Instant Ebook Access, One Click Away – Begin at ebooknice.

com

(Ebook) Data Science and Innovations for

Intelligent Systems: Computational Excellence and
Society 5.0 (Demystifying Technologies for
Computational Excellence) by Kavita Taneja
(editor), Harmunish Taneja (editor), Kuldeep Kumar
(editor), Arvind Selwal (editor), Eng Lieh Ouh
https://ebooknice.com/product/data-science-and-innovations-
(editor) ISBN 9780367676278, 0367676273
for-intelligent-systems-computational-excellence-and-
society-5-0-demystifying-technologies-for-computational-
excellence-33975008

OR CLICK BUTTON

DOWLOAD EBOOK

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

 Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...

Start reading on any device today!

(Ebook) Vagabond, Vol. 29 (29) by Inoue, Takehiko ISBN 9781421531489, 1421531488

https://ebooknice.com/product/vagabond-vol-29-29-37511002

ebooknice.com

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason; Viles, James ISBN
9781459699816, 9781743365571, 9781925268492, 1459699815, 1743365578, 1925268497

https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

ebooknice.com

(Ebook) Boeing B-29 Superfortress ISBN 9780764302725, 0764302728

https://ebooknice.com/product/boeing-b-29-superfortress-1573658

ebooknice.com

(Ebook) 29, Single and Nigerian by Naijasinglegirl ISBN 9781310004216, 1310004218

https://ebooknice.com/product/29-single-and-nigerian-53599780

ebooknice.com
(Ebook) Jahrbuch für Geschichte: Band 29 ISBN 9783112622223, 3112622227

https://ebooknice.com/product/jahrbuch-fur-geschichte-band-29-50958290

ebooknice.com

(Ebook) Harrow County 29 by Cullen Bunn, Tyler Crook

https://ebooknice.com/product/harrow-county-29-53599548

ebooknice.com

(Ebook) Organometallic Chemistry, Volume 29 by M. Green ISBN 0854043284

https://ebooknice.com/product/organometallic-chemistry-volume-29-2440106

ebooknice.com

(Ebook) Analysis of Legal Argumentation Documents: A Computational Argumentation

Approach (Translational Systems Sciences, 29) by Hayato Hirata, Katsumi Nitta ISBN
9789811929274, 9811929270

https://ebooknice.com/product/analysis-of-legal-argumentation-documents-a-
computational-argumentation-approach-translational-systems-sciences-29-44646302

ebooknice.com

(Ebook) Communication Yearbook 29 by Pamela J. Kalbfleisch (ed.) ISBN 9780805855814,

0805855815

https://ebooknice.com/product/communication-yearbook-29-5293350

ebooknice.com
 DATA SCIENCE AND
INNOVATIONS FOR
INTELLIGENT SYSTEMS
Demystifying Technologies for Computational
Excellence: Moving Towards Society 5.0
Series Editors: Vikram Bali and Vishal Bhatnagar

This series encompasses research work in the field of Data Science, Edge Computing, Deep
Learning, Distributed Ledger Technology, Extended Reality, Quantum Computing,
Artificial Intelligence, and various other related areas, such as natural-language processing
and technologies, high-level computer vision, cognitive robotics, automated reasoning,
multivalent systems, symbolic learning theories and practice, knowledge representation and
the semantic web, intelligent tutoring systems, AI and education.
The prime reason for developing and growing out this new book series is to focus on the
latest technological advancements-their impact on the society, the challenges faced in im-
plementation, and the drawbacks or reverse impact on the society due to technological
innovations. With the technological advancements, every individual has personalized access
to all the services, all devices connected with each other communicating amongst them-
selves, thanks to the technology for making our life simpler and easier. These aspects will
help us to overcome the drawbacks of the existing systems and help in building new systems
with latest technologies that will help the society in various ways proving Society 5.0 as one
of the biggest revolutions in this era.

Data Science and Innovations for Intelligent Systems

Computational Excellence and Society 5.0
Edited by Kavita Taneja, Harmunish Taneja, Kuldeep Kumar, Arvind Selwal, and
Ouh Lieh
Artificial Intelligence, Machine Learning, and Data Science Technologies
Future Impact and Well-Being for Society 5.0
Edited by Neeraj Mohan, Ruchi Singla, Priyanka Kaushal, and Seifedine Kadry
Transforming Higher Education Through Digitalization
Insights, Tools, and Techniques
Edited by S. L. Gupta, Nawal Kishor, Niraj Mishra, Sonali Mathur, and Utkarsh Gupta
A Step Towards Society 5.0
Research, Innovations, and Developments in Cloud-Based Computing Technologies
Edited by Shahnawaz Khan, Thirunavukkarasu K., Ayman AlDmour, and Salam
Salameh Shreem
Computing Technologies and Applications
Paving Path Towards Society 5.0
Edited by Latesh Malik, Sandhya Arora, Urmila Shrawankar, Maya Ingle, Indu Bhagat

For more information on this series, please visit: https://www.routledge.com/Demystifying-

Technologies-for-Computational-Excellence-Moving-Towards-Society-5.0/book-series/
CRCDTCEMTS
 DATA SCIENCE AND
INNOVATIONS FOR
INTELLIGENT SYSTEMS
Computational Excellence and
Society 5.0

Edited by
Kavita Taneja, Harmunish Taneja
Kuldeep Kumar, Arvind Selwal, and
Eng Lieh Ouh
MATLAB® is a trademark of The MathWorks, Inc. and is used with permission. The
MathWorks does not warrant the accuracy of the text or exercises in this book. This book’s
use or discussion of MATLAB® software or related products does not constitute
endorsement or sponsorship by The MathWorks of a particular pedagogical approach or
particular use of the MATLAB® software.
First edition published 2022
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
and by CRC Press
2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN
© 2022 selection and editorial matter, Kavita Taneja, Harmunish Taneja, Kuldeep Kumar,
Arvind Selwal and Eng Ouh Lieh; individual chapters, the contributors
CRC Press is an imprint of Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the author
and publisher cannot assume responsibility for the validity of all materials or the
consequences of their use. The authors and publishers have attempted to trace the copyright
holders of all material reproduced in this publication and apologize to copyright holders
if permission to publish in this form has not been obtained. If any copyright material has not
been acknowledged please write and let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted,
reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other
means, now known or hereafter invented, including photocopying, microfilming, and
recording, or in any information storage or retrieval system, without written permission
from the publishers.
For permission to photocopy or use material electronically from this work, access
www.copyright.com or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood
Drive, Danvers, MA 01923, 978-750-8400. For works that are not available on CCC please
contact [email protected]
Trademark notice: Product or corporate names may be trademarks or registered trademarks
and are used only for identification and explanation without intent to infringe.

ISBN: 978-0-367-67627-8 (hbk)

ISBN: 978-0-367-67628-5 (pbk)
ISBN: 978-1-003-13208-0 (ebk)

DOI: 10.1201/9781003132080

Typeset in Times
by MPS Limited, Dehradun
Contents
Editors ......................................................................................................................vii
Contributors ..............................................................................................................ix

Chapter 1 Quantum Computing: Computational Excellence for

Society 5.0 ........................................................................................... 1
Paul R. Griffin, Michael Boguslavsky, Junye Huang,
Robert J. Kauffman, and Brian R. Tan

Chapter 2 Prediction Models for Accurate Data Analysis: Innovations in

Data Science ......................................................................................33
Balwinder Kaur, Anu Gupta, and R. K. Singla

Chapter 3 Software Engineering Paradigm for Real-Time Accurate

Decision Making for Code Smell Prioritization...............................67
Randeep Singh, Amit Bindal, and Ashok Kumar

Chapter 4 Evaluating Machine Learning Capabilities for Predicting Joining

Behavior of Freshmen Students Enrolled at Institutes of Higher
Education: Case Study from a Novel Problem Domain..................95
Pawan Kumar and Manmohan Sharma

Chapter 5 Image Processing for Knowledge Management and Effective

Information Extraction for Improved Cervical Cancer
Diagnosis .........................................................................................111
S. Jaya and M. Latha

Chapter 6 Recreating Efficient Framework for Resource-Constrained

Environment: HR Analytics and its Trends for Society 5.0..........139
Kamakshi Malik, Rakesh K. Wats, and Aman Khera

Chapter 7 Integration of Internet of Things (IoT) in Health Care

Industry: An Overview of Benefits, Challenges, and
Applications ..................................................................................... 165
Afshan Hassan, Devendra Prasad, Meenu Khurana,
Umesh Kumar Lilhore, and Sarita Simaiya

v
vi Contents

Chapter 8 Cloud, Edge, and Fog Computing: Trends and Case

Studies.............................................................................................. 181
Eng Lieh Ouh, Stanislaw Jarzabek, Geok Shan Lim, and
Ogawa Masayoshi

Chapter 9 A Paradigm Shift for Computational Excellence from

Traditional Machine Learning to Modern Deep
Learning-Based Image Steganalysis ...............................................209
Neelam Swarnkar, Arpana Rawal, and Gulab Patel

Chapter 10 Feature Engineering for Presentation Attack Detection in

Face Recognition: A Paradigm Shift from Conventional to
Contemporary Data-Driven Approaches.........................................241
Deepika Sharma and Arvind Selwal

Chapter 11 Reconfigurable Binary Neural Networks Hardware Accelerator

for Accurate Data Analysis in Intelligent Systems........................ 261
A. Kamaraj and J. Senthil Kumar

Chapter 12 Recommender System: Techniques for Better Decision making

for Society 5.0 .................................................................................281
Asha Rani, Kavita Taneja, and Harmunish Taneja

Chapter 13 Implementation of Smart Irrigation System Using Intelligent

Systems and Machine Learning Approaches .................................299
Raghuraj Singh, Ashutosh Deshwal, and Kuldeep Kumar

Chapter 14 Lightweight Cryptography Using a Trust-Based System for

Internet of Things (IoT) ..................................................................319
Amit Chadgal and Arvind Selwal

Chapter 15 Innovation in Healthcare for Improved Pneumonia Diagnosis

with Gradient-Weighted Class Activation Map Visualization.......339
Guramritpal Singh Saggu, Keshav Gupta, and
Palvinder Singh Mann

Index......................................................................................................................365
Editors
Kuldeep Kumar works at the Department of Computer Science and Engineering,
National Institute of Technology Jalandhar, India. He received his Ph.D. degree in
computer science from the National University of Singapore in 2016. Prior to
joining the institute, he worked for two years in the Birla Institute of Technology
and Science, Pilani, India. He received his Ph.D. in computer science from the
School of Computing, National University of Singapore (NUS SoC), Singapore in
2016. He has several publications in reputed international journals/conferences.
https://orcid.org/0000-0003-1160-9092

Eng Lieh Ouh received his Ph.D. in computer science (software engineering) from
the National University of Singapore. He is involved in several large-scale infor-
mation technology industry projects for a decade at IBM Singapore and Sun
Microsystems before joining academia as an educator. His research areas are software
reuse, software architecture design, design thinking, and software analytics. He has
experiences delivering courses for the postgraduate students, undergraduate students, and
industry participants in the software engineering areas including design thinking, practical
software architecture design, security engineering, and mobile development. He received
multiple teaching excellence awards and industry projects recognition awards throughout
his career. https://orcid.org/0000-0001-7759-348X

Arvind Selwal works as an assistant professor in the Department of Computer Science

and Information Technology, Central University of Jammu, India. His research
interests include machine learning, biometric security, digital image processing, and
soft computing. He has contributed more 25 research articles in reputed international
journals and he has authored a book titled Fundamentals of Automata Theory and
Computation. He is an active member of the Computer Society of India (CSI) and he
is undertaking two research projects from DRDO, New Delhi, India. https://orcid.org/
0000-0002-1075-6966

Harmunish Taneja is an assistant professor at DAV College, Sector 10, Panjab

University, Chandigarh. He received his Ph.D. in computer science and applications
from Kurukshetra University, Kurukshetra. He has more than 21 years of teaching
and research experience. He has guided 05 M.Phil. students and more than 90 PG
students of various universities. Four students are pursuing their Ph.D. under his
guidance. He is a reviewer of many reputed journals and has been a review committee
member of many conferences. His research interests are information computing,
mobile ad hoc networks, image processing, data science, recommender systems, and
system simulation. He has published and presented over 55 papers in international
journals/conferences of repute. He has also authored books of computer science and
applications. https://orcid.org/0000-0003-1745-0748

vii
viii Editors

Kavita Taneja is an assistant professor at Panjab University, Chandigarh. She

received her Ph.D. in computer science and applications from Kurukshetra
University, Kurukshetra, India. She has published and presented over 60 papers in
national/international journals/conferences and has had best paper awards in many
conferences including IEEE, Springer, Elsevier, ACM, and many more. She is a
reviewer of many reputed journals and has been a technical program committee
member of many conferences. She has also authored and edited computer books. She
has more than 18 years of teaching experience in various technical institutions and
universities. She is also member of BoM, Academic Council, and Board of Studies
of many universities and institutions. She has guided scholars of Ph.D./M.Phils.,
more than 100 PG students of various universities, and currently four students are
pursuing a Ph.D. under her guidance at Panjab University. Her teaching and research
activities include mobile ad hoc networks, simulation and modeling, and wireless
communications. https://orcid.org/0000-0002-9348-3587
Contributors
Amit Kumar Bindal is an associate professor in the Department of Computer
Science & Engineering, M. M. Engineering College, M. M. (Deemed to be University)
Mullana, Ambala, Haryana, India. He received Ph.D. from Maharishi Markan-
deshwar University, M. Tech (Computer Engineering) from Kurukshetra University
and a B. Tech. in computer engineering from Kurukshetra University Kurukshetra. He
has been in teaching and research and development since 2005. He has published about
60 research papers in international, national journals, and refereed international con-
ferences. His current research interests are in wireless sensor networks, underwater
wireless sensor networks, sensors, and IoT etc. https://orcid.org/0000-0002-2624-8077

Michael Boguslavsky is presently the head of AI and is leading the AI team developing
ML credit analytics for a start-up trade finance platform. The team builds predictive
models for SME credit risk and develops trade credit event prediction models based on
supply chain graph flows analysis. Prior to this he was an advisor for Blackstone
Alternative Asset Management, advising the leading alternative asset manager on fund
structuring, portfolio optimization, risk modelling, and placement for alternative, private,
and illiquid credit, hedge funds, infrastructure, and emerging market debt funds. He has
also worked in banks and asset management companies over 15 years as acting head of
EMEA Equity Derivatives Structuring and head of ALM and Quantitative Analytics, as
well as being a trader and quantitative analyst. He holds a Ph.D. in mathematics from the
University of Amsterdam; a Ph.D. in computer sciences from the Russian Academy of
Sciences, Moscow; and a master’s degree in mathematics and applied mathematics.
https://orcid.org/0000-0001-8423-9403

Amit Chadgal completed the M.Tech. in 2020 at the Department of Computer Science
and Technology, Central University of Jammu. Prior to M.Tech., he completed a B.
Tech. in 2018 at the Department of Information Technology, National Institute of
Technology Srinagar, India. The research topic of his M. Tech. was “light-weight
cryptography techniques in IoT” at the Department of Computer Science and
Technology, Central University of Jammu. https://orcid.org/0000-0002-9013-308X

Ashutosh Deshwal is currently pursuing a Ph.D. in computer science and engineering

at Thapar Institute of Engineering, Patiala, Punjab, India, since July 2019. His areas of
interest include blockchain, machine learning, Internet of Things, embedded system,
and product analysis. https://orcid.org/0000-0001-8444-1572

Paul R. Griffin is a member of the faculty of Singapore Management University (SMU)

teaching postgraduate and undergraduate students in IT and FinTech as an associate
professor of information systems. He gained a Ph.D. at the Imperial College London
in 1997 on quantum well solar cells and thermophotovoltaics and is now researching
disruptive technologies applications and impact. In particular, his research covers
decentralized solutions and the application of quantum computing focusing on

ix
x Contributors

financial applications. With a number of projects ongoing for consensus, trade

finance, and portfolio optimization, he has been advising companies since 2014 and
presenting at events, judging hackathons and moderating panel discussions on
FinTech. Prior to SMU he was leading application development on global,
regional, and local projects for over 15 years in the United Kingdom and Asia in
the finance industry. During this time, as well as leading internal IT development
teams, he worked on outsourcing, off-shoring projects, and IT support. With a Black
Belt in Six Sigma, he is keen on quantifying quality and ensuring efficiency in current
processes and during change. https://orcid.org/0000-0003-2294-5980

Anu Gupta, has been working as a professor in computer science and applications
at the Panjab University, Chandigarh since July 2015, where she has been working as
a faculty member since 1998. She has the experience of working on several platforms
using a variety of development tools and technologies. Her research interests include
open-source software, software engineering, cloud computing, and data mining. She
is a life member of the “Computer Society of India” and “Indian Academy of
Science.” She has published more than 25 research papers in various journals and
conferences. https://orcid.org/0000-0002-1403-5023

Keshav Gupta is a pre-final year student at the Indian Institute of Information

technology, pursuing an integrated postgraduate in information technology. He has been
associated with machine learning and data science for the better part of the last three
years. He also has experience working with various machine learning based start-ups.
He has worked as a research scholar in his institute for research project related
multimodal AI and has formidable ranks in various data science challenges and
hackathons. His research interests include computer vision, natural language processing,
and time series forecasting analysis. https://orcid.org/0000-0002-5763-0690

Afshan Hassan is a Ph.D. research scholar in Chitkara University, Punjab, India.

She has completed her B.Tech. (Computer Science & Engineering) degree from
Islamic University of Science & Technology, J&K, India in 2014 and M.Tech.
(Computer Science & Engineering) degree from Chandigarh Group of Colleges,
Landran, India in 2019. She is a gold medalist in B.Tech. (CSE). She has several
research publications in Scopus – Indexed journals of high repute. Her research
interests include wireless sensor networks, ad hoc networks, machine learning,
and IoT security. https://orcid.org/0000-0001-5133-2927

Junye Huang is a quantum developer advocate at IBM. He is part of the Qiskit

community team whose mission is building an open, diverse, and inclusive quantum
community. He is focusing on promoting quantum education in the Asia Pacific
region. He organized the first Qiskit university hackathon in the world and is a guest
lecturer for two quantum computing courses at National University of Singapore. In
addition, he is supporting regional and global educational initiatives such as the IBM
Quantum Challenge and the Qiskit Global Summer School. His passion for quantum
computers drives him to create educational games for quantum computers such as
QPong, a quantum version of Pong which he created at the first Qiskit camp. QPong
Contributors xi

was subsequently ported to a physical Quantum arcade machine and toured around
Europe, including the EU Quantum Flagship Event in Helsinki in October 2019. He is
a Ph.D. candidate in experimental low temperature physics at the National University
of Singapore. https://orcid.org/0000-0001-7896-0595

Stan Jarzabek received a M.Sc. and Ph.D. from Warsaw University. He has been a
professor at Bialystok University of Technology since 2015; in 1992–2015 he was
an associate professor at the Department of Computer Science, National University
of Singapore; in 1990–92 he was a research manager of CSA Research Ltd in
Singapore. Before, Stan taught at McMaster University, Canada, and worked for an
industrial research institute in Warsaw. Stan’s research interest is software engineering
(software reuse and maintenance), and in recent years mHealth – the use of mobile
technology for psychotherapy support, patient monitoring, and data collection/analysis.
https://orcid.org/0000-0002-7532-3985

S. Jaya is current Ph.D. research scholar (Full Time) in the Department of Computer
Science at Sri Sarada college for Women (Autonomous), Salem-16, Tamilnadu, India.
She has published three papers and one chapter in IGI Global. Her area of interest is
digital image processing. https://orcid.org/0000-0003-0361-8052

A. Kamaraj received his B.E. degree in Electronics and Communication Engineering

from Bharathiar University, Coimbatore, Tamil Nadu, India in 2003. He completed his
post graduation from Anna Universiy, Chennai in the field of VLSI Design in 2006.
He completed his Ph.D. at Anna University, Chennai in 2020. Currently he is an
associate professor in the Department of Electronics and Communication Engineering,
MepcoSchlenk Engineering College, Sivakasi, India. His research interests include
digital circuits and logic design, reversible logic and synthesis, and advanced
computing techniques. During his 14 years of teaching, he has published 21 papers
in international journals and 23 papers in national and international conferences. He
has filed two patents and was granted one copyright. He has been member of IETE and
ISTE. https://orcid.org/0000-0001-6952-2374

Robert J. Kauffman holds the Endowed Chair in Digitalization at the Copenhagen

Business School, sponsored by Danske Bank, Copenhagen Airport A/S, and the Danish
Society for Education and Business. He earlier served at Singapore Management
University’s School of IS as professor of IS Management and Associate Dean (Research,
Faculty). He was the W.P. Carey Chair in IS at Arizona State until 2011, and professor
and director of the MIS Research Center, Carlson School of Management, University
of Minnesota until 2007. His graduate degrees are from Cornell and Carnegie Mellon.
His research focuses on senior management issues in strategy, information, technology,
economics and society, digital commerce, the FinTech Revolution, and resource and
energy sustainability. His past employment was in international lending, payment
networks, trade services, investments and trading, technology innovation, and other
financial services. His funded research has spanned these and hospitality services,
digital entertainment, IT consulting, airlines, and agribusiness. https://orcid.org/0000-
0002-3757-0010
xii Contributors

Balwinder Kaur, is a research scholar in computer science and applications,

Panjab University, Chandigarh where she has been working as an assistant
professor since 2011. She has working experience of more than 14 years. She
received her MCA degree from Kurukshetra University. Her research interest is in
educational data mining. https://orcid.org/0000-0001-2222-5903

Aman Khera received his Ph.D. (Business Laws) from Panjab University, Chandigarh
in the year 2014 and Ph.D. (Business Management & Commerce) from Panjab
University, Chandigarh in the year 2018. He did his MBA (HRM) at Punjab Technical
University in the year 2004 and LLB at Panjab University Chandigarh in the year 2009
and LLM at Kurukshetra University, Kurukshetra in the year 2011. He has seven years
of corporate experience in HR. He has published 23 research papers in both national and
international journals and has presented various papers at national and international
conferences. He has guided one Ph.D. student and is guiding two Ph.D. students.
Presently he is working as an assistant professor in University Institute of Applied
Management Sciences (UIAMS), Panjab University, Chandigarh since 2011. His area of
specialization is HRM and business law. https://orcid.org/0000-0003-4909-2332

Meenu Khurana has a Ph.D. in computer science and engineering at Chitkara

University Institute of Engineering and Technology, Chitkara University,
Punjab, India. She has won scores of academic awards during her graduate
and post-graduate studies. She is a certified programmer for Java 2 Platform
from Sun Systems, USA and also a certified programmer from Brain bench,
USA. She has over 26 years of experience in industry, academics, research, and
administration to her credit. Her areas of expertise are mobile ad-hoc networks,
vehicular ad-hoc networks, wireless technologies, network security, sensor networks,
machine learning, cloud computing, curriculum designing and development, and
pedagogical innovation. Her areas of interest include artificial intelligence, machine
learning, and algorithms. She has guided research candidates in areas of expertise.
https://orcid.org/0000-0001-6515-7939

Ashok Kumar is an ex-professor in the Department of Computer Science &

Engineering, M. M. Engineering College, M. M. (Deemed to be University) Mullana,
Ambala, Haryana, India and former professor of Kurukshetra University. He was in
teaching and research and development for more than 40 years. He has published
many research papers in international, national journals, and refereed international
conferences. His current research interests are in software engineering and digital
image processing. https://orcid.org/0000-0003-1240-7054

Pawan Kumar is pursuing Ph.D. from Lovely Professional University, Punjab,

India. He is serving as an assistant professor in the School of Computer Applications,
Lovely Professional University, Punjab, India. He has an experience of more than 14
years in academics. He has qualified UGC-NET and GATE in 2012. His research
interests include machine learning and data analytics. He is a lifetime member of
Computer Society of India (CSI) and Indian Science Congress Association (ISCA).
https://orcid.org/0000-0003-1698-3286
Contributors xiii

J. Senthil Kumar received his B.E. degree in electronics and communication

engineering from M.S. University, Madurai, Tamil Nadu, India in 2003. He has
completed his post graduation from Anna Universiy, Chennai in the field of Communi-
cation Systems in 2005. He has completed his Ph.D. at Anna University, Chennai in
2017. Currently he is an associate professor in the Department of Electronics and
Communication Engineering, MepcoSchlenk Engineering College, Sivakasi, India. His
research interests include robotics, Internet of Things, and embedded systems. During
his 15 years of teaching career, he has published 24 papers in international journals and
nine papers in national and international conferences. He has published five patents and
granted one copyright. He has completed three sponsored research projects worth of 78
Lakhs. He has been member of IETE and ISTE. http://orcid.org/0000-0002-9516-0327

M. Latha is currently working as associate professor of computer science in Sri

Sarada College for Women (Autonomous), Salem-16, Tamilnadu, India. She has
24 years of teaching experience. Her area of research includes software engineering
and digital image processing. Her h-index is 5 and i10 index is 4. https://orcid.org/
0000-0002-2648-765X

Umesh Kumar Lilhore is an associate professor in Chitkara University Institute of

Engineering and Technology, Chitkara University, Punjab, India. He is Ph.D. in
computer science engineering and M. Tech. in computer science and engineering. He
has research publications in SCI-Indexed international journals of high repute. His
research area includes AI, machine learning, computer security, computational
intelligence, and information science. https://orcid.org/0000-0001-6073-3773

Kamakshi Malik is working as an assistant professor in the Department of Management

and Commerce in DAV College, Chandigarh. She has a total teaching experience of
more than a decade. Her research interests include employee engagement, organi-
zational behavior, and human resource management and is currently pursuing a Ph.D.
from IKG, Punjab Technical University, Jalandhar. She has published 16 research and
review articles. She has also presented her work at various national and international
conferences.

Palvinder Singh Mann received his bachelor’s degree (B.Tech.) with honors
(Institute Gold Medal) in information technology from Kurukshetra University,
Kurukshetra, Haryana, India, M.Tech. in computer science and engineering from IKG
Punjab Technical University, Jalandhar, Punjab, India, and Ph.D. in computer science
and engineering from IKG Punjab Technical University, Jalandhar, Punjab, India.
Currently, he is working as an assistant professor at DAV Institute of Engineering and
Technology, Jalandhar, Punjab, India. He has published more than 50 research papers
in various international journals, international conferences, and national conferences. His
research interests include wireless sensor networks, computational intelligence, and
digital image processing. https://orcid.org/0000-0002-9859-6193

Ogawa Masayoshi graduated from Singapore Management University with degrees

in economics and information systems. He is passionate about developing greater
xiv Contributors

understanding of the world and never passes up the opportunity to learn. Currently, he
is a pricing actuarial executive in the life insurance industry. https://orcid.org/0000-
0002-1605-6477

Gulab Kumar Patel received his master's in mathematics and computing from Indian
Institute of Technology, Guwahati, Assam, India, in 2017. Currently, he is pursuing his
doctoral work on queuing theory in the faculty of Mathematical Sciences from Indian
Institute of Technology (Banaras Hindu University), Varanasi, Uttar Pradesh, India. His
area of research interests fall into realms of deep learning, statistics, and queuing theory.
https://orcid.org/0000-0002-5641-8532

Devendra Prasad is a professor in Chitkara University Institute of Engineering and

Technology, Chitkara University, Punjab, India. He has received his M.Tech. (Computer
Science and Engineering) and Ph.D. (Computer Science and Engineering). He has
supervised several M.Tech. and Ph.D. students. He has research publications in SCI-
Indexed international journals of high repute. His research interest includes network
security and fault-tolerant mobile ad-hoc, wireless sensor networks, and machine
learning. https://orcid.org/0000-0002-1771-4670

Asha Rani is an assistant professor at Gujranwala Guru Nanak Khalsa College,

Ludhiana (Punjab). She is a research scholar in the Department of Computer
Science and Applications Panjab University, Chandigarh, India. She has published
and presented over 12 papers in national/international journals/conferences. She has
more than seven years of teaching experience in various institutions. Her teaching and
research activities include recommendation systems, multi-criteria decision making
methods, and data mining techniques. https://orcid.org/0000-0001-5481-2141

Arpana Rawal is chairman, Board of Studies; faculty of computer science and

engineering; former head, Department of Information Technology; professor, Department
of Computer Science and Engineering at Bhilai Institute of Technology; Durg affiliated to
Chhattisgarh Swami Vivekanand Technical University, Bhilai, India. Her graduation is in
the discipline of computer technology from Nagpur University (1994) and post-
graduation in the same discipline from Pt. Ravi Shankar Shukla University, Raipur
(2003). She was awarded a doctoral degree in the faculty of computer science and
engineering in 2013. She has teaching experience of more than 20 years and research
experience of 15 years with current areas of research interest in innovative machine
learning techniques to combat high obfuscation levels of plagiarism detection, natural
language processing, text mining, temporal sequence mining, educational data mining,
machine-assisted recommender systems, recognition of text document images, universal
image steganalysis, and database forensics. https://orcid.org/0000-0002-6819-6498

Guramritpal Singh Saggu is pursuing an integrated M.Tech. in information technology

from Indian Institute of Information Technology, Gwalior. He has been working and
exploring the domains of data science and machine learning for the past few years and
has prior experience of working with multiple organizations in developing their machine
learning and backend pipelines. He has worked as a research scholar in his institute for
Contributors xv

research project related multimodal AI and has had formidable ranks in various data
science challenges and hackathons. His research interests include computer vision, natural
language processing, and time series forecasting analysis. https://orcid.org/0000-0002-761
9-3553

Lim Geok Shan graduated summa cum laude from Singapore Management University
in 2020 with a bachelor of science (Information Systems) and a second major in
advanced business technology (Information Security and Assurance). During her time as
an undergraduate, she conducted a systematic literature review of software multitenancy
architecture. She also worked as an undergraduate teaching assistant for software
engineering and database management classes. She has a keen interest in the field of
software engineering and security. https://orcid.org/0000-0003-1965-4648

Deepika Sharma received her master's degree in computer applications (MCA)

from the Department of Computer Science and Information Technology, Central
University of Jammu, Jammu and Kashmir, India. Presently, she is pursuing her
Ph.D. from Central University of Jammu in the area of biometrics security. Her
research interests include biometrics, machine learning, and deep learning.
https://orcid.org/0000-0003-1376538X

Manmohan Sharma is serving as an associate professor in school of computer

applications, Lovely Professional University, Punjab, India. He has a vast experience of
more than 20 years in the field of academics, research, and administration with different
universities and institutions of repute such as Dr. B.R. Ambedkar University and
Mangalayatan University. He has been awarded his doctorate degree from Dr. B.R.
Ambedkar University, Agra in 2014 in the field of wireless mobile networks. His areas of
interest include wireless mobile networks, ad hoc networks, mobile cloud computing,
recommender systems, data science, and machine learning, etc. A large number of
research papers authored and co-authored, published in international or national journals
of repute and conference proceedings come under his credits. He is currently supervising
six doctoral theses. Three M.Phil. degrees have already awarded under his supervision.
He has guided more than 1,000 PG and UG projects during his service period under the
aegis of various universities and institutions. He worked as a reviewer of many
conference papers and member of the technical program committees for several
technical conferences. He is a member of various professional/technical societies
including Computer Society of India (CSI), Association of Computing Machines
(ACM), Cloud Computing Community of IEEE, Network Professional Association
(NPA), International Association of Computer Science and Information Technology
(IACSIT), and Computer Science Teachers Association (CSTA). https://orcid.org/
0000-0001-9445-5898

Sarita Simaiya, is an associate professor at Chitkara University Institute of Engineering

and Technology, Chitkara University, Punjab, India. She is Ph.D. in computer science
engineering and an M.Tech. in computer science and engineering. She has research
publications in SCI-Indexed international journals of high repute. Her research areas
xvi Contributors

include artificial intelligence, machine learning, computer security, and wireless sensor
networks. https://orcid.org/0000-0001-7686-8496

Ravinder K. Singla has been a professor of computer science and applications

at Panjab University, Chandigarh since July 2004 where he has been a faculty member
since 1988. He has held many positions including dean of university instructions, dean
research, director, computer center, chairperson of the Department of Computer Science
and Applications; coordinator, TIFAC-DST; project leader; and programmer/analyst. His
research interests include scientific computing, Linux networking, mobile computing,
open-source software, and software cost estimation. He is also a member of the editorial
board of the Panjab University Research Journal (Science)-New Series. He is a life
member of the Computer Society of India. He has published more than 70 research
papers in various journals and conferences. https://orcid.org/0000-0001-9120-0267

Raghuraj Singh is currently pursuing a Ph.D. in computer science and engineering

at Dr. Br. Ambedkar National Institute of Technology, Jalandhar, Punjab, India,
since July 2019. His areas of interest include software fault prediction, machine
learning, and soft computing and its applications in modern devices. https://
orcid.org/0000-0002-0634-1485

Randeep Singh is a research scholar in the Department of Computer Science and

Engineering, M. M. Engineering College, M. M. (Deemed to be University) Mullana,
Ambala, Haryana, India. Randeep Singh received an M.Tech. from Kurukshetra
University. Randeep Singh has been in teaching and research and development since
2008. He has published about 15 research papers in international and national journals
and refereed international conferences. His current research interest is in software
engineering. https://orcid.org/0000-0003-2371-4148

Neelam Swarnkar received her diploma in engineering in information technology in

2003 from Government Polytechnic College, Durg, Chhatisgarh. She received her
B.E. degree in information technology in 2006 from Bhilai Institute of Technology,
Durg, Chhattisgarh, India and received and M.E. degree in computer technology and
applications in 2010 from Shri Shankaracharya College of Engineering and Tech-
nology, Durg, Chhattisgarh, India. Currently, she is working towards a Ph.D. at the
Bhilai Institute of Technology Research Centre, Durg under Chhattisgarh Swami
Vivekananda Technical University, Bhilai, Chhattisgarh, India. Her research interests
are information security, steganography, and steganalysis. https://orcid.org/0000-
0002-9940-4034

Brian R. Tan focuses on the study of value-creation in his research. He obtained his
Ph.D. from the University of Seattle, Washington, and has served on the faculty at the
Nanyang Business School at Nanyang Technological University (NTU) and the
National University of Singapore (NUS) Business School. His industry background
comes from management consulting at the Boston Consulting Group, and B.R.I.T.
Management Consulting. He also serves as an advisor with a start-up marketing
agency in Vietnam called K2, Kreative Kommunications. His efforts at integrating
Contributors xvii

academia with industry have led to the development of a thinking framework called
the “Wisdom Approach.” It utilizes quantum mechanics principles with broad neuro-
science foundations and supports teaching the cognitive aspect of wisdom. The
application component of the “Wisdom Approach” has been used to teach MSc and
MBA students at NUS. Brian is currently applying it to create better thinking
machines and artificial wisdom. https://orcid.org/0000-0002-0275-846X

Rakesh Wats, is a professor and head of the Media Engineering Department at

NITTTR, Chandigarh. He has a master's in civil engineering and doctorate in quality
management. He holds his expertise in education, educational management, curriculum
development, and media design and development. He has three modules, one MOOC
in addition to more than 75 research publications to his credit. He has working
experience of more than three decades in academics, research, and industry. https://
orcid.org/0000-0001-8985-6777
 1 Quantum Computing:
Computational
Excellence for Society 5.0
Paul R. Griffin1, Michael Boguslavsky2,
Junye Huang3, Robert J. Kauffman4, and
Brian R. Tan5
1
School of Information Systems Singapore Management
University, Singapore
2
Tradeteq, London, United Kingdom
3
IBM Quantum, Singapore
4
Copenhagen Business School Denmark, and School of
Information Systems, Singapore Management University,
Singapore
5
B.R.I.T. Management Consulting, Singapore

CONTENTS
1.1 Introduction ......................................................................................2
1.2 Quantum Computing Fundamentals ................................................ 4
1.2.1 Key Concepts ....................................................................... 4
1.2.2 Hardware, Software, Alorithms, and Workflow .................6
1.3 Quantum Computing Needs and Service Industry Applications....9
1.3.1 Business Needs and Concerns .............................................9
1.3.2 Decision Problem Framing and Computation...................11
1.3.3 The Range of Business Problem Areas That Can Be
Addressed ...........................................................................12
1.3.4 The Unique Role of Quantum Computing in Financial
Services Applications.........................................................14
1.4 Application Framework.................................................................. 16
1.4.1 Algorithm Design...............................................................17
1.4.2 Software Development.......................................................18
1.4.3 Hardware for Quantum Computing...................................21
1.4.4 Integration with Other IT Systems in the Firm ................22
1.5 Case: Implementing a Quantum Neural Network
for Credit Risk................................................................................22
1.5.1 Credit Risk Assessment .....................................................22
1.5.2 Algorithm Design for a Quantum Neural
Network (QNN)..................................................................24

DOI: 10.1201/9781003132080-1 1
2 Data Science and Innovations for Intelligent Systems

1.5.3 Software Design for QNN .................................................25

1.5.4 Hardware for Quantum Credit Scoring .............................26
1.5.5 Issues for Moving from a Stand-Alone to an
Integrated System.............................................................. 27
1.6 Conclusion ......................................................................................27
Acknowledgments....................................................................................................28
Notes ........................................................................................................................28
References................................................................................................................28
Appendix A: Glossary of Terms ........................................................................31

1.1 INTRODUCTION
Quantum computing is a key part of building an intelligent systems infrastructure
for Society 5.0 and can be used in the future across the main pillars of fintech,
healthcare, logistics, and artificial intelligence (AI). Intelligent systems based on
data science are machines that are sufficiently advanced to be able to perceive and
react to external events. Quantum computers offer various avenues to go beyond
systems using classical computers and extend computational excellence beyond its
current state. Digital innovation underpins the concept of Society 5.0 for a better
future with an inclusive, sustainable, and knowledge-intensive society that uses
information computing. A key to realizing this society is to utilize gargantuan
volumes of data in real-time in intelligent systems. The sharing of information in
Society 4.0 has been insufficient and integration of data problematic, whereas
Society 5.0 integrates cyberspace and physical space. For example, in Society 5.0,
the huge amount of data from physical Internet of Things (IoT) devices are required
to be analyzed, processed, and fed back to robotic devices interacting with people in
various forms. In Japan alone, the next 15 years is expected to see a growth in IoT
and robotics of US$20 bn and US$70 bn, respectively (JapanGov News, 2019).
However, the aim of Society 5.0 is to balance economic development and solutions
for social issues to bring about a human-centered society. This chapter shows how
quantum computing can be applied to many current challenges and open up new
opportunities with innovative ways that align better with human thinking.
Classical computing has brought society great benefits over many years from the
abacus 3,300 years ago through to modern computing from Alan Turing in the 20th
century – to the latest smartphones we are now familiar with. Computers have enabled
products and services that humans cannot provide alone such as increasing productivity,
enhancing communication, storing vast amounts of data, sorting, organizing, and
searching through information amongst many more. However, many problems still exist
such as data security, scalability, manageability, and interoperability. Furthermore,
Moore’s Law increases in computing power is now beginning to fail (Loeffler, 2018).
Stefan Filipp, a quantum scientist at IBM Research, has stated that to “continue the pace
of progress, we need to augment the classical approach with a new platform, one that
follows its own set of rules. That is quantum computing” (Singh, 2019). Using the
advantages of quantum over classical computing, it is possible to increase computing
capacity beyond anything that classical computers can achieve.
Quantum Computing 3

Quantum computing was suggested in the 1980s by Manin (1980) and Feynman
(1982). In the past few years, it has become a reality and accessible to everyone,
with IBM putting the first quantum computer on the cloud in 2016. Now, in 2021,
there are dozens of quantum computers online with processing capabilities much
better than the first one. While there is little doubt that quantum computers can
outperform classical computers for some processes, such as unstructured search
problems (Grover, 1996), it is not clear whether and how quantum computing will
be advantageous for a particular business need or, indeed, worth the effort to in­
vestigate further.
This chapter is aimed at providing a framework to assess the likelihood that
quantum computing will be an area that is worthwhile to get involved in for par­
ticular business opportunities and challenges. The main differences for quantum
computing are superposition and entanglement. Traditional computers use bits of
either 0 or 1. In contrast, quantum computers use qubits existing in a state that is
best described as the probability of being either 0 or 1. This is called the super­
position of states (Nielsen & Chuang, 2010).1 Qubits also exhibit entanglement,
whereby they may be spatially nearby or far apart, may interact with one another at
certain times, and yet are not able to be characterized as being independent of one
another. The result is that two qubits may work together as if they were one larger
qubit. This is fundamentally different from bits that are always kept separate in
classical computing. However, current quantum computers are noisy and have an
insufficient number of qubits to be able to show provable advantages. And even the
widely publicized Google experiment (Arute et al., 2019) is still held to be con­
tentious (Pednault, 2020). Even more contentious are annealer-type quantum
computers (Rønnow et al., 2014). These will not be covered in this chapter as a
result, and we will focus on gate-type quantum computers instead.
Considering the Society 5.0 issues of data volumes, real-time processing and
linking data, we present a framework to assess what business needs may potentially
be addressed by quantum computing and how quantum computing is different to
classical systems. There are four areas of concern: the data, the processing, the
infrastructure, and the environment. (See Figure 1.1.) First, data may be complex,

FIGURE 1.1 The main areas of concern in the quantum application framework.
4 Data Science and Innovations for Intelligent Systems

natively quantum or probabilistic, for example, chemical reactions involving

quantum particles and financial derivatives pricing including many predictor vari­
ables. Second, processing may involve a range of different data analytics ap­
proaches, such as simulation, machine learning (ML), optimization, and AI. Third,
current quantum computing infrastructure solutions vary in their types and speci­
fications, with the overall marketplace in the throes of rapid technological in­
novation. Last, the environment may have legal regulations in terms of what can be
done, how data can be used, and how physical and human resources can be used
while protecting private information.
The remainder of this chapter is laid out as follows. Section 1.2 reviews the main
differences between quantum computing and classical computing. It introduces the
concepts of qubits, quantum states, and quantum operations. It also gives overviews of
different approaches to quantum hardware, as well as the end-to-end process of running a
quantum algorithm. Available software and debugging challenges are also discussed,
along with some common quantum algorithms. Section 1.3 looks at current managerial
concerns and related business problems that quantum computing can potentially address.
Section 1.4 offers a framework showing the areas that need to be addressed when
considering whether to build a quantum computing solution. Using this framework,
Section 1.5 provides an example of its application for a quantum neural network (QNN)
solution for the credit rating of small and medium-size enterprises (SMEs). Section 1.6
looks at the future of quantum computing, covering improvement areas for theoretical
development, hardware, and integration with analytics software.
While there is no expectation that quantum computers will outperform current
classical models in the near term, exploring the requirements and limitations of
current quantum computers can be useful for thinking through how to develop a
future system to meet business objectives when quantum computing reaches a
suitable level of maturity.

1.2 QUANTUM COMPUTING FUNDAMENTALS

This section reviews the fundamental properties of quantum computers, and the
end-to-end workflow from data preparation to analyzing the quantum circuit output.
We further discuss the available software, with a focus on Qiskit as an example.
Hardware and available algorithms are also assessed.

1.2.1 KEY CONCEPTS

The fundamental concepts of superposition, entanglement, interference, qubits,
quantum gates, the concept of Bloch's sphere, and adiabatic annealers are key to
understanding quantum computers. We will explain these concepts here.
Superposition. The fundamental processing and storage unit of quantum
computers is the quantum bit, also called a qubit. A qubit can represent 0 or 1 like
a classical bit. However, it can also represent 0 and 1 in a superposition state.
A classical bit is either in the state 0 or 1. In contrast, a qubit is in a state which can
be characterized by a general formula, |ψ〉 = a|0〉 + b|1〉, where a and b are prob­
ability amplitudes represented by complex numbers. Although a qubit can be in a
Quantum Computing 5

superposition, when it is measured or read out in a computer language, the result

can only be 0 or 1 and not both at the same time. However, by measuring a qubit
multiple times, the probability of obtaining 0 or 1 is obtained. The probability of 0
is | a |2 and 1 is | b |2, and due to the conservation of the total probability being equal
to 1, | a |2 + | b |2 = 1.
Entanglement. Another strange phenomenon related to qubits is entanglement. It
refers to the correlation of different qubits in a system. If we extend the general
formula for a qubit state to two qubits, we have: |ψ〉 = a|00〉 + b|01〉 + c|10〉 + d|11〉.
Qubits are entangled if the state of one qubit is correlated with by the other. For
example, when b = c = 0, the state becomes a|00〉 + d|11〉. In this state, the two
qubits’ states are strongly correlated so that when the first qubit is |0〉 the second
qubit is always also |0〉, and when the first qubit is |1〉 the second qubit is always |1〉.
Interference. Quantum states can interfere with each other due to the phase
difference between the probability amplitudes. Quantum interference is similar to
wave interference and can be understood in the same way. For example, when two
waves act together in phase, their probability amplitudes sum to create a higher
probability, and when the probability amplitudes are out of phase, their amplitudes
cancel out. (See Figure 1.2.) This phenomenon can be utilized by quantum
algorithms to speed up their calculations. For example, the Grover (1996) search
algorithm uses interference to increase the probability of finding the right answer
and reduce the probability of the wrong ones.

FIGURE 1.2 Interference pattern from a simulated double-slit experiment with electrons.
Source: Alexandre Gondran, distributed under a CC-BY-SA-4.0 license.
6 Data Science and Innovations for Intelligent Systems

FIGURE 1.3 An example of a quantum circuit.

Quantum gates. Quantum algorithms can be represented by quantum circuits.

Quantum circuits consist of quantum gates like the AND gate and the XOR gate
from classical logic circuits. A common set of universal gates which can perform
any quantum operation includes the Clifford gates and the T gates. This set of
universal gates is composed of CNOT, H, S, and T gates.
Hadamard (H) gates map |0〉 to |0〉 + |1〉 and |1〉 to |0〉 − |1〉. H gates are ubi­
quitous in quantum algorithms for generating superpositions. S and T gates are part
of the phase shift gates family. They leave |0〉 unchanged and map |1〉 to eiϕ |1〉
(with ϕ = π/2 and π/4 for the S and T gates, respectively). Phase gates are important
for manipulating phases of quantum states to achieve intended interference.
Controlled NOT (CNOT) gates act on two qubits, unlike the H, S, and T gates,
which act on a single qubit only. CNOT gates perform a NOT operation on the
second qubit whenever the first qubit is in the state |1〉 and can be described as
mapping the state |a, b〉 to the state |a, a ⊕ b〉. CNOT is important for generating
entanglement between qubits.
An example of a quantum circuit is shown next. (See Figure 1.3.) This circuit has
an H gate on qubit-0 and four CNOT gates between qubit-0 (the control) and each
of the remaining qubits (the target). Finally, the qubits are measured, and the
outcomes are stored in classical registers.

1.2.2 HARDWARE, SOFTWARE, ALORITHMS, AND WORKFLOW

We introduce a range of available software and hardware, followed by a general
workflow for a quantum program’s execution. The IBM Quantum Experience and
Qiskit are used together as an example to illustrate a typical workflow.
Available hardware. As of January 2021, there were a handful of quantum
computing hardware providers. They include: IBM Quantum Experience,
Rigetti Quantum Cloud Services, AWS Braket, Microsoft Azure Quantum,
Xanadu Quantum Cloud, D-Wave Leap, Quantum Inspire (from QuTech), and
Origin Quantum Cloud. All except AWS Braket and Azure Quantum have their
own hardware systems. AWS Braket has three different external hardware
Quantum Computing 7

providers: D-Wave (annealer), IonQ (trapped ions), and Rigetti (super­

conducting qubits), while Azure Quantum has IonQ (trapped ions), Honeywell
(trapped ions), and Quantum Circuits (superconducting qubits). IBM Quantum
Experience, Quantum Inspire, and Origin Quantum Cloud have systems that
are open for the public free of charge, while the others charge for access.
IBM Quantum Experience (IQX) is the leading provider of quantum computing
services. In 2016, IQX put the first quantum computer on the cloud. Since then, the
platform has grown tremendously and now has more than 10 quantum systems (of
up to 15 qubits) with free access and an additional more than 10 premium access
quantum systems (with up to 65 qubits) for the IBM Quantum Network’s partners.
IQX has more than 250,000 users who collectively run more than 1 billion quantum
circuits each day, and have published more than 250 related research papers
(IBM.com, 2020). The IBM Quantum Network has more than 200 partners in in­
dustry and academia. IBM also released its quantum hardware roadmap in the
annual Quantum Summit in September 2020, and laid out the steps toward building
quantum systems with more than a million qubits. IBM aims to release processors
with 127 qubits in 2021, 433 qubits in 2022, and 1,121 qubits in 2023 over the next
three years (Gambetta, 2020). The quantum systems will be large enough to in­
vestigate the implementation of quantum error correction (QEC), to open the door
to the practical implementation of many quantum algorithms.
Available software. There is a range of software development kits (SDKs) available
for writing and running quantum programs, including Qiskit from IBM, Cirq from
Google, QDK from Microsoft, Forest from Rigetti, and ProjectQ from ETH Zurich
(LaRose, 2019). All of these SDKs, except QDK, are based on Python, which allows
easy integration of the Python ecosystem’s capabilities for scientific computing and ML.
Many companies also work on quantum software packages for specific domain appli­
cations to interface with the SDKs mentioned previously. Notable examples include:
Qiskit Aqua for chemistry, ML and optimization, from IBM; PennyLane for ML from
Xanadu; OpenFermion for chemistry, from Google; and TensorFlow Quantum for ML,
from Google.
Available algorithms. The most famous quantum algorithm is attributable to
Shor (1997), who showed an exponential quantum speed-up related to the best-
known classical algorithms for factorization. It is likely to threaten the existing
cryptography infrastructure, however, it also requires a large number of qubits and
gate operations, which also require QEC capabilities. The best estimate is that it will
require on the order of millions of physical qubits (Gidney & Eker, 2019). Other
famous textbook algorithms such as Grover’s algorithm also require QEC cap­
abilities, which will be difficult to implement in NISQ devices within the next three
to five years.
There is a class of quantum-classical hybrid algorithms called variational
quantum eigensolver algorithms (VQE, a kind of the more general variational
quantum algorithms, VGAs). These are suitable for implementation on NISQ
devices (Peruzzo et al., 2014). Such algorithms contain both quantum circuits
and classical procedures that are invoked iteratively. The quantum circuits in
each iteration have a small number (< 100) of qubits and a small number (also
< 100) of quantum gates, and these can be run with current NISQ devices. The
8 Data Science and Innovations for Intelligent Systems

results of the quantum circuits are then fed to classical procedures for calcu­
lation and optimization to determine the parameters of a quantum circuit in the
next iteration of the algorithm. This hybrid approach enables harnessing the
power of quantum computers in the NISQ era before QEC becomes widely
available.
Quantum computing workflow stages and examples. A typical quantum program
workflow includes three stages: data loading, data processing execution, and data
extraction.

• In the data loading stage, data are loaded into the memory of classical
computer and converted to the states represented by qubits.
• In the data processing execution stage, a qubit’s state is changed by the
application of quantum gates. A compiler converts a logical circuit to a circuit
that can be executed on a physical quantum processor, considering physical
qubit connections and native gate sets.
• In the data extraction stage, all qubit states are read out by measurements.
Due to the probabilistic nature of quantum states, multiple measurements are
usually needed to sample the probability distribution in order to obtain a
meaningful understanding of the solutions.

We next will look at Qiskit. It consists of four different modules: Terra for
writing and running quantum circuits; Aer for high-performance simulation;
Ignis for analyzing and mitigating noise; and Aqua for quantum algorithms and
applications. Using Qiskit and the IQX, we illustrate an example of how a
quantum program is executed on a quantum computer via the cloud. (See
Figure 1.4.)
For this circuit, we don't need to load any data as the qubits are initialize to |0〉 by
default. In the circuit execution stage, the program written in Python first needs to
go through a compiler to convert it to a circuit that can run on a five-qubit quantum
processor, such as the ibmq_vigo quantum processor. The H gate in Figure 1.3 is
converted to a U2 gate that is native to IBM quantum processors. Such processors
are superconducting devices that are usually not connected to all of the other qubits.
However, the circuit written in Qiskit does not take this into account. So, it’s the
compiler’s job to convert the logical circuit in Figure 1.3 to a circuit that can be run
on the actual hardware. (See Figure 1.4 again.)

FIGURE 1.4 An example of a compiled quantum circuit.

Quantum Computing 9

If a CNOT gate is applied to qubits that are not directly connected, the compiler
will need to apply a SWAP gate to move the quantum states of the qubits to two
qubits that are connected. SWAP gates are implemented by using three CNOT
gates, each of which generally has a ten times higher error rate than single qubit
gates have. In the NISQ era, quantum processors do not have QEC and any errors
may make a quantum program fail completely.
In Qiskit, this optimization procedure is implemented by a transpiler (i.e., a
source-to-source compiler or transcompiler) consisting of a series of passes
that optimize circuits based on different algorithms. The transpiled circuit is
converted to a quantum object (Qobj) file and then sent to the IQX system
(McKay et al., 2018). The Qobj file’s contents control the electronics necessary
to convert the program into microwave pulses transmitted to the quantum
processor inside a cryogenic dilution refrigerator, which ensures that a low
temperature is maintained. At the end of a quantum program, all the qubits will
be read out, and the circuits and measurements will have run multiple times to
sample the measurement outcome instead of only capturing a single-point
measure.

1.3 QUANTUM COMPUTING NEEDS AND SERVICE INDUSTRY

APPLICATIONS
We next consider a variety of business needs, and a new perspective on operational,
managerial, and strategic problem-solving and decision-making that will benefit
from the application of the quantum computing approach in the services industries.
In addition to sharing our perspective on the reasons for the upwelling of interest in
the new computational methods, we also provide some characteristics of appro­
priate business applications in operations, and the travel and hospitality services.
We further offer a deeper reading of several prominent problem areas in con­
temporary financial services.

1.3.1 BUSINESS NEEDS AND CONCERNS

Organizations operate in complex and dynamic economies driven by increasing
social connectivity, the free flow of decentralized information, and technological
advances that allow tracking of individuals using a network perspective. To deal
with this increase in complexity and making good decisions (Nijs, 2015), managers
in different kinds of organizations should be encouraged to move from a linear
transaction logic and toward a networked logic that emphasizes value.
From the perspective of a networked logic of value (i.e., business value across
business units, processes, product families, business partners, and smart networks in
global value chains), and to make good decisions effectively, there is a need for
businesses and organizations to capture and understand information from different
sources. Information from the internal environment, such as employee-based or
patent-related data, can help an organization to orchestrate its resources so it can
develop higher-level capabilities and competencies. Likewise, information from the
external environment, such as customer preference data, competitor information, or
10 Data Science and Innovations for Intelligent Systems

macro-environment trends, can aid an organization in decisions related to pricing,

market segmentation, customer targeting, and product positioning. Understanding
how to utilize both internal and external information types is critical for organi­
zations to achieve sustainable advantage, driven by a networked logic of value.
To obtain relevant and meaningful information for decisions, organizations must
consider four areas of managerial concern and related business problems. (See
Table 1.1.)
Relevant information can only be obtained if the data used is accurate and
captures what is being measured in a meaningful way. Data pertaining to real-life
situations, such as customer preferences, competitor actions or employee data are
inherently unstable, complex and dynamic. They often occur as range estimates
(rather than just as point estimates); they also exhibit stochastic variation or are not
easily measured or estimated. They interact with each other in complex probabil­
istic, interdependent and constrained ways (Ménard, Ostolic, Patel, & Volz, 2020).

1. Data and information. It is necessary to use well-specified and reliable data

inputs to obtain good information for decision-making, lest such a decision
process will devolve to GIGO thinking – or “garbage in, garbage out.” This
acronym implies that bad inputs will result in bad output, and this principle
applies more generally to all analysis and logic that cannot support mean­
ingful evidence-based conclusions.
2. Explanatory power and processing. Considerations about respondent bias
(Furnham, 1986) and the explanatory power of the solution methodology used
should also be considered. In addition, the nature of processing is often linked
with the types of data to be analyzed, as well as the computational activities
that are required to achieve a solution for a complex problem.

TABLE 1.1
Concerns and Related Business Problems to Consider
Managerial Concerns Examples of Related Business Problems

Data and information Conversion of data with regard to numerous forms to information/
knowledge; text, sentiments, natural language, images/videos,
map/spatial representations; and firm/competition/market sources
Explanatory power and Power of varied solutions/approaches; robustness/vulnerability to
processing data problems; effectiveness of handling different kinds of
complex problems; trying to match human-related conditions
Infrastructure and capabilities Tech infrastructure; computational/staff support; human limitations
in working on NP-hard vs. soft organizational problems; and
limit to value in firm-level data gathering/problem models
Environment and regulation Business sustainability, fair allocation of resources; appropriate use
of physical/human resources; ensure compliance; and protect
personal info (customers/partners/employees)
Quantum Computing 11

3. Infrastructure and capabilities. Organizations also must make judgment calls

on the type of resources to utilize to develop capabilities and competencies in
information gathering and interpretation. Resource types under consideration
include the organization’s choice of IT systems, its processes, systems, and
staff capabilities.
4. Environmental and regulatory constraints. Since organizations operate within
a larger societal context, they must develop systems within environmental
sustainability and other constraints. Thus, governmental regulations on per­
sonal information protection and privacy laws, financial constraints, and se­
curity considerations are all important.

1.3.2 DECISION PROBLEM FRAMING AND COMPUTATION

Quantum computing is relevant in such situations, not just because of its potential
computing supremacy or its advantage for some kinds of problems, but because of
the way in which it achieves solutions via optimization, simulation, and unique
distillation approaches (Arute et al., 2019). To understand how quantum computing
can achieve supremacy, we should understand its source of power. The power from
quantum computing comes from the qubit, exhibiting the characteristics of a range
estimate and with multiple qubits, it is possible to express additional levels of
correlation and stochasticity, to characterize problems of higher complexity (Arora
& Barak, 2009). To create computational power, computer scientists have devel­
oped processes that leverage the nature of the qubit, through the integration of its
stochastic range and point estimates, in a manner that makes it possible (with ad­
ditional hardware developments) to obtain problem solutions with faster parallelism
and speed. In contrast, classical computers operate in a linear manner, which slows
them down. Both quantum computing and classical computing are competitive on
problems of small to modest size.
There are three implications based on the source of power and differences in
computing methods:

1. Quantum computing, like traditional computing, supports parallelism. A key

contrast is that classical computers use more linear approaches to processes of
computation though.
2. Suitably large quantum machine learning (QML) systems may lead to radical
advances in the creation of thinking machines and AI that approaches the
capabilities of humans.
3. Quantum computers are suitable for problems that are complex, combina­
torial, and stochastic, where judgment in decision making is important. A
wisdom-oriented approach towards thinking, by considering range-estimate
inputs and requiring judgment becomes more relevant when compared to an
intelligence-based approach (Jeste et al., 2010).

With these characteristics in mind, we consider some business problems in different

service industries to which quantum computing seems uniquely suited.
12 Data Science and Innovations for Intelligent Systems

1.3.3 THE RANGE OF BUSINESS PROBLEM AREAS THAT CAN BE ADDRESSED

The business issues to which quantum computing methods apply possess char­
acteristics that often are observed. (See Table 1.2.) We caution the reader to re­
cognize, however, that quantum computing in late 2020 is at the apex of its hype
cycle, but there are not yet “killer applications” or completed commercial tools that
will permit applications of extreme complexity.
First, most problems to which the methods are appropriate involve considerable
computational complexity for traditional computers to handle – especially problems
that are non-deterministic and of non-polynomial time-complexity (NP-hard). In

TABLE 1.2
Problems that Quantum Computing Addresses: Optimization of Complexity,
Stochastic Drivers, Real-Time Computation, Intelligent Simulation, and
Rugged Landscape Analytics
Problem Characteristics Explanation Examples

Complexity in optimization and Solves NP-hard models for shortest- Traveling salesman
simulation. times, shortest-routes, lowest problem (TSP).
costs. Traffic flow optimization.
Airline route scheduling.
Stochastic modeling and solution Leverages quantum (qubit) Product delivery
considerations are prominent. entanglement and probabilistic networks.
superposition of 0/1 bits. Taxi routing in congested
traffic.
Service system control
with measure errors.
Real-time solutions are required Applies quantum computing speed Perishable goods revenue
by business. for app-specific improvements vs. yield management.
Moore’s Law limits. Real-time financial
portfolio risk
management.
Problems conceptualized for Landscapes with uplifting mountains Terrorist network
rugged landscape quantum and settling low points for member identification.
computing. optimization. Voice, speech, and facial
recognition.
Autonomous and
driverless vehicle
routing.
Intelligence needed for obtaining Wisdom-based problem Smart mobility platform
solutions appropriate for representation applied to multi- apps for people.
individuals. factor and soft model choices. Pollution mitigation,
sustainability controls.
Genomic data analytics
for personal care.
Quantum Computing 13

addition, stochastic modeling is required in many problem settings, such as taxi

routing in congested city streets, and the control of complex systems when volatility
and measurement errors are present. Third, an increasingly important problem
characteristic is business solutions that must produce a computational result in real-
time or near real-time in order to support a business process. Intraday revenue yield
pricing for perishable goods (e.g., airline seats, hotel rooms and rental cars) is an
exemplar. Fourth, other problems require consideration of “rugged landscapes,”
which make it hard for ML solutions that do hill-climbing optimization to succeed.
Finally, a fifth class requires machine intelligence to find solutions that are tailored
to individuals, such as healthcare genomics-based treatment decisions. Ideally, what
is required is a more wisdom-based problem representation, so it is possible to find
good solutions in the presence of hard and soft constraints, and objective and
subjective goals.
In the routing optimization domain of NP-hard problems, the well-known tra­
veling salesman problem (TSP) stands out as one to which heuristic methods have
been successful, for example, to address the 1.9 million-city World TSP Tour
(http://www.math.uwaterloo.ca/tsp/world/index.html). More value-laden quantum
computing use cases with strong industrial relevance are related to the airline in­
dustry. Othmani, Ettl, and Guonaris (2020) have identified areas in the progression
of quantum computing since the 1900s development of quantum science, then
quantum-ready proofs-of-concept, and commercial advantage for solving real-world
problem in the industry. The applications they point to include: contextually en­
hanced service personalization; untangling operational disruptions (irregular ops
management) in air travel, that requires dynamic updating and optimization of crew,
take-off slots, and equipment, all while trying to address customer satisfaction and
cost over-run concerns. Quantum computing offers promise for solving the vexing
matter of overcoming the fragmentation of the overall problem. According to the
authors, quantum computing is able to address data and variable range-uncertainty,
while re-aggregating the fragmented optimization sub-problems that will produce a
more efficient and competitive advantage.
In the past 10 years, applied research and corporate practices with data analytics
and problems that can be studied from the large-scale availability of the digital
traces of people as consumers, social media users, users of mobile phones, and the
tracking of people who exercise has given rise to new directions in predictive
modeling and “living analytics” (as with our work at the Living Analytics Research
Centre (LARC, https://larc.smu.edu.sg/) of Singapore Management University). It
stands to reason that new directions in data analytics are now able to support in­
novative work in other areas that are open to innovations with quantum computing
methods. Many business and social problems are naturally modeled in ways that
play to the strengths of the quantum paradigm. Examples are the modeling and
recognition of complex voice, facial expressions and emotions, and other image
recognition problems. Even fast computation by traditional computers has not been
sufficient due to limitations in how such problems have been approached in com­
putational terms.
Another direction lies in creating the computational basis for doing things such
as terrorist network member identification and driverless, autonomous vehicle
14 Data Science and Innovations for Intelligent Systems

routing (Burkacky, Pautasso & Mohr, 2020). These involve the conceptualization of
related problems with rugged landscape optimization (i.e., discontinuous modeling
surfaces), so that the typical methods of hill climbing in optimization and ML are
reduced in their power, especially with dynamic changes in their content and en­
vironment. We note the innovations based on the “changing landscape” analogy for
problem identification and quantum computational methods. Finally, there are many
such problems that require a smart approach, for “wicked” operational, social, and
healthcare problems, like smart mobility platform design for changing urban traffic
and transportation opportunities (Akrout, 2020), and policy analytics for pollution
and sustainability, as well as genomic data analytics.

1.3.4 THE UNIQUE ROLE OF QUANTUM COMPUTING IN FINANCIAL SERVICES

APPLICATIONS
Among the various settings noted above, it’s beneficial to consider problem areas
that can potentially be treated in the financial services space by quantum com­
puting approaches. They include offering enhanced power, flexibility, and
representational authenticity to achieve smart, nearly real-time solutions for
complex optimization, stochastic modeling, and intelligent choice problems.
Financial markets are essentially complex systems that exhibit a high degree of
randomness, for example, in the movement of equities’ and derivatives’ market
prices, as well as foreign exchange rate pairs and interest rates. Their volatilities
have been observed, and numerous applications and problem contexts have been
analyzed in university and industry research. They have used portfolio con­
struction models and optimization, neural networks (NNs), and Monte Carlo si­
mulation, and ML and AI approaches. These further leverage configural patterns
that are hard for human analysts to identify in big data, due to the presence of
stochastic variation and the distributional mechanics of asset returns. In all these
cases, the large size of the data sets used to calibrate changing risk levels in day-
to-day operations typically require a huge amount of computational power for
understanding the necessary relationships in the data so meaningful forecasts can
be made (Lee, 2020).
Quantum computing is attractive for financial applications, like many-asset,
multi-market portfolio construction and risk management controls. It also is re­
levant for establishing intraday trading paths to find ways to successfully exploit
cross-asset or cross-market arbitrage opportunities. Other domains include near
real-time, irrevocable settlement of securities trades, and the voluminous and vo­
latile payments in continuous matching systems that maintain adequate settlement
system participants’ liquidity. These exhibit knowledge about the degree of the
cross-state independence and correlations of asset price realizations from dis­
tributions that are the basis for the evolving lattice of assets prices in a market,
which could utilize the property of quantum entanglement, such that it is present.
Finance practitioners know that a measurement of the state of such systems at
any time is inherently random, though there is underlying cross-correlations, which
result in the cointegrated evolution of asset prices and value-at-risk, a popular loss
Quantum Computing 15

estimate at some level of statistical certainty over time. These states can be de­
scribed in mathematical finance as wave functions, which quantum computing
hardware has the potential to speed up the simulations in comparison to more speed-
limited classical computer hardware and algorithms. Another consideration going
forward is that the extent to which computational speed and power are likely to
become available with lower electricity costs, and beneficial implications for more
sustainable electricity consumption by financial institutions.
An application to securities trading settlement transactions, involving the ex­
change of a delivered financial instrument for an irrevocable cash payment, has
been reported by Barclays Bank and IBM Research (Braine, Egger, Glick, &
Woerner, 2019). The authors utilize quantum algorithms to make a mixed-binary
math programming model’s optimization faster. This is done for value maximiza­
tion of continuously submitted batches of securities trades, with settlement pro­
cessing that considers counter-party credit, collateral facilities, and regulatory
compliance rules as objective function constraints. Traditional computation uses
problem representations that support simulated annealing, which mimics a physical
process of heating some material and lowering its temperature gradually. This re­
duces material defects, and the overall energy of the system. The analogy for fi­
nance is related to how computational algorithms approximate globally optimal
solutions in a complex and constrained system, based on the weighted sum of
settled transactions. The basic math programming formulation is transformed into
an unconstrained model with a lambda-penalty function (like Kuhn-Tucker quad­
ratic optimization) to support a solution with one qubit for each of the securities
transactions to be settled, and a wave function for each possible settlement trans­
action that together form a batch for actual settlement. Quantum computation le­
verages unique features of the stochastic optimization, while reducing the liquidity
risk for a participant that cannot settle its net funds position, and credit risk for
system participants, should there be a fall in the value of traded securities.2
Another finance problem that deserves comment was presented by Woerner &
Egger (2019). It is on quantum risk analytics, in which the quantum amplitude
estimation (AE) algorithm is estimated for an unknown parameter. The problem the
authors applied their approach to is the estimation of the uncertain price evolution
for a U.S. Treasury bill, with a daily trading value of about US$500 bn and ag­
gregate government debt as of 2016 of approximately US$14.8 tr. T-bill prices are
subject to yield curve risk over time (i.e., as a fixed income security, from changing
interest rates across the bills’ maturities). For their analysis, the authors used
constant maturity treasury (CMT) rates, to calculate the daily risk of a one-bill
portfolio toy problem.3 They did this for one-day changes in CMT based on the
distributions of three correlated, underlying explanatory principal components
called shift, twist, and butterfly. They are known to account for 96% of CMT’s daily
variations. Among them, only the first two were retained due to the low correlation
(quantum entanglement) of the third component with them (i.e., due to its in­
dependence and thus lack of suitability for quantum analysis).
Three qubits were used to represent shift uncertainty, and two were used to
represent twist uncertainty. The authors’ approach was intended to deliver a
quadratic speed-boost in comparison to traditional Monte Carlo simulation for
16 Data Science and Innovations for Intelligent Systems

T-bill diffusion in two-asset portfolios, which adds to forecasting problem com­

plexity. They admitted that their sample simulation was not able to be to scaled up
to more qubits for estimation, since they lacked the requisite quantum computing
hardware to demonstrate the power-gain beyond massively-parallel, traditional
computer hardware.4
We next turn to a discussion of a new framework for the implementation of
quantum computing solutions, with special attention given to more technical aspects
than have been discussed so far.

1.4 APPLICATION FRAMEWORK

In this section, we provide a framework describing the areas to consider when
applying quantum computing to a business need. This is based on the assumption
that advantage may be gained from using quantum computing, and a deeper look
into what that entails is required. Taking into account what has been discussed in
Section 1.2 on the basics of quantum computing and in Section 1.3 on current
business needs, we now describe a framework for applying quantum computing to
particular business problems. We will cover the four areas of concern presented in
Figure 1.1 (data, processing, technical, and governance issues) by looking at the
data preparation, algorithm, quantum circuit design, and some issues around in­
tegration. (See Figure 1.5.) These implementation areas are likely to be iterative as
development continues towards the final solution and the state-of-the-art continues
to evolve in all areas of quantum computing.

FIGURE 1.5 The iterative development cycle.

Quantum Computing 17

1.4.1 ALGORITHM DESIGN

Once the business needs have been defined, we must consider designing a suitable
algorithm for the problem(s) involved in the business need. This can be one of the
hardest steps as it involves both a knowledge of the business and a grasp of
quantum algorithms. However, there are a number of algorithms already available
that have been well established and used on a range of problems. Some of the
algorithms can be found in online tutorials with sample code, and there are also
consultancies available to advise potential users further. If there is nothing available
that is suitable, then a new algorithm can be designed. It is most likely that even
well-described algorithms will require modification for the business need. There are
also companies working on embedding quantum algorithms into common data
analytics packages, such as MATLAB, to utilize the power of quantum computing
while hiding the complexity of quantum algorithms (http://horizonquantum.com/).
We next list a number of common problems that could exploit quantum prop­
erties and provide some algorithms examples that are suitable to explore. (See
Table 1.3.)

TABLE 1.3
Problem Type, Quantum Properties, and Sample Algorithms
Problem Type Quantum Property Sources of Sample Algos
Black-box and oracle problems, Interference between qubits Deutsch and Jozsa (1992).
discrete logarithm problems, provides an efficient Bernstein and Vazirani (1997).
integer factorization, boson quantum Fourier transform Quantum phase estimation
sampling problems. operation. ( Simon, 1997).
Shor’s algorithm ( Shor, 1997).
Searching an unstructured data set, Amplitude amplification Grover (1996).
BQP-complete problems, using conversion between
decision making, modelling and probability amplitude and
simulation, linear systems. phase.
Element distinctness, triangle- Quantum walks display Ambainis (2007).
finding, rugged landscape. exponential speed-up.
Graph theory problems, minimize Hybrid quantum/classical Quantum approximation
the energy expectation to find algorithms. optimization algo (QAOA)
the ground state energy. ( Farhi, Goldstone, &
Gutmann, 2014).
Variational quantum
eigensolver VQE) ( Peruzzo
et al., 2014).
ML and optimzation including: Uses quantum superposition. QAOA ( Farhi et al., 2014).
least squares fitting, semidefinite VQE ( Peruzzo et al., 2014).
programming (SDP), NNs, and
combinatorial optimization.
18 Data Science and Innovations for Intelligent Systems

The entire scope of the software that needs to be developed can be broken into
various components: preparing and loading the data with classical code; creating a
quantum circuit using qubits and quantum gates; executing the quantum circuit with
classical code using quantum libraries; and processing the output of the quantum
computer with classical code.

1.4.2 SOFTWARE DEVELOPMENT

With an algorithm decided, we now must develop the classical application that runs
the algorithm with the quantum circuit. The classical code may just run the quantum
circuit once or it may run it many times, and then rebuild the quantum circuit
depending on the output. An example is when a hybrid neural network (NN) uses
classical gradient descent to optimize the quantum weights in a hidden layer, or via
a variational design where the quantum gates are rebuilt after each run.
Data preparation. The data to be input to the algorithm will need to be prepared.
This will depend on the type of data and the type of quantum circuit. Depending on
the quantum circuit, the data will have to be loaded in different ways. Data can be
loaded either as binary data, as qubit rotations, or as quantum states.
The first way, binary data loading, is similar to a classical bit, where each one is
encoded directly onto the pure state of the qubit. A classical 0 is encoded as a |0〉
state and 1 as a |1〉 state. This is simple to understand and, assuming that the default
state of a qubit is |0〉, then the circuit will have either a NOT gate immediately
applied to the input qubit for a |1〉 state or nothing, so the qubit stays in a |0〉 state.
However, this method has the lowest data density of information encoding and is
not utilizing the potential range of probability amplitude and phase amplitude su­
perposition of the qubit.
Encoding classical values directly onto qubit probability rotation angles is
possible using rotation gates, either on the probability amplitude or the phase
amplitude after scaling the values between 0 and π radians. Going beyond π actually
decreases the amplitude, for example 0 = 2π. Further note also that the probabilities
have an angular dependence on the rotation angle. To apply a probability of P to the
qubit will need an angle θ, where θ (radians) = arcos(2P - 1), using an Ry gate for
amplitude probability and an Rx gate for phase probability. This method is simple to
apply and is easy to use with simulators and real backends. However, it still does
not utilize the possibility of applying values to all quantum states.
Classical values can also be applied onto the qubit quantum states. In IQX, this is
implemented using the statevector object and the initialize function of the quantum
circuit object. This can utilize the full availability of states and map complex data
directly onto the qubits. However, there is a need to normalize the statevector data
for real computers before initialization. Some suggestions for how to handle the
different data types include: binary – use binary; integers – scale use rotation or
state; floats – scale and use rotation or state; complex – use state; text – convert to
integer, scale and use rotation or state; image – convert to integer, scale and use
rotation or state; and objects – serialize and convert to integer, and also scale and
use rotation or state.
Quantum Computing 19

Creating the quantum circuit. The code to create a quantum circuit defines the
registers, adds single and multiple qubits gates and measurement instructions.
Registers are of two types: quantum and classical. Quantum registers contain qubits
and classical registers contain classical bits. Classical registers will be used to
collect the results of the quantum circuit execution and may be used to interact with
the quantum circuit during the execution on simulators, such as the IF statement in
Qiskit (Foy, 2019).
Qubits are initialized in the ground |0〉 state. The first operation is the a rotation
gate to encode a classical value or a Hadamard gate to put the qubit into a 50-50
probability superposition. Single- and multi-qubit gates are then used to operate on
qubits affecting the whole state of the quantum register. So, measurements are
performed on the qubits in the quantum registers. The measured qubits may be the
same as the input qubits or they may be ancilla used to collect the final circuit
output.
Executing the quantum circuit. Two architectures are now considered: (1) a static
quantum circuit that is executed once and has all the logic in it, and (2) a variational
approach where the quantum circuit is modified after each run most likely to
converge upon a result.

• Static quantum circuit: The quantum circuit is designed and executed on a

quantum computer. Data are prepared in the classical code and initialized onto
the input qubits of the quantum circuit. The code then defines a quantum
backend which executes the quantum circuit, and then waits until a result is
returned. (See Figure 1.6.) Before attempting to run the circuit, the circuit
depth should be checked to ensure that it does not exceed the coherence time
of the quantum computer, the amount of time a quantum state can continue in

FIGURE 1.6 The combination of classical and quantum code to execute a quantum circuit.
The shaded area is the extra step for variational quantum circuits.
20 Data Science and Innovations for Intelligent Systems

the same stable form. (See the glossary in Appendix Table A1 for additional
details.) It is likely that the high-level quantum circuit gate operations will
need to be decomposed into the native gates of the quantum computer to
understand the actual circuit depth that will be executed on the backend. Also,
it is necessary to check the scaling of the algorithm’s quantum depth with the
quantum width, since algorithms can scale significantly.
• Variational quantum circuits: The entire system couples the quantum and
classical code. The classical code recreates the quantum circuit after each time
it is executed, depending on the results from the quantum circuit. (See
Figure 1.6.)

The execution of a quantum circuit takes milliseconds, but the whole time for the
call to the quantum computer can take time to come back, especially if the circuit
jobs are queued and there is a significant queue depth. In this case, the calling code
needs to have a wait or callback function to continue the processing after the call is
complete.
Processing the output of the quantum computer. When a quantum computer
provides an output, it needs to be retrieved by a classical computer and processed
to be used by downstream systems. Measurement destroys the quantum state on
the qubit and is usually performed at the end of a quantum circuit’s execution.
Measurements are only performed for the probability amplitude on one axis (the
basis) of the Bloch sphere, which on IBM machines is the z-axis. The results are
acquired from executing the quantum circuit a number of times, called shots,
which default to 1,024. Each time a quantum circuit is executed it only returns a
bit (0/1), depending on the probability amplitude for each state in the quantum
register. For example, in Table 1.4 there are two qubits that have four states in
total. The counts are how many times the circuit gave a 1 in that state. These
counts can be calculated to obtain the probability of each state occurring in the
output. (See Table 1.4.)
To find the probability of each qubit being in the |1〉 state sum the probabilities
for that qubit across all the states. In the above example, with the probability of q0
being |1〉, and 0.33 + 0.36 = 0.69, whereas q1 is 0.31. The probability outputs can
then be analyzed by classical code.

TABLE 1.4
The Count and Probability Example for a 2-Qubit Circuit
State Count of 1 Probability

00 160 16%
01 336 33%
10 158 15%
11 370 36%
Quantum Computing 21

1.4.3 HARDWARE FOR QUANTUM COMPUTING

The quantum hardware used is also important to the success of the solution. As well
as the quantity of the input data that needs to be processed, the fidelity, qubit
coupling and coherence time of the qubits are also important. Noise in the qubits
and gates will also need to be considered however, noise may be a useful for some
business cases such as decision making.
Circuit width and number of physical qubits. Inspecting an algorithm will give
a good idea about the quantity of logical qubits required. From the number of
logical qubits plus the number of qubits needed for correcting errors, the total
number of physical qubits can be determined. For example, if the problem ap­
proach is to do optimization, and there are 10 features to optimize on and another
3 qubits are needed for error correction, then at least 13 physical qubits are
needed.
Circuit depth and coupling. The algorithm that has been designed will give an
idea of the circuit depth. But the qubit coupling of the quantum device will also
affect the circuit depth by potentially adding more gates to move information
around. For example, on the IQX, with the ibmq_vigo processor device, there is no
direct coupling between qubit 0 and 4. So, if qubit 0 is entangled with qubit 4,
there has to be more gates (SWAP gates) used to move the quantum information
between the qubits. (See Figure 1.7.) Each gate adds to the circuit depth and time
spent by the circuit that could be used elsewhere. Matching the circuit design to
the coupling can increase performance. As the circuit’s mismatch to the coupling
increases, the number of gates increases and the coherence times can run over
what is desired.
Device adjustments. Finally, on some backends such as IQX, it is possible to
adjust the qubit interactions at the lowest level of the physical interactions. In IQX,
for example, it is possible to adjust the microwave pulse that controls the qubit
rotation. This can also help with reducing errors. Finally, it also is worth noting that

FIGURE 1.7 Coupling map for ibmq_vigo v1.2.1.

22 Data Science and Innovations for Intelligent Systems

qubits can have more than two states. While the fidelity of even the third state is not
sufficient for computation, this property would increase the computing power fur­
ther than now with mn states, where m is the number of states in a qubit and n is the
number of qubits, if a quantum register were possible.

1.4.4 INTEGRATION WITH OTHER IT SYSTEMS IN THE FIRM

Currently, as quantum computing is still being used for research purposes, there is
no integration with companies’ internal IT systems. However, when a significant
quantum advantage has been shown for business processes, then the quantum
computer must be integrated into business flows so its value can be realized. It will
then be important to consider data privacy and proprietary data processing issues. In
light of this, current cloud-based quantum computers need to be assessed on how
they transmit, store and delete data and quantum circuits that are used. Simulations
can likely be performed locally for developing and testing quantum circuits, but to
gain quantum advantage, real quantum computers will need to be used. Though it is
possible to buy a real quantum computer now, the initial and maintenance costs are
known to be quite high.

1.5 CASE: IMPLEMENTING A QUANTUM NEURAL NETWORK

FOR CREDIT RISK
We next return to the quantum computing application framework described earlier,
to illustrate a current QML project that the first two coauthors are currently
working on.

1.5.1 CREDIT RISK ASSESSMENT

The business problem of quantitative credit risk assessment is not new. The first
systematic data-driven approach to this problem dates back at least to 1968 when
Altman (1968) published his pioneering Z-score formula for predicting bankruptcy.
That work used linear discriminant analysis to predict the likelihood of bankruptcy
from four observed financial ratios of a company. Since then, this formula has been
refined many times with thousands of academic papers and hundreds of business
applications introducing new features more sophisticated prediction models using
larger company data sets. The vast majority of these modeling improvements are
still focused on accounting data. This means that the model inputs are changing
very slowly (e.g., once a year for a typical non-listed company). In addition, the
input feature set is quite narrow – with no more than a dozen accounting data fields
that are available for most corporates, and the input features are observed with a
significant delay. These constraints limit the complexity of sensible models and the
use of powerful modern ML approaches would result in heavily over-engineered
solutions.
Our interest in credit scoring comes from the need for risk assessment in trade
finance. Trade financing often exposes lenders to the risks of SMEs, as many of
them operate in emerging markets with little timely and reliable information
Quantum Computing 23

available. This means that classical accounting-based scoring approaches have

limited utility in many cases. However, rapid digitization of trade finance means
that a wealth of new information is available to lenders. The credit decisions can be
made based on the information available at the level of individual transactions.
Many trade finance users are engaged in high-intensity flows of relatively homo­
genous transactions. Many lenders have access to transaction data of dozens or even
hundreds of thousands of borrowers. Moreover, these observed transactions are
often inter-linked, either geographically or along industry verticals. Transaction
data credit scoring makes the credit assessment problem with too few features
available into a problem with a thousand or even millions of feature observations
per day. Tradeteq uses sophisticated graph ML algorithms running on cloud GPU
farms to calibrate and optimize these models.
The ongoing introduction of blockchains and IoT into the supply chains means
that potentially relevant data flows are about to intensify by another three or four
orders of magnitude, with real-time shipment location and conditions tracking
available for more and more goods in transit. Thus, transaction credit scoring will
soon become more taxing for existing computing systems.
Of course, classical computing and ML, in particular, are developing quite
quickly, with rapid progress at all levels of the technological stack – from hardware
to algorithms. However, to be ready for the increasing intensity of the data flows,
we need to explore all available options for future ML systems, including the new
QML systems.
Credit risk assessment problems are similar to certain problems that were already
explored on quantum systems, for example, classification and regression ML and
optimization problems. However, one still needs to ascertain to which extent pe­
culiarities of credit risk problems are well suited for quantum or hybrid quantum
and classical ML architectures. Credit risk ML problems often exhibit strongly
imbalanced data sets with asymmetric noise, where one class is much less frequent
and noisier than others. Supply chain graph topology matters a lot for the risk and
needs to be reflected in a transaction risk model. It remains to be seen how well
quantum systems will be able to handle these problem features though. The current
state of quantum systems and QML algorithms means that they are quite far behind
the capabilities of modern classical ML architectures. Even for relatively simple
company credit scoring, our current models use over 300 features per company. On
current quantum systems with limited number of qubits, it is not possible to process
this number of features. and reaching these levels will take years. Theoretically,
quantum systems may offer large advantages as they can process a very large
number of possibilities simultaneously. In the future it is expected that quantum
systems will reach a scale to be useful for these problems. To be ready for that, we
need to build expertise and community knowledge about QML techniques and their
capabilities.
We hope that quantum computing may become a good fit for these problems
because, as explained earlier, the problems are dealing with high and growing
volumes of complex data. So, model recalibration needs to be done quickly.
Quantum algorithms have already been developed and tested for a number of si­
milar problems, including classification ML, optimization, and graph ML. Our data
24 Data Science and Innovations for Intelligent Systems

and models are already residing in computing clouds, and we have already solved
the accompanying security and access problems. Thus, connectivity to quantum
systems is not going to add massive architectural complexity. Overall, quantum
neural networks (QNNs) alongside classical processing seem to be a good match for
the business needs.

1.5.2 ALGORITHM DESIGN FOR A QUANTUM NEURAL NETWORK (QNN)

The literature for QNNs contains two basic designs, one for a simple 2-qubit per­
ceptron similar to what is offered by Entropica Labs (2020) and another involving a
hybrid quantum/classical model that uses PyTorch and Qiskit (Jupyter Book
Community, 2020).5
2-qubit perceptron. Here, q0 and q1 are the input qubits and q2 is the output qubit
from a Toffoli gate. Ry1 and Ry2 are the input values as rotations, and Ry3 and Ry4
are the neural network weights to be trained. (See Figure 1.8.) The circuit is re­
created with new input and weight rotations and re-run for each data row. The
output of the circuit is compared to the expected 0/1 output (binary in this case for
“defaulted or not”). The best set of weights is the optimum accuracy from this
routine. The main steps of the algorithm are: Data is prepared and loaded; a
quantum circuit is created; the quantum circuit is executed; the output of the
quantum circuit is processed.
The hybrid approach for this routine involved these steps. (See Figure 1.9.) Data is
prepared and loaded; the data is input to the classical first layer; a quantum circuit
is created (see Figure 1.10) based on the output of the first layer; the quantum circuit is
executed; the output from the quantum circuit is used as input to the classical last layer.

FIGURE 1.8 The 2-qubit perceptron quantum circuit.

FIGURE 1.9 Hybrid algorithm using a quantum circuit as a hidden layer in a back-
propagation NN.
Quantum Computing 25

FIGURE 1.10 The quantum circuit used in the hidden layer.

Finally, the output of the last layer is fed back to the input layer adjusted for back-
propagation.
Our approach was to first assess two features of credit data using these simple (1-
or 2-qubit) designs and then to expand them to more features. There are other
algorithms written, including variational approaches (e.g., Tacchino et al., 2019) in
which the states hold weight information, as well as approaches that can also
capture the training accuracy in the phase (e.g., Liao et al., 2019). As the im­
plementation of these models is not easy, it is important to start small and increase
complexity after getting meaningful results on smaller problems.

1.5.3 SOFTWARE DESIGN FOR QNN

With the algorithm decided, the application needed to be coded for data preparation
and loading, executing the quantum circuit, and performing output processing.
Qiskit was chosen as the we already had some experience and access to the IBM Q
machines. First, the sample code was executed to ensure that all worked as ex­
pected, and then the code was modified to use the credit rating data.
Data preparation. Before the data is encoded onto the qubits, it is important to
cap outliers. The cut-off points are determined empirically, and the step is per­
formed classically. This step enables consistent scaling of feature values between 0
and π radians. Additionally, this step ensures that when the model is deployed for
prediction, new data points will not take on values out of the usual range.
The data are encoded as amplitude rotations on the input qubits. These are coded
into the circuit that is re-built at each data row. Rotation was used so far for the
perceptron and hybrid models as it is straightforward and maps relatively easily to
data features. State was not used due to the normalization issue with the scaling
needing to be done across the data set not at each data row. With other models, such
as Liao et al. (2019) with the whole data set applied in one shot, then this would be
the best approach allowing for enormous data inputs and processing. The data set is
going to be huge in the end state, but for now a smaller set of 2,000 rows of data
with 40 features can be prepared.

• For the 2-qubit perceptron (see Figure 1.8), classical data preparation is used
at first. The values for each row of data are put onto a z-rotation of the input
qubits, and two more rotations are made after an entanglement for the neuron
weights. A range of weight values are tested for each row of data, the result is
stored outside the quantum circuit, and the best set of weights chosen as the
final best solution.
26 Data Science and Innovations for Intelligent Systems

• For the hybrid approach, there is 1 qubit in a hidden layer between two
classical layers (Jupyter Book Community, 2020). The output from the first
classical layer is applied as a rotation to the qubit in the hidden layer. The
output of the hidden layer then is used as the input to the second classical
layer.

The quantum circuit’s design can be further modified:

• For the simple perceptron, different single and multiple gates can be selected,
and then the ROC (receiver operating characteristic) and AUC (area under the
curve) metrics can be tested for each design.
• For the hybrid approach, the circuit starts in a simple way with 1 qubit and is
then expanded with multiple qubits and entanglement operations.

Executing the quantum circuit. We next consider how the quantum circuit will
be run.

• For the simple perceptron, the same quantum circuit is executed for each
combination of weight values for each data row and the accuracy of the output
is stored. The most accurate set of weights is then chosen as the best NN
model and the test data set is measured for the ROC and AUC metrics.
• For the hybrid approach, the quantum circuit is embedded with the hidden
layer of a classical NN model. The quantum-related code is coded into a
QuantumClass with Qiskit and a QuantumClassicalClass with PyTorch.

Processing the output. The output of the both models is a single probability of
classification, and a threshold is applied to give a final binary result for the pre­
diction of whether a company will default or not. For the simple perceptron model
the output qubit probability of 1 is used. For the hybrid model the output of the
output classical layer provides the classification output. In both instances, the
outputs are analyzed and visualized using standard statistics packages, such as
Sklearn (sclkit-learn.org).

1.5.4 HARDWARE FOR QUANTUM CREDIT SCORING

For the credit-scoring use case, we expect to need a large number of qubits to be
available to enable good use of a range of features. IQX was chosen given the
availability of machines with a relatively large number of qubits for the quantum
circuits, effective software development with Qiskit, and good support. One issue
with these models is if there is the use of queuing on the backends compared to
having a dedicated time slot. Having a queue means that executing the quantum
circuit multiple times can significantly increase the overall time if there is a long
queue. This is not such a big problem for a proof-of-concept, but would not be
feasible for production.
Coupling, meanwhile, is not a problem for small circuits. But, as the size of the
NN grows in terms of features, weights and entanglement, the circuit will need to
Another Random Scribd Document
with Unrelated Content
 XVIII.
Pinocchio ritrova la Volpe e il Gatto, e va con loro a seminare
le quattro monete nel Campo dei miracoli.

Come potete immaginarvelo, la Fata lasciò che il burattino

piangesse e urlasse una buona mezz'ora a motivo di quel suo naso
che non passava più dalla porta di camera: e lo fece per dargli una
severa lezione e perchè si correggesse dal brutto vizio di dire bugie,
il più brutto vizio che possa avere un ragazzo. Ma quando lo vide
trasfigurato e cogli occhi fuori della testa dalla gran disperazione,
allora, mossa a pietà, battè le mani insieme, e a quel segnale
entrarono in camera dalla finestra un migliaio di grossi uccelli
chiamati Picchi, i quali, posatisi tutti sul naso di Pinocchio,
cominciarono a beccarglielo tanto e poi tanto, che in pochi minuti
quel naso enorme e spropositato si trovò ridotto alla sua grandezza
naturale.
— Quanto siete buona, Fata mia, — disse il burattino,
asciugandosi gli occhi — e quanto bene vi voglio!
— Ti voglio bene anch'io, — rispose la Fata — e se tu vuoi
rimanere con me, tu sarai il mio fratellino e io la tua buona
sorellina....
— Io resterei volentieri.... ma il mio povero babbo?
— Ho pensato a tutto. Il tuo babbo è stato digià avvertito: e
prima che faccia notte, sarà qui.
 — Davvero? — gridò Pinocchio, saltando dall'allegrezza. — Allora,
Fatina mia, se vi contentate, vorrei andargli incontro! Non vedo l'ora
di poter dare un bacio a quel povero vecchio, che ha sofferto tanto
per me!
— Va' pure, ma bada di non ti sperdere. Prendi la via del bosco, e
sono sicura che lo incontrerai. —
Pinocchio partì: e appena entrato nel bosco, cominciò a correre
come un capriòlo. Ma quando fu arrivato a un certo punto, quasi in
faccia alla Quercia grande, si fermò, perchè gli parve di aver sentito
gente fra mezzo alle frasche. Difatti vide apparire sulla strada,
indovinate chi?... la Volpe e il Gatto, ossia i due compagni di viaggio
coi quali aveva cenato all'osteria del Gambero rosso.
— Ecco il nostro caro Pinocchio! — gridò la Volpe, abbracciandolo
e baciandolo. — Come mai sei qui?
— Come mai sei qui? — ripetè il Gatto.
— È una storia lunga — disse il burattino — e ve la racconterò a
comodo. Sappiate però che l'altra notte, quando mi avete lasciato
solo nell'osteria ho trovato gli assassini per la strada....
— Gli assassini?... Oh povero amico! E che cosa volevano?
 — Ecco il nostro caro Pinocchio! — gridò la Volpe, abbracciandolo.

— Mi volevano rubare le monete d'oro.

— Infami!... — disse la Volpe.
— Infamissimi! — ripetè il Gatto.
— Ma io cominciai a scappare — continuò a dire il burattino — e
loro sempre dietro: finchè mi raggiunsero e m'impiccarono a un
ramo di quella quercia.... —
E Pinocchio accennò la Quercia grande, che era lì a due passi.
— Si può sentir di peggio? — disse la Volpe. — In che mondo
siamo condannati a vivere! Dove troveremo un rifugio sicuro noialtri
galantuomini? —
Nel tempo che parlavano così, Pinocchio si accorse che il gatto
era zoppo dalla gamba destra davanti, perchè gli mancava in fondo
tutto lo zampetto cogli unghioli; per cui gli domandò:
— Che cosa hai fatto del tuo zampetto? —
Il gatto voleva rispondere qualche cosa, ma s'imbrogliò. Allora la
Volpe disse subito:
— Il mio amico è troppo modesto, e per questo non risponde.
Risponderò io per lui. Sappi dunque che un'ora fa abbiamo
incontrato sulla strada un vecchio lupo, quasi svenuto dalla fame,
che ci ha chiesto un po' d'elemosina. Non avendo noi da dargli
nemmeno una lisca di pesce, che cosa ha fatto l'amico mio, che ha
davvero un cuore di Cesare? Si è staccato coi denti uno zampetto
delle sue gambe davanti e l'ha gettato a quella povera bestia, perchè
potesse sdigiunarsi. —
E la Volpe, nel dir così si asciugò una lagrima.
Pinocchio, commosso anche lui, si avvicinò al Gatto,
sussurrandogli negli orecchi:
— Se tutti i gatti ti somigliassero, fortunati i topi!
— E ora che cosa fai in questi luoghi? — domandò la Volpe al
burattino.
— Aspetto il mio babbo, che deve arrivare qui di momento in
momento.
— E le tue monete d'oro?
— Le ho sempre in tasca, meno una che la spesi all'osteria del
Gambero rosso.
— E pensare che, invece di quattro monete, potrebbero diventare
domani mille e duemila! Perchè non dài retta al mio consiglio?
Perchè non vai a seminarle nel Campo dei miracoli?
— Oggi è impossibile: vi anderò un altro giorno.
— Un altro giorno sarà tardi! — disse la Volpe.
 — Perchè?
— Perchè quel campo è stato comprato da un gran signore, e da
domani in là non sarà più permesso a nessuno di seminarvi i denari.
— Quant'è distante di qui il Campo dei miracoli?
— Due chilometri appena. Vuoi venire con noi? Fra mezz'ora sei
là: semini subito le quattro monete: dopo pochi minuti ne raccogli
duemila, e stasera ritorni qui con le tasche piene. Vuoi venire con
noi? —
Pinocchio esitò un poco a rispondere, perchè gli tornò in mente la
buona Fata, il vecchio Geppetto e gli avvertimenti del Grillo-parlante;
ma poi finì col fare come fanno tutti i ragazzi senza un fil di giudizio
e senza cuore; finì, cioè, col dare una scrollatina di capo, e disse alla
Volpe e al Gatto:
— Andiamo pure; io vengo con voi. —
E partirono.
Dopo aver camminato una mezza giornata arrivarono a una città
che aveva nome «Acchiappacitrulli.» Appena entrato in città,
Pinocchio vide tutte le strade popolate di cani spelacchiati, che
sbadigliavano dall'appetito, di pecore tosate, che tremavano dal
freddo, e di galline rimaste senza cresta e senza bargigli, che
chiedevano l'elemosina d'un chicco di granturco, di grosse farfalle
che non potevano più volare, perchè avevano venduto le loro
bellissime ali colorite, di pavoni tutti scodati, che si vergognavano a
farsi vedere, e di fagiani che zampettavano cheti cheti, rimpiangendo
le loro scintillanti penne d'oro e d'argento, ormai perdute per
sempre.
In mezzo a questa folla di accattoni e di poveri vergognosi,
passavano di tanto in tanto alcune carrozze signorili con entro o
qualche Volpe, o qualche Gazza ladra, o qualche uccellaccio di
rapina.
— E il Campo dei miracoli dov'è? — domandò Pinocchio.
— È qui a due passi. —

Vide tutte le strade popolate di cani spelacchiati.

Detto fatto traversarono la città, e, usciti fuori delle mura, si

fermarono in un campo solitario che, su per giù, somigliava a tutti gli
altri campi.
— Eccoci giunti; — disse la Volpe al burattino — ora chinati giù a
terra, scava con le mani una piccola buca nel campo, e mettici
dentro le monete d'oro. —
 Pinocchio obbedì. Scavò la buca, ci pose le quattro monete d'oro
che gli erano rimaste: e dopo ricoprì la buca con un po' di terra.
— Ora poi — disse la Volpe — va' alla gora qui vicina, prendi una
secchia d'acqua e annaffia il terreno dove hai seminato. —
Pinocchio andò alla gora, e perchè non aveva lì per lì una
secchia, si levò di piedi una ciabatta, e riempitala d'acqua, annaffiò
la terra che copriva la buca. Poi domandò:
— C'è altro da fare?
— Nient'altro; — rispose la Volpe — ora possiamo andar via. Tu
poi ritorna qui fra una ventina di minuti, e troverai l'arboscello già
spuntato dal suolo e coi rami tutti carichi di monete. —
Il povero burattino, fuori di sè dalla gran contentezza, ringraziò
mille volte la Volpe e il Gatto, e promise loro un bellissimo regalo.
— Noi non vogliamo regali; — risposero que' due malanni — a
noi ci basta di averti insegnato il modo di arricchire senza durar
fatica, e siamo contenti come pasque. —
Ciò detto salutarono Pinocchio, e augurandogli una buona
raccolta, se ne andarono per i fatti loro.
 XIX.
Pinocchio è derubato delle sue monete d'oro, e per gastigo si
busca quattro mesi di prigione.

Il burattino, ritornato in città, cominciò a contare i minuti a uno a

uno: e quando gli parve che fosse l'ora, riprese subito la strada che
menava al Campo dei miracoli.
E mentre camminava con passo frettoloso, il cuore gli batteva
forte e gli faceva tic, tac, tic, tac, come un orologio da sala, quando
corre davvero. E intanto pensava dentro di sè:
— E se invece di mille monete ne trovassi su i rami dell'albero
duemila?... E se invece di duemila, ne trovassi cinquemila? e se
invece di cinquemila, ne trovassi centomila? O che bel signore,
allora, che diventerei!... Vorrei avere un bel palazzo, mille cavallini di
legno e mille scuderie, per potermi baloccare, una cantina di rosolii e
di alchermes, e una libreria tutta piena di canditi, di torte, di
panettoni, di mandorlati e di cialdoni colla panna. —
 Tirò fuori una mano di tasca e si dette una lunghissima grattatina di capo.

Così fantasticando, giunse in vicinanza del campo, e lì si fermò a

guardare se per caso avesse potuto scorgere qualche albero coi rami
carichi di monete: ma non vide nulla. Fece altri cento passi in avanti,
e nulla; entrò sul campo.... andò proprio su quella piccola buca,
dove aveva sotterrato i suoi zecchini, e nulla. Allora diventò
pensieroso, e, dimenticando le regole del Galateo e della buona
creanza, tirò fuori una mano di tasca e si dette una lunghissima
grattatina di capo.
 In quel mentre sentì fischiarsi negli orecchi una gran risata: e
voltandosi in su, vide sopra un albero un grosso pappagallo, che si
spollinava le poche penne che aveva addosso.
— Perchè ridi? — gli domandò Pinocchio con voce di bizza.
— Rido, perchè nello spollinarmi mi son fatto il solletico sotto le
ali. —
Il burattino non rispose. Andò alla gora e riempita d'acqua la
solita ciabatta, si pose nuovamente ad annaffiare la terra, che
ricopriva le monete d'oro.
Quand'ecco che un'altra risata, anche più impertinente della
prima, si fece sentire nella solitudine silenziosa di quel campo.
— Insomma, — gridò Pinocchio arrabbiandosi — si può sapere,
Pappagallo mal educato, di che cosa ridi?
— Rido di quei barbagianni, che credono a tutte le scioccherie e
che si lasciano trappolare da chi è più furbo di loro.
— Parli forse di me?
— Sì, parlo di te, povero Pinocchio: di te che sei così dolce di
sale, da credere che i denari si possano seminare e raccogliere nei
campi, come si seminano i fagiuoli e le zucche. Anch'io l'ho creduto
una volta, e oggi ne porto le pene. Oggi (ma troppo tardi!) mi son
dovuto persuadere che per mettere insieme onestamente pochi soldi
bisogna saperseli guadagnare o col lavoro delle proprie mani o
coll'ingegno della propria testa.
— Non ti capisco — disse il burattino, che già cominciava a
tremare dalla paura.
— Pazienza! Mi spiegherò meglio — soggiunse il Pappagallo. —
Sappi dunque che, mentre tu eri in città, la Volpe e il Gatto sono
tornati in questo campo: hanno preso le monete d'oro sotterrate, e
poi sono fuggiti come il vento. E ora, chi li raggiunge è bravo! —
Pinocchio restò a bocca aperta, e non volendo credere alle parole
del Pappagallo, cominciò colle mani e colle unghie a scavare il
terreno che aveva annaffiato. E scava, scava, scava, fece una buca
così profonda, che ci sarebbe entrato per ritto un pagliaio: ma le
monete non c'erano più.
Preso allora dalla disperazione, tornò di corsa in città e andò
difilato in tribunale, per denunziare al giudice i due malandrini, che
lo avevano derubato.
Il giudice era uno scimmione della razza dei Gorilla: un vecchio
scimmione rispettabile per la sua grave età, per la sua barba bianca
e specialmente per i suoi occhiali d'oro, senza vetri, che era costretto
a portare continuamente, a motivo d'una flussione d'occhi, che lo
tormentava da parecchi anni.
 Pinocchio, alla presenza del giudice raccontò per filo e per segno l'iniqua frode.

Pinocchio, alla presenza del giudice, raccontò per filo e per segno
l'iniqua frode, di cui era stato vittima; dette il nome, il cognome e i
connotati dei malandrini, e finì chiedendo giustizia.
Il giudice lo ascoltò con molta benignità; prese vivissima parte al
racconto: s'intenerì, si commosse: e quando il burattino non ebbe
più nulla da dire, allungò la mano e sonò il campanello.
A quella scampanellata comparvero subito due can mastini vestiti
da giandarmi.
Allora il giudice, accennando Pinocchio ai giandarmi, disse loro:
— Quel povero diavolo è stato derubato di quattro monete d'oro:
pigliatelo dunque, e mettetelo subito in prigione. —
 Il burattino, sentendosi dare questa sentenza fra capo e collo,
rimase di princisbecco e voleva protestare: ma i giandarmi, a scanso
di perditempi inutili, gli tapparono la bocca e lo condussero in
gattabuia.
E lì v'ebbe a rimanere quattro mesi: quattro lunghissimi mesi: e
vi sarebbe rimasto anche di più, se non si fosse dato un caso
fortunatissimo. Perchè bisogna sapere che il giovane Imperatore che
regnava nella città di Acchiappacitrulli, avendo riportato una bella
vittoria contro i suoi nemici, ordinò grandi feste pubbliche, luminarie,
fuochi artificiali, corse di barberi e di velocipedi, e in segno di
maggiore esultanza, volle che fossero aperte anche le carceri e
mandati fuori tutti i malandrini.
— Se escono di prigione gli altri, voglio uscire anch'io — disse
Pinocchio al carceriere.
 — Quel povero diavolo è stato derubato di quattro monete d'oro: pigliatelo
dunque, e mettetelo subito in prigione.

— Voi no, — rispose il carceriere — perchè voi non siete del bel
numero....
— Domando scusa; — replicò Pinocchio — sono un malandrino
anch'io.

Gli tapparono la bocca e lo condussero in gattabuia.

— In questo caso avete mille ragioni, — disse il carceriere; e

levandosi il berretto rispettosamente e salutandolo, gli aprì le porte
della prigione e lo lasciò scappare.
 XX.
Liberato dalla prigione, si avvia per tornare a casa della Fata;
ma lungo la strada trova un serpente orribile, e poi rimane
preso alla tagliuola.

Figuratevi l'allegrezza di Pinocchio quando si sentì libero. Senza

stare a dire che è e che non è, uscì subito fuori della città e riprese
la strada che doveva ricondurlo alla Casina della Fata.
A cagione del tempo piovigginoso, la strada era diventata tutta
un pantano e ci si andava fino a mezza gamba.
Ma il burattino non se ne dava per inteso.
Tormentato dalla passione di rivedere il suo babbo e la sua
sorellina dai capelli turchini, correva a salti come un can levriero, e
nel correre le pillacchere gli schizzavano fin sopra il berretto. Intanto
andava dicendo fra sè e sè:
— Quante disgrazie mi sono accadute... E me le merito! perchè io
sono un burattino testardo e piccoso..., voglio far sempre tutte le
cose a modo mio, senza dar retta a quelli che mi voglion bene e che
hanno mille volte più giudizio di me!... Ma da questa volta in là,
faccio proponimento di cambiar vita e di diventare un ragazzo
ammodo e ubbidiente!... Tanto ormai ho bell'e visto che i ragazzi, a
essere disubbidienti, ci scapitano sempre e non ne infilano mai una
per il su' verso. E il mio babbo mi avrà aspettato?... Ce lo troverò a
casa della Fata? È tanto tempo, pover'uomo, che non lo vedo più, e
che mi struggo di fargli mille carezze e di finirlo dai baci! E la Fata mi
perdonerà la brutta azione che le ho fatta?... E pensare che ho
ricevuto da lei tante attenzioni e tante cure amorose.... e pensare
che se oggi son sempre vivo, lo debbo a lei! Ma si può dare un
ragazzo più ingrato e più senza cuore di me?... —
Nel tempo che diceva così, si fermò tutt'a un tratto spaventato, e
fece quattro passi indietro.
Che cosa aveva veduto?
Aveva veduto un grosso serpente, disteso attraverso alla strada,
che aveva la pelle verde, gli occhi di fuoco e la coda appuntata che
gli fumava come una cappa di camino.
Impossibile immaginarsi la paura del burattino: il quale,
allontanandosi più di mezzo chilometro, si mise a sedere sopra un
monticello di sassi, aspettando che il serpente se ne andasse una
buona volta per i fatti suoi e lasciasse libero il passo della strada.
Aspettò un'ora; due ore: tre ore: ma il serpente era sempre là, e
anche di lontano, si vedeva il rosseggiare de' suoi occhi di fuoco e la
colonna di fumo che gli usciva dalla punta della coda.
Allora Pinocchio, figurandosi di aver coraggio, si avvicinò a pochi
passi di distanza, e facendo una vocina dolce, insinuante e sottile,
disse al serpente:
— Scusi, signor Serpente, che mi farebbe il piacere di tirarsi un
pochino da una parte, tanto da lasciarmi passare? —
Fu lo stesso che dire al muro. Nessuno si mosse.
Allora riprese colla solita vocina:
— Deve sapere, signor Serpente, che io vado a casa, dove c'è il
mio babbo che mi aspetta e che è tanto tempo che non lo vedo
più!... Si contenta dunque, che io seguiti per la mia strada? —
 Aspettò un segno di risposta a quella domanda: ma la risposta
non venne: anzi il serpente, che fin allora pareva arzillo e pieno di
vita, diventò immobile e quasi irrigidito. Gli occhi gli si chiusero, e la
coda gli smesse di fumare.
— Che sia morto davvero? — disse Pinocchio, dandosi una
fregatina di mani dalla gran contentezza; e senza mettere tempo in
mezzo, fece l'atto di scavalcarlo, per passare dall'altra parte della
strada. Ma non aveva ancora finito di alzare la gamba, che il
serpente si rizzò all'improvviso come una molla scattata: e il
burattino, nel tirarsi indietro spaventato, inciampò e cadde per terra.

Cadde così male, che restò col capo conficcato nel fango della strada....
 E per l'appunto cadde così male, che restò col capo conficcato nel
fango della strada e colle gambe ritte su in aria.
Alla vista di quel burattino, che sgambettava a capofitto con una
velocità incredibile, il serpente fu preso da una tal convulsione di risa
che ridi, ridi, ridi, alla fine, dallo sforzo del troppo ridere, gli si
strappò una vena sul petto: e quella volta morì davvero.
Allora Pinocchio ricominciò a correre per arrivare a casa della
Fata avanti che si facesse buio. Ma lungo la strada, non potendo più
reggere ai morsi terribili della fame, saltò in un campo coll'intenzione
di cogliere poche ciocche d'uva moscadella. Non l'avesse mai fatto!
Appena giunto sotto la vite, crac.... sentì stringersi le gambe da
due ferri taglienti, che gli fecero vedere quante stelle c'erano in cielo.
Il povero burattino era rimasto preso a una tagliuola appostata là
da alcuni contadini per beccarvi alcune grosse faine, che erano il
flagello di tutti i polli del vicinato.
 XXI.
Pinocchio è preso da un contadino, il quale lo costringe a far
da can di guardia a un pollaio.

Pinocchio, come potete figurarvelo, si dètte a piangere, a

strillare, a raccomandarsi: ma erano pianti e grida inutili, perchè lì
all'intorno non si vedevano case, e dalla strada non passava anima
viva.
Intanto si fece notte.
Un po' per lo spasimo della tagliuola che gli segava gli stinchi, e
un po' per la paura di trovarsi solo e al buio in mezzo a quei campi, il
burattino principiava quasi a svenirsi: quando a un tratto, vedendosi
passare una lucciola di sul capo, la chiamò e le disse:
— O lucciolina, mi faresti la carità di liberarmi da questo
supplizio?...
— Povero figliuolo! — replicò la lucciola, fermandosi impietosita a
guardarlo. — Come mai sei rimasto colle gambe attanagliate fra
cotesti ferri arrotati?
— Sono entrato nel campo per cogliere due grappoli di quest'uva
moscadella, e....
— Ma l'uva era tua?
— No....
— E allora chi t'ha insegnato a portar via la roba degli altri?...
— Avevo fame....
 — La fame, ragazzo mio, non è una buona ragione per potersi
appropriare la roba che non è nostra...
— È vero, è vero! — gridò Pinocchio piangendo — ma un'altra
volta non lo farò più. —
A questo punto il dialogo fu interrotto da un piccolissimo rumore
di passi, che si avvicinavano. Era il padrone del campo che veniva in
punta di piedi a vedere se qualcuna di quelle faine, che gli
mangiavano di nottetempo i polli, fosse rimasta presa al trabocchetto
della tagliuola.
E la sua maraviglia fu grandissima quando, tirata fuori la lanterna
di sotto al pastrano, s'accòrse che, invece di una faina, c'era rimasto
preso un ragazzo.
— Ah, ladracchiòlo! — disse il contadino incollerito — dunque sei
tu che mi porti via le galline?
— Io no, io no! — gridò Pinocchio, singhiozzando. — Io sono
entrato nel campo per prendere soltanto due grappoli d'uva!
— Chi ruba l'uva è capacissimo di rubare anche i polli. Lascia fare
a me, che ti darò una lezione da ricordartene per un pezzo. —
 — Ah, ladracchiòlo! dunque sei tu che mi porti via le galline?

E aperta la tagliuola, afferrò il burattino per la collottola e lo

portò di peso fino a casa, come si porterebbe un agnellino di latte.
Arrivato che fu sull'aia dinanzi alla casa, lo scaraventò in terra: e
tenendogli un piede sul collo, gli disse:
— Oramai è tardi e voglio andare a letto. I nostri conti gli
aggiusteremo domani. Intanto, siccome oggi m'è morto il cane che
mi faceva la guardia di notte, tu prenderai subito il suo posto. Tu mi
farai da cane di guardia. —
 — Tu puoi andare a cuccia in quel casotto di legno.

Detto fatto, gl'infilò al collo un grosso collare tutto coperto di

spunzoni d'ottone, e glielo strinse in modo, da non poterselo levare
passandoci la testa di dentro. Al collare c'era attaccata una lunga
catenella di ferro: e la catenella era fissata nel muro.
— Se questa notte — disse il contadino — cominciasse a piovere,
tu puoi andare a cuccia in quel casotto di legno, dove c'è sempre la
paglia che ha servito di letto per quattro anni al mio povero cane. E
se per disgrazia venissero i ladri, ricordati di stare a orecchi ritti e di
abbaiare. —
Dopo quest'ultimo avvertimento, il contadino entrò in casa
chiudendo la porta con tanto di catenaccio: e il povero Pinocchio
rimase accovacciato sull'aia più morto che vivo, a motivo del freddo,
della fame e della paura. E di tanto in tanto cacciandosi
rabbiosamente le mani dentro il collare, che gli serrava la gola,
diceva piangendo:
— Mi sta bene!... Pur troppo mi sta bene! Ho voluto fare lo
svogliato, il vagabondo.... ho voluto dar retta ai cattivi compagni, e
per questo la fortuna mi perseguita sempre. Se fossi stato un
ragazzino per bene, come ce n'è tanti; se avessi avuto voglia di
studiare e di lavorare, se fossi rimasto in casa col mio povero babbo,
a quest'ora non mi troverei qui, in mezzo ai campi, a fare il cane di
guardia alla casa di un contadino. Oh se potessi rinascere un'altra
volta!... Ma oramai è tardi e ci vuol pazienza! —
Fatto questo piccolo sfogo, che gli venne proprio dal cuore, entrò
dentro il casotto e si addormentò.
 XXII.
Pinocchio scuopre i ladri, e in ricompensa di essere stato
fedele vien posto in libertà.

Ed era già più di due ore che dormiva saporitamente, quando

verso la mezzanotte fu svegliato da un bisbiglio e da un pissi—pissi
di vocine strane, che gli parve di sentir nell'aia. Messa fuori la punta
del naso dalla buca del casotto, vide riunite a consiglio quattro
bestiole di pelame scuro che parevano gatti. Ma non erano gatti:
erano faine, animaletti carnivori, ghiottissimi d'uova e di pollastrine
giovani. Una di queste faine, staccandosi dalle sue compagne, andò
alla buca del casotto, e disse sottovoce:
— Buona sera, Melampo.
— Io non mi chiamo Melampo — rispose il burattino.
— O dunque chi sei?
— Io sono Pinocchio.
— E che cosa fai costì?
— Faccio il cane di guardia.
— O Melampo dov'è? dov'è il vecchio cane, che stava in questo
casotto?
— È morto questa mattina.
— Morto? povera bestia!... Era tanto buono!... Ma giudicandoti
dalla fisonomia, anche te mi sembri un cane di garbo.
— Domando scusa, io non sono un cane!...
 — O chi sei?

Una di queste faine, staccandosi dalle sue compagne andò alla buca del casotto.

— Io sono un burattino.
— E fai da cane di guardia?
— Pur troppo: per mia punizione!...
— Ebbene, io ti propongo gli stessi patti che avevo col defunto
Melampo, e sarai contento.
— E questi patti sarebbero?
— Noi verremo una volta la settimana, come per il passato, a
visitare di notte questo pollaio e porteremo via otto galline. Di
queste galline, sette le mangeremo noi, e una la daremo a te, a
condizione, s'intende bene, che tu faccia finta di dormire e non ti
venga mai l'estro di abbaiare e di svegliare il contadino.
— E Melampo faceva proprio così? — domandò Pinocchio.
— Faceva così, e fra noi e lui, siamo andati sempre d'accordo.
Dormi dunque tranquillamente, e stai sicuro che prima di partire di
qui ti lasceremo sul casotto una gallina bell'e pelata per la colazione
di domani. Ci siamo intesi bene?
— Anche troppo bene!... — rispose Pinocchio: e tentennò il capo
in un certo modo minaccioso, come se avesse voluto dire: — Fra
poco ci riparleremo!... —
Quando le quattro faine si credettero sicure del fatto loro,
andarono difilato al pollaio, che rimaneva appunto vicinissimo al
casotto del cane; e aperta a furia di denti e di unghioli la porticina di
legno, che ne chiudeva l'entrata, vi sgusciarono dentro, una dopo
l'altra. Ma non erano ancora finite d'entrare, che sentirono la
porticina richiudersi con grandissima violenza.
Quello che l'aveva richiusa era Pinocchio; il quale, non contento
di averla richiusa, vi passò davanti per maggior sicurezza una grossa
pietra, a guisa di puntello.
E poi cominciò ad abbaiare: e, abbaiando proprio come se fosse
un cane di guardia, faceva colla voce: bù—bù-bù-bù.
A quella abbaiata, il contadino saltò il letto, e preso il fucile e
affacciatosi alla finestra, domandò:
— Che c'è di nuovo?
— Ci sono i ladri! — rispose Pinocchio.
— Dove sono?
— Nel pollaio.
— Ora scendo subito. —
E difatti, in men che si dice amen, il contadino scese: entrò di
corsa nel pollaio, e dopo avere acchiappate e rinchiuse in un sacco le
quattro faine, disse loro con accento di vera contentezza:
— Alla fine siete cascate nelle mie mani! Potrei punirvi, ma sì vil
non sono! Mi contenterò, invece, di portarvi domani all'oste del
vicino paese, il quale vi spellerà e vi cucinerà a uso lepre dolce e
forte. È un onore che non vi meritate, ma gli uomini generosi come
me non badano a queste piccolezze!... —
Quindi, avvicinatosi a Pinocchio, cominciò a fargli molte carezze,
e fra le altre cose, gli domandò:
— Com'hai fatto a scoprire il complotto di queste quattro
ladroncelle? E dire che Melampo, il mio fido Melampo, non s'era mai
accorto di nulla!... —
Il burattino, allora, avrebbe potuto raccontare quel che sapeva;
avrebbe potuto, cioè, raccontare i patti vergognosi che passavano tra
il cane e le faine; ma ricordandosi che il cane era morto, pensò
subito dentro di sè: — A che serve accusare i morti?... I morti son
morti, e la miglior cosa che si possa fare è quella di lasciarli in
pace!...
— All'arrivo delle faine sull'aia, eri sveglio o dormivi? — continuò
a chiedergli il contadino.
— Dormivo: — rispose Pinocchio — ma le faine mi hanno
svegliato coi loro chiacchiericci, e una è venuta fin qui al casotto per
dirmi: «Se prometti di non abbaiare, e di non svegliare il padrone,
noi ti regaleremo una pollastra bell'e pelata!» Capite, eh? Avere la
sfacciataggine di fare a me una simile proposta! Perchè bisogna
sapere che io sono un burattino, che avrò tutti i difetti di questo
mondo: ma non avrò mai quello di star di balla e di reggere il sacco
alla gente disonesta!
— Bravo ragazzo! — gridò il contadino, battendogli sur una
spalla. — Cotesti sentimenti ti fanno onore: e per provarti la mia
grande soddisfazione, ti lascio libero fin d'ora di tornare a casa. —
E gli levò il collare da cane.
 XXIII.
Pinocchio piange la morte della bella Bambina dai capelli
turchini: poi trova un Colombo, che lo porta sulla riva del
mare, e lì si getta nell'acqua per andare in aiuto del suo
babbo Geppetto.

Appena Pinocchio non sentì più il peso durissimo e umiliante di

quel collare intorno al collo, si pose a scappare attraverso ai campi, e
non si fermò un solo minuto finchè non ebbe raggiunta la strada
maestra, che doveva ricondurlo alla casina della Fata.
Arrivato sulla strada maestra, si voltò in giù a guardare nella
sottoposta pianura, e vide benissimo, a occhio nudo, il bosco dove
disgraziatamente aveva incontrato la Volpe e il Gatto: vide, fra
mezzo agli alberi, inalzarsi la cima di quella Quercia grande, alla
quale era stato appeso ciondoloni per il collo; ma, guarda di qui,
guarda di là, non gli fu possibile di vedere la piccola casa della bella
Bambina dai capelli turchini.
Allora ebbe una specie di tristo presentimento; e datosi a correre
con quanta forza gli rimaneva nelle gambe, si trovò in pochi minuti
sul prato, dove sorgeva una volta la Casina bianca. Ma la Casina
bianca non c'era più. C'era, invece, una piccola pietra di marmo,
sulla quale si leggevano in carattere stampatello queste dolorose
parole:
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

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

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com