................................................................................................................................................... 406

Foreword

The field of artificial intelligence evolves with the objective of developing intelligent machines that
could think the way humans do. Perception-generated feelings and feeling-driven actions are two of the
most important aspects of human intelligence. These require a knowledge system in which subjective
impression and emotions of human’s responses to the surroundings could be described and processed.
Psychological impression or human emotions are unlikely to be fully captured, but these may be apparent
at the subconscious level. Kansei Engineering has been developed to measure, analyze, and represent
such information which is defined as Kansei. Research into Kansei involves determining which sensory
attributes elicit particular subjective responses from people, and then designing a product, a machine or
a system using the attributes which elicit the desired responses.
Among these technologies involved in Kansei Engineering, soft computing is undoubtedly the most
popular and important approach of computational intelligence where artificial neural networks, fuzzy
logic, rough set theory, evolutionary computation and other hybrid techniques are used to the knowledge
representation and decision making of humans. Due to its tolerance to imprecision, uncertainty and partial
truth, soft computing deals well with human related systems. Since the process of solving real world
problems by inducing the concept of Kansei often interacts with human and involves human’s affections
with uncertainty, imprecision and dependence on the context, the recent advances in soft computing and
Kansei engineering, and the fusion of these two fields are considered to be critical. The book, Kansei
Engineering and Soft Computing: Theories and Practice, here KESC for short, is one of the first texts
to focus on the topics related to these fields.
As soft computing steps into the field of Kansei engineering, one needs to go from the development
and enhancement of modeling, algorithm development, computer simulation, and implementation in
useful prototypes for rigorous testing and evaluation. Applications of soft computing to kansei related
real world problems are sufficiently complex. This book, KESC, edited by Dr. Ying Dai, Dr. Basabi
Chakraborty, and Dr. Minghui Shi, provides a comprehensive overview of the theories, methodologies,
and recent developments in the field of Kansei engineering towards this end. The book covers such
important topics as the framework of soft computing, up-to-date soft computing techniques and their
applications in Kansei engineering, the practice in measuring, analyzing and representing Kansei, and
the Kansei based product design.
KESC is a book well-organized, clearly presented and illustrated by many examples and many experi-
ments. It is a must reading for ones interested in acquiring an understanding of what Kansei engineering
is, developing an ability to employ it for human centric systems, and utilizing the soft computing tech-
xvi

niques for real world problems with uncertainty and imprecision. The editors and contributors deserve
our thanks and congratulations for producing a book, which adds so much and so importantly to the
advancement of Kansei engineering and soft computing.

Nanning Zheng, Ph.D.

IEEE Fellow
Professor and Director
Institute of Artificial Intelligence and Robotics,
Xi’an Jiaotong University, China

Nanning Zheng (IEEE SM’93-F’06) graduated from the Department of Electrical Engineering, Xi’an Jiaotong University,
Xi’an, China, in 1975, and received the MS degree in information and control engineering from Xi’an Jiaotong University in
1981 and the PhD degree in electrical engineering from Keio University, Yokohama, Japan, in 1985. He jointed Xi’an Jiaotong
University in 1975, and he is currently a Professor and the Director of the Institute of Artificial Intelligence and Robotics, Xi’an
Jiaotong University. His research interests include computer vision, pattern recognition and image processing, and hardware
implementation of intelligent systems. Dr. Zheng became a member of the Chinese Academy of Engineering in 1999, and he is
the Chinese Representative on the Governing Board of the International Association for Pattern Recognition. He also serves
as an executive deputy editor of the Chinese Science Bulletin.
xvii

Preface

With the increasing concern regarding human factors in system development, the concepts of human-
ized technology and human-related systems have become the focus of more and more research. Kansei
engineering and soft computing are the most representative research fields in this area.
The word kansei (derived from the Japanese 感性 [kansei]), refers to human feelings, such as impres-
sions, affect, and emotions, derived by observing the surrounding environment. Such feelings can arouse
people to act instinctively—for example, just as happiness results in laughter, enjoying something will
motivate an individual to obtain it. In this way, the concept of kansei consists of two distinct aspects:
perception-generated feelings and feeling-driven action. Both aspects include issues of uncertainty,
diversity, and dependence on the environment. Kansei engineering aims to solve these problems while
bringing together culture and technology through the establishment of an information society based on
the concept of kansei—in other words, the harmonization of the social, cultural, and natural sciences and
technology with human skills, and the creation and promotion of human happiness [1]. With this goal in
mind, kansei engineering focuses on the qualitative and quantitative evaluation of kansei, including the
measurement and analysis of the impressions, emotions, and sense perceptions of groups and individu-
als, as well as the development of procedures for the representation, design and creation of products,
machines and systems that embody the kansei of groups and individuals. The theories and technologies
of soft computing are therefore essential developmental tools in kansei engineering.
Soft computing, a consortium of tools and techniques suitable for dealing with the uncertainty and
imprecision of human-centric computing, comprises artificial neural networks, fuzzy logic, the rough set
theory, evolutionary computation and other hybrid techniques. Recently, soft computing tools have been
developed to solve many real-world problems involving human behavior. Until now, kansei engineering
and soft computing have developed as two independent fields.
This book offers the reader a comprehensive review of kansei engineering, soft computing techniques,
and the fusion of these two fields from a variety of viewpoints. After introducing the traditional tech-
nologies, the book’s focus shifts to the solution of real-world problems through the concept of kansei
and the effective utilization of soft computing techniques, while such real world problems often involve
uncertainty and imprecision, and are dependence on the context. Cutting-edge research on the measure-
ment of kansei and its application in areas such as design, production, and healthcare is also introduced.
The book aims to reach professionals, researchers and students in the field of kansei information
processing and soft computing, both in academia and industry. It will also serve as a reference book
for professionals, researchers and students interested in studying existing kansei engineering and soft
computing techniques from theoretical and practical viewpoints and obtaining insight into the applica-
tion of kansei research in humanized technology or human-related systems.
xviii

The book is divided into two sections and comprises of 19 chapters. Section 1, “Basic Concepts,
Framework and Techniques”, consisting of 9 chapters, introduces the framework of soft computing,
current soft computing techniques, the general process of constructing kansei retrieval systems, and the
implementation of kansei factors in dynamic systems. Section 2, “Measurement, Analysis, and Represen-
tation of Kansei”, consisting of 10 chapters, covers many critical kansei issues, such as the psychological
and physiological measurement of kansei, representing kansei by analyzing the relationship between
human’s feelings and physical attributes of entities, and the incorporation of kansei into product design.
The first chapter provides an introduction to the three essential paradigms of soft computing: neural
network, fuzzy logic and evolutional computation. These paradigms are integrated to provide a framework
for flexible information processing, which is useful for processing kansei-related issues. In particular, the
notions, methodologies, and some algorithms of neural networks, fuzzy set theory, and heuristic search
techniques, which are utilized throughout the book, are mainly introduced and discussed. Furthermore,
methods for applying soft computing techniques to real-world problems involving uncertainty, including
kansei-related issues, are presented.
Chapter 2 presents the utility and efficiency of the Particle Swarm Optimization (PSO) technique, a
recently developed soft computing tool, in solving real-world humanistic problems. This chapter pro-
poses novel PSO algorithms for clustering high-dimensional data and automatically determining the
number of clusters in data mining applications, supported by the results of simulation experiments with
synthetic data sets. This chapter is useful in handling optimization and search problems while dealing
with kansei databases and categorizing concepts related to kansei.
Chapter 3 introduces basic text mining techniques that help to analyze interviews and question-
naires, including keyword extraction, word graphs, clustering of text and association rule mining. In
addition, a case study using text mining for the analysis of interviews and questionnaires revealing the
opinions, concerns and needs of subjects is presented. This chapter is useful for learning the basics of
text mining and how to use it to extract information such as the concerns and needs of subjects from
questionnaire data.
Chapter 4 demonstrates an application of a human cognitive behavior model in developing an auto-
matic biometric authentication system. Using set estimation, the authors developed a novel threshold
selection technique for identifying individuals who are learned by the system, and differentiating humans
from no-humans. This chapter deals with an important subset of cognitive system design involving hu-
man factors and demonstrates the possibility of successful design of such systems through simulation
experiments in a real-world environment.
Chapter 5 introduces DNA computing, another emerging computational technique encompassing
computer science, biological science and engineering. DNA computing, a variant of biomolecular com-
putation, is now widely accepted as a new computing model for future computing devices. This chapter
describes the theoretical framework and formal models of DNA computation and presents several DNA
computing models used to solve real-world problems. DNA encoding may play a crucial role in designing
successful human-centric computational paradigms that might be useful in kansei engineering.
Chapter 6 summarizes the general flow of kansei retrieval systems and presents the structure of a
typical kansei database. Indexing technologies for kansei retrieval are then described and discussed. In
particular, in order to speed up the retrieval process, the author proposes an original adaptive R*-tree
method that is quite appropriated for the kansei database. This chapter not only helps readers to understand
the mechanisms and methodology of kansei retrieval, but also to realize the importance of considering
efficiency when constructing a kansei database and performing kansei retrieval.
xix

Human moods are influenced by the multiple forms of media that surround us. Detection of emo-
tions elicited by various media has emerged as an area of active research area in the past few decades.
Chapter 7 surveys advances in this area, including a general overview of research on affective analysis of
multimedia contents, recent research on detecting emotional semantics from images, videos and music,
three typical archetypal systems related to the three fields of images, videos and music, several critical
problems, and strategies for problem resolution. This chapter is helpful in understanding state-of-the-art
research on kansei analysis regarding multimedia and the critical hurdles to overcome in constructing
a kansei-based multimedia retrieval system.
Chapter 8 reviews the fundamental concepts of non-smooth dynamical systems, together with
examples showing the diversity of their nonlinear behaviors, and introduces a Takagi-Sugeno fuzzy
modeling concept, demonstrating how it could be extended to represent a non-smooth system and ap-
plied to a stability analysis to predict the onset of structural instability in the evolution of a dynamical
system. This interesting topic implies the potentiality of inferring the bifurcation of kansei evaluation in
processes such as product design, painting, or music video production, because they can be considered
non-smooth dynamic systems following fuzzy decision rules.
Humanoid and android robots are expected to be one of the greatest new industries of the 21st cen-
tury. We can therefore anticipate the presence of more and more robots not only in factories but also in
our daily lives. Chapter 9 discusses the future relationship between humans and humanoid and android
robots. How will robots change human society? Will the robots endanger us? How should we regulate
the development of humanoid robots? This chapter is valuable for researchers and engineers who aim
to explore the potential of robotic technology from the perspective of kansei.
Chapter 10 and Chapter 11 present the typical processing flow of kansei engineering by measuring
and analyzing subjects’ psychological states. After reviewing kansei engineering research on music,
which is classified into six categories (kansei evaluation methodology, music psychological research,
physiological measurement, music theoretical research, kansei music system and recommendation
system), Chapter 10 presents approaches for kansei research on melody and rhythm from the perspec-
tive of music theory. Methods for analyzing relations between modes, melodic ranges, or rhythms and
kansei evaluation based on principal component analysis are introduced, and how the arrangements of
modes affects human impressions and feelings is revealed and applied in a real-time melody recogni-
tion system. This chapter helps readers to understand the basis of kansei engineering and how various
approaches are applied.
To investigate the relationship between people’s perceptions of sidewalk environments and their
component elements, Chapter 11 adopts factor analysis and the rough sets approach to determine the
most important attributes to people’s perceptions, minimal attribute sets without redundancy, and a series
of decision rules that represent the relationships between perceptions and the physical components of
sidewalk environments. The analytical approach promotes better understanding of people’s perceptions
of sidewalk environments and then establishes a useful and constructive framework for discussion of
walking environment design and management.
As an example of the application of kansei engineering, the development of a kansei communication
robot, Ifbot, is presented in Chapter 12. Ifbot communicates with people by expressing emotions through
facial expressions. Here, the authors present their development approach, including the association of
facial expressions with human emotions using an auto-associative neural network, a soft computing tool.
In addition, methods for generating expressive faces that convey human emotion in order to enhance the
quality of human-robot communication are discussed. A method for creating personality through facial
xx

expressions has also been proposed. This study and its application represent an important development
in the cooperation between kansei engineering and soft computing.
Chapter 13 describes how several color sequences (temporal or spatial) affect human impressions,
and demonstrates how to develop a color sequence that reminds viewers of the natural world. A number
of experiments have investigated the different effects of spatial (or temporal) ordered color sequences
on the naturalness by calculating the projected route area and route complexity of a hexagonal diagram
comprising six color sequences. Then, a simple fuzzy model of the colors used to give an impression of
naturalness is achieved. The proposed techniques are also useful for investigating the effects of color
order on other impressions, which play an important role in product design.
Chapter 14 examines how a particular narrative can become the narrative of an entire community,
and thereby influence or control the behavior of all members of that community. This research extracts
its hypothesis from the narrative theory of scenery, which holds that narratives are distinct with different
work experiences and can be shared by people in the same community. These shared normative scenery
narratives subsequently influence community members’ behavior. Normative scenery narratives can
be viewed as products of the general kansei of farmers in a community. This chapter helps readers to
understand how general kansei is formed through sharing of group members’ experiences and feelings,
and how it controls the behaviors of the group.
An overview of the emerging approaches for incorporating kansei (human feelings) into system de-
sign is presented in Chapter 15. Three approaches involving the relationship between psychophysiology
and the design process have been studied. Tools and methods are developed with a psychological basis
with respect to human inspiration, behavior and mental images of the design process and requirements.
Each of the proposed approaches is supported by real-world examples and applications. This chapter is
highly educational in developing the field of kansei design.
Chapter 16 represents another application of a human cognitive action model in designing a person
authentication system. This chapter focuses on handwriting as a primary coordinated activity of human
movement and, with the help of soft computing tools, presents a framework for tacit handwriting skill
analysis for the extraction of embedded knowledge. A technique for detecting human identity by analyz-
ing handwriting captured by a computer writing pad is also presented and is supported by simulation
experiments with benchmark data.
Chapter 17 and 18 introduce methods for measuring physiological signals and utilizing these signals
to assess human mental states, such as stress, mood, or feelings. Chapter 17 explores salivary biomark-
ers and their potential to reveal the degree of stress accumulation, while Chapter 18 presents bioelectric
signals and their application in estimating human mood states in combination with PCA, HMM and NN
approaches . The contents of these two chapters demonstrate the possibility of measuring and analyzing
kansei using physiological signals.
Chapter 19 presents a 4-year research project (2008-2011) currently underway at Xiamen University,
China, to build China’s first artificial brain. The project takes an “evolutionary engineering” approach,
effectively evolving tens of thousands of neural net modules (or “agents” in the sense of Minsky’s
“Society of Mind”) and connecting them to make artificial brains. These modules are evolved rapidly
and are then connected according to the artificial brain designs of human “brain architects” (BAs). The
artificial brain will eventually contain thousands of pattern recognizer modules and hundreds of decision
modules that, when suitably combined, will be able to control the hundreds of behaviors of a robot. As
a general research report, this chapter is worth reading to understand the construction of a system that
is expected to develop functions rivaling the human brain in complexity.
xxi

We are confident that professionals, researchers and students in the fields of kansei information pro-
cessing and soft computing will be able to use this book to learn more about the ways in which kansei
research can be applied to different environments.

Ying Dai
Iwate Prefectural University, Japan

Basabi Chakraborty
Iwate Prefectural University, Japan

Minghui Shi
Xiamen University, China

EndnotE

1
Japan Society of Kansei Engineering, http://www.jske.org/
xxii

Acknowledgment

We wish to acknowledge all the people who have helped us directly or indirectly in completing the book.
First, we would like to thank all chapter contributors for their innovative proposals, serious preparation
and enthusiastic cooperation in the completion of all necessary chapters. We would also like to thank the
reviewers, who took the time to review their assigned chapters in earnest and provided many constructive
comments for improving the quality of the book. We are also grateful to Elizabeth Ardner and Christine
Bufton from IGI Global for their technical support during all the phases of the book development process.

Ying Dai
Iwate Prefectural University, Japan

Basabi Chakraborty
Iwate Prefectural University, Japan

Minghui Shi
Xiamen University, China
Section 1
Basic Concepts, Frameworks
and Techniques
1

Chapter 1
Soft Computing and
its Applications
Siddhartha Bhattacharyya
The University of Burdwan, India

Ujjwal Maulik
Jadavpur University, India

Sanghamitra Bandyopadhyay
Indian Statistical Institute, India

ABStRACt
Soft Computing is a relatively new computing paradigm bestowed with tools and techniques for handling
real world problems. The main components of this computing paradigm are neural networks, fuzzy logic
and evolutionary computation. Each and every component of the soft computing paradigm operates either
independently or in coalition with the other components for addressing problems related to modeling,
analysis and processing of data. An overview of the essentials and applications of the soft computing
paradigm is presented in this chapter with reference to the functionalities and operations of its con-
stituent components. Neural networks are made up of interconnected processing nodes/neurons, which
operate on numeric data. These networks posses the capabilities of adaptation and approximation. The
varied amount of uncertainty and ambiguity in real world data are handled in a linguistic framework
by means of fuzzy sets and fuzzy logic. Hence, this component is efficient in understanding vagueness
and imprecision in real world knowledge bases. Genetic algorithms, simulated annealing algorithm
and ant colony optimization algorithm are representative evolutionary computation techniques, which
are efficient in deducing an optimum solution to a problem, thanks to the inherent exhaustive search
methodologies adopted. Of late, rough sets have evolved to improve upon the performances of either
of these components by way of approximation techniques. These soft computing techniques have been
put to use in wide variety of problems ranging from scientific to industrial applications. Notable among
these applications include image processing, pattern recognition, Kansei information processing, data
mining, web intelligence etc.

DOI: 10.4018/978-1-61692-797-4.ch001

Copyright © 2011, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Soft Computing and its Applications

1. IntRoduCtIon discovery from databases, expert systems, induc-

tive reasoning and pattern recognition. The main
The field of Soft Computing is a synergistic inte- advantage of rough set theory in data analysis
gration of essentially three computing paradigms, is that it does not need any preliminary or addi-
viz. neural networks, fuzzy logic and evolutionary tional information about data – like probability
computation entailing probabilistic reasoning in statistics, or basic probability assignment in
(belief networks, genetic algorithms and chaotic Dempster-Shafer theory, grade of membership or
systems) to provide a framework for flexible the value of possibility in fuzzy set theory.
information processing applications designed
to operate in the real world. Bezdek [Bezdek92]
referred to this synergism as computational intelli- 2. nEuRAl nEtwoRkS
gence [Kumar2004]. Soft computing technologies
are robust by design, and operate by trading off A neural network is a powerful data-modeling
precision for tractability. Since they can handle tool that is able to capture and represent complex
uncertainty with ease, they conform better to real input/output relationships similar to a human brain.
world situations and provide lower cost solutions. Artificial neural networks resemble the human
The three components of soft computing differ brain in the following two ways:
from one another in more than one way. Neural
networks operate in a numeric framework, and are • A neural network acquires knowledge
well known for their learning and generalization through learning.
capabilities. Fuzzy systems [Zadeh65] operate in • A neural network’s knowledge is stored
a linguistic framework, and their strength lies in within inter-neuron connection strengths
their capability to handle linguistic information known as synaptic weights.
and perform approximate reasoning. The evolu-
tionary computation techniques provide powerful The true power and advantage of neural net-
search and optimization methodologies. All the works lie in their ability to represent both linear
three facets of soft computing differ from one and non-linear relationships and in their ability
another in their time scales of operation and in the to learn these relationships directly from the data
extent to which they embed a priori knowledge. being modeled.
Of late, rough set theory has come up as a
new mathematical approach to model imperfect Artificial neural network
knowledge, crucial to addressing problems in ar-
eas of artificial intelligence. Apart from the fuzzy An artificial neural network [Haykin99], as the
set theory pointed out in the previous paragraph, name suggests, is a parallel and layered intercon-
rough set theory proposed by Pawlak [Pawlak82] nected structure of a large number of artificial
presents still another attempt to handle real world neurons, each of which constitutes an elementary
uncertainties. The theory has attracted attention computational primitive. The distributed repre-
of many researchers and practitioners all over sentation of the interconnections through massive
the world, who have contributed essentially to parallelism achieved out of the inherent network
its development and applications. The rough set structure, bestows upon such networks properties
approach seems to be of fundamental importance of graceful degradation and fault tolerance. These
to artificial intelligence and cognitive sciences, network structures differ from one to another in
especially in the areas of machine learning, knowl- the topology of the underlying interconnections
edge acquisition, decision analysis, knowledge as well as on the target problem they are put to.

2
Soft Computing and its Applications

Since the essence of neural network operation is network. These functions recognize specific
based on the behavior of human brain, these net- range of input signals and selectively tune the
works require a form of training or learning ability. neurons to respond to the input signals according
Once these are trained with the different aspects some learning algorithm. Most of these activa-
of the problem at hand, they can be used to solve tion functions take an input as an infinite range
similar problems given the immense generaliza- of activations (-∞, +∞) and squashes/transforms
tion capabilities embedded therein. Depending on them in the finite range [0, 1] or {-1, 1} [Le-
the type of learning procedure adopted, different ondes98]. Thus, these functions are able to map
neural network architectures have evolved from the input information into bipolar excitations.
time to time [Haykin99, Kumar2004]. Though these functions may vary from neuron
In the most general form, an artificial neural to neuron within the network, yet most network
network is a layered structure of neurons. It com- architectures are field-homogeneous i.e. all the
prises seven essential components [Kumar2004], neurons within a layer are characterized by the
viz., (i) neurons, (ii) activation state vector, (iii) same signal function. Some of the common neural
activation function, (iv) connection topology, network signal functions [Kumar2004, Haykin99]
(v) activity aggregation rule, (vi) learning rule include (i) binary threshold, (ii) bipolar threshold,
and (vii) environment. These components are (iii) linear, (iv) liner threshold, (v) sigmoid, (vi)
discussed in the following sections. hyperbolic tangent, (vii) Gaussian, (viii) stochastic
[Kumar2004] etc.
Neurons
Connection Topology
Neurons are the processing units of a neural net-
work. There are basically three types of neurons This refers to the interconnection topology of
viz. input, hidden and output. The input neurons the neural network architectures. These connec-
are designated to accept stimuli from the external tions may be either excitatory (+) or inhibitory
world. The output neurons generate the network (-) or absent (0). These connections, or synapses
outputs. The hidden neurons, which are shielded basically house the memory of the network. The
from the external world, are entrusted with the behavior of neural network architecture is decided
computation of intermediate functions necessary by its connection topology.
for the operation of the network.
Activity Aggregation Rule
Activation State Vector
This rule aggregates the activities of the neurons
Neural network models operate in a real n- at a given layer. It is usually computed as the
dimensional vector space Rn. The activation state inner product of the input vector and the neuron
vector, X = (x1, x2, …, xn)T ∈ Rn, is a vector of the fan-in interconnection strength (weight) vector.
activation levels xi of the individual neurons of An activation rule thereby, determines the new
the network. This state vector acts as the driving activation level of a neuron based on its current
force for a neural network. activation and external inputs.

Activation Function Learning Rule

The characteristic activation functions are used The neural network learning rules define an
to supplement the learning process of a neural architecture-dependent procedure to encode

3
Soft Computing and its Applications

pattern information into inter-neuron intercon- update their interconnection weights dur-
nections. This is a data driven process executed ing the process of self-organization.
by modifying the connection weights. Two types
of learning are in vogue; viz. supervised learning Environment
and unsupervised learning.
The operational environment of neural networks
• Supervised learning: Supervised learning can be either deterministic (noiseless) or stochastic
encodes a behavioristic pattern into a neu- (noisy). A neural network N is a weighted directed
ral network by attempting to approximate graph, where the nodes are connected as either
the function that best describes the data set
employed. For an input vector, Xk ∈ Rn (a • a feedforward architecture, in which
real n-dimensional vector space) related the network has no loops. Examples in-
to an output vector Dk ∈ Rp (in a real p- clude the perceptron, multilayer percep-
dimensional vector space), a supervised tron [Duda73], support vector machines
learning algorithm aims at deriving an un- [Cortes95] and radial basis function
known mapping function f: Rn→ Rp. The networks [Broomhead88], Kohonen’s
algorithm tries to reduce the error (Dk-Sk) Self Organizing Feature Map (SOFM)
in the system response, where Sk is the ac- [Kohonen95] etc., or
tual response of the system by employing • a feedback (recurrent) architecture, in
the desired output response of the system which loops occur in the network because
Dk (also referred to as the teaching input) of feedback connections. Examples in-
and the associate of Xk. Thus, input-output clude the Hopfield network [Hopfield84],
sample pairs are used to train/teach the net- BSB model [Hui92], Boltzmann machines
work through a simple form of error cor- [Ackley85], bidirectional associative
rection learning or gradient descent weight memories [Kosko88], adaptive resonance
adaptation. Hence, the system generates an theory [Carpenter95] etc.
output Dk in response to an input Xk. The
learning process achieves an association The following sections discuss the basic phi-
between Dk and Xk, when a stimulus X’k losophy of the supervised learning environment
close to Xk elicits a response S’k sufficiently with reference to the simple artificial neuron and
close to Dk. the multilayer perceptron.
• Unsupervised learning: This paradigm
simply provides the system with an input Simple Artificial neuron
Xk, and allows it to self organize/self super-
vise its parameters to generate internal pro- The basic computational element of an artificial
totypes of the sample vectors. Such a para- neural network model is often referred to as a node
digm attempts to represent the entire data or unit. It receives inputs from some other units,
set by employing a smaller number of pro- or perhaps from an external source. The basic
totypical vectors. These prototypes are in a function of a single neuron is to add up its inputs,
state of continual updating as newer system and to produce an output if this sum is greater than
inputs enter into the system. This is often some value, known as the threshold value. The
driven by a complex competitive-cooper- basic neuron model, also known as “perceptron”
ative process where the individual neurons after Frank Rosenblatt [Haykin99, Rosenblatt58],
compete and cooperate with each other to is a binary classifier that maps real-valued vectored

4
Soft Computing and its Applications

Figure 1. A basic neuron

the simplest case, if f is the identity function the
unit’s output is just its net input. Thus, the node
acts as a linear unit.
A variant of this architecture compares the com-
puted sum to a certain value in the neuron called
the threshold value. The thresholding process is
accomplished by comparison of the computed
sum to a predefined threshold value. If the sum is
greater than the threshold value, then the network
outputs a 1. On the other hand, if the sum is lesser
than the threshold value, the network generates
inputs to a single binary output value. It generally an output of 0. An example threshold function is
possesses the following features: shown graphically in Figure 2.
Equivalently, the threshold value can be sub-
1. The output from a neuron is either on or off. tracted from the weighted sum, and the resulting
2. The output depends only on the inputs. A value then compared to zero. Depending on
certain number of neurons must be on (or whether the difference yields a positive or nega-
activated) at any one time in order to make tive result, the network outputs a 1 or 0. The
a neuron fire. The efficiency of the synapses output yi of the ith neuron can then be written as:
at coupling the incoming signal into a neuron
n
can be modeled by having a multiplicative yi = f (∑ wij y j − k ) (2)
factor on each of the inputs to the neuron. j =1

This multiplicative factor, often referred

to as the interconnection weight w, can be where, k the neuron’s bias or offset, and f is a
modified so as to model synaptic learning. step function (actually known as the Heaviside
A more efficient synapse, which transmits function) and
more of the signal, has a correspondingly
larger weight, whilst a weak synapse has a 1 for x > 0
smaller weight. Its output, in turn, can serve f ( x) =  (3)
0 for x ≤ 0
as input to other units. The basic neuron can 
be represented as shown in Figure 1.
Thus the threshold function produces only a
Expressed mathematically, if there are n inputs 1 or a 0 as the output, so that the neuron is either
with n associated weights on the input lines, then activated or not.
the ith unit computes some function f of the
weighted sum of its inputs, given by Learning Algorithm
n
yi = f (∑ wij y j ) (1) Since the single layer perceptron is a supervised
j =1 neural network architecture, it requires learning
of some a priori knowledge base for its operation.
The following algorithm illustrates the learning
where, wij refers to the weight from jthto the ith unit.
paradigm for a single layer perceptron. It com-
The function f is the unit’s activation function. In
prises the following steps.

5
Soft Computing and its Applications

Figure 2. A threshold function

• Initialization of the interconnection error for a specific choice of perceptron model

weights and thresholds randomly. parameters. Minimizing the loss function would
• Calculating the actual outputs by taking yield more accurate predicted outputs by the per-
the thresholded value of the weighted sum ceptron. This is solved by an iterative numerical
of the inputs. technique called gradient descent [Haykin99,
• Altering the weights to reinforce correct Snyman2005]. It comprises the following steps.
decisions and discourage incorrect deci-
sions, i.e. reducing the error. Choose some (random) initial values
for the network parameters
The weights are however, unchanged if the Repeat until G:=0
network makes the correct decision. Also the Calculate the gradient G of the
weights are not adjusted on input lines, which error function with respect to each
do not contribute to the incorrect response, since parameter
each weight is adjusted by the value of the input Change the parameters slowly in the
on that line xi, which would be zero (see Table 1). direction of the greatest rate of
In order to predict the expected outputs, a loss decrease of the error, i.e., in the
(also called objective or error) function E can be negative direction of G i.e., -G
defined over the model parameters to ascertain the End
error in the prediction process. A popular choice
for E is the sum-squared error given by Computation of the Gradient

E= ∑ (y i
− d i )2 (4) During the training process of neural networks
i by the gradient descent mechanism, the gradient
G of the loss function (E) with respect to each
In words, it is the sum of the squared difference weight wijof the network is computed [Haykin99,
between the target value di and the perceptron’s Snyman2005]. This gradient indicates as to how
prediction yi (calculated from the input value xi) the small changes in the network weights affect
computed over all points i in the data set. For a the overall error E. Let, the loss function be rep-
linear model, the sum-squared error is a quadratic resented for each p training sample, as
function of the model parameters. The loss func-
tion E provides an objective measure of predictive

6
Soft Computing and its Applications

Table 1.

Begin
Initialize interconnection weights and threshold
Set wi(t=0), (0<= i <= n), to small random values Interconnection weights from input i at time t
Set w0 := k k is the bias in the output node
Set x0:=1 x is the input to the network

Present the inputs and the desired output to the network

Present x0, x1, x2, x3, ..., xn x0, x1, x2, x3, ..., xn are the inputs to the network
Present d(t) d(t) is the desired output

Calculate the actual output

n
y (t ) := f (∑ wi (t ) xi (t ))
i =1

Adapt interconnection weights

wi (t + 1) := wi (t ) ± xi (t )
End

E= ∑E p
(5) The first factor of equation 8 can be obtained
p
by differentiating equation 6.

where, δE
= −(d o − yo ) (9)
δyo
1
Ep = ∑ (dop − yop )2
2 o
(6)
Using yo = ∑w oj
y j , the second factor be-
j

o refers to the range of the output units of the comes

network. Then,
δyo δ
δE δ ∂E p
δwoi
=
δwoi
∑w oj
y j = yi (11)
G=
δwoi
=
δwoi
∑ E = ∑ δw
p
(7) j
p p oi

Hence,
Generalizing for all training samples and
decomposing the gradient into two factors using δE
chain rule, one gets = −(d o − yo ) yi (12)
δwoi

δE δE δyo
= (8) The gradient G for the entire data set can be
δwoi δyo δwoi
obtained by summing at each weight the contribu-

7
Soft Computing and its Applications

tion given by equation 12 over all the data points. classes is much more complex, the single layer
Then, a small proportion μ (called the learning rate) model fails abruptly.
of G is subtracted from the weights to achieve the
required gradient descent. Multilayer Perceptron

The Gradient Descent Algorithm In order to overcome the shortcomings of the single
layer perceptron model, the first and foremost
The loss function minimization procedure during way out is to resort to a multilayer model with the
the training of a neural network involves the com- threshold function slightly smoothed out to provide
putation of the gradient of the loss function with some information about the nonlinearly separable
time. The algorithm shown in Table 2 illustrates inputs [Haykin99]. This means that the network
the steps in determining the gradient for attain- will be able to adjust the weights as and when
ing the minimum of the loss function [Haykin99, required. A possible multilayer neural network
Snyman2005]. model comprising of three layers of nodes, viz.,
The algorithm terminates once the minimum the input layer node, the hidden layer node and the
of the error function, i.e., G=0 is reached. At this output layer node along with their characteristic
point the algorithm is said to have converged. activation functions and inter-connection weights,
An important consideration is the learning rate is shown in Figure 3. In the figure, an extra node
(μ), which determines by how much the weights with a nonlinear activation function has been
are changed at each step. If μ is too small, the inserted between input and output. Since such a
algorithm will take a long time to converge. Con- node is “hidden” inside the network, it is com-
versely, if μ is too large, the algorithm diverges monly referred to as a hidden unit. The hidden unit
leading to imprecise learning. also has a weight from the bias unit. In general,
However, the single layer perceptrons suffer all non-input neural network units have such bias
from several limitations. They learn a solution weights. For simplicity however, the bias unit and
if there is a possibility of finding it. They can weights are usually omitted from neural network
separate linearly separable classes easily enough, diagrams. The sole output layer node shown in
but in situations where the division between the the figure is characterized by a linear activation
function. Since the input layer node acts only as a

Table 2.

Begin
Initialize wij to small random values wij are the interconnection weights
Repeat until done
For each weight wij set Dwij:=0
For each data point (x, t)p
Set input units to x
Compute value of output units
For each weight wij set Dwij:= Dwij+(di - yi)yj
For each weight wij set wij:=wij+m Dwij m is the learning rate
End

8
Soft Computing and its Applications

Figure 3. Schematic diagram of a multilayer perceptron

receptor of information from the external world, Learning Algorithm

it is not driven by such activation mechanism.
As stated before, neural networks are often Since the target values for the hidden layers of
driven by nonlinear activation functions. The multilayer networks are known, the gradient
typical standard nonlinear hyperbolic tangent descent algorithm used for the training of linear
characteristic activation function is shown in networks (discussed earlier) cannot be applied to
Figure 4. Such a network model exhibits nonlin- train multilayer neural networks. This inherent
ear behavior since theoretical results indicate that problem led to the sudden downfall of the neural
given enough hidden units. networking paradigm after the 1950s until the
A multilayer neural network architecture like error backpropagation algorithm [Haykin99,
the one shown in Figure 5 can approximate any Rumelhart86, Chauvin95], or in short, backprop
nonlinear function to any required degree of ac- came to the rescue.
curacy. Hence, a multilayer neural network In principle, backprop provides a way to train
model is able to classify nonlinearly separable networks with any number of hidden units arranged
datasets. However, too many hidden layers can in any number of layers. The basic requirement
degrade the network’s performance [Kumar2004]. of the backprop algorithm is to ensure that the
network connection pattern must not contain any

Figure 4. Hyperbolic tangent function

9
Soft Computing and its Applications

Figure 5. Schematic diagram of a multilayer perceptron with multiple hidden nodes

cycles. Networks that respect this constraint are deh65] explains the varied nature of ambiguity
called feedforward networks and their connection and uncertainty that exist in the real world. This
pattern forms a directed acyclic graph or dag. is in sheer contradiction to the concept of crisp
For the purpose of training a multilayer feedfor- sets, where information is more often expressed
ward neural network by gradient descent, a training in quantifying propositions.
dataset consisting of pairs (x, d) is considered. The underlying concept behind the notion of
Here, vector x represents an input pattern to the fuzzy sets [Zadeh65] is that each and every observa-
network and vector d is the corresponding target tion, exists with a varied degree of containment in
output. The corresponding learning algorithm the universe of discourse. This degree of contain-
employing backprop is illustrated as shown in ment is referred to as the membership value of the
Table 3. observation. A fuzzy set is a mapping from an input
Neural networks are powerful and robust tools universe of discourse into the interval [0, 1] that
for the retrieval of incomplete data, finding pat- describes the membership of the input variable.
terns in datasets, and to mimic the human behavior This is referred to as fuzzification. The reverse
when it comes to the analysis and interpretation mechanism to revert to the crisp world is termed
of data. As such, they find wide use in processing, as defuzzification. Thus, it can be inferred that
retrieval and recognition of data patterns.
Whereas, crisp sets quantify quantities, fuzzy sets
qualify qualities.
3. Fuzzy SEtS And Fuzzy logIC
Fuzzy logic is a collection of conditional
It may be pointed out that much of the informa- statements or fuzzy rules, which form the basis
tion available in the real world exhibit vagueness, of the linguistic reasoning framework, which
imprecision and uncertainty. In fact, the fuzzy embodies representation of shallow knowledge.
sets approach fits in with the linguistic modes The fundamental atomic terms in this linguistic
of reasoning that are natural to human beings. or natural language-reasoning framework are
The fuzzy set theory, introduced by Zadeh [Za- often modified with adjectives or “linguistic

10
Soft Computing and its Applications

Table 3.

Begin
dE dE
Define d j := the error signal for unit j; Dwij := − the (negative) gradient for weight wij.
d net j d wij
Let Ai={j:$ wij} represents the set of preceding nodes to unit i and Pj={i: $ wij} represents the set of succeeding nodes
to unit j of the network.

Computation of the gradient

d E d netij
Dwij := − the first factor is the error of unit i.
d netij d wij
d netij d
The second factor,
d wij
:=
d wij
∑w
k∈ Ai
ik yk := y ' j

Dwij := d i y ' j
So,

Forward activation
Remark: The activity of the input units is determined by the network’s external input x. For all other units, the activity
is propagated forward as

yi := fi ( ∑ wij y j )
j∈ Ai

Remark: The activity of all the preceding nodes Ai to unit i must be known before calculating the activity of i.

Calculation of output error

1
E := ∑ (d o − yo ) 2
2 o ; So, d :=d -y the error for output unit o.
o o o

Error backpropagation
Remark: The output error is propagated back for deriving the errors of the hidden units in terms of the succeeding
nodes.
d E d neti d y j
d j := − ∑ the first factor is the error of node i.
i∈Pj d neti d y j d net j
d neti d d yj d f j (net j )
The second factor,
d yj
:=
d yj
∑w
k∈ Ak
ik yk := wij ; The third factor,
d net j
:=
d net j
:= f '(net j )

If, the hidden units are characterized by the tanh activation function, then
∑
f '(neth ) := 1 − yh2 ; So, d j := f '( net j ) d i wij
i∈Pj

End

hedges”. These linguistic hedges have the effect which is similar to natural language expressions,
of modifying the membership function for a basic can be written as
atom. The general form of a fuzzy rule [Zadeh65],

11
Soft Computing and its Applications

IF premise (antecedent) THEN conclusion (con- that represents distinct α-cuts of a given fuzzy set
sequent) A, is called a level set of A, i.e.,

It is generally referred to as the IF-THEN rule Λ A = {α | µ A ( x) = α, x ∈ U } (15)

based form. It typically expresses an inference
such that if a fact (premise, hypothesis or ante-
The support SA ∈ [0, 1], of such a fuzzy set A
cedent) is known, then another fact (conclusion
is defined as
or consequent) can be derived.
n µ A ( xi )
Fuzzy Set theoretic Concepts S A = {∑ : xi ∈ X ∀µ A ( xi ) > 0}
i =1 xi
As already stated, all the elements in the uni- (16)
verse of discourse X exhibit varying degrees of
membership. This membership or containment The core (CA) of a fuzzy set A represents all
of an element in a fuzzy set A is decided by a those constituent elements whose membership
characteristic membership function, μA(x) ∈ values are equal to unity, i.e.,
[0,1]. The closer is the membership value of an
element to unity, the stronger is the containment C A = { xi ∈ U | µ A ( xi ) = 1} (17)
of the element within the fuzzy set. Similarly, a
lower membership value implies a weaker con-
The bandwidth (BWA) of a fuzzy set A is ex-
tainment of the element within the set. A fuzzy
pressed as
set A, characterized by a membership function
μA(xi) and comprising elements xi,i = 1,2,3, ..., n,
BWA = { xi ∈ U | µ A ( xi ) ≥ 0.5} (18)
is mathematically expressed as

µ A ( xi ) Figure 6 provides a graphical representation

A= ∑ xi
; i = 1, 2, 3, ..., n (13) of the core, bandwidth, α-level and support of a
i
fuzzy set.
The maximum of all the membership values in
a fuzzy set A is referred to as the height (hgtA) of
where, å represents a collection of elements. the fuzzy set. If hgtA is equal to 1, then the fuzzy
i
set is referred to as a normal fuzzy set. If hgtA is
The resolution of a fuzzy set A is determined
less than 1, then it is referred to as a subnormal
by the α-cut (or α-level set) of the fuzzy set. It
fuzzy set. A normal fuzzy set is a superset of several
is a crisp set Aα containing all the elements of
nonempty subnormal fuzzy subsets.
the universal set U, that have a membership in A
A subnormal fuzzy subset (As) can be con-
greater than or equal to α, i.e.
verted to its normalized equivalent by means of
the normalization operator given by [Bhattacha-
Aα = {xi ∈ U | µ A ( xi ) ≥ α}, α ∈ [0, 1] (14) ryya2008]

If Aα = {x ε U| μA(x)>α}, then Aα is referred As ( x)

to as strong α-cut. The set of all levels α ∈ [0, 1] Norm A ( x) = (19)
s
hgt A
s

12
Soft Computing and its Applications

Figure 6. Representation of fuzzy set concepts

Fuzzy Cardinality

The scalar cardinality of a fuzzy set A is the sum-

mation of the membership grades of all elements
of x in A. It is given by [Zadeh65, Ross95]

| A |= ∑µ A
( x) (25)
x∈U

where,U is the universe of discourse. When a

fuzzy set A has a finite support, its cardinality
can be defined as a fuzzy set. This fuzzy cardinal-
ity is denoted by |Af| and is defined by Zadeh as
[Zadeh65, Ross95]
The corresponding denormalization operation
is given by [Bhattacharyya2008] α
| Af |= ∑|A |
(26)
α ∈Λ A α

DeNorm A ( x) = hgt A ( x ) × Norm A ( x ) (20)

s s s

where, α is the cut-off value, Aα is the α-level set of

In general, for a subnormal fuzzy subset with the fuzzy set and ΛA is the corresponding level set.
support, SAs ∈ [L, U], the normalization and the
denormalization operators are expressed as [Bhat- Fuzzy operators
tacharyya2008]
Several operators are used to form the “linguistic
As ( x) − L hedges” in fuzzy logic. These are (i) Concentra-
Norm A ( x ) = ; DeNorm A ( x ) = L + (U − L) × Norm A ( x )
s
U −L s s
tion (ii) Dilation and (iii) Intensification operators
(21) [Ross95, Bhattacharyya2006].

Fuzzy Set theoretic operations • Concentration: This operator tends to con-

centrate the elements of a fuzzy set by re-
The fuzzy union, intersection and complement ducing the degree of membership of those
operations [Zadeh65] on two fuzzy sets A, B for elements that are “partly” in the set. It is
an element x in the universe of discourse X, are expressed as
defined as
µ A ( x) = µ A ( x)2 for 0 ≤ µ A ( x) ≤ 1 (27)
Union : µ A∪ B ( x) = max[µ A ( x), µ B ( x)] (22)

• Dilation: This operator dilates or stretches

Intersection : µ A∩ B ( x) = min[µ A ( x), µ B ( x)] a fuzzy set by increasing the membership
(23) of elements that are “partly” in the set. It is
expressed as
Complement : µ A ( x) = 1 − µ A ( x) (24)

13
Soft Computing and its Applications

Figure 7. Fuzzy operators

Due to the ability of handling uncertainties,

µ A ( x) = d × µ A ( x) for 0 ≤ µ A ( x) ≤ 1
fuzzy sets and fuzzy logic play a crucial role in
(28) decision-making processes. Fuzzy sets and fuzzy
logic find application in modeling imprecise da-
where, d is the amount of dilation tasets, designing intelligently controlled systems,
quantifying the varied amount of ambiguities in
• Intensification: This operator acts as a the domains of signal processing and computer
combination of the concentration and dila- vision, etc. They are also used to quantify the
tion operators. It is expressed as inherent vagueness/ambiguities often encountered
in real life situations.
 2µ A ( x)2 for 0 ≤ µ A ( x) ≤ 0.5

INT ( A) = 
1 − 2[1 − µ A ( x)]2 for 0.5 < µ A ( x) ≤ 1

(29) 4. HEuRIStIC SEARCH
tECHnIquES
Thus, intensification increases the contrast
between the elements, which have more than This computing paradigm employs several search
half-membership and those elements, which have and optimization algorithms based on Darwinian
less than half-membership. Figure 7 illustrates laws of biological evolution. Several evolutionary
the operations of concentration, dilation and in- algorithms are in vogue. These include (a) genetic
tensification for fuzzy linguistic hedges [Short, programming (GP), which evolve programs, (b)
Medium, Long] on a typical fuzzy set A. evolutionary programming (EP), which focuses
The resultant hedges are represented by [Some- on optimizing continuous functions without
what Short, Indeed Medium, Very Long]. It is recombination, (c) evolutionary strategies (ES),
evident from the figure that the fuzzy operators which focuses on optimizing continuous functions
tend to smooth out the hedges represented by the with recombination, and (d) genetic algorithms
membership functions [Short, Medium, Long]. (GAs), which focuses on optimizing general com-
The steeper slopes of the hedges [Short, Medium, binatorial problems. Most of these evolutionary
Long] are smoothed out to gradual variations. algorithms are characterized by a population of
This change is attributed to the manipulation of trial solutions and a collection of operators to act
the membership values of these hedges. on the population. The basic philosophy behind

14
Soft Computing and its Applications

these algorithms is to search the population space • Stochastic optimization, which is an um-
by the application of the embedded operators so as brella set of methods that includes simu-
to arrive at an optimal solution space. Generally, lated annealing and numerous other
two types of operators are used, viz. reproduc- approaches.
tion and evolution. The reproduction operator is
guided by a selection mechanism. The evolution The following sections discuss the operational
operator includes the crossover and mutation procedures of three popular heuristic search tech-
operators. The search technique is implemented niques, viz., genetic algorithms, simulated anneal-
through a series of iterations, whereby the differ- ing and ant colony optimization.
ent operators are applied in a loop on the initial
population. Each iteration is referred to as a gen- genetic Algorithms
eration. Each generation produces a new solution
space of parent individuals, which are selectively Genetic algorithms (GAs) [Goldberg89, Davis91,
chosen for participating in the next generation of Michal92, Bandyopadhyay2007a] are efficient,
the optimization procedure. The selection of the adaptive and robust multi-point search and op-
participating parents for the next generation is timization techniques guided by the principles
decided by a figure of merit, often referred to as of evolution and natural genetics. They provide
the objective function. This objective function is parallel near optimal and solutions of an objec-
entrusted with the evaluation of the fitness of the tive or fitness function in complex, large and
candidate solutions in a particular generation to multimodal landscapes. GAs are modeled on
qualify for the next generation of operations. Other the principles of natural genetic systems, where
notable and related search techniques include: the genetic information of each individual or
potential solution is encoded in structures called
• Quantum annealing [Apolloni89, chromosomes. They use some domain or problem
Das2005], which uses “quantum fluctua- dependent knowledge for directing the search in
tions” instead of thermal fluctuations to get more promising areas of the solution space; this
through high but thin barriers in the target is known as the fitness function. Each individual
function. or chromosome has an associated fitness function,
• Tabu search [Glover97], which normally which indicates its degree of goodness with respect
moves to neighboring states of lower en- to the solution it represents. Various biologically
ergy, but takes uphill moves when it finds inspired operators like selection, crossover and
itself stuck in a local minimum; and avoids mutation are applied on the chromosomes to yield
cycles by keeping a “taboo list” of solu- potentially better solutions.
tions already seen.
• Ant colony optimization (ACO) Basic Principles and Features
[Colorni91, Dorigo92], which uses many
ants (or agents) to traverse the solution A GA essentially comprises a set of individual
space and find locally productive areas. solutions or chromosomes (called the population)
• Harmony search, which mimics musicians and some biologically inspired operators that
in improvisation process where each mu- create a new (and potentially better) population
sician plays a note for finding a best har- from an old one. The different steps of a GA can
mony all together. be represented as follows.

15
Soft Computing and its Applications

Initialize the population Genetic Operators

Do
Encode the strings and compute The frequently used genetic operators are the se-
their fitness values lection, crossover and mutation operators. These
Reproduce/select strings to create are applied to a population of chromosomes to
new mating pool yield potentially new offspring. The operators are
Generate new population by cross- described in the following sections.
over and mutation
Loop while (not_termination) Selection
The selection/reproduction process copies indi-
The components of GAs are described in the vidual strings (called parent chromosomes) into a
following sections. tentative new population (known as mating pool)
for genetic operations. The number of copies that
Encoding Strategy and Population an individual receives for the next generation
is usually taken to be directly proportional to
To solve an optimization problem, GAs start with its fitness value thereby mimicking the natural
the chromosomal representation of a parameter selection procedure to some extent. This scheme
set which is to be encoded as a finite size string is commonly called the proportional selection
over an alphabet of finite length. For example, scheme. Roulette wheel parent selection, stochas-
the string tic universal selection, and binary tournament
selection [Goldberg89, Michal92] are some of
1 0 0 1 1 0 1 0 the most frequently used selection procedures.
In the commonly used elitist model of GAs, the
is a binary chromosome (string of 0’s and 1’s) best chromosome seen up to the last generation
of length 8. Each chromosome actually refers to is retained either in the population, or in a loca-
a coded possible solution. A set of such chromo- tion outside it.
somes in a generation is called a population, the
size of which may be constant or may vary from Crossover
one generation to another. The chromosomes in the The main purpose of crossover is to exchange
initial population are either generated randomly information between randomly selected parent
or using domain specific information. chromosomes by recombining parts of their
genetic information. It combines parts of two
Evaluation Technique parent chromosomes to produce offspring for
the next generation. Single point crossover is
The fitness function is chosen depending on the one of the most commonly used schemes. Here,
problem to be solved, in such a way that the strings first of all, the members of the selected strings in
(possible solutions) representing good points in the mating pool are paired at random. Then each
the search space have high fitness values. This is pair of chromosomes is subjected to crossover
the only information (also known as the payoff with a probability μc where an integer position
information) that GAs use while searching for k (known as the crossover point) is selected uni-
possible solutions. formly at random between 1 and l-1 (l>1 is the
string length). Two new strings are created by
swapping all characters from position (k+1) to l.

16
Soft Computing and its Applications

For example, let the two parents and the crossover the population size, probabilities of performing
points be as shown below. crossover (usually kept in the range 0.6 to 0.8) and
mutation (usually kept below 0.1) and the termina-
1 0 0 1 1 | 0 1 0 tion criteria. Moreover, one must decide whether
0 0 1 0 1 | 1 0 0 to use the generational replacement strategy where
the entire population is replaced by the new popula-
After crossover the offspring will be the fol- tion, or the steady state replacement policy where
lowing: only the less fit individuals are replaced. Most of
such parameters in GAs are problem dependent,
1 0 0 1 1 1 0 0 and no guidelines for their choice exist in the lit-
0 0 1 0 1 0 1 0 erature. Therefore, several researchers have also
kept some of the GA parameters variable and/or
Some other common crossover techniques are adaptive [Baker85, Srinivas94].
two-point crossover, multiple point crossover, The cycle of selection, crossover and muta-
shuffle-exchange crossover and uniform crossover tion is repeated a number of times till one of the
[Davis91]. following occurs:

Mutation 1. average fitness of a population becomes more

The main objective of mutation is to introduce or less constant over a specified number of
genetic diversity into the population. It may so generations,
happen that the optimal solution resides in a por- 2. desired objective function value is attained
tion of the search space, which is not represented by at least one string in the population,
in the population’s genetic structure. Hence, the 3. number of generations is greater than some
algorithm will therefore be unable to attain the predefined threshold.
global optima. In such a scenario, only mutation
can possibly direct the population to the optimal Simulated Annealing
section of the search space by randomly altering
the information in a chromosome. Mutating a Simulated annealing (SA) [Kirkpatrick83,
binary gene involves simple negation of the bit, Cerny85] is a probabilistic metaheuristic search
while that for real coded genes are defined in a technique useful for finding a good approximation
variety of ways [Eshelman93, Michal92]. to the global minimum of a given function in a
For example in binary bit-by-bit mutation every large search space. It is generally efficient with a
bit in a chromosome is subject to mutation with a discrete search space. In situations, which demand
probability μm. The result of applying the bit-by-bit an acceptably good solution in a fixed amount of
mutation on positions 3 and 7 of a chromosome time, SA has been found to be more effective than
is shown below. other exhaustive search techniques.
The term “simulated annealing” is derived
1 0 0 1 1 0 1 0 from the common annealing process in metallurgy
1 0 1 1 1 0 0 0 where subsequent heating and controlled cooling
of a material are performed to increase the size
Parameters of a Genetic Algorithm of its crystals. Heating excites the material atoms
and causes them to wander randomly in higher
There are several parameters in GAs that have energy states. Subsequent slow cooling settles
to be tuned by the user. Some among these are

17
Soft Computing and its Applications

them to configurations with lower internal energy which implies that the system may migrate to the
than the initial one. new state even when it has a higher energy than
Similarly, in the SA algorithm, each step re- the current one. This prevents the process from
places the current solution by a random neighbor- becoming stuck in a local minimum. When T goes
ing solution. The probability of choosing such a to zero, the acceptance probability tends to zero if
neighboring solution depends on the difference g’ > g. P however, attains a positive value if g’ < g.
between the corresponding energy function Thus, the system favors transitions that go
values and a global parameter T referred to as to lower energy values, and avoid those that go
the temperature. This temperature is gradually higher for sufficiently small values of T. When
decreased during the cooling process. Thus, the T becomes 0, the procedure will ensure making
current solution changes almost randomly when the migration only if it goes to lower energy.
T is large, but the rate of change of states goes Thus, it is clear that the evolution of the state
down as T is reduced. depends crucially on the temperature T. Roughly
speaking, the evolution of a state ρ is sensitive to
Overview of SA coarser changes of energy variations when T is
large, and to finer changes of energy variations
The basic objective of SA is to minimize the system when T is small.
internal energy function F(ρ), where ρ corresponds
to each point of the search space. Thus, it aims to The Annealing Schedule
bring the system, from an arbitrary initial state, to Another essential feature of the SA method is that
a state with the minimum possible energy. the temperature (T) should be gradually reduced as
For this purpose, SA considers some user the simulation proceeds [Kirkpatrick83, Cerny85].
specified neighbors ρ’ of the current state ρ, and Initially, T is set to a high value (i.e. ∞), and then
probabilistically decides between migrating the it is decreased at each step according to some
system to either state ρ’ or to retain in state ρ. annealing schedule. The user generally specifies
The probabilities are chosen such that the system this schedule for the decrement of T. However, it
ultimately migrates to lower energy states. This must be ensured that the schedule should be such
step is repeated until the system reaches a state that it would end up with T = 0 towards the end
that is a good approximation to the required one, of the annealing process.
or until a prespecified limit to the approximation Thus, the system is expected to migrate initially
has been reached. The following sections highlight towards a broader region of the search space con-
the important aspects of simulated annealing. taining good solutions ignoring smaller features of
the energy function in the process. Subsequently,
Acceptance Probabilities it would drift towards the lower energy regions
This is the probability of migrating from the that become narrower and narrower. Finally, the
current state ρ to a candidate new state ρ’. This system migrates downhill according to the steepest
is decided by an acceptance probability function descent heuristic. However, the pure version of SA
P(g, g’, T), where g = F(ρ) and g’ = F(ρ’) and T does not keep track of the best solution obtained
is the temperature (mentioned earlier). The ac- in terms of the lower energy levels attained at
ceptance probability is usually chosen such that any point of time.
the probability of allowing a transition decreases
when the difference (g’− g) increases. This means SA Pseudocode
that smaller uphill migrations are more likely than The pseudocode shown in Table 4 implements
the larger ones. P must be nonzero when g’ > g, the simulated annealing heuristic, as described

18
Soft Computing and its Applications

above, starting from state ρ0 and continuing to a implementation of the algorithm. In other words,
maximum of kmax steps or until a state with energy the diameter of the search graph must be small to
gmax or less is found. facilitate faster transitions between states. Hence,
the choice of the search graph diameter is an es-
Selection of Operating sential criterion for successful operation of the
Parameters of SA simulated annealing algorithm.

Several parameters need to be specified for the Transition Probabilities

application of SA. These include the state space, The migration from the current state ρ to the state
the energy function F, the candidate generator ρ’ is governed by another probability, viz., the
procedure, the acceptance probability function transition probability. The transition probability
P and the annealing schedule. The performance depends on the current temperature (Tc), on the
of SA depends on the suitability of the choice of order of candidate transitions and on the accep-
these parameters. The following sections throw tance probability function P.
some light on the selection of these parameters.
Efficient Candidate Generation
Search Graph Diameter It is evident that the current state is expected to have
Considering all the possible states of a simulated much lower energy than a random state after a few
annealing process to be the vertices of a graph with iterations of the SA algorithm. This observation is
the edges representing the candidate transitions, important in the selection of the candidate genera-
simulated annealing may be modeled as a search tor function. In practice, the generator function is
graph which aims to provide a sufficiently short sensitive towards those candidate migrates where
path from the initial state to any intermediate state the energy of the destination state ρ’ is likely to
or the global optimum state. This implies that the be similar to that of the current state. This implies
search space must be small enough for an efficient

Table 4.

Initialize state, energy and “best” solutions

r:=r0; g:=F(r); rbest:=r; gbest:=g initial and best states, initial and best energy
k:=0 starting point
Do
Neighborhood selection and energy computation

rnew:=neighbor(r) new neighboring state

gnew:=F(rnew) new energy
if gnew < gbest then
rbest:=rnew; gbest:=gnew best state and energy
if P(g, gnew, Tc(k/kmax)) > j then Tc is current temperature and j is a random number
r:=rnew; g:=gnew updated states and energy
k :=k + 1 next step
Loop while k < kmax and g > gmax

19
Soft Computing and its Applications

that those candidate states ρ’ for which P[F(ρ), gbest, respectively and the annealing schedule is
F(ρ’), T] is large should be opted for first. restarted. The decision to restart could be based
on a fixed number of steps, or based on the current
Avoidance of Getting Stuck to Local Minima energy being too high from the best energy so far.
Another aspect of the selection of the candidate
generator function is to reduce the number of local Ant Colony optimization
minima, which may come up during the anneal-
ing process. Otherwise, the SA algorithm may be The ant colony optimization algorithm (ACO) is
trapped in these minima with a high probability another probabilistic computational search tech-
for a very long time. The probability of occurrence nique useful for finding the best possible paths
of such traps is proportional to the number of in search graphs. It is a member of the family of
states achieved in the local minimal state of SA. ant colony algorithms, which are referred to as
The time of trapping is exponential on the energy swarm intelligence methods. Initially proposed
difference between the local minimal state and its by Marco Dorigo in 1992 in his PhD thesis [Co-
surrounding states. These requirements, however, lorni91, Dorigo92], the first algorithm was aimed
can be met by resorting to slighter changes to the at searching for an optimal path in a graph based
candidate generator function. on the behaviors exhibited by ants while searching
for food out of their colony.
Cooling Schedule
The simulated annealing algorithm assumes that Overview of ACO
the cooling rate is always low enough such that
the probability distribution of the current state In their quest for food, ants generally start search-
remains near the thermodynamic equilibrium at ing randomly through all possible paths that lead
all times. But the time required for attaining the to food. Once food is found, they return to their
equilibrium state (referred to as the relaxation time) colony leaving behind pheromone trails. If the
after a change in temperature strongly depends on following ants make the path, they do not hover
the nature of the energy function, the current tem- randomly, instead follow the trail to find food.
perature (Tc) as well as on the candidate generator. This pheromone trail however, starts to evapo-
Hence, there is no basis for selecting an ideal rate with time, thereby reducing its attractive
cooling rate for the algorithm. It should be esti- strength. The more time it takes for an ant to travel
mated and adjusted empirically for a particular down the path and back again, the more time the
problem. pheromones have to evaporate. A short path, by
However, this problem has been taken care of comparison, gets marched over faster, and thus
by the thermodynamic simulated annealing algo- the pheromone density remains high as it is laid
rithm, which adjusts the temperature at each step on the path as fast as it can evaporate.
based on the energy difference between the two The phenomenon of pheromone evaporation
states according to the laws of thermodynamics prevents the convergence of the algorithm to a
instead of applying any cooling schedule. local optimum. In absence of any pheromone
evaporation the following ants would always be
Restarting of SA attracted to the paths traversed by the first/leading
Sometimes it is better to revert back to a solution ant thereby leading to a constrained solution space.
that was significantly better rather than always This procedure adopted by real world ants is
migrate from the current state. This is called adapted in implementing the ant colony optimiza-
restarting. To do this ρ and g are set to ρbest and tion algorithm, which always leads to the short-

20
Discovering Diverse Content Through
Random Scribd Documents
No sugar in your tay—
Wor-rkin’ on th’ U. Pay. Ra-a-ailway!”
And—“Down! Down!” “Whang! Whang! Whangity-whang!”
The track-laying and the grading gangs were red-shirted, blue-
shirted, gray-shirted; with trousers tucked into heavy boots—and
many of the trousers were the army blue. For though the men were
mainly Irish, they were Americans and two-thirds had fought in the
Union armies during the Civil War. Some also had fought in the
Confederate armies.
There were ex-sergeants, ex-corporals, and ex-privates by the
scores, working shoulder to shoulder. In fact, the whole U. P. corps
was like an army corps. Chief Engineer Dodge had been a major-
general in the East and on the Plains; Chief Contractor “Jack”
Casement had been brigadier general; about all the way-up men had
been generals, colonels, majors, what-not; while the workers under
them were ready at a moment to drop picks and shovels and sledges
and transits, and grabbing guns “fall in” as regular soldiers.
This meant a great deal, when the Indians were fighting the road.
This past winter the engineers doing advance survey work had been
told by Chief Red Cloud of the Sioux that they must get out and stay
out of the country—but there they were there again. Nobody could
bluff those surveyors: fellows like “Major” Marshall Hurd who had
served as a private of engineers through the war, and Tom Bates,
and young Percy Browne, and their parties.
All the survey parties—some of them 500 miles in the lead—moved
and worked, carrying guns; the graders’ camps were little forts; the
track-builders marched to their jobs, and stacked their rifles while
they plied their tools. At night the guns were arranged in racks in the
boarding-cars, to be handy. The construction-trains’ cabooses were
padded with sand between double walls, and loop-holed, and even
the passenger trains were supplied with rifles and revolvers, in cars
and cabs. General Dodge called his private car, in which he shuttled
up and down the line, his “traveling arsenal.”
This was the arrangement, from the end o’ track back to beginning,
360 miles, and on ahead to the last survey camp. The Central Pacific
was not having such trouble.
“An’ lucky for it, too,” as said Paddy Miles. “For betwixt the yaller an’
the red, sure I’d bet on the red. Wan Injun could lick all the Chinymen
on this side the Paycific. But there’s niver an Injun who can lick an
Irishman, b’ gosh!”
However, today everything seemed peaceful. Usually a detachment
of soldiers, or a company of the Pawnee Indian scouts under Major
Frank M. North, their white-scout commander, were camped near by,
guarding the track-laying. But the soldiers were elsewhere, on a
short cross-country trip, and the Pawnees (Company A) were up at
Fort Sedgwick, near old Julesburg, fifteen or twenty miles west.
The air was very clear. The graders working on the roadbed five
miles away might be seen. The long trains of huge wagons, hauling
supplies, wended slowly out to refit them. On this section there were
100 teams and 2,000 men, scattered along; on the next section there
were another thousand men, doing the first grading according to the
stakes set by the engineers. And eastward there were the trains and
the stations, all manned, and other gangs fixing the rough places in
the track.
Of all this Terry felt himself to be rather a small part—just riding old
Jenny back and forth, with the little rail-truck, while his father imitated
with the engine of the boarding-train. Of course, his father had a bad
knee (which the war had made worse), and driving an engine was
important; but he himself envied his chum, George Stanton. George
was out with his father on railroad survey under Mr. Tom Bates—
probably fighting Injuns and shooting buffalo and bear, too. That also
was man’s work, while riding an old yellow mule over the track was
boy’s work.
Every truck-load of forty rails carried the track forward about 560
feet. To that steady “Down! Down!” and “Whangity-whang!” end o’
track reached out farther and farther from the piles of iron thrown off
by the construction-train, and from the boarding-train that waited for
the construction-train to back in with another supply.
So while cleaning up the piles, Terry and little Jimmie Muldoon had
to travel farther and farther with their loads. Then in due time the
construction-train would come puffing up, the boarding-train, with
Terry’s father leaning from the cab, would move on as close to end o’
track as it dared, the construction-train would follow and with a great
noise dump its cargo of jangling iron, and retreat again; the
boarding-train would back out, to clear the track for the trucks; and
Terry and Jimmie would start in on short hauls, for a spell.
The supply of iron at the last dump was almost exhausted. The
construction-train was hurrying in, with more. Engine Driver Ralph
Richards and his stoker, Bill Sweeny, were climbing lazily into the
cab of old No. 119, ready to pull on up as soon as Jimmie Muldoon’s
truck left with the final load. Terry had his eye upon the track, to see
it emptied——
“Drill, my paddies, drill!
Drill, you tarriers, drill!
O, it’s work all day,
No sug——”
Hark! A sudden spatter of shots sounded—a series of shouts and
whoops—the whistle of the boarding-train was wide open—up the
grade the graders were diving to cover like frightened prairie-dogs—
and out from the sandhills not a quarter of a mile to the right there
boiled a bevy of wild horsemen, charging full tilt to join with another
bevy who tore down diagonally past the graders themselves.
Sioux? Or Cheyennes? The war had begun, for 1867!
CHAPTER II
A LITTLE INTERRUPTION

The Indians had chosen exactly the right time, for them. They had
awaited the moment when the main body of track-layers were
farthest separated from the boarding-train and the stacks of arms;
they had seen that there were no soldiers on guard; and here they
came, with a rush, at least 500 of them.
“Fall in, men! Lay down! Down wid yez!”
Terry tumbled off his yellow mule in a jiffy. Dropping spade and
sledge, ducking and lunging, the men were scurrying along the
roadbed, seeking shelter. Only the squad of tampers and ballasters
following end o’ track to settle the ties were near the first gun stacks;
Terry joined their flat line. The Springfield carbines were passed
rapidly, but there were not enough.
“Stiddy, boys!” bawled Pat. He had been a top sergeant in the
regular army before the war. “Hug the ground. The word from
headquarters is ‘Niver retrate.’ Sure, if we haven’t guns we can foight
wid picks. Wait for orders, now.”
Down dashed the Indians, at reckless speed: one party straight from
the north, one party obliquing from the west. The engines of both
trains were shrieking furiously. All up the grade the wagons were
bunching, at a gallop, with military precision; the laborers were
rushing in squads to corral in them and in the low dug-outs beside
the roadbed.
The party of Indians from the westward split; one half veered in, and
racing back and forth there, pelted the road embankment with a
storm of bullets and arrows. The graders replied, but it was hard to
land on those weaving, scudding figures.
The other half of the party tore on, heading to unite with the second
party and cut off the boarding-train. That was it! The Indians wanted
the boarding-train and supplies.
Hurrah! The boarding-train was coming on, regardless. It was
manned by only Engine Driver Richards and Fireman Sweeny, a
brakeman and the cooks; but no matter. Like a great demon it was
coming on, whistling long shrieks and belching pitchy smoke.
The Sioux (some Cheyennes, too) were close upon it. They began to
race it, whooping and shooting. The windows of dining-car and
caboose replied with jets of white, as the cook and the brakeman
bravely defended. Stoker Bill shot from his side of the cab. The train
gathered way slowly; the ponies easily kept up with it—their riders,
swerving in, flung themselves free of the saddles, clung to the steps
and ladders and vaulted the couplings; clung like ants and dragged
and writhed, as if they could hold it back!
They charged the engine; even cast their ropes at the smoke-stack;
swarmed to the tender and from there shot into the cab. Terry’s heart
welled into his mouth, with fear for his father. Suddenly there was a
great gush of white steam—Engineer Richards had opened the
cylinder cocks, and the cloud of scalding vapor surged back,
sweeping the tender. Out popped sprawly brown figures, to land
head over heels upon the right-of-way, and blindly scramble for
safety.
Hurrah! Bully for Engineer Richards! And the construction-train was
coming too. No! Look at it! From Terry’s neighbors a groan of disgust
issued.
“The dirty cowards! Bad cess to ’em! Turnin’ tail, they are!”
For the construction-train was standing still, on the track, and the
engine was making off, back across the wide plains, leaving a trail of
smoke and a good-by shriek.
“Niver mind! We’ve a train of our own. Yis, an’ hearts to match it.”
“’Tis all right, boys. He’s only goin’ to the tiligraph,” Pat shouted. “It’s
word to the troops at Sidgwick he has up his sleeve. The Pawnees’ll
be wid us in a jiffy—an’ then watch them red rascals skedaddle. ’Asy,
’asy,” continued Pat, “till the train’s widin reach of us. Stiddy. We
mustn’t get scattered, like.”
The boarding-train was jolting and swaying on the newly laid rails;
but what ailed it, besides? Aha!
“Settin’ the brakes! Settin’ the brakes, they are!”
And sure enough. These Cheyennes and Sioux were wise. For a
year and a half they had been watching the white man’s iron horses
and big thunder wagons advancing onward into the buffalo country;
and they had learned a number of new wrinkles. They were no
longer afraid of the strange “medicine.” For here they were, boldly
tackling the cars, laying hands upon them, climbing aboard—and
setting the brakes!
Their almost naked figures, outlined against the sky, atop the cars,
tugged and hauled at the brake wheels. The brake-shoes ground
harshly; fumy smoke floated from underneath, as the locked car-
wheels slid on the rails; the engine, with throttle open, roared vainly.
Out from the cab darted Fireman Bill Sweeny, mounted the tender
and, skipping to the first car, revolver in hand, hung to the ladder
while he raked the tops beyond.
“Sharp-shooters give it to ’em!” Pat yelped. The carbines of the
track-layer gang banged hopefully.
The Indians ducked and swung off to the farther side. The brakeman
was out of the caboose. He lay flat upon one end of the train, the
fireman lay flat upon the other end; and hitching along they began to
kick the brakes free. The galloping Indians peppered at them, but
failed to hit them.
“Be ready, lads,” Pat ordered. “Skirmishes wid the guns, first. The rist
of us wid the picks. We’ll run for it, and meet the train. Jist a minute,
now.” And—“Oh, the divils!” he added. “Charge!”
A squad of the Indians, dismounted, had thrown a tie across the
track. A wild volley from the carbines had not stayed them. Engineer
Richards, plunged in his own steam cloud, evidently did not see the
tie; he came on, pushing Jimmie Muldoon’s loaded truck before him;
the white horse tried to bolt and fell with a broken neck just as the
rope parted; the smoke-stack was atilt, and spitting smoke and
steam from a dozen bullet-holes; but twitched by the roaring engine,
the train moved faster and faster.
Up sprang the men, with a yell. The line of skirmishers, carbines
poised, charged—charged in splendid order, like soldiers, aiming,
firing, running. With picks and sledges and even spades the other
men also charged, behind the skirmishers; bending low and
shouting, yes, laughing in their excitement.
“The tie! Look out for the tie on the track!” they hallooed.
Terry had nothing to carry, and he was fast on his feet. Never had he
sprinted so, before. The first thing he knew, he was through the
skirmishers and legging on by himself, while the bullets hummed by
him and every instant the distance between tie and truck was
lessening. All his eyes and thoughts were on that tie. If the engine—
his father’s engine—rammed it with the rail-loaded truck—wow!
He lost his hat—he heard whoops and shouts and excited Shep’s
wild barking—the Indians on his side were swerving off, before the
carbine bullets—but the engine was thundering down upon him, he
saw his father’s astonished grimy face peering from the cab and he
glimpsed the cars behind spewing naked figures. Then he dived for
the tie. He barely had time to lift one end when the truck struck the
tie, hurled it to the left and him to the right; but they both fell clear, for
as he picked himself up the box-cars were rumbling by, jerking to the
sharply braked engine.
All was hurly-burly with the Indians scooting and screeching, the
men scrambling and cheering, catching at the steps and braces,
running alongside until the train stopped, and clutching the guns
passed out from doors and windows.
The dining-car door slid back; the sweaty faces of the cook and
cookee grinned down; the brakeman leaped off——
“Fall in, now! Fall in wid yez!” were Pat’s orders. “Take your
distances ben’ath—two men to each pair o’ wheels. An’ them that
hasn’t guns lay flat inside.”
Terry had no notion of lying flat inside. He plunged like a rabbit under
the dining-car (bewildered Shep at his heels), for a place between
the rails; found none, and dodged on, trying not to step on anybody
or be in the way. He arrived at the tender, and had to come out.
“Get in here! Quick!” It was his father, sighting him. Terry hoisted
himself into the engine, while several bullets rang upon the metal
grasped by his hands. He lurched to the fireman’s seat and huddled
there, to gain breath and grin. With a running leap Shep followed, to
curl close in a corner, safe, he believed, from all that racket.
“Well, where were you going?” his father demanded.
“Just looking for a good place,” Terry panted.
“You’ve found it, and you’d better stick. ’Tisn’t healthy, outside. What
were you doing on the track ahead of me? Didn’t I hit something?”
“A tie, dad. They’d laid a tie across the track.”
“Oho! Good for you. But you took a big chance. Did you reach it?”
“I got one end up.”
“If I’d hit it plumb, reckon some of those rails would have been driven
into the boiler. I couldn’t see plain, on account the steam and the
truck. The crooked stack bothered me, too. Anyhow, here’s one train
they don’t capture.”
“They can’t take it, can they, dad?”
“Not on your life, Terry. Not while there’s a cartridge for a gun or an
Irishman to swing a pick, or an ounce of steam in the boiler of old
119. If worst comes to worst we can run back and forth, ’twixt here
and that construction-train.”
Terry jumped down and crawled to peek out between engine and
tender.
“No, we can’t, dad. They’re piling ties on the track ’way behind!”
“I declare! They’re too smart. They even set the brakes on me, and
tried to rope the engine stack, like they would a horse’s neck! So
they think they have us corralled, do they?”
That was so. The pesky Indian had daringly charged to the farthest
pile of ties—a spare pile—tied ropes, and at a gallop dragging the
ties to safer distance were erecting a barricade upon the track.
Evidently they meant business, this time. It was to be a fight to a
finish. All up the graded roadbed the U. P. men were fighting off the
red bandits—fighting from the dug-outs and the embankment and
the wagon corrals; they had no chance to sally to the boarding-train.
And here at the boarding-train Paddy Miles’ track-layers were
fighting.
Part of the Indians dashed around and around in a great circle,
whooping gleefully and shooting at long distance. “Blamed if they
haven’t got better guns than we have,” remarked Terry’s father, as
now and then a bullet pinged viciously against the boiler-iron of
engine or tank. Others, dismounted, crept steadily forward, like
snakes, firing from little hollows and clumps of brush.
The Paddy Miles sharp-shooters, snug beneath the cars, and
protected by the rails and the car-wheels, stanchly replied. The
heavy Springfield balls kicked up long spurts of sand and ’dobe dust;
once in a while a pony rider darted in, for closer shot—sometimes he
got away with it, and sometimes his horse lunged headlong, to lie
floundering while the rider himself ran hunched, for shelter. Then the
men cheered and volleyed at him; maybe bowled him over, but not
always.
Terry’s father had lighted his pipe; and there he sat, on his seat, with
his gun poked out of the window, to get a shot when he might. He
was as cool as a cucumber, and ready for any kind of business. This
was not his first scrape, by any means. He had been a gold-seeker
in the rush of Fifty-nine, to the Pike’s Peak diggin’s of Colorado; and
he had served in the Union Army of the Civil War. Only his crippled
knee had put him into the cab—but brave men were needed here,
the same as elsewhere, these days.
“Where did the other engine go, dad?” Terry asked.
“To the nearest wire. There’s a spur station and operator ten miles
back, you know. Sedgwick has the word, by now; and so has North
Platte. Pretty soon we’ll see the Pawnees coming from the one
direction and the general himself from the other; and that’ll put an
end to this fracas.”
Terry exclaimed.
“They’re shooting fire arrows!”
Cleverly worming along, several of the Indians had posted
themselves near enough to use their bows. They launched arrow
after arrow, with bunches of flaming dried grass and greasy rags—
yes, as like as not old waste—tied to the heads; and these plumped
into car top and car side.
“The confounded rascals!” growled Engineer Richards.
Fireman Bill Sweeny hurdled from the first car down to the tender.
He was sweat-streaked and grim, and bleeding at the shoulder. He
grabbed a bucket, soused it into the tank, and away he staggered.
“Train’s afire, Ralph,” he yelled back. “Don’t shove out——” and he
was gone.
Forward bustled other men, with buckets; dipped into the tank and
sped for the rear again. Matters were getting serious. The
Springfields seemed unable to ferret out the bow-wielders. There
was a cheer, and Pat Miles led a charge. Out from beneath the cars
there rushed a line of skirmishers, while behind them the carbines
barked, supporting them. Up from their coverts sprang the fire-arrow
Indians, and bolted. Giving them a volley the skirmish dropped and
dug in.
A line was thrown out on the other side of the train, also. This made
the Indians furious; their horsemen raced madly up and down,
showing only an arm and a leg, or suddenly firing from the saddle
and hanging low again. At the best they were difficult marks. They
had plenty of ammunition, and rifles that outranged the stubby
carbines.
“Fire’s squelched except the last car; that’s a-burnin’,” gasped Stoker
Bill, lurching in and sinking breathless upon his seat. “Don’t back up.
Say, kid, help me tie this shoulder, will you?”
“Hurt bad, Bill?” Engineer Richards queried, keenly.
“Nope. Just perforated a trifle.”
“Anybody else hurt?”
“None particular. But I sure thought this kid was a goner, though. Did
you see him?”
“Where?”
“When he reached for that tie?”
“Didn’t see him or the tie either, till too late. I knew I hit something.”
“Well, I happened to be squinting up this way, and I saw him just as
he heaved an end clear of the track. Next thing, you sent him one
way and the tie the other. He’s an all-right boy.”
“Guess he is,” laughed Terry’s father. “He’ll get promoted off that old
yellow mule, first thing we know.”
“Wish General Dodge would let me go out on a survey,” Terry
blurted. “Like George Stanton.”
“I’ll speak to the general about it,” said Fireman Bill, with a wink at
his cab partner.
But Engineer Richards did not notice. He was peering behind, out of
his window.
“Hi! Here comes the other engine,” he uttered. “Yes, and the
headquarters car for a trailer! The old man (that was Major-General
Dodge, of course) is inside it, I’ll bet a hat!”
They all looked. Far down the track an engine, twitching a single car,
was approaching. By her trail of dense wood smoke and the way she
bounced on the little curves and bumps, she was making good time,
too.
“Chief boss is on the job, sure,” quoth Bill.
“Usually is,” added Terry’s father. “Always has been. Nothing
happens from one end of line to t’other, but he’s there.”
The fighting track-layers had seen, and began to cheer afresh. Away
galloped a portion of the enemy, to pester the reinforcements. But
the engine came right on, until it halted at the end of the
construction-train. Out from the headquarters car issued man after
man—springing to the ground, guns in their hands, until they
numbered some twenty.
The first was a straight, well-knit figure in broad-brimmed black
slouch hat and ordinary civilian clothes. There appeared to be two or
three men in regular city clothes with him; the rest were dressed
more rough and ready, like trainmen and workmen.
The Indians were circling and yelling and shooting, at long distance.
The slouch hat led forward at a run. From the construction-train the
handful of train crew leaped out; they had been housed, waiting, on
defense, but helpless to do much. All ran forward. The slouch hat
man pointed and gave orders; the train crew jumped at the pile of
ties, while the other men rapidly deployed, in accurate line—
advancing as if in uniform, and yielding not an inch.
The ties were scattered in a twinkling; the engine pushed—the train
moved slowly up track, with the slouch hat’s men clearing either side
of the track, at a trot, fire, and trot again. The train crew closed the
rear. The engine whistled triumphantly; Terry’s father yanked the
whistle cord of No. 119, and by blast after blast welcomed the new-
comers.
In spite of the frantic Indians the trains joined. But the fighting was
not over. It had only been extended into a longer line. Terry could
stay quiet no more. He simply had to be out into the midst of things.
With General Dodge, the chief engineer and noted army man, on
deck, there would be a change of program.
“I’m going, dad,” he announced. Not waiting for answer, out he
tumbled, so quickly that Shep did not know it. For Shep was sound
asleep.
CHAPTER III
“TRACK’S CLEAR”

The few carbine barrels jutting here and there from behind the car-
wheels were silent, as hugging the side of the train Terry boldly
stepped over them; the skirmish lines were doing the shooting. Half
way down the train a knot of men were holding a council.
They were Chief Engineer Dodge (the figure in the black slouch hat)
and three men in city clothes, and Pat Miles. But before Terry might
steal nearer, fresh cheers arose.
“The Pawnees! Here they come! Hooray for the Pawnees!”
The men underneath the cars began to squirm out, and stand, to
yell. Over a swale up the graded right-of-way there appeared a
mounted force—looked like soldiers—cavalry—one company, two
companies, deploying in broad front; and how they did come!
The graders yonder were waving hats, and cheering; the Cheyennes
and Sioux hemming them in dug their heels into their ponies and
bending low fled before the charge. The General Dodge council had
moved out a few paces, to watch. The general swung his hat, also.
“Now for it!” he shouted. “Form your men, Pat. Blair, you wanted to
see some fighting. Take one company and advance to the left.
Simpson, you take another detachment and advance to the right.
White, you and I and Pat will guard the train with the train crews and
the reserve. We’ll put those rascals between two fires.”
“Fall in! Fall in wid yez!” Pat bawled, running. The words were
repeated. “Yez’ve thray gin’rals, a major an’ meself to lead yez,”
bawled Pat.
“Come on, men,” cried the general named Blair, to his detachment;
he climbed through between the cars; his men followed him and
away they went, in extended order, picking up the skirmishers as
they proceeded.
In the other direction ran General Simpson’s detachment, and out
across the plain. But the Indians did not stand. With answering yells
they scattered, and occasionally firing backward at the Pawnees
they scoured away—the Pawnees, separating into their two
companies, pursuing madly.
And a funny sight it was, too; for as the Pawnees rode, they kept
throwing off their uniforms, until pretty soon they were riding in only
their trousers.
“B’ gorry!” Pat panted, as he and the general halted near Terry. “The
only thing I have ag’in them Pawnees is, that when they come
there’s nothin’ left for the Irish.” He turned on the general, and
saluted—coming to a carry arms, with his left arm stiffly across his
red-shirted chest. “Track’s clear, gin’ral.”
“So it seems,” laughed General Dodge. “Simpson and Blair might as
well come in. Now let’s see what the damage is.” His sharp eyes fell
on Terry, standing fascinated. “What’s this boy doing out here? He
ought to be under cover.”
“Sure, he’s bigger’n he looks,” apologized Pat. “If ye could have
seen him lift at a tie when the engine was jest onto it——! He earned
a brevet—but I thought he was under the wheels entoirely.”
“That’s the kind of work that counts—but I’ll have to hear about it
later,” answered the general. “Now let’s check up the damage, and
get the men out again. Where’s General Casement?”
“He’s on up at Julesburg, sorr; him and Mr. Reed, too. But I’m
thinkin’ they’ll both be back in a jiffy.”
General “Jack” Casement was the chief contractor—the head boss
of the whole construction. Mr. S. B. Reed was the general
superintendent of building. Yes, they doubtless would arrive on the
jump.
The two companies of the construction gang were brought in, for the
Pawnees had chased the Sioux and Cheyennes out of sight. Before
they came in, themselves, General Dodge and Foreman Pat had
made their inspection. Three men badly wounded, here; several
slightly wounded; one car burned, other cars, and the engine, riddled
and scarred.
But within half an hour all the unhurt men had stacked their guns,
had resumed their tools, and were out on the grade, ready to start in,
just as though there had been no fight.
Jenny the yellow mule had a bullet hole through her ear; Jimmie
Muldoon’s white horse was dead; but speedily he and Terry were
mounted again, waiting for the construction-train to finish unloading,
and for the boarding-train to back out and clear.
That was the system of the U. P., building across the plains into the
Far West.
“Hey, Jimmie! Where were you?” hailed Terry.
“I got behind the cook’s stove,” piped little Jimmie, blushing as red as
his hair. “But I came out and handed ca’tridges. Weren’t you afraid?”
“I dunno. I guess I was too excited.”
“You done well, anyhow,” praised Jimmie, with disregard of grammar.
General Dodge went on up the grade, inspecting. The three men in
city clothes, with him, were General J. H. Simpson, of the United
States Engineer Corps; General Frank P. Blair, who had been one of
the youngest major-generals in the Civil War; and Congressman H.
M. White, who was called “Major” and “Doctor.” They formed a board
of inspectors, or commissioners, sent out by the Government to
examine every twenty or forty miles of the road, when finished, and
accept it.
The United States was lending money for the building of the first
railroad across continent, and naturally wished to see that the money
was being well spent.
The commissioners traveled in a special coach, called the “Lincoln”
coach because it had been made for President Abraham Lincoln,
during the War. The railroad had bought it, for the use of officials.
Now it was back at North Platte, the terminus. When the
commissioners heard of the fight, they had volunteered to come
along with General Dodge and help out.
“Drill, my paddies, drill!
Drill, you tarriers, drill!
O, it’s work all day,
No sugar in your tay,
Wor-rkin’ on th’ U. Pay. Ra-a-ailway!”
The construction-train had dumped its iron, the boarding-train had
backed out, and Jimmie and Terry again plied back and forth, with
the rails.
The Pawnees returned, in high feather like a lot of boys themselves.
They certainly were fighters. Major Frank M. North, a white man, was
their commander. He had lived among them, and spoke their
language, and they’d follow him to the death. He had enlisted four
companies—drilled them as regular cavalry, according to army
regulations; they were sworn into the United States Army as scouts,
and were deadly enemies to the Sioux and the Cheyennes. The
Sioux and Cheyennes feared them so, that it was said a company of
North’s Pawnees was worth more than a regiment of regular
soldiers. When these Pawnees sighted an enemy, they simply threw
off their clothes and waded right in.
The two companies, A and B, made camp on the plains, a little
distance off, near the Platte River. Major North and Chief
Petalesharo—who was the war-chief and son of old Petalesharo,
known as “bravest of the braves”—cantered forward to the track. The
major wore buckskin and long hair, like a frontiersman. Petalesharo
wore army pants with the seat cut out, and the legs sewed tight,
same as leggins.
“Take any hair, major?” was the call.
“Yes; there are three or four fresh scalps in the camp yonder. But
most of the beggars got away too fast.”
“Say, Pete! Heap fight, what?”
Petalesharo smiled and grunted, with wave of hand.
“He says the Sioux ponies have long legs,” called Major North.
“Where’s the general? He was here, wasn’t he?”
“Yes; he’s up ahead, with the graders.”
The major—young and daring and very popular—rode on with Chief
“Pete,” as if to report to General Dodge.
They all came back together, after a time—and the newly laid track
was advancing to meet them. Already the boarding-train had moved
up a notch. The Pawnees from the camp were scattered along,
watching the progress. The way with which the white man’s road
grew, before their eyes, seemed to be a constant marvel to them.
“Faith, we’ll build our two miles this day in spite o’ the Injuns,”
cheered the sweaty Pat, everywhere at once and urging on the
toiling men.
The three commissioners were as interested as the Pawnees; they
hung around, while Chief Engineer Dodge, General Jack Casement
and Supervising Engineer Reed (who had arrived horseback)
conferred in the headquarters car.
General Simpson and Dr. White had seen the track-laying gang at
work last year, but this was young General Blair’s first trip out. Now
while he was here, three-quarters of a mile of track was laid before
the call for supper sounded; and as the men rushed to meet the
train, Engineer Richards unhooked and gave the three
commissioners a ride on the cow-catcher to the very end o’ track, to
show them how well the rails had been put down.
In honor of the commissioners, after supper there was a parade of
the Pawnees, under Major North and the white captains Lute North
and Mr. Morse, Lieutenants Beecher and Matthews, and Chief
Petalesharo.
A great parade it was, too—“Might call it a dress p’rade, and ag’in ye
might call it an undress p’rade,” as Foreman Pat remarked. The
Pawnees were in all kinds of costume: some wore cavalry blouses
and left their legs naked; some wore cavalry trousers with the seats
cut out, and left their bodies naked; some wore large black campaign
hats of Civil War time, with brass bugles and crossed muskets and
crossed cannon, on the front; some wore nothing but breech clouts,
and brass spurs on their naked heels; but they kept excellent line
and wheeled and trotted at word of command.
They broke up with a wild yell, and away they went, careening over
the plain, whooping and prancing and shooting, and taking scalps—
chasing the “Sioux.”
“The gin’ral wants to see you,” ordered Pat, of Terry. “Ye’ll find him in
his car yon. Now stand on your feet an’ take off your hat an’ do the
polite, an’ mebbe it’s promoted you’ll be.”
So Terry, with Shep close following, trudged down the line of box-
cars, to the Chief Engineer’s “traveling arsenal.” He was curious to
see the inside of it. This was the general’s home, in which he toured
up and down the line, from Omaha to the end o’ track, caring not a
whit for the Indians.
It was fitted up inside with bunks and a desk and racked guns, and a
forward compartment which was dining-room and kitchen, ruled by a
darky cook. When the general was not traveling in his car, he was
out overseeing the surveys far beyond the railroad; he had explored
through the plains and mountains to Salt Lake long before the
railroad had started at Omaha.
The whole party were in the car; the three commissioners (General
Simpson was a famous explorer, too), and General Casement, and
Superintendent Reed, sitting with General Dodge. Terry removed his
dusty hat, and stood in the doorway. Shep stuck his black nose past
his legs, to gaze and sniff.
“Hello, my boy,” General Dodge greeted.
“Pat Miles said you wanted to see me, sir.”
“That’s right. Come in, dog and all. Gentlemen, this is Terry
Richards. They tell me he risked his life to save the boarding-train
from being wrecked during the Indian attack. I move that we all
shake hands with him.”
Terry, considerably flustrated, had his hand shaken, all ’round.
“Well, what’s your job, Terry?” asked General Dodge. He was a
handsome man, every inch a soldier, but with a very kind eye above
a dark, trimmed beard. Nobody could feel afraid of General Dodge.
“I help bring up the rails, on a truck. I ride a yellow mule, sir.”
“You’re rather a big boy to be doing that.”
“Yes, sir; but that’s my job. Somebody has to do it. The men have got
to have rails.”
“Very necessary, in building a railroad,” laughed General Blair.
“We did almost two miles today,” informed Terry. “We’d have done
two miles sure if the Injuns hadn’t tried to stop us.”
“That’s the right spirit,” approved General Simpson.
“General Casement is responsible for it,” quickly spoke Chief
Engineer Dodge. “His men are trained to the minute, either to work
or to fight. But the Union Pacific Company doesn’t overlook
individual acts of bravery. What would you like to do instead of riding
that yellow mule, Terry?”
“I’d like to be out in front, exploring with the engineers, sir.”
“Oh, you would!” General Dodge’s eyes kindled. Evidently he liked
that kind of work, himself. “Why? It’s the most dangerous job of all—
away out in the Indian country, with only a handful of men and
maybe no help except your own guns.”
“I think I’d like it, though,” stammered Terry. “If I could be any use,
George Stanton’s out there somewhere.”
“Who’s George Stanton?”
“He’s another boy. He’s my pardner. We were station hands on the
Overland [that was the stage line] before we joined the railroad.”
“Where is George?”
“I don’t know, exactly. He went out with his father in Mr. Bates’
survey party, as a sort of a cub to learn engineering. I guess he cuts
stakes.”
“Oh, I see. The Bates party are bound from Utah, to run a line this
way. But they’ll not be back before winter. Probably none of the
survey parties will turn up before winter. I’m afraid it’s too late for a
job with the engineers in the field, this year. Maybe you’ll have to
stick to your old mule, and haul rails for General Casement.”
“Well, if there’s nothing better I can do,” agreed Terry. “It’s fun to help
the track go forward, anyhow. We’ll beat the Central folks.”
“Yes, siree!” General Casement declared. He was a great little man,
this General “Jack” Casement: a wiry, nervy, snappy little man, not
much more than five feet tall, peaceful weight about 135 pounds and
fighting weight about a ton—“an’ sure there’s sand enough in him to
ballast the tracks clear to Californy,” Pat asserted. He had a brown
beard and a bold blue eye and a voice like a whip-crack. His brother
“Dan” Casement was smaller still, outside, but just as big inside.
They two were commanders of the grading and track-laying outfits.
“There’s one more party to go out yet,” General Dodge suddenly
said; “and that’s mine. If General Casement will lend you to me,
maybe I’ll have a place for you. We’ll see if we can’t find the Bates
party, and George Stanton.” And he added, with a smile, to the other
men: “A fellow can always use a boy, around camp, you know,
gentlemen.”
“Golly! I’d sure like to go, sir,” Terry blurted.
“Were you ever farther west?”
“Yes, sir. I helped drive stage, when I was working for the Overland.
And George and I had a pass to Salt Lake, but George broke his leg
up on the divide, in the mountains, so we quit and came back.”
“How did you happen to get a pass?”
“Just for something we did. We brought a stage through, when the
driver was near frozen. ’Twasn’t much, though. But we were glad to
get a pass. We’d never been west over the line.”
“How far east have you been over this line?” asked the general,
keenly.
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.