Linear Programming Computation 2nd Edition Ping-

Qi Pan install download

https://ebookmeta.com/product/linear-programming-computation-2nd-
edition-ping-qi-pan/

Download more ebook from https://ebookmeta.com

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

Advances in Optimization and Linear Programming 1st

Edition Ivan Stanimirovi■

https://ebookmeta.com/product/advances-in-optimization-and-
linear-programming-1st-edition-ivan-stanimirovic/

Linear Programming Foundations and Extensions 5th

Edition Robert J. Vanderbei

https://ebookmeta.com/product/linear-programming-foundations-and-
extensions-5th-edition-robert-j-vanderbei/

Linear Programming for Project Management Professionals

1st Edition Partha Majumdar

https://ebookmeta.com/product/linear-programming-for-project-
management-professionals-1st-edition-partha-majumdar/

Bioinformatics and Computational Biology 1st Edition

Hamid R. Arabnia

https://ebookmeta.com/product/bioinformatics-and-computational-
biology-1st-edition-hamid-r-arabnia/
Advances in Postharvest Management of Cereals and
Grains 1st Edition Dirk E. Maier (Editor)

https://ebookmeta.com/product/advances-in-postharvest-management-
of-cereals-and-grains-1st-edition-dirk-e-maier-editor/

Abdication of the Sovereign Self The Psycholinguistics

of Invalid Synthetic Propositions 1st Edition Andrew
Spano

https://ebookmeta.com/product/abdication-of-the-sovereign-self-
the-psycholinguistics-of-invalid-synthetic-propositions-1st-
edition-andrew-spano/

Hybrid Forgotten (Hybrid Shifters Book 10) 1st Edition

Romy Lockhart

https://ebookmeta.com/product/hybrid-forgotten-hybrid-shifters-
book-10-1st-edition-romy-lockhart-2/

Lock the Door Jane Holland

https://ebookmeta.com/product/lock-the-door-jane-holland/

Antimicrobial Photodynamic Therapy 1st Edition Busi

Siddhardha Ram Prasad

https://ebookmeta.com/product/antimicrobial-photodynamic-
therapy-1st-edition-busi-siddhardha-ram-prasad/
Introduction to Positive Media Psychology 1st Edition
Arthur A Raney Sophie H Janicke Bowles Mary Beth Oliver
Katherine R Dale

https://ebookmeta.com/product/introduction-to-positive-media-
psychology-1st-edition-arthur-a-raney-sophie-h-janicke-bowles-
mary-beth-oliver-katherine-r-dale/
Ping-Qi PAN

Linear
Programming
Computation
Second Edition
Linear Programming Computation
Ping-Qi PAN

Linear Programming
Computation

Second Edition
Ping-Qi PAN
Department of Mathematics
Southeast University
Nanjing, China

ISBN 978-981-19-0146-1 ISBN 978-981-19-0147-8 (eBook)

https://doi.org/10.1007/978-981-19-0147-8

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore
Pte Ltd. 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Dedicated to my parents, Chao Pan and
Yinyun Yan; my wife, Minghua Jiang; my son,
Yunpeng; my granddaughter, Heidi
Preface to First Edition

Linear programming (LP) founded by Dantzig (1948–1951b) might be one of the

most well-known and widely used mathematical tools in the world. As a branch
of optimization, it serves as the most important cornerstone of operations research,
decision science, and management science.
This branch emerged when the American mathematician George B. Dantzig
created the LP model and the simplex method in 1947. The computer, emerging
around the same period, propelled the development of LP and the simplex method
toward practical applications. As a basic branch, LP orchestrated the birth of
a number of new branches, such as nonlinear programming, network flow and
combinatorial optimization, stochastic programming, integer programming, and
complementary theory, etc., and activated the whole field of operations research.
A prominent feature of LP is given by its broad applications. Closely related
to LP, many individuals have made pioneering contributions in their respective
fields. In the field of economics, in particular, Russian-American economist Wassily
Leontief took the 1973 Nobel Economic Prize for his epoch-making contribution
to quantitative analysis of economic activities. The academician L. V. Kantorovich
of the former Soviet Academy of Science and American economist Professor T. C.
Koopmans won the 1975 Nobel Prize for their optimal allocation theory of resources
using LP. The same prize was also given to Professors K. Arrow, P. Samuelson, H.
Simon, and L. Herwricz, several decades later when they paid close attention to LP
at the starting days of their professional careers.
The simplex method has also achieved great success in practice. As it is known,
its applications to many fields, such as economy, commerce, production, science and
technology, and defense and military affairs, etc, have brought about astonishing
economic and social benefits. It is recognized as one of The Ten Algorithms in the
Twenty Century (IEEE2002; see Cipra (2000).
After more than 70 years since its birth, LP is now a relatively mature but still
developing branch. Nevertheless, there exist great challenges. The importance of
large-scale sparse LP models is nowadays enhanced further by globalization. The
everyday practice calls upon the research community to provide more powerful
solution tools just to keep up with the ever-increasing problem sizes. Therefore, this

vii
viii Preface to First Edition

book does not only present fundamental materials but also attempts to reflect the
state of the art of LP. It has been my long-lasting belief that research, in operations
research/management science, in particular, should be of practical value, at least
potentially. The author, therefore, focuses on theories, methods, and implementation
techniques that are closely related to LP computation.
This book consists of two parts. Part I mainly covers fundamental and conven-
tional materials, such as the geometry of the feasible region, the simplex method,
duality principle, and dual simplex method, implementation of the simplex method,
sensitivity analysis, parametric LP, variants of the simplex method, decomposition
method, and interior-point method. In addition, integer linear programming (ILP),
differing from LP in nature, is also discussed not only because ILP models can be
handled by solving a sequence of LP models but also because they are so rich in
practice and form a major application area of LP computations.
Part II presents published or unpublished new results achieved by the author
himself, such as pivot rule, dual pivot rule, simplex phase-1 method, dual simplex
phase-I method, reduced simplex method, generalized reduced simplex method,
deficient-basis method, dual deficient-basis method, face method, dual face method,
and pivotal interior-point method. The last chapter contains miscellaneous topics,
such as some special forms of the LP problem, approaches to intercepting for primal
and dual optimal sets, practical pricing schemes, relaxation principle, local duality,
decomposition principle, and ILP method based on the generalized reduced simplex
framework.
To make the material easier to follow and understand, the algorithms in this book
were formulated and illustrative examples were worked out wherever possible. If
the book is used as a textbook for upper-level undergraduate/graduate course, (Part
I) may be suitable to be as basic course material.

Acknowledgments

At this very moment, I deeply cherish the memory of Professor Xuchu He, my
former mentor at Nanjing University, aroused my interest in optimization. I honor
the Father of linear programming and the simplex method, Professor George. B.
Dantzig for the encouragement given during the 16th International Symposium on
Mathematical Programming, Lausanne EPFL., in August 1997. I am very grateful
to Professor Michael Saunders of Stanford University for his thoughtful comments
and suggestions given in the past. He selflessly offered the MINOS 5.51 package
(the latest version of MINOS at that time) and related materials, from which my
work and this book benefited greatly. MINOS has been the benchmark and platform
of our computational experiments. I thank Professor R. Tyrrell Rockafellar of the
University of Washington, Professor Thomas F. Coleman of Cornell University, and
Professor Weide Zheng of Nanjing University for their support and assistance. I
would also like to thank the following colleagues for their support and assistance:
Professor James V. Burke of the University of Washington, Professor Liqun
Preface to First Edition ix

Qi of Hong Kong Polytechnic University, Professor Lizhi Liao of Hong Kong

Baptist University, Professors Leena Suhl and Dr. Achim Koberstein of Paderborn
University, Professor Uwe Suhl of Freie University, and Dr. Pablo Guerrero-Garcia
of the University of Malaga.
This book was supported partially by projects 10871043 and 70971136 of the
National Natural Science Foundation of China.

Nanjing, China Ping-Qi Pan

December 2012
Preface to Second Edition

Since its birth, linear programming (LP) has achieved great success in many fields,
such as economy, commerce, production, science, and technology, and brought
about amazing economic and social benefits. After 70 years of development, LP
now becomes a relatively mature branch in OR/MS. Nevertheless, the academic
community faces a major challenge of the growing demand, which needs more
powerful and robust tools to deal with large-scale and difficult problems.
To meet the challenge, this book draws materials from a practical point of
view, focusing on theories, methods, and implementation techniques that are
closely related to LP computation. Its first edition consists of two parts. Roughly
speaking, Part I covers fundamental materials of LP, and Part II includes pub-
lished/unpublished new results achieved by the author.
Little attention was received via the author’s ResearchGate Web page until Shi,
Zhang, and Zhu published their book review in the European Journal of Operational
Research (June 2018). Since then, the book quickly attracted considerable attention
from academic communities around the world. As of November 6, 2022, the top ten
chapter reads are as follows:

Chapter Reads
1. Duality Principle and Dual Simplex Method (Part I) 15419
2. Simplex Feasible-Pointx Method (Part II) 7861
3. Integer Linear Programming (ILP) (Part I) 7340
4. Implementation of the Simplex Method (Part I) 3284
5. Dual Simplex Phase-I Method (Part II) 2613
6. Simplex Method (Part I) 2576
7. Variants of the Simplex Method (Part I) 2328
8. Geometry of the Feasible Region (Part I) 1951
9. Pivot Rule (Part II) 1844
10. Simplex Phase-I Method (Part I) 1506

xi
xii Preface to Second Edition

The preceding data are consistent with corresponding downloads from the
Springer website.
Surprisingly enough, the chapter ”Dual Principle and Dual Simplex Method” not
only received the maximum number of reads as a whole but also almost weekly,
even if it is devoted to such a classical topic. This might be because the discussions
are comprehensive, compared with popular literature.
Proposed for the first time, the Feasible-Point Simplex Method, which appeared
as the final section of the chapter ”Pivotal Interior-Point Method,” received the
second-highest reads as a whole. It achieved this in a short period of 10 weeks.
Indeed, supported by the solid computational results, the method itself might exceed
all expectations.
The final chapter of Part I, ”Integer Linear Programming (ILP),” usually received
the second-highest reads weekly. This happened, I guess, because many researchers
are interested in the newly proposed controlled-branch method and controlled-
cutting method, which have potential applications for developing new ILP solvers.
The author’s main contributions seem to have not been fully valued by now,
including most pivot rules, reduced and D-reduced simplex methods, deficient-basis
and dual deficient-basis methods, face and dual face methods, and the decomposi-
tion principle, included in Part II. This is somewhat surprising since these methods
are supported by solid computational results, except for the newly introduced
reduced and D-reduced simplex methods, and the decomposition principle. In the
author’s view, the reduced simplex method is particularly noteworthy. It is the
first simplex method that searches along the objective-edge. In contrast to existing
simplex methods, it first determines pivot row and then pivot column, and does
so without any selection essentially. As a dual version of it, the D-reduced simplex
method shares similar attractive features, through determining pivot column first and
pivot row later. As for the decomposition principle, it allows for solving arbitrarily
large-scale (even dense) LP problems, in essence, giving a glimmer of light on some
other types of separable large-scale problems. Time will likely clarify the value of
these methods vs other LP solvers.
We are also optimistic about the face and dual face methods. Without exploiting
sparsity, the original methods were implemented using orthogonal transformation
with favorable computational results reported. In the second edition, we added
two chapters devoted to new methods with LU factorization for sparse computing.
Indeed, this is a very natural development which the author cannot help himself
from breaking into.
Some changes on chapters and sections were made with corrections and improve-
ments, and the whole second edition was organized into two parts. Roughly speak-
ing, Part I (Foundations) contains Part I of the first edition, and Part II (Advances)
includes Part II. The Simplex Feasible-Point Algorithm was improved, and removed
from the chapter ”Pivotal Interior-Point Method” to form an independent chapter
with its new title “Simplex Interior-Point Method,” since it actually represents a
new class of interior-point algorithms, which can be transformed from the traditional
simplex algorithms. The title of the original chapter was changed to “Facial Interior-
Point Method,” since the remaining algorithms represent another new class of
Preface to Second Edition xiii

interior-point algorithms, which can be transformed from the normal interior-point

algorithms. In particular, the chapter “Integer LP” was rewritten in great gains of
the introduction of the objective cutting. Another exciting improvement was the
reduced simplex method. The original derivation of its prototype was presented in
a chapter with the same title, and then transformed into the so-called “improved”
one in another chapter. Fortunately, inspired by the bisection simplex method, we
recently found a quite concise new derivation, so we can introduce it in a single
chapter now.
Finally, I am grateful to many individuals for their valuable comments and
suggestions on the first edition of this book. In particular, I appreciate comments
and encouragement given by Dr. Y.-Y Shi, Professor L.-H Zhang, and Professor
W.-X Zhu.

Nanjing, China Ping-Qi Pan

November 2022

References

1. Cipra BA (2000) The best of the 20th century: editors name top 10 algorithms.
SIAM News 33:1–2
2. Murtagh BA, Saunders MA (1998) MINOS 5.5 user’s guid. Technical report SOL
83-20R, Department of Engineering Economics Systems & Operations Research,
Stanford University, Stanford
3. Shi Y-y, Zhang L-H, Zhu W-X (2018) A review of linear programming compu-
tation by Ping-Qi Pan. European Journal of Operational Research 267(3):1182–
1183
Contents

Part I Foundations
1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3
1.1 Error of Floating-Point Arithmetic . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4
1.2 From Real-Life Issue to LP Model . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 6
1.3 Illustrative Applications . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9
1.4 Standard LP Problem.. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 14
1.5 Basis and Feasible Basic Solution .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 17
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 22
2 Geometry of Feasible Region .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 23
2.1 Feasible Region as Polyhedral Convex Set . . . . .. . . . . . . . . . . . . . . . . . . . 24
2.2 Interior Point and Relative Interior Point . . . . . . .. . . . . . . . . . . . . . . . . . . . 28
2.3 Face, Vertex, and Extreme Direction . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 31
2.4 Representation of Feasible Region . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 42
2.5 Optimal Face and Optimal Vertex . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 44
2.6 Graphic Approach .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 46
2.7 Heuristic Characteristic of Optimal Solution . . .. . . . . . . . . . . . . . . . . . . . 47
2.8 Feasible Direction and Active Constraint . . . . . . .. . . . . . . . . . . . . . . . . . . . 50
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 54
3 Simplex Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 57
3.1 Simplex Algorithm: Tableau Form . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 57
3.2 Getting Started.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 68
3.3 Simplex Algorithm .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 74
3.4 Degeneracy and Cycling . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 80
3.5 Finite Pivot Rule. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 84
3.6 Notes on Simplex Method . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 91
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 93
4 Implementation of Simplex Method .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 95
4.1 Miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 95
4.2 Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 97

xv
xvi Contents

4.3 LU Factorization of Basis . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 99

4.4 Sparse LU Factorization of Basis . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 102
4.5 Updating LU Factors .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 106
4.6 Crash Procedure for Initial Basis . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109
4.7 Harris Rule and Tolerance Expending.. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 111
4.8 Pricing for Reduced Cost . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 113
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 117
5 Duality Principle and Dual Simplex Method . . . . . . . .. . . . . . . . . . . . . . . . . . . . 119
5.1 Dual LP Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 119
5.2 Duality Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 122
5.3 Optimality Condition.. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 125
5.4 Dual Simplex Algorithm: Tableau Form . . . . . . . .. . . . . . . . . . . . . . . . . . . . 129
5.5 Dual Simplex Algorithm .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 132
5.6 Economic Interpretation of Duality: Shadow Price .. . . . . . . . . . . . . . . . 136
5.7 Dual Elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 137
5.8 Bilevel LP: Intercepting Optimal Set . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 142
5.9 Notes on Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 146
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 147
6 Primal-Dual Simplex Method . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 149
6.1 Mixed Two-Phase Simplex Algorithm . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 149
6.2 Primal-Dual Simplex Algorithm.. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 152
6.3 Self-Dual Parametric Simplex Algorithm .. . . . . .. . . . . . . . . . . . . . . . . . . . 158
6.4 Criss-Cross Algorithm Using Most-Obtuse-Angle Rule . . . . . . . . . . . 162
6.5 Perturbation Primal-Dual Simplex Algorithm . .. . . . . . . . . . . . . . . . . . . . 166
6.6 Notes on Criss-Cross Simplex Algorithm .. . . . . .. . . . . . . . . . . . . . . . . . . . 170
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 170
7 Sensitivity Analysis and Parametric LP. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 173
7.1 Change in Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 173
7.2 Change in Right-Hand Side .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 176
7.3 Change in Coefficient Matrix . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 178
7.3.1 Dropping Variable . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 178
7.3.2 Adding Variable . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 180
7.3.3 Dropping Constraint .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 182
7.3.4 Adding Constraint . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 184
7.3.5 Replacing Row/Column .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 189
7.4 Parameterizing Objective Function .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 192
7.5 Parameterizing Right-Hand Side . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 196
8 Generalized Simplex Method. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 199
8.1 Generalized Simplex Algorithm .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 200
8.1.1 Generalized Phase-I . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 207
8.2 Generalized Dual Simplex Algorithm: Tableau Form . . . . . . . . . . . . . 207
8.2.1 Generalized Dual Phase-I . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 212
Contents xvii

8.3 Generalized Dual Simplex Algorithm .. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 213

8.4 Generalized Dual Simplex Algorithm with Bound-Flipping .. . . . . . 220
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 223
9 Decomposition Method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 225
9.1 D-W Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 226
9.1.1 Starting-Up of D-W Decomposition .. .. . . . . . . . . . . . . . . . . . . . 231
9.2 Illustration of D-W Decomposition . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 232
9.3 Economic Interpretation of D-W Decomposition.. . . . . . . . . . . . . . . . . . 238
9.4 Benders Decomposition .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 240
9.5 Illustration of Benders Decomposition .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 248
9.6 Dual Benders Decomposition.. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 254
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 259
10 Interior-Point Method.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 261
10.1 Karmarkar Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 262
10.1.1 Projective Transformation .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 263
10.1.2 Karmarkar Algorithm . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 264
10.1.3 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 266
10.2 Affine Interior-Point Algorithm . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 271
10.2.1 Formulation of the Algorithm .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 272
10.2.2 Convergence and Starting-Up . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 274
10.3 Dual Affine Interior-Point Algorithm . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 276
10.4 Path-Following Interior-Point Algorithm . . . . . . .. . . . . . . . . . . . . . . . . . . . 279
10.4.1 Primal–Dual Interior-Point Algorithm.. . . . . . . . . . . . . . . . . . . . 281
10.4.2 Infeasible Primal–Dual Algorithm .. . . .. . . . . . . . . . . . . . . . . . . . 284
10.4.3 Predictor–Corrector Primal–Dual Algorithm . . . . . . . . . . . . . 286
10.4.4 Homogeneous and Self-Dual Algorithm . . . . . . . . . . . . . . . . . . 289
10.5 Notes on Interior-Point Algorithm . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 291
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 295
11 Integer Linear Programming (ILP) . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 299
11.1 Graphic Approach .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 300
11.1.1 Basic Idea Behind New ILP Solvers .. .. . . . . . . . . . . . . . . . . . . . 302
11.2 Cutting-Plane Method .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 303
11.3 Branch-and-Bound Method .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 306
11.4 Controlled-Cutting Method .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 313
11.5 Controlled-Branch Method . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 315
11.5.1 Depth-Oriented Strategy . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 316
11.5.2 Breadth-Oriented Strategy . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 319
11.6 ILP: with Reduced Simplex Framework . . . . . . . .. . . . . . . . . . . . . . . . . . . . 323
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 337
xviii Contents

Part II Advances
12 Pivot Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 341
12.1 Partial Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 342
12.2 Steepest-Edge Rule. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 344
12.3 Approximate Steepest-Edge Rule . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 346
12.4 Largest-Distance Rule.. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 348
12.5 Nested Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 350
12.6 Nested Largest-Distance Rule . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 352
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 353
13 Dual Pivot Rule .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 355
13.1 Dual Steepest-Edge Rule. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 355
13.2 Approximate Dual Steepest-Edge Rule . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 359
13.3 Dual Largest-Distance Rule . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 360
13.4 Dual Nested Rule .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 362
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 363
14 Simplex Phase-I Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 365
14.1 Infeasibility-Sum Algorithm .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 365
14.2 Single-Artificial-Variable Algorithm . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 370
14.3 Perturbation of Reduced Cost. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 376
14.4 Using Most-Obtuse-Angle Column Rule . . . . . . .. . . . . . . . . . . . . . . . . . . . 380
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 383
15 Dual Simplex Phase-l Method . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 385
15.1 Dual Infeasibility-Sum Algorithm .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 385
15.2 Dual Single-Artificial-Variable Algorithm .. . . . .. . . . . . . . . . . . . . . . . . . . 389
15.3 Perturbation of the Right-Hand Side . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 395
15.4 Using Most-Obtuse-Angle Row Rule . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 398
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 401
16 Reduced Simplex Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 403
16.1 Reduced Simplex Algorithm.. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 404
16.2 Dual Reduced Simplex Algorithm . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 412
16.2.1 Dual Reduced Simplex Phase-I:
Most-Obtuse-Angle.. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 418
16.3 Perturbation Reduced Simplex Algorithm . . . . . .. . . . . . . . . . . . . . . . . . . . 422
16.4 Bisection Reduced Simplex Algorithm . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 426
16.5 Notes on Reduced Simplex Algorithm .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 430
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 430
17 D-Reduced Simplex Method. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 433
17.1 D-Reduced Simplex Tableau . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 434
17.2 Dual D-Reduced Simplex Algorithm.. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 437
17.2.1 Dual D-Reduced Phase-I . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 442
17.3 D-Reduced Simplex Algorithm .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 445
17.3.1 D-Reduced Phase-I: Most-Obtuse-Angle Rule. . . . . . . . . . . . 452
Contents xix

17.4 Bisection D-Reduced Simplex Algorithm . . . . . .. . . . . . . . . . . . . . . . . . . . 456

Reference .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 460
18 Generalized Reduced Simplex Method. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 461
18.1 Generalized Reduced Simplex Algorithm . . . . . .. . . . . . . . . . . . . . . . . . . . 462
18.1.1 Generalized Reduced Phase-I: Single Artificial
Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 469
18.2 Generalized Dual Reduced Simplex Algorithm . . . . . . . . . . . . . . . . . . . . 476
18.2.1 Generalized Dual Reduced Phase-I .. . .. . . . . . . . . . . . . . . . . . . . 481
18.3 Generalized Dual D-Reduced Simplex Algorithm . . . . . . . . . . . . . . . . . 484
19 Deficient-Basis Method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 497
19.1 Concept of Deficient Basis. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 498
19.2 Deficient-Basis Algorithm: Tableau Form . . . . . .. . . . . . . . . . . . . . . . . . . . 500
19.3 Deficient-Basis Algorithm . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 504
19.3.1 Computational Results . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 508
19.4 On Implementation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 509
19.4.1 Initial Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 509
19.4.2 LU Updating in Rank-Increasing Iteration . . . . . . . . . . . . . . . . 510
19.4.3 Phase-I: Single-Artificial-Variable .. . . .. . . . . . . . . . . . . . . . . . . . 511
19.5 Deficient-Basis Reduced Algorithm .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 516
19.5.1 Phase-I: Most-Obtuse-Angle Rule . . . . .. . . . . . . . . . . . . . . . . . . . 524
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 530
20 Dual Deficient-Basis Method . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 531
20.1 Dual Deficient-Basis Algorithm: Tableau Form . . . . . . . . . . . . . . . . . . . . 531
20.2 Dual Deficient-Basis Algorithm . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 536
20.3 Dual Deficient-Basis D-Reduced Algorithm: Tableau Form .. . . . . . 540
20.4 Dual Deficient-Basis D-Reduced Algorithm .. . .. . . . . . . . . . . . . . . . . . . . 549
20.5 Deficient-Basis D-Reduced Gradient Algorithm:
Tableau Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 552
20.6 Deficient-Basis D-Reduced Gradient Algorithm . . . . . . . . . . . . . . . . . . . 563
Reference .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 564
21 Face Method with Cholesky Factorization.. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 565
21.1 Steepest Descent Direction . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 566
21.2 Updating of Face Solution . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 569
21.3 Face Contraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 570
21.4 Optimality Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 571
21.5 Face Expansion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 572
21.6 Face Algorithm .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 573
21.6.1 Face Phase-I: Single-Artificial-Variable .. . . . . . . . . . . . . . . . . . 574
21.6.2 Computational Results . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 577
21.7 Affine Face Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 578
21.8 Generalized Face Algorithm .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 582
21.9 Notes on Face Method . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 584
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 585
xx Contents

22 Dual Face Method with Cholesky Factorization . . . .. . . . . . . . . . . . . . . . . . . . 587

22.1 Steepest Ascent Direction.. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 588
22.2 Updating of Dual Face Solution . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 591
22.3 Dual Face Contraction . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 591
22.4 Optimality Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 593
22.5 Dual Face Expansion .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 594
22.6 Dual Face Algorithm .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 595
22.6.1 Dual Face Phase-I . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 596
22.6.2 Computational Results . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 600
22.7 Dual Face Algorithm via Updating (B T B)−1 . .. . . . . . . . . . . . . . . . . . . . 601
23 Face Method with LU Factorization .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 607
23.1 Decent Search Direction . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 607
23.2 Updating of Face Solution . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 609
23.3 Pivoting Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 612
23.4 Optimality Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 614
23.5 Face Algorithm: Tableau Form . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 615
23.6 Face Algorithm .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 618
23.7 Notes on Face Method with LU Factorization . .. . . . . . . . . . . . . . . . . . . . 624
Reference .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 624
24 Dual Face Method with LU Factorization . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 625
24.1 Key of Method.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 626
24.2 Ascent Search Direction . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 630
24.3 Updating of Dual Face Solution . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 632
24.4 Pivoting Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 634
24.5 Optimality Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 636
24.6 Dual Face Algorithm: Tableau Form . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 637
24.7 Dual Face Algorithm .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 638
24.8 Notes on Dual Face Method with LU Factorization .. . . . . . . . . . . . . . . 640
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 641
25 Simplex Interior-Point Method . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 643
25.1 Column Pivot Rule and Search Direction . . . . . . .. . . . . . . . . . . . . . . . . . . . 644
25.2 Row Pivot Rule and Stepsize . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 645
25.3 Optimality Condition and the Algorithm.. . . . . . .. . . . . . . . . . . . . . . . . . . . 647
25.4 Computational Results . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 649
26 Facial Interior-Point Method . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 651
26.1 Facial Affine Face Interior-Point Algorithm . . . .. . . . . . . . . . . . . . . . . . . . 651
26.2 Facial D-Reduced Interior-Point Algorithm . . . .. . . . . . . . . . . . . . . . . . . . 655
26.3 Facial Affine Interior-Point Algorithm . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 666
Reference .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 671
27 Decomposition Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 673
27.1 New Decomposition Method . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 673
27.2 Decomposition Principle: “Arena Contest” . . . . .. . . . . . . . . . . . . . . . . . . . 677
Contents xxi

27.3 Illustration on Standard LP Problem . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 678

27.4 Illustration on Bounded-Variable LP Problem ... . . . . . . . . . . . . . . . . . . . 680
27.5 Illustration on ILP Problem .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 681
27.6 Practical Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 683

Appendix A On the Birth of LP and More . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 685

Appendix B MPS File .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 689
Appendix C Test LP Problems .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 695
Appendix D Empirical Evaluation for Nested Pivot Rules . . . . . . . . . . . . . . . 703
Appendix E Empirical Evaluation for Primal and Dual Face
Methods with Cholesky Factorization .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 721
Appendix F Empirical Evaluation for Simplex Interior-Point
Algorithm .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 727
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 730

Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 733
About the Book

This book represents a real breakthrough in the field of linear programming (LP).
Being both thoughtful and informative, it focuses on reflecting and promoting the
state of the art by highlighting new achievements in LP. This new edition was orga-
nized into two parts. The first part addresses foundations of LP, including geometry
of feasible region, simplex method, implementation of simplex method, duality
and dual simplex method, primal-dual simplex method, sensitivity analysis and
parametric LP, generalized simplex method, decomposition method, interior-point
method, and integer LP method. The second part mainly introduces contributions
of the author himself, such as efficient primal/dual pivot rules, primal/dual Phase-
I methods, reduced/D-reduced simplex methods, generalized reduced simplex
method, primal/dual deficient-basis methods, primal/dual face methods, and new
decomposition principle.
Many important improvements were made in this edition. The first part includes
new results, such as mixed two-phase simplex algorithm, dual elimination, fresh
pricing scheme for reduced cost, and bilevel LP model and intercepting of optimal
solution set. In particular, the chapter ”Integer LP Method” was rewritten in
great gains of the objective cutting for new ILP solvers controlled-cutting/branch
methods, as well as an attractive implementation of the controlled-branch method. In
the second part, the “simplex feasible-point algorithm” was rewritten and removed
from the chapter “Pivotal Interior-Point Method” to form an independent chapter
with the new title “Simplex Interior-Point Method,” as it represents a class of new
interior-point algorithms transformed from traditional simplex algorithms. The title
of the original chapter was then changed to “Facial Interior-Point Method,” as
the remaining algorithms represent another class of new interior-point algorithms
transformed from normal interior-point algorithms. Without exploiting sparsity, the
original primal/dual face methods were implemented using Cholesky factorization.
To cope sparse computation, two new chapters for ones with LU factorization were
added to the second part. Another exciting improvement was the reduced simplex
method. In the first edition, the derivation of its prototype was presented in a chapter
with the same title, and then converted into the so-called “improved” one in another
chapter. Fortunately, the author recently found a quite concise new derivation, so

xxiii
xxiv About the Book

he can now introduce the distinctive and very promising new simplex method in a
single chapter.
This book is a rare work in LP. With a focus on computation, it contains many
novel ideas and methods, supported by sufficient numerical results. Being clear
and succinct, its content deduces in a fresh manner, from simple to profound.
In particular, a larger number of examples were worked out to demonstrate
algorithms. So, it is an indispensable tool for undergraduate/graduate students,
teachers, practitioners, and researchers in LP and related fields.
About the Author

Ping-Qi Pan is a professor and doctoral supervisor of the Department of Math-

ematics at Southeast University, Nanjing, China. He was a visiting scholar at the
University of Washington (1986–1987) and visiting scientist at Cornell University
(1987–1988). His research interest focuses on mathematical programming and
operations research, especially large-scale linear optimization. He was standing
council member of the Mathematical Programming Society of China and standing
council member of the Operation Research Society of China. Professor Pan has
received the honorary title of Outstanding Scientific-Technical Worker of Jiangsu
Province of China. He was nominated as Top 100 scientist of 2012 by the Interna-
tional Biographical Centre, Cambridge, England. He won the Lifetime Achievement
Award by Who’s Who in the World in 2017.

xxv
Notation

In this book, in general, uppercase English letters are used to denote matrices,
lowercase English letters used to denote vectors, and lowercase Greek letters do
denote reals. Set is designated by uppercase English or Greek letters. Unless
indicated otherwise, all vectors are column ones.
LP Linear programming
ILP Integer linear programming
Rn Euclidean space of dimension n
R Euclidean space of dimension 1
0 Origin of Rn (or the zero vector)
ei Unit vector with the ith component 1 (dimension will be clear from the
context; the same below)
e Vector of all ones
I Unit matrix
A Coefficient matrix of the linear program; also stands for index set of A’s
columns, i.e., A = {1, . . . , n}
m Number of rows of A
n Number of columns of A
b Right-hand side of the equality constraint
c Cost vector, i.e., coefficient vector of the objective function
B Basis (matrix). Also the set of basic indices
N Nonbasic columns of A. Also the set of nonbasic indices
aj The j th column of A
ai j The entry of the ith row and j th column of A
vj The j th component of vector v
v Euclidean norm of vector v
max(v) The largest component of vector v
AJ Submatrix consisting of columns corresponding to index set J
vI Subvector consisting of components corresponding to row index set I
AT Transpose of A
X Diagonal matrix whose diagonal entries are components of vector x

xxvii
xxviii Notation

∇f (x) Gradient of function f (x)

⊂  is a subset of 
∪ Union set, i.e., {τ | τ ∈  or τ ∈ }
∩ Intersection set, i.e., {τ | τ ∈ , τ ∈ }
\ Complementary set, i.e., {τ | τ ∈ , τ ∈ }
∅ Empty set
| Such that. For example, {x | Ax = b} means the set of all x such that
Ax = b holds
Far less than
Far greater than
O(α) Implies a number less than and equal to kα, where k is a fixed integer
constant independent of α
|τ | Absolute value of τ if τ is a real, or cardinality of τ if τ is a set
sign (t) Sign of real t
range H Column space of matrix H
null H Null space of matrix H
dim  Dimension of set 
Cnm Number of combinations of taking m from n elements
int P Interior of set P
α The largest integer no more than α
α The smallest integer no less than α
This book involves the following two simplex FORTRAN codes:
MINOS The sophisticated smooth optimization package developed by Murtagh
and Saunders (1998) at Department of Management Science and Engi-
neering of Stanford University. Based on the simplex method, the LP
option of this sparsity-exploiting code was used as the benchmark and
platform for empirical evaluation of new LP methods
RSA The author’s private code based on the revised two-phase simplex
algorithm without exploiting sparsity. It uses the Harris two-pass row
rule
 Part I
Foundations
Chapter 1
Introduction

As intelligent creatures, human beings plan and carry out activities with pre-set
objectives. Early human ancestors relied on their experience only, whereas their
modern-day descendants make their decisions based on scientific tools.
The word optimization means to achieve the best results with minimum effort.
For example, how would one allocate limited resources (labor, money, materials,
etc.) to maximize returns or minimize costs while meeting a specific demand.
People collect information and data and express real-life problems with appropriate
forms mathematically using numbers, functions, and equations, etc., through a so-
called mathematical modeling process. Then, optimization methods/software are
applied to deal with these models, consequently offering quantitative bases for
sound decision making. This process has been revolutionized by the advent of the
electronic computer.
The linear programming (LP) branch emerged when the American mathemati-
cian George B. Dantzig founded the LP model and the simplex method in 1947 (for
the birth of LP, see Appendix A). The LP model has a simple structure, involving
linear functions and equalities/inequalities; somewhat surprisingly, it turns out to be
one of the most important and widely used models in the world. This seems to be in
response to the words of Lao Tzu:1
The Great Road Is the Simplest!
Of course, “simplest” does not necessarily mean “easy.” The challenge is the huge
scale of LP models that are difficult to handle. After 70 years of development,
LP methods/software are now capable to solve models with considerably large
scales. Even 30 years ago, for example, an LP model with 12,753,313 variables was
solved successfully to provide an optimal scheme for an airline crew scheduling
application (Bixby et al. 1992). Nevertheless, the academic community still faces
major challenges to deal with large-scale and difficult models. Therefore, it might

1 Lao Tzu (about 571–471 BC) was an ancient Chinese philosopher.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 3
P.-Q. PAN, Linear Programming Computation,
https://doi.org/10.1007/978-981-19-0147-8_1
4 1 Introduction

be time to review the foundations as well as advances in LP and to provide a book

for undergraduate/graduate students, teachers, researchers, and practitioners in LP
and the related fields.
In this chapter, we first introduce primitive notation and concepts of numerical
algebra that are closely related to LP, including errors of floating-point arithmetic,
the mathematical model of LP, illustrative applications, the standard LP problem,
the basis matrix, and the basic solution.

1.1 Error of Floating-Point Arithmetic

As a basic tool for modern numerical computation, a computer is indispensable.

LP without it is not imaginable. In using a computer, however, numerical errors are
unavoidable because it can only represent a proper subset F of real numbers, and
as a result, data can only be stored and calculated with finite digits. On one hand,
primary data may be out of F and expressed approximately; on the other hand,
numerical results yielded from subsequent operations of arithmetic may have to be
rounded. Errors arising therefrom are called rounding errors.
The set F of the floating-point values represented by a computer is characterized
by 4 positive integers: the base β, precision (mantissa digits) t, and exponent range
[L, U ] (Forsythe et al. 1977, pp 10–29). More precisely, a computer can represent
all nonzero numbers of the following form:

f l(x) = ±.α1 . . . αt β e , 0 ≤ αi < β, i = 1, . . . , t, α1 = 0; L ≤ e ≤ U,

(1.1)

which together with 0 constitute F , and where α1 . . . αt is called mantissa part and
β e is the exponent part.
We use the notation

m = β L−1 , M = β U (1 − β −t ).

It is clear that for any floating-point number f l(x) ∈ F , it holds that m ≤ |x| ≤ M.
Thus, F is a set of non-uniformly distributed rational numbers. Define the set

S = {x ∈ R | m ≤ |x| ≤ M}.
1.1 Error of Floating-Point Arithmetic 5

Then, for any 0 = x ∈ S, the approximate number f l(x), yielded from the
computer, only holds mantissa of t digits, via the following two methods:
(i) Rounding method: It is based on the (t + 1)th digit of the mantissa of x. The
digits at this position and after are all taken off if this digit is less than a half of
β; in the other case, the digits are taken off after a unit is added to the previous
digit. The associated absolute error bound is then

1 e−t
|f l(x) − x| ≤ β . (1.2)
2
(ii) Truncation method: The (t + 1)th figure and after are all taken off. Thereby,
f l(x) is the floating-point number nearest to x in F , whose absolute values are
no more than |x|, and hence, the associated absolute error bound is

|f l(x) − x| ≤ β e−t . (1.3)

The t is a key quantity to numerical computations. To see this, define

⎧ 1 1−t
⎨ 2β rounding
= β 1−t truncating (1.4)
⎩
.

Then, from either (1.2) or (1.3) it follows that

|f l(x) − x| ≤ β e−1 .

Since

|x| ≥ .α1 β e > 0,

it holds that
 
 f l(x) − x  β e−1
 ≤
 x  .α β e = /α1 ≤ ,
1

that is

f l(x) = x(1 + δ), |δ| ≤ .

Therefore,  is just a relative error bound of the approximate number f l(x). The
so-called machine precision, , is the smallest positive value in F , satisfying

f l(1 + ) = 1.

Thereby, one may get to know the precision of a computer.

6 1 Introduction

Assume a, b ∈ F , and represent any of the four arithmetic operations (+, −, ∗, /)

by . It is clear that if ab ∈ S, then the computer cannot handle and display
overflow (when |ab| > M) or underflow (when 0 < |ab| < m) and terminate the
computation (some computers and compilers place value 0 in the case of underflow).
In normal cases, for the computed value f l(ab) of ab, it is clear that

f l(ab) = (ab)(1 + δ), |δ| < .

Therefore, the relative error bound for a single arithmetic operation is , a small
number. Unfortunately, the situation may be entirely contrary when executing an
algorithm or a series of operations.
Usually, the algorithm’s performance may be affected by rounding or truncation
errors greatly. Due to error accumulations, final results could be entirely meaning-
less, as they could be far from the theoretical value. It is therefore important to take
care of numerical errors in computations. In this aspect, the following simple rules
should be followed:
1. Avoid using a number close to 0 as the denominator or divisor.
2. Avoid subtracting a number from another that is close to it.
3. Avoid operating numbers with a very different order of magnitude, so that “large
swallows small whole.”
4. Avoid results close to m or M.
5. Reduce the number of operations, especially multiplications or divisions.
When δ 1, for instance, calculating from the right-hand side of each equation
below gives a more accurate result than from the left-hand side.
√ √ √ √
x + δ − x = δ/( x + δ + x),
cos(x + δ) − cos x = −2 sin(x + δ/2) sin(δ/2).

Due to different errors introduced in computations, algorithms, which are

equivalent theoretically, are often not equivalent numerically. An algorithm is said
to be of good (numerical) stability if the resulting numerical error is under control
and has little impact on the results. Unstable algorithms are unreliable.
Usually, it is not an easy task to accurately evaluate the error bound of results
computed by algorithms. Instead, the so-called backward error analysis is applied.
Interested readers are referred to Wilkinson (1971).

1.2 From Real-Life Issue to LP Model

In practice, there are usually many schemes to be chosen by a decision maker, and
different schemes may lead to widely diverging outcomes, as is crucial in fierce
competition. This situation provides a wide stage for the decision maker to play.
1.2 From Real-Life Issue to LP Model 7

Making decisions just based on experience cannot be mentioned in the same breath
with mathematical tools.
The formation of a mathematical model is the first step toward the application
of LP. Based on full and reliable data gathered, a good model comes from one’s
knowledge, skill, experience, and wisdom. This topic is beyond the scope of this
book, and this section only gives the reader a taste of simple examples.
Example 1.2.1 A manufacturer, who has 1100 tons of log, is committed by contract
to supply 470 tons of lath to a company. There is a 6% loss in log when it is
processed into laths. In addition, the production costs have risen since the signing
of the contract, while the selling price of lath remains the same. In fact, he can
make more profit by selling the log as raw material to another firm. What can the
manufacturer do to have the maximum profit while still honoring the contract?
This problem can be solved algebraically. Let x be the number of tons of logs the
manufacturer can sell as raw material, and let y be the number of logs available to
produce lath. It is clear that

x + y = 1100,

where x should be maximized to gain the maximum profit. On the other hand, since
6% out of y is lost during the manufacturing process, the actual amount of lath
produced is

y − 0.06y,

which, associated with the contract, must be equal to 470 tons, i.e.,

y − 0.06y = 470.

Consequently, we have the following system of linear equations:


y − 0.06y = 470
(1.5)
1100 − y = x,

which has a unique solution

x = 600, y = 500.

Therefore, the manufacturer should sell 600 tons of logs as raw material and use the
rest 500 tons for the production of lath.
In the preceding, the manufacturer only has a unique scheme to choose. Usually,
decision makers face multiple or even unlimited choices, as shown in the following
example.
8 1 Introduction

Example 1.2.2 A manufacturer produces lath and sheet piles. Profit per lath is
2 dollars, and profit per sheet pile is 5 dollars. The equipment capability of the
manufacturer can produce at most 6000 lath or 4000 sheet piles per day. According
to the sales contract, he can sell 3000 sheep piles at most, and the sales volume of
laths and sheet piles cannot exceed 5000 per day. How should he arrange to produce
lath and sheep piles to gain a maximum profit?
Let x be the amount of lath, and let y be that of sheep piles to be produced per day
(in thousand as the unit). For values of each pair of variables (x, y) corresponding
to a decision, the associated profit is f (x, y) = 2x + 5y thousand dollars. So, what
the manufacturer wants to do is to maximize the function value of f (x, y).
It is clear that the yielded profit would be infinitely large if there was no
restriction on the production. It is not the case, of course. According to the
equipment capability, the average rate of producing lath and sheet piles is 6 thousand
and 4 thousand per day, respectively. Therefore, variables x and y should satisfy
inequality x/6 + y/4 ≤ 1, or equivalently,

2x + 3y ≤ 12.

According to the selling limitation, in addition, there is another inequality

x + y ≤ 5,

and finally,

y ≤ 3.

In addition, the amount of output should be nonnegative. As the amount is large, we

use the restriction x, y ≥ 0 instead for simplicity and finally obtain integer results
by rounding.
The preceding analysis can be put into the following so-called mathematical
model:

max f (x, y) = 2x + 5y
s.t. −2x − 3y ≥ −12
− x − y ≥ −5 (1.6)
−y ≥ −3
x, y ≥ 0,

where x and y are called decision variables, and f (x, y) is called objective function;
the inequalities are called constraints, (x, y) satisfying constraints is a feasible
solution, and the set of all feasible solutions is the feasible region. Solving a model
is trying to find its optimal solution, that is, a feasible solution maximizing the
objective function over the feasible region.
1.3 Illustrative Applications 9

Fig. 1.1 The feasible region y

of Example 1.2.2

B y=3
A
2x+3y=12
C

P x+y=5

O D x

In contrast with Example 1.2.1, which has a single feasible solution, Exam-
ple 1.2.2 involves infinitely many feasible solutions. In fact, the problem has the
following feasible region:

P = {(x, y) | 2x + 3y ≤ 12, x + y ≤ 5, y ≤ 3, x ≥ 0, y ≥ 0}.

Since there is a 1-to-1 correspondence between pairs of real numbers and points
in a rectangular coordinate system, the feasible region P can be represented
geometrically by Fig. 1.1, where x (thousand) represents the number of laths and
y (thousand) that of sheep piles, and where each boundary is labeled by an equation
corresponding to the closed half-plane. The feasible region P is just the intersection
of these half-planes, within which each point corresponds to a feasible solution.
Therefore, the problem now is how to find a point in P corresponding to the
maximum possible value of function 2x + 5y. Since there are infinitely many points
in P , it is impossible to pick up and compare them one by one, not to mention much
more challenging large-scale problems with hundreds of thousands of variables and
constraints. Although this simple instance can be solved by a graphical approach
(Sect. 2.6), dealing with general LP problems requires sophisticate mathematical
tools.

1.3 Illustrative Applications

LP has a very wide range of applications, touching upon almost all fields related
to decision making and management. This section only brings up a few illustrative
instances. Strictly speaking, some of these instances involve variables of nonneg-
ative integer values and hence belong to the so-called integer/mixed-integer LP
models, although the latter requirement is ignored here for simplicity.
10 1 Introduction

Example 1.3.1 (Production Planning) A factory produces furniture A, B, C. Each

furniture production goes through 3 procedures: component processing, electroplat-
ing, and assembling. The production capacity of each procedure per day is converted
into effective working hours. Below are the required effective working hours and
getatable profit for each piece of the furniture.

Procedure Hours per piece Available hours per day

Product A Product B Product C
Component processing 0.025 0.05 0.3 400
Electrofacing 0.20 0.05 0.1 900
Assembling 0.04 0.02 0.20 600
Profit (dollars/piece) 1.25 1.5 2.25

Now how can the factory achieve the highest profit?

Answer Let x1 , x2 , and x3 be yields of furniture A,B,C, respectively, and let f
be the total profit. To gain the highest profit, construct the following model:

max f = 1.25x1 + 1.5x2 + 2.25x3

s.t. 0.025x1 + 0.05x2 + 0.3x3 ≤ 400
0.20x1 + 0.05x2 + 0.1x3 ≤ 900
0.04x1 + 0.02x2 + 0.20x3 ≤ 600
x1 , x2 , x3 ≥ 0.

An optimal solution obtained by the simplex method is

x1 = 2860, x2 = 6570, x3 = 0,

and the associated objective function value is f = 13430. That is to say, the factory
should produce 2860 pieces of product A, 6570 pieces of product B, and no product
C, achieving the highest profit 13430 dollars.
Example 1.3.2 (Transportation) A company has 8800, 7200, and 5700 containers
at ports A, B, C, respectively. These containers should be transported to plants 1,
2, 3, 4, 5, whose working ability is 3200, 5300, 4100, 6200, and 2900 containers,
1.3 Illustrative Applications 11

respectively. The following table lists the freight rate (dollars/container) of transport
service from the ports to the plants:

Port 1 2 3 4 5
A 55 150 78 38 90
B 25 60 25 43 35
C 102 80 52 74 60

How should the company arrange to achieve the least total freight charges?
Answer This is a balanced transportation problem, as the total number of
containers is equal to the total number of containers, i.e.,

8800 + 7200 + 5700 = 3200 + 5300 + 4100 + 6200 + 2900 = 21700.

Let xij , i = 1, 2, 3; j = 1, . . . , 5, be the number of containers that are

transported from ports A, B, C to plants 1–5, and let f be the total freight charges.
The model is as follows:

min f = 55x11 + 150x12 + 78x13 + 38x14 + 90x15

+25x21 + 60x22 + 25x23 + 43x24 + 35x25
+102x31 + 80x32 + 52x33 + 74x34 + 60x35
s.t. x11 + x12 + x13 + x14 + x15 = 8800
x21 + x22 + x23 + x214 + x25 = 7200
x31 + x32 + x33 + x34 + x35 = 5700
x11 + x21 + x31 = 3200
x12 + x22 + x32 = 5300
x13 + x23 + x33 = 4100
x14 + x24 + x34 = 6200
x15 + x25 + x35 = 2900
xij ≥ 0, i = 1, 2, 3; j = 1, . . . , 5.

An optimal solution obtained by the simplex method is

x11 = 2600, x14 = 6200, x21 = 600, x23 = 4100, x25 = 2500, x32 = 5300, x35 = 400,

and all the other variables take value 0. Then the company pays the minimum total
freight 1031600 dollars.
12 1 Introduction

Example 1.3.3 (Burdening) To feed animals, someone needs 30 grams of mineral,

700 grams of protein, and 100 milligrams of vitamin per day, at least. There are
5 kinds of feed to buy. The nutrition and price per kilogram for each feed are as
follows:

Fodder Mineral (g.) Protein (g.) Vitamin (mg.)) Price (dollars1/kg.)

A 0.5 18.0 0.8 0.8
B 2.0 6.0 2.0 0.3
C 0.2 1.0 0.2 0.4
D 0.5 2.0 1.0 0.7
E 1.0 3.0 0.5 0.2

Find a buying scheme to have a minimum cost to serve animals’ needs.

Answer Let xj (kg.), j = 1, . . . , 5, be the purchase quantities of the 5 kinds of
feed, respectively, and let f be the total cost. The goal is to determine the values of
these decision variables to achieve the minimum cost. The model is as follows:

min f = 0.8x1 + 0.3x2 + 0.4x3 + 0.7x4 + 0.2x5

s.t. 0.5x1 + 2.0x2 + 0.2x3 + 0.5x4 + x5 ≥ 30
18.0x1 + 6.0x2 + x3 + 2.0x4 + 0.3x5 ≥ 700
0.8x1 + 2.0x2 + 0.2x3 + x4 + 0.5x5 ≥ 100
xj ≥ 0, j = 1, . . . , 5.

An optimal solution is obtained by the simplex method, i.e.,

x1 = 25.6, x2 = 39.7, x3 = x4 = x5 = 0,

with minimum objective function value f = 32.28. So, he should buy 25.6 kg. of
feed A, 39.7 kg. of feed B, and none of the other 3 feeds with the minimum cost
32.28 dollars.
Example 1.3.4 (Human Resource Arrangement) A round-the-clock supermarket
arranges salespersons. Every day falls into 6 time intervals, 4 h each. Every
salesperson starts work at the beginning of the intervals and then continuously works
for 8 h. The required numbers of salesperson for each of the intervals are as follows:

Interval 1 2 3 4 5 6
2:00–6:00 6:00–10:00 10:00–14:00 14:00–18:00 18:00–22:00 22:00–2:00
Number 7 15 25 20 30 7
1.3 Illustrative Applications 13

How to arrange to achieve the least number of salespersons per day?

Answer Let xj , j = 1, . . . , 6, be the number of salespersons who start work at
the beginning of interval j , and let f be the total number of salespersons. The goal
is to determine the variable values to achieve the minimum number of salespersons
working per day. The model is as follows:

min f = x1 + x2 + x3 + x4 + x5 + x6
s.t. x1 + x6 ≥ 7
x1 + x2 ≥ 15
x2 + x3 ≥ 25
x3 + x4 ≥ 20
x4 + x5 ≥ 30
x5 + x6 ≥ 7
xj ≥ 0, j = 1, . . . , 6.

An optimal solution obtained by the simplex method is

x2 = 25, x4 = 30, x6 = 7, x1 = x3 = x5 = 0,

and the associated objective function value is f = 62. So, the business should
arrange 25 salespersons working from 6:00, 30 ones working from 14:00, and 7
from 22:00. This scheme requires only 62 salespersons.
Example 1.3.5 (Laying-off) A shop is asked to make 450 steel frames. Each frame
needs 1.5, 2.1, and 2.9 meters long angle steel. Each raw material is 7.4 meters long.
Find the optimum laying-off way.
Answer A direct way is to make a frame using raw material to get the 3 kinds
of angle steel. If so, there would be a need for 450 raw materials, each of which
would leave an 0.9 meters long offcut. To obtain the optimum scheme, consider the
following ways, each of which would leave an offcut no more than 0.9 meters long.

Length (meters) Way

A B C D E F
2.9 1 2 0 1 0 1
2.1 0 0 2 2 1 1
1.5 3 1 2 0 3 1
Total 7.4 7.3 7.2 7.1 6.6 6.5
Offcut 0.0 0.1 0.2 0.3 0.8 0.9
14 1 Introduction

These ways should be mixed and matched to gain the minimum total length of
offcuts. Let xj , j = 1, . . . , 6, be numbers of raw materials taken, corresponding
to ways A, B, C, D, E, F , and let f be the total length of offcuts. Construct the
following model:

min f = 0.1x2 + 0.2x3 + 0.3x4 + 0.8x5 + 0.9x6

s.t. x1 + 2x2 + x4 + x6 = 450
2x3 + 2x4 + x5 + x6 = 450
3x1 + x2 + 2x3 + 3x5 + x6 = 450
xj ≥ 0, j = 1, . . . , 6.

The optimal solution given by the simplex method is

x1 = 135, x2 = 45, x4 = 225, x3 = x5 = x6 = 0,

with objective function value f = 72. Therefore, the shop should use 135 raw
materials for way A, 45 for way B, and 225 for way D, with a total length of 72
meters of offcuts. The minimum total number of used raw materials is 405.
The sizes of the preceding problems are very small, which are just for the reader
to get a taste of LP applications. Nowadays, real-world problems are usually very
large. Even so, small problems still help to convey ideas and illustrate algorithms.

1.4 Standard LP Problem

Some of LP problems in the previous section seek for maximizing the objective
function, while others seek for minimizing it. Besides, some of their constraints
are equalities, while others are inequalities of “≥” or “≤.” For convenience, we
introduce the following problem in a so-called standard form:

min f = c1 x1 + c2 x2 + · · · + cn xn
s.t. a11x1 + a12 x2 + · · · + a1n xn = b1
a21x1 + a22 x2 + · · · + a2n xn = b2
.. (1.7)
.
am1 x1 + am2 x1 + · · · + amn xn = bm
xj ≥ 0, j = 1, . . . , n.
1.4 Standard LP Problem 15

All kinds of LP problems can be transformed into standard forms:

1. Any maximization problem can be transformed into a minimization problem by
introducing the negative objective function instead.
Using f = −f  , e.g., the objective

max f  = c1 x1 + c2 x2 + · · · + cn xn

can be replaced by

min f = −c1 x1 − c2 x2 − · · · − cn xn .

It is clear that this does not matter to the optimal solution but changes the sign of
the optimal objective value only.
2. Any inequality can be transformed into equality by introducing an extra nonneg-
ative variable.
Introducing a so-called slack variable xk+1 ≥ 0, e.g., the “≤” type of inequality

α1 x1 + α2 x2 + · · · + αk xk ≤ β

can be transformed into equality

α1 x1 + α2 x2 + · · · + αk xk + xk+1 = β,

and the “≥” type of inequality

α1 x1 + α2 x2 + · · · + αk xk ≥ β

can be transformed into the equality

α1 x1 + α2 x2 + · · · + αk xk − xk+1 = β.

3. Nonnegative constraints on variables can be obtained by introducing auxiliary

variables.
Using variable transformation xk = −xk , e.g., nonpositive constraint xk ≤ 0, can be
converted into xk ≥ 0. A free variable, say xk , can be eliminated by the following
transformation:

xk = xk − xk+1 , xk , xk+1 ≥ 0.

16 1 Introduction

Example 1.4.1 Transform the following LP problem into the standard form:

max f  = −2x1 + x2 − 5x3

s.t. −3x1 + x2 + x3 ≤ 2
x1 − 7x2 + 4x3 ≥ −3
x1 − 3x2 − 2x3 ≤ 1
x1 , x2 ≥ 0.

Answer Maximizing f  is transformed into minimizing f = −f  . The free

variable x3 is substituted by x3 = x3 − x4 , x3 , x4 ≥ 0. The first 3 inequalities are
transformed into equalities by introducing slack variables x5 , x6 , x7 , respectively.
The resulting standard problem is as follows:

min f = 2x1 − x2 + 5x3 − 5x4

s.t. −3x1 + x2 + x3 − x4 + x5 = 2
x1 − 7x2 + 4x3 − 4x4 − x6 = −3
x1 − 3x2 − 2x3 + 2x4 + x7 = 1
xj ≥ 0, j = 1, . . . , 7.

Using vector and matrix notation

A = (aij ), x = (x1 , . . . , xn )T , c = (c1 , . . . , cn )T , b = (b1 , . . . , bm )T ,

we put the standard problem (1.7) into the compact form below:

min f = cT x
(1.8)
s.t. Ax = b, x ≥ 0,

where A ∈ Rm×n , c ∈ Rn , b ∈ Rm , m < n. It is assumed that rank (A) = m,

except indicated otherwise.
It goes without saying that each row or column of coefficient matrix A does
not vanish. f = cT x is called objective function, and c is price (cost). The system
Ax = b is called equality constraint, and x ≥ 0 nonnegative constraint.
Usually, theories/methods are developed to deal with standard problems,
although they are essentially applicable to general LP problems.
Definition 1.4.1 A solution fulfilling the constraints is a feasible solution, and the
set of all feasible solutions is the feasible region, i.e.,

P = {x ∈ Rn | Ax = b, x ≥ 0}. (1.9)

A feasible solution with the minimum objective function value over the feasible
region is an optimal solution.
1.5 Basis and Feasible Basic Solution 17

The minimization problem is called (lower) unbounded if its objective value in

the feasible region can be arbitrarily small.

1.5 Basis and Feasible Basic Solution

The standard LP problem (1.8) involves the linear system of equations, Ax = b,

as equality constraint. Note that it is a so-called underdetermined system, as the
number of equations is less than that of unknowns.
The set of all solutions is called solution set of the system. Systems are equivalent
if they have the same solution set. There are two basic types of equivalent transfor-
mations into systems. In this respect, the validity of the following propositions is
evident.
Proposition 1.5.1 A system, resulting from multiplying any equation of it by a
nonzero, is equivalent to the original.
Proposition 1.5.2 A system, resulting from adding a multiple of any equation to
another, is equivalent to the original.
Any of the preceding operations is called elementary (row) transformation.
The second type of elementary transformation is especially important since it can
eliminate a nonzero entry of A. Using a series of such kinds of transformations, e.g.,
the Gauss–Jordan elimination, converts a linear system into a so-called canonical
form that is readily solvable.
Let us bring up an example of 3 × 5 standard LP problem:

min f = x1 + 2x2 − x4 + x5
s.t. 2x1 + x2 + 3x3 + 2x4 = 5
x1 − x2 + 2x3 − x4 + 3x5 = 1 (1.10)
x1 − 2x3 − 2x5 = −1
xj ≥ 0, j = 1, . . . , 5.

For convenience, we strip the augmented coefficient matrix from the linear system
and put it into the following tableau:

x1 x2 x3 x4 x5 RHS
2 1 3 2 5
1 −1 2 −1 3 1
1 −2 −2 −1

where empty cells stand for value 0 (the same below).

18 1 Introduction

The preceding tableau and the system itself are regarded as the same. Carrying
out elementary transformations with the tableau amounts to doing the same thing
with the system itself.
A so-called pivot is a nonzero entry in the tableau, which determines a Gauss–
Jordan elimination process. As an example, take the first entry 2 in the first column
as pivot. Multiply by 1/2 the first row to turn the pivot to 1, leading to the following
tableau:

x1 x2 x3 x4 x5 RHS
1 1/2 3/2 1 5/2
1 −1 2 −1 3 1
1 −2 −2 −1

Then add −1 times of the first row to the second and third row, respectively, to
eliminate entries in the first column except the pivot, giving the following tableau:

x1 x2 x3 x4 x5 RHS
1 1/2 3/2 1 5/2
−3/2 1/2 −2 3 −3/2
−1/2 −7/2 −1 −2 −7/2

Take the second entry in the second column as the next pivot. Multiply by −2/3
the second row to turn the pivot to 1, leading to the following tableau:

x1 x2 x3 x4 x5 RHS
1 1/2 3/2 1 5/2
1 −1/3 4/3 −2 1
−1/2 −7/2 −1 −2 −7/2

Add −1/2 and 1/2 times of the second row to the first and third rows,
respectively, to eliminate entries in the second column except the pivot, giving the
following tableau:

x1 x2 x3 x4 x5 RHS
1 5/3 1/3 1 2
1 −1/3 4/3 −2 1
−11/3 −1/3 −3 −3
1.5 Basis and Feasible Basic Solution 19

Take the third entry of the third column as pivot. Multiply by −3/11 the third
row to turn the pivot to 1, leading to the following tableau:

x1 x2 x3 x4 x5 RHS
1 5/3 1/3 1 2
1 −1/3 4/3 −2 1
1 1/11 9/11 9/11

Add −5/3 and 1/3 times of the third row to the first and second rows,
respectively, to eliminate entries in the third column except the pivot, giving the
following tableau:

x1 x2 x3 x4 x5 RHS
1 2/11 −4/11 7/11
1 15/11 −19/11 14/11
1 1/11 9/11 9/11

Consequently, the preceding tableau includes a 3 × 3 unit matrix, and Gauss–

Jordan elimination is completed. The corresponding so-called canonical system can
be written as
⎧
⎨ x1 = 7/11 − (2/11)x4 + (4/11)x5
x = 14/11 − (15/11)x4 + (19/11)x5 (1.11)
⎩ 2
x3 = 9/11 − (1/11)x4 − (9/11)x5,

which is equivalent to the original system (Propositions 1.5.1 and 1.5.2). However,
the preceding is so simple that itself exhibits general solutions of the system if x4
and x5 are viewed as independent variables.
In particular, setting zero to both x4 and x5 leads to

x1 = 7/11, x2 = 14/11, x3 = 9/11, x4 = x5 = 0,

which is said to be basic solution. In fact, this solution can be directly read from the
RHS side of the last tableau.
Nevertheless, Gauss–Jordan elimination does not always go smoothly if pivots
are determined in the natural order. The process breaks down if some pivot happens
to be zero. Even if all pivots are nonzero, the outcome may still be useless if some
pivot is too close to zero (Sect. 1.1).
20 1 Introduction

Table 1.5.1 Canonical xj1 xj2 ··· xjr xjr+1 ··· xjn RH S
tableau
1 āi1 jr+1 ··· āi1 jn b̄i1
1 āi2 jr+1 ··· āi2 jn b̄i2
.. .. .. .. ..
. . . . .
1 āir jr+1 ··· āir jn b̄ir
āir+1 jr+1 ··· āir+1 jn b̄ir+1
.. .. .. ..
. . . .
āim jr+1 ··· āim jn b̄im

Let us turn to the so-called Gauss–Jordan elimination with complete pivoting.

Put the augmented matrix (A|b) to the following tableau:

x1 x2 ··· xn RHS
a11 a12 ··· a1n b1
a21 a22 ··· a2n b2
.. .. .. .. ..
. . . . .
am1 am2 ··· amn bm

Without loss of generality, suppose that the first r pivots are in the natural order
and eliminations are performed such that the r × r northwest corner submatrix of A
becomes a unit one, as shown in Tableau 1.5.1.
At the next step, select āpq from entries within the non-pivot rows and columns
such that

|āp q | = max{|āi j | | i = ir+1 , . . . , im ; j = jr+1 , . . . , jn }. (1.12)

Assume that it is nonzero. Taking it as the (r + 1)-th pivot, we multiply row p by

1/ap q and add −ai q , (i = ir+1 , . . . , im ; i = p) times of the row to corresponding
rows, respectively. To complete this step, we perform row and column permutations
such that the (r + 1) × (r + 1) northwest submatrix becomes unit.
The process terminates if r = m, or r < m but āp q = 0, that is, the (m − r) ×
(n − r) southeast submatrix vanishes. The final tableau is called canonical tableau.
It is clear that if r = m, or r < m but b̄it , ∀t = r + 1, . . . , m, vanish, and hence all
non-pivot rows can be dropped from the tableau, system Ax = b is consistent.
According to linear algebra, such a key positive integer r equals the rank of A.
Thus, the following can be claimed.
Theorem 1.5.1 The system Ax = b (m < n) has solutions if and only if r = m,
or r < m but all entries in the non-pivot rows of the right-hand side vanish. In the
latter case, it has infinitely many solutions.
1.5 Basis and Feasible Basic Solution 21

The index set B = {j1 , . . . , jr } of pivot columns is said basic, and N = A\B =
{jr+1 , . . . , jn } said nonbasic. The associated variables are said basic and nonbasic
correspondingly.
Recall that for the standard LP problem, it is usually assumed that rank A = m.
That is to say, Ax = b is always consistent, and the final canonical tableau offers a
set of solutions to Ax = b, i.e.,

xB = b̄ − N̄ xN ,
xN = xN ,

where the nonbasic variables of xN on the right-hand side are viewed as indepen-
dent.
Definition 1.5.1 The solution resulting from setting the nonbasic variables to zero
is a basic solution. If its basic components are all nonnegative, it is a feasible basic
solution.
Thus, the canonical tableau corresponds to the basic solution

x̄B = b̄, x̄N = 0,

which is a feasible basic solution if x̄B ≥ 0.

Theorem 1.5.2 A feasible solution is a feasible basic solution if and only if the
columns of A corresponding to its positive components are linearly independent.
Proof Necessity. It is clear that all positive components of a feasible basic solution
are basic ones and hence correspond to linearly independent columns.
Sufficiency. Assume that a feasible solution has s positive components corre-
sponding to linearly independent columns. Since rank A = m, it holds that s ≤ m.
If s = m, then the feasible solution is clearly a basic solution. If, otherwise, s < m,
there exist additional m − s columns of A, which and the s columns together
constitute a linearly independent set of columns. Clearly, the solution is just the
basic one associated with the canonical form, corresponding to the set taken as basic
columns. 

Definition 1.5.2 A nonsingular m × m submatrix of A is a basis. It is a feasi-
ble/optimal basis if the corresponding basic solution is feasible/optimal.
There exists a basis for the standard LP problem since A has full row rank r = m.
Therefore, Ax = b has an infinite number of solutions. The concept of basis plays
a central role in the simplex methodology.
By Gauss–Jordan elimination with column/row permutations, any basis can be
converted into m × m unit matrix, and hence, (A|b) converted into a canonical
tableau. It is to say, a basis corresponds to a canonical tableau, and hence a
unique basic solution. But it is not a 1-to-1 correspondence: a basic solution could
correspond to multiple bases, which is closely related to the so-called degeneracy
(Sect. 3.4).
22 1 Introduction

Based on the fact that the objective function is continuous over a closed feasible
region, we conclude this section with the following claim:
Proposition 1.5.3 The answer to any LP problem must be one of the following three
cases:
(i) Infeasible problem: There exists no feasible solution.
(ii) Unbounded problem: There exists a feasible solution, but the feasible value is
lower unbounded over the feasible region.
(iii) There exists an optimal basic solution.

References

Bixby RE, Gregory JW, Lustig IJ, Marsten RE, Shanno DF (1992) Very large-scale linear
programming: a case study in combining interior point and simplex methods. Oper Res 40:885–
897
Forsythe GE, Malcom MA, Moler CB (1977) Computer methods for mathematical computations.
Prentice-Hall, Englewood Cliffs
Wilkinson JH (1971) Modern error analysis. SIAM Rev 13:548–568
Chapter 2
Geometry of Feasible Region

The feasible region P , defined by Definition 1.4.1, is of great importance to the LP

problem. Theories and methods of LP are closely related to P , without exception.
In this chapter, we introduce its special structure in geometry, including P as a
polyhedral convex set, interior point, relative interior point, face, vertex, extreme
direction, representation of P , optimal face and optimal vertex, graphic approach to
LP, heuristic characteristic of an optimal solution, and feasible direction and active
constraint.
In general, P can be degenerate or even empty. The following lemma (Farkas
1902) renders a sufficient and necessary condition for nonempty P (for the proof,
see the end of Sect. 5.2). Readers can refer to the literature on the basic concepts
and operations of Rn space. Rn space.
Lemma 2.0.3 (Farkas) Assume A ∈ Rm×n and b ∈ Rm . The feasible region P is
nonempty if and only if

b T y ≥ 0, ∀ y ∈ {y ∈ Rm | AT y ≥ 0}.

Figure 2.1 serves as a geometric explanation for the preceding Lemma, where
a1 , a2 , a3 are the columns of A ∈ R2×3 .
The area Y between the vectors v and w corresponds to all vectors that form an
angle no more than π/2 with each column vector of A. It is seen that each y ∈ Y
forms an angle no more than π/2 with vector b1 , but not with b2 . Therefore, P is
nonempty when b = b1 but empty when b = b2 .
Thereafter in this book, discussions will be based on the following:
Assumption 1 The feasible region P is nonempty and infinite.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 23
P.-Q. PAN, Linear Programming Computation,
https://doi.org/10.1007/978-981-19-0147-8_2
Discovering Diverse Content Through
Random Scribd Documents
exhilaration of body and mind, they now experience a sad reverse, in which they find
much more pleasing music in the prudent advice of the apothecary, than in all the Odes
of Anacreon. The cry is not then,

Let us drain the nectar’d bowl,

Let us raise the song of soul—&c.

But,

Let us drain the saline dose,

Let’s expel these humours gross.

Now, though poets have favoured us with many a canto on the raptures inspired by
flowing bowls and sparkling goblets, they have rarely condescended to give us one line, if
it were only by way of note, on the “state of the stomach” on the morning after one of
those “nights and suppers of the gods.” Such a detail indeed was never intended for the
divine art of poesy. It is a job not at all calculated for the lover of agreeable fiction, and
hence the world hears little on the subject. Those after-reckonings are nevertheless
serious, though unpalatable things. Pleasure here acts much like a tavern host, who
remembers most accurately all the good things he provides, though his guests are both
apt and willing to forget them. Every item is carefully put down, and must be paid for. I
shall only say, that fortunate is he who takes warning in time. I might moralise on this
theme in good set phrase, but the ground has been so well and so frequently beaten by
others, that I forbear. With respect to such articles as opium and spirits, the “spirit of the
age,” as I have already intimated, runs quite in an opposite direction to that of
indulgence; and it is wisely considered that as those who can be temperate in the use of
such ticklish commodities must owe a great deal to a happy temperament of constitution,
and be few in number, whilst the greater part of mankind are not so felicitously moulded,
the rule of teetotalism, viz., entire abstinence, is on the whole the safest and best. But
there is one article in our pharmacopœia of stimulants, upon which there seems to be
some difference of opinion, and with regard to which I should wish to record my humble
opinion. I allude to the nicotian leaf-tobacco.
Now, I regret to say that a long and attentive study of the subject compels me to
pronounce an unfavourable sentence on this article. Whatever value it may possess as a
medicine—and that, in the present state of our knowledge, is not much—I must say that,
as an instrument of luxury in ordinary use, it is unwholesome and injurious. To the
physician it may be satisfactory to ascertain in what way, precisely, the injurious effect is
produced; but it may suffice others to learn from experience and observation what is the
actual result. It is obvious that tobacco causes an excitement of the nervous system, and
thus disturbs the course of nature; but nature never is, and never can be, disturbed with
impunity. To apply a stimulus to the system for which there is no natural demand, is to
cause a waste of nervous energy, of which nature has need for her own legitimate
purposes, and therefore to inflict an injury upon her, greater or less according to the
amount of that uncalled for expenditure. To keep such an unnatural stimulus in constant
action, is tantamount to the creation of a constitutional derangement of the functions, or
the introduction of an actual disease into the body; and nobody will pretend to say that
this is not injurious. To my simple apprehension, it is anti-hygeian practice with a
vengeance. I am no physician, but I believe this to be the true theory of our subject,
regarded in a physiological point of view, and it is decisive against the nicotian habit,
however small the quantity of the article used may be. People are rather indisposed to
believe that an “agreeable” sensation can be an “unwholesome” one; but unfortunately
for poor humanity, and the popularity of us sages, nothing in nature is more certain than
the possibility of such a conjuncture. It is not only certain, but, alas, commonly known by
experience, that an agreeable thing may be unwholesome, and a pleasant sensation
anything at all but a symptom of healthful action.
Again, people are apt to suppose that no injury is done to their health, because they are
not sensible of the wound at the moment; but this also is a notion which we must class
among vulgar errors. It is a matter of demonstration, not merely of hypothesis, that we
may sustain most grievous injury of which we are not instantly sensible; nay, that so long
a time may elapse after the impression has been imparted, that we become unable to
trace the effect to its cause; and yet the relation of cause and effect stands sure,
however ignorant or unconscious we may be of it. As an illustration of this position, I
shall mention a case which came under my own observation. I was once acquainted with
a gentleman, who at eighty years of age was what would be called a stout, healthy old
fellow. He was certainly of a most robust constitution, and had never addicted himself to
any habit “calculated to shorten life,” as they say at the Insurance Offices, saving and
excepting that of taking snuff. Well, it has been said to me, “See how your anti-nicotian
theory is set at defiance by this hearty old fellow. If tobacco be a slow poison, it must be,
as was said of tea, very slow indeed, or how should we have such an exemplary
octogenarian as this, ‘o’er all its ills victorious?’ He has been taking snuff all his life, and
yet, you perceive, is nothing the worse for it.” Now, I did not perceive any such thing, but
was well aware that the contrary was the case. I was of opinion, and am now fully
convinced of the fact, that he suffered extremely, nay, intensely, from the habit, without
himself or others being at all aware of it. I do not speak of a nose and face perpetually
begrimed with snuff—of a waistcoat and inexpressibles embrowned and powdered all
over with it—of the expenditure of pocket-handkerchiefs, and waste of time in nose-
blowing—everlasting sneezing and coughing, &c.: such matters are mere trifles in the
estimate of your professed snuff-takers; but I do speak of an habitual depression of
spirits, and frequently an access of the most miserable melancholy, to which this
gentleman was subject, and which I attribute to his inveterate habit of snuff-taking, and
to no other cause. He would complain bitterly of his wretchedness on those occasions,
and ascribe it to the skyey influences—the humidity of our climate, the fogs, and I know
not what besides; but it was nothing but “the snuff.” Such intelligence would have
doubtless been very unwelcome; for this very snuff—this actual fons et origo malorum,
ay, “more snuff”—was his most favourite remedy and consolation under these distressing
visitations! So much for our ignorance of causes.
The late Doctor Adam Clarke was a great enemy to the tobacco leaf, and published a
strong paper in condemnation of it. He takes high ground upon the subject.
“That it is sinful to use it, as most do,” he says, “I have no doubt—if destroying the
constitution, and vilely squandering away the time and money which God has given for
other purposes, may be termed sinful. I have observed some whole families, and very
poor ones too, who have used tobacco in all possible ways, and some of them for more
than half a century. Now, suppose the whole family, consisting of four, five, or six, to
have used but 1s. 6d. worth a-week, then, in the mere article of tobacco, nearly £200
sterling is totally and irrecoverably lost in the course of fifty years. Were all the attending
expenses, such as appropriate implements, neglect of business, and other concomitants,
taken into account, probably four times the sum would be too small an estimate.”
Captain Scott, in his interesting work “Rambles in Egypt and Candia,” says—
“All the Arab race are addicted to the use of the pipe, and to this pernicious habit may be
traced the origin of most of their vices, and a great proportion of their misery.” And
again, in a note, he observes—“Nothing tends so much as the pernicious and universal
habit of smoking to retard all improvement amongst the natives of the East, producing
habitual indolence, and occasioning an irreparable loss of time.” He calls it elsewhere the
“predominant vice of Mahomedanism.” Now, with such warning and such examples
before me, I own that I cannot contemplate the possibility of my countrymen becoming a
nation of smokers, without the utmost pain. I would wish to put all parties, but especially
the young, on their guard against the insidious and seductive approaches of the habit.
The elegant pipe, the splendid snuff-box, and all the curious conveniences of tube, light,
tobacco-pouch, and so on, are so many lures to the unwary; and many, by simply
nibbling at these captivating baits, have been gradually led on, and at last turned into
confirmed consumers. There is a temptation in the furniture of our fashionable snuff and
cigar shops—“divans,” as they are called, which it is hard to resist. It would seem almost
worth while to “consume,” for the sake of encompassing oneself with such beautiful toys;
but I class all such resorts in the same category with the gin-palaces of London. Look to
the end—observe what a confirmed habit of snuffing or smoking is—how wasteful, how
enervating, how every way pernicious! The tyranny of it is dreadful. No man knows it
thoroughly but he who has once been its slave. The craving of the nose once accustomed
to be fed, for snuff—of the throat and fauces once seasoned to the use, for smoke—and
of the teeth and gums once used to be drawn, for the reiterate chew—oh, it is dreadful!
—and I say there is no remedy against the evil but teetotalism.
I have said nothing on those popular stimulants, tea and coffee, for, as generally used, I
think they are both innocent, as they are certainly agreeable beverages. Let not my fair
countrywomen, however, when they indulge in the “cup that cheers but not inebriates”—I
mean the Howqua, or any other tea-mixture—aim at celebrity for preparing it over
strong; for in this state, like other stimulants that we have been considering, I have no
doubt that it is bad for weak nerves.
F.

People with one Idea.—There are people who have but one idea: at least if they have more,
they keep it a secret, for they never talk but of one subject. There is Major C——; he has
but one idea, or subject of discourse, Parliamentary Reform. Now, Parliamentary Reform
is (as far as I know) a very good thing, a very good idea, and a very good subject to talk
about; but why should it be the only one? To hear the worthy and gallant Major resume
his favourite topic is like law-business, or a person who has a suit in Chancery going on.
Nothing can be attended to, nothing can be talked of but that. Now it is getting on, now
again it is standing still; at one time the Master has promised to pass judgment by a
certain day, at another he has put it off again, and called for more papers; and both are
equally reasons for speaking of it. Like the piece of pack-thread in the barrister’s hands,
he turns and twists it all ways, and cannot proceed a step without it. Some schoolboys
cannot read but in their own book; and the man of one idea cannot converse out of his
own subject. Conversation it is not, but a sort of recital of the preamble of a bill, or a
collection of grave arguments for a man’s being of opinion with himself. It would be well
if there was any thing of character, any thing of eccentricity in all this; but that is not the
case. It is a political homily personified, a walking common-place we have to encounter
and listen to. It is just as if a man was to insist on your hearing him go through the fifth
chapter of the Book of Judges every time you meet, or like the story of the Cosmogony in
the Vicar of Wakefield. It is a tune played on a barrel-organ. It is a common vehicle of
discourse into which people get and are set down when they please, without any pains or
trouble to themselves. Neither is it professional pedantry or trading quackery: it has no
excuse. The man has no more to do with the question which he saddles on all his hearers
than you have. This is what makes the matter hopeless. If a farmer talks to you about his
pigs or his poultry, or a physician about his patients, or a lawyer about his briefs, or a
merchant about stock, or an author about himself, you know how to account for this; it is
a common infirmity; you have a laugh at his expense, and there is no more to be said.
But here is a man who goes out of his way to be absurd, and is troublesome by a
romantic effort of generosity. You cannot say to him, “All this may be interesting to you,
but I have no concern in it;” you cannot put him off in that way. He has got possession of
a subject which is of universal and paramount interest, and on that plea may hold you by
the button as long as he chooses. His delight is to harangue on what nowise regards
himself; how then can you refuse to listen to what as little amuses you? The business
admits of no delay. The question stands first on the order of the day—takes precedence
in its own right of every other question. Any other topic, grave or gay, is looked upon in
the light of impertinence, and sent to Coventry. Business is an interruption to it, pleasure
a digression from it. As Cicero says of study, it follows the man into the country, it stays
with him at home; it sits with him at breakfast, and goes out with him to dinner. It is like
a part of his dress, of the costume of his person, without which he would be at a loss
what to do. If he meets you in the street, he accosts you with it as a form of salutation; if
you see him at his own house, it is supposed you come upon that. If you happen to
remark, “it is a fine day,” or “the town is full,” it is considered as a temporary compromise
of the question; you are suspected of not going the whole length of the principle. Is not
this a species of sober madness more provoking than the real? Has not the theoretical
enthusiast his mind as much warped, as much enslaved by one idea, as the
acknowledged lunatic, only that the former has no lucid intervals? If you see a visionary
of this class going along the street, you can tell as well what he is thinking of and will say
next as the man that fancies himself a tea-pot or the Czar of Muscovy. The one is as
inaccessible to reason as the other: if the one raves, the other dotes!—Hazlitt’s Table-
Talk.

Comfortable Circumstances favour Foresight.—It is a most remarkable fact, totally at variance

with what might a priori be expected, but confirmed by the universal experience of
mankind, that the dominion of reason over the passions, the habit of foresight, and the
power of forming a systematic plan for the conduct of life, are just in proportion to the
degree in which the danger of immediate or the pressure of actual suffering has been
removed from mankind. The savage who has no stock whatever for his support—who is
in danger of immediate starvation, if his wonted supplies from the chase or his herds
were to fail—is totally regardless of the future in every part of the world; while the rich
man, whose subsistence and affluence are almost beyond the reach of chance, is
incessantly in disquietude about the manner in which his subsequent life is to be spent.
The certain prospect of instant death to himself and all that are dear to him, from the
occurrence of a probable event, is unable to draw the attention of the one from the
enjoyments of the moment; while the slight and improbable chance of a diminution in the
smallest articles of future comfort, renders the other indifferent to the means of present
enjoyment which are within his reach.—Alison’s Principles of Population.

Appreciation.—After all, it is appreciation rather than praise that is delightful. An artist, for
instance, how tired he must be of hearing his pictures called “beautiful, exquisite!”—of
being told for the one hundredth time that he has surpassed himself; but let any one
point out to him one of his own thoughts on the canvass, which he supposed likely to
escape the general eye, and how grateful it is!

Printed and published every Saturday by Gunn and Cameron, at the

Office of the General Advertiser, No. 6, Church Lane, College Green,
Dublin.—Agents:—R. Groombridge, Panyer Alley, Paternoster Row,
London; Simms and Dinham, Exchange Street, Manchester; C. Davies,
North John Street, Liverpool; Slocombe & Simms, Leeds; Fraser and
Crawford, George Street, Edinburgh; and David Robertson, Trongate,
Glasgow.
*** END OF THE PROJECT GUTENBERG EBOOK THE IRISH PENNY
JOURNAL, VOL. 1 NO. 37, MARCH 13, 1841 ***

Updated editions will replace the previous one—the old editions

will be renamed.

Creating the works from print editions not protected by U.S.

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

START: FULL LICENSE

THE FULL PROJECT GUTENBERG LICENSE
 PLEASE READ THIS BEFORE YOU DISTRIBUTE OR USE THIS WORK

To protect the Project Gutenberg™ mission of promoting the

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

Section 1. General Terms of Use and

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

1.B. “Project Gutenberg” is a registered trademark. It may only

be used on or associated in any way with an electronic work by
people who agree to be bound by the terms of this agreement.
There are a few things that you can do with most Project
Gutenberg™ electronic works even without complying with the
full terms of this agreement. See paragraph 1.C below. There
are a lot of things you can do with Project Gutenberg™
electronic works if you follow the terms of this agreement and
help preserve free future access to Project Gutenberg™
electronic works. See paragraph 1.E below.
1.C. The Project Gutenberg Literary Archive Foundation (“the
Foundation” or PGLAF), owns a compilation copyright in the
collection of Project Gutenberg™ electronic works. Nearly all the
individual works in the collection are in the public domain in the
United States. If an individual work is unprotected by copyright
law in the United States and you are located in the United
States, we do not claim a right to prevent you from copying,
distributing, performing, displaying or creating derivative works
based on the work as long as all references to Project
Gutenberg are removed. Of course, we hope that you will
support the Project Gutenberg™ mission of promoting free
access to electronic works by freely sharing Project Gutenberg™
works in compliance with the terms of this agreement for
keeping the Project Gutenberg™ name associated with the
work. You can easily comply with the terms of this agreement
by keeping this work in the same format with its attached full
Project Gutenberg™ License when you share it without charge
with others.

1.D. The copyright laws of the place where you are located also
govern what you can do with this work. Copyright laws in most
countries are in a constant state of change. If you are outside
the United States, check the laws of your country in addition to
the terms of this agreement before downloading, copying,
displaying, performing, distributing or creating derivative works
based on this work or any other Project Gutenberg™ work. The
Foundation makes no representations concerning the copyright
status of any work in any country other than the United States.

1.E. Unless you have removed all references to Project

Gutenberg:

1.E.1. The following sentence, with active links to, or other

immediate access to, the full Project Gutenberg™ License must
appear prominently whenever any copy of a Project
Gutenberg™ work (any work on which the phrase “Project
Gutenberg” appears, or with which the phrase “Project
Gutenberg” is associated) is accessed, displayed, performed,
viewed, copied or distributed:

This eBook is for the use of anyone anywhere in the United

States and most other parts of the world at no cost and
with almost no restrictions whatsoever. You may copy it,
give it away or re-use it under the terms of the Project
Gutenberg License included with this eBook or online at
www.gutenberg.org. If you are not located in the United
States, you will have to check the laws of the country
where you are located before using this eBook.

1.E.2. If an individual Project Gutenberg™ electronic work is

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

1.E.3. If an individual Project Gutenberg™ electronic work is

posted with the permission of the copyright holder, your use and
distribution must comply with both paragraphs 1.E.1 through
1.E.7 and any additional terms imposed by the copyright holder.
Additional terms will be linked to the Project Gutenberg™
License for all works posted with the permission of the copyright
holder found at the beginning of this work.

1.E.4. Do not unlink or detach or remove the full Project

Gutenberg™ License terms from this work, or any files
 containing a part of this work or any other work associated with
Project Gutenberg™.

1.E.5. Do not copy, display, perform, distribute or redistribute

this electronic work, or any part of this electronic work, without
prominently displaying the sentence set forth in paragraph 1.E.1
with active links or immediate access to the full terms of the
Project Gutenberg™ License.

1.E.6. You may convert to and distribute this work in any binary,
compressed, marked up, nonproprietary or proprietary form,
including any word processing or hypertext form. However, if
you provide access to or distribute copies of a Project
Gutenberg™ work in a format other than “Plain Vanilla ASCII” or
other format used in the official version posted on the official
Project Gutenberg™ website (www.gutenberg.org), you must,
at no additional cost, fee or expense to the user, provide a copy,
a means of exporting a copy, or a means of obtaining a copy
upon request, of the work in its original “Plain Vanilla ASCII” or
other form. Any alternate format must include the full Project
Gutenberg™ License as specified in paragraph 1.E.1.

1.E.7. Do not charge a fee for access to, viewing, displaying,

performing, copying or distributing any Project Gutenberg™
works unless you comply with paragraph 1.E.8 or 1.E.9.

1.E.8. You may charge a reasonable fee for copies of or

providing access to or distributing Project Gutenberg™
electronic works provided that:

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

• You provide a full refund of any money paid by a user who

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

• You provide, in accordance with paragraph 1.F.3, a full refund of

any money paid for a work or a replacement copy, if a defect in
the electronic work is discovered and reported to you within 90
days of receipt of the work.

• You comply with all other terms of this agreement for free
distribution of Project Gutenberg™ works.

1.E.9. If you wish to charge a fee or distribute a Project

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

1.F.

1.F.1. Project Gutenberg volunteers and employees expend

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

1.F.2. LIMITED WARRANTY, DISCLAIMER OF DAMAGES - Except

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

1.F.3. LIMITED RIGHT OF REPLACEMENT OR REFUND - If you

discover a defect in this electronic work within 90 days of
receiving it, you can receive a refund of the money (if any) you
paid for it by sending a written explanation to the person you
received the work from. If you received the work on a physical
medium, you must return the medium with your written
explanation. The person or entity that provided you with the
defective work may elect to provide a replacement copy in lieu
of a refund. If you received the work electronically, the person
or entity providing it to you may choose to give you a second
opportunity to receive the work electronically in lieu of a refund.
If the second copy is also defective, you may demand a refund
in writing without further opportunities to fix the problem.

1.F.4. Except for the limited right of replacement or refund set

forth in paragraph 1.F.3, this work is provided to you ‘AS-IS’,
WITH NO OTHER WARRANTIES OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR ANY PURPOSE.

1.F.5. Some states do not allow disclaimers of certain implied

warranties or the exclusion or limitation of certain types of
damages. If any disclaimer or limitation set forth in this
agreement violates the law of the state applicable to this
agreement, the agreement shall be interpreted to make the
maximum disclaimer or limitation permitted by the applicable
state law. The invalidity or unenforceability of any provision of
this agreement shall not void the remaining provisions.

1.F.6. INDEMNITY - You agree to indemnify and hold the

Foundation, the trademark owner, any agent or employee of the
Foundation, anyone providing copies of Project Gutenberg™
electronic works in accordance with this agreement, and any
volunteers associated with the production, promotion and
distribution of Project Gutenberg™ electronic works, harmless
from all liability, costs and expenses, including legal fees, that
arise directly or indirectly from any of the following which you
do or cause to occur: (a) distribution of this or any Project
Gutenberg™ work, (b) alteration, modification, or additions or
deletions to any Project Gutenberg™ work, and (c) any Defect
you cause.

Section 2. Information about the Mission

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

Volunteers and financial support to provide volunteers with the

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

Section 3. Information about the Project

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

The Foundation’s business office is located at 809 North 1500

West, Salt Lake City, UT 84116, (801) 596-1887. Email contact
links and up to date contact information can be found at the
Foundation’s website and official page at
www.gutenberg.org/contact
 Section 4. Information about Donations to
the Project Gutenberg Literary Archive
Foundation
Project Gutenberg™ depends upon and cannot survive without
widespread public support and donations to carry out its mission
of increasing the number of public domain and licensed works
that can be freely distributed in machine-readable form
accessible by the widest array of equipment including outdated
equipment. Many small donations ($1 to $5,000) are particularly
important to maintaining tax exempt status with the IRS.

The Foundation is committed to complying with the laws

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

While we cannot and do not solicit contributions from states

where we have not met the solicitation requirements, we know
of no prohibition against accepting unsolicited donations from
donors in such states who approach us with offers to donate.

International donations are gratefully accepted, but we cannot

make any statements concerning tax treatment of donations
received from outside the United States. U.S. laws alone swamp
our small staff.

Please check the Project Gutenberg web pages for current

donation methods and addresses. Donations are accepted in a
number of other ways including checks, online payments and
credit card donations. To donate, please visit:
www.gutenberg.org/donate.

Section 5. General Information About

Project Gutenberg™ electronic works
Professor Michael S. Hart was the originator of the Project
Gutenberg™ concept of a library of electronic works that could
be freely shared with anyone. For forty years, he produced and
distributed Project Gutenberg™ eBooks with only a loose
network of volunteer support.

Project Gutenberg™ eBooks are often created from several

printed editions, all of which are confirmed as not protected by
copyright in the U.S. unless a copyright notice is included. Thus,
we do not necessarily keep eBooks in compliance with any
particular paper edition.

Most people start at our website which has the main PG search
facility: www.gutenberg.org.

This website includes information about Project Gutenberg™,

including how to make donations to the Project Gutenberg
Literary Archive Foundation, how to help produce our new
eBooks, and how to subscribe to our email newsletter to hear
about new eBooks.