Atkinson - Resources in Numerical Analysis
Atkinson - Resources in Numerical Analysis
Atkinson - Resources in Numerical Analysis
Kendall E. Atkinson
University of Iowa
Introduction
I. General Numerical Analysis
A. Introductory Sources
B. Advanced Introductory Texts with Broad Coverage
C. Books With a Sampling of Introductory Topics
D. Major Journals and Serial Publications
1. General Surveys
2. Leading journals with a general coverage in numerical analysis.
3. Other journals with a general coverage in numerical analysis.
E. Other Printed Resources
F. Online Resources
II. Numerical Linear Algebra, Nonlinear Algebra, and Optimization
A. Numerical Linear Algebra
1. General references
2. Eigenvalue problems
3. Iterative methods
4. Applications on parallel and vector computers
5. Over-determined linear systems.
B. Numerical Solution of Nonlinear Systems
1. Single equations
2. Multivariate problems
C. Optimization
III. Approximation Theory
A. Approximation of Functions
1. General references
2. Algorithms and software
3. Special topics
4. Multivariate approximation theory
5. Wavelets
B. Interpolation Theory
1. Multivariable interpolation
2. Spline functions
C. Numerical Integration and Differentiation
1. General references
2. Multivariate numerical integration
IV. Solving Differential and Integral Equations
A. Ordinary Differential Equations
B. Partial Differential Equations
C. Integral Equations
V. Miscellaneous Important References
VI. History of Numerical Analysis
INTRODUCTION
Numerical analysis is the area of mathematics and computer science that creates, analyzes, and
implements algorithms for solving numerically the problems of continuous mathematics. Such
problems originate generally from real-world applications of algebra, geometry, and calculus,
and they involve variables that vary continuously; these problems occur throughout the natural
sciences, social sciences, engineering, medicine, and business. During the second half of the
twentieth century and continuing up to the present day, digital computers have grown in power
and availability. This has led to the use of increasingly realistic mathematical models in science
and engineering, and numerical analysis of increasing sophistication has been needed to solve
these more sophisticated mathematical models of the world. The formal academic area of
numerical analysis varies, from quite foundational mathematical studies to the computer science
issues involved in the creation and implementation of algorithms.
In this chapter, we place more emphasis on the theoretical mathematics involved in studying
numerical analysis while also discussing more briefly the resources associated with the computer
science aspects of the subject. The implementation of numerical algorithms is affected by
physical characteristics of the computers being used for the computation, and we consider this in
our presentation. In addition, the purpose of most numerical analysis research is to develop
actual computer codes to solve real problems; and thus the development of computer software to
implement numerical algorithms is an important part of the subject. With the growth in
importance of using computers to carry out numerical procedures in solving mathematical
models of the world, an area known as "scientific computing" or "computational science" has
taken shape during the 1980s and 1990s. This new area looks at the use of numerical analysis
from a computer science perspective. It is concerned with using the most powerful tools of
numerical analysis, computer graphics, symbolic mathematical computations, and graphical user
interfaces to make it easier for a user to set up, solve, and interpret complicated mathematical
models of the real world. We will give a few resources for scientific computing, but numerical
analysis is the focus of this presentation.
Following is a selection of texts that together provide an overview of numerical analysis, given
in order from introductory to specialist.
A Quarteroni, R Sacco, and F Saleri. Numerical Mathematics, Springer-Verlag, New York,
2000. This is a current introduction to most topics in numerical analysis, including the numerical
solution of partial differential equations. The text is suitable for a beginning graduate student in
mathematics.
K Atkinson and W Han. Theoretical Numerical Analysis, Springer-Verlag, New York, 2001.
This introduces a wide variety of functional analysis tools for studying numerical methods.
These are applied to problems in approximation theory, ordinary and partial differential
equations, integral equations, and nonlinear variational inequalities.
G Golub and C Van Loan. Matrix Computations, 3rd ed., John Hopkins University Press, 1996.
This is an excellent reference book for a wide variety of topics in numerical linear algebra at all
levels of the subject. It covers the theoretical framework for a variety of linear algebra problems,
and it also considers the effects of computer arithmetic and serial vs. parallel computers.
A Iserles. A First Course in the Numerical Analysis of Differential Equations, Cambridge
University Press, 1996. This is a very nice introductory presentation of the theoretical
framework for the numerical analysis for both ordinary and partial differential equations.
I. General Numerical Analysis
A. Introductory Sources
The best general introductions to numerical analysis are beginning graduate-level
textbooks that cover most of the commonly recognized topics within numerical analysis.
A few of my current favorites, listed alphabetically, are the following. Each of them
gives an introduction to most of what are considered basic topics in numerical analysis. I
include both current texts and classical texts that are still important references.
M Allen III and E Issacson. Numerical Analysis for Applied Science, John Wiley, New
York, 1998. A very nice introduction.
K Atkinson. An Introduction to Numerical Analysis, 2nd ed., John Wiley, New York,
1989. I am the author of this text, and I believe it gives a good introduction to most
introductory level topics in numerical analysis.
W Gander and J Hrebicek. Solving Problems in Scientific Computing Using Maple and
Matlab, 3rd ed., Springer, 1997.
W Gautschi. Numerical Analysis: An Introduction, Birkhuser, Boston, 1997.
P Henrici. Elements of Numerical Analysis, John Wiley, New York, 1964. This is a
classic text by a master of the subject. It contains well-written discussions of a broad set
of topics. It is dated in some respects, but still contains much that is useful and
interesting.
E Isaacson and H Keller. Analysis of Numerical Methods, corrected reprint of the 1966
original, Dover Pub., New York, 1994. This is a classic text covering many topics not
covered elsewhere. For example, this text contains a very good introduction to finite
difference methods for approximating partial differential equations.
R Kress. Numerical Analysis, Springer-Verlag, New York, 1998.
A Quarteroni, R Sacco, and F Saleri. Numerical Mathematics, Springer-Verlag, New
York, 2000. This was cited earlier in the introduction as giving a very good introduction
to the current state of numerical analysis.
There are over fifty journals devoted to various aspects of numerical analysis; and in
addition, many other mathematics journals contain articles on numerical analysis. Also,
many areas of business, engineering, and the sciences are closely tied to numerical
analysis, and much of the published research in those areas is about numerical methods of
solving problems for the area of application. We highlight here some of the major
journals and serial publications in numerical analysis.
1. General surveys
Acta Numerica (ISSN 0962-4929) published by Cambridge University Press, is an
annual collection of survey articles in a variety of areas of numerical analysis,
beginning in 1992. Articles in this annual publication are intended to contain new
research results and to also survey the state of the field as regards current research at
its leading edge. If you are seeking information on the current state of art in some
area of numerical analysis, I recommend you begin by looking at the table of contents
of all of the issues of this annual publication. This is an excellent collection of
articles!
The State of the Art in Numerical Analysis is the name of a research conference held
every ten years in the United Kingdom, and it is also the name of the accompanying
volume of the survey talks presented at the conference. The most recent such
conference was in 1996; and there have been four such conferences. The most recent
volume is referenced as follows.
I Duff and G Watson, editors. The State of the Art in Numerical Analysis, Oxford
University Press, Oxford, 1997.
I strongly recommend the articles in these volumes as introductions to many areas of
numerical analysis.
SIAM Review (ISSN 0036-1445) is the flagship journal for the Society of Industrial
and Applied Mathematics, and it is published quarterly. The first part of the journal
contains survey articles, and a number of these are survey articles on the current state
of the research in some area of numerical analysis.
2. Leading journals with a general coverage in numerical analysis.
SIAM (Society of Industrial and Applied Mathematics) has a number of journals
which are devoted to various aspects of numerical analysis. The SIAM Journal on
Numerical Analysis (ISSN 0036-1429) is probably the most popular and referenced
journal in numerical analysis. Its coverage is all of numerical analysis. Other SIAM
journals that deal with special areas of numerical analysis are SIAM Journal on
Scientific Computing (ISSN 1064-8275), SIAM Journal on Matrix Analysis (ISSN
0895-4798), and SIAM Journal on Optimization (ISSN 1052-6234). These are all
well-referenced journals. In addition, important review articles for topics in
software. We list here a few online resources that contain information on numerical
analysis and associated software; and there are many sites we do not list.
sci.math.num-analysis. This is a popular Usenet bulletin board devoted to numerical
analysis.
Netlib (http://www.netlib.org/). This is the most extensive online library of numerical
analysis software. It is operated jointly by Oak Ridge National Laboratory and the
University of Tennessee. Many important software projects are available through this
site; for example, LINPACK, EISPACK, and LAPACK are important numerical linear
algebra packages which now form the core of most numerical analysis libraries in the
linear algebra. The Collected Algorithms of the ACM are also available here, while first
being published in the ACM Transactions on Mathematical Software.
NIST GAMS (http://gams.nist.gov/). This acronym denotes the National Institute for
Science and Technology Guide to Available Mathematical Software. It is an excellent
guide to numerical analysis software, some of which is free through Netlib. In addition,
the site http://math.nist.gov/ provides links to other areas of numerical computation.
The Mathematical Atlas.
(http://www.math.niu.edu/~rusin/known-math/index/tour_na.html). This is a general
look at all of mathematics, and numerical analysis is one of the options provided. The
sponsor is Northern Illinois University.
Scientific Computing FAQ
(http://www.mathcom.com/corpdir/techinfo.mdir/scifaq/index.html). The acronym
translates to "Frequently Asked questions". It lists a large amount of information on
numerical analysis and associated areas within Scientific Computing. The sponsor is
MathCom Inc.
II. Numerical Linear Algebra, Nonlinear Algebra, and Optimization
This refers to problems involving the solution of systems of linear and nonlinear equations
and the related problem of optimizing a function of several variables; often the number of
variables is quite large.
A. Numerical Linear Algebra
Many problems in applied mathematics involve solving systems of linear equations, with
the linear system occurring naturally in some cases and as a part of the solution process in
other cases. Linear systems are usually written using matrix-vector notation, Ax=b with A
the matrix of coefficients for the system, x the column vector of the unknown variables,
and b a given column vector. Solving linear systems with up to n=1000 variables is now
considered relatively straightforward in most cases. For small to moderate sized linear
systems (say n 1000), the favorite numerical method is Gaussian elimination and its
variants; this is simply a precisely stated algorithmic variant of the method of elimination
J Dongarra, I Duff, D Sorensen, & H van der Vorst. Solving Linear Systems on Vector
and Shared Memory Computers, SIAM Pub., 1991. This discusses the general
problems involved in solving linear systems on parallel and vector pipeline
computers.
5. Over-determined linear systems
Another important type of linear system to solve is the "over-determined linear
system". The most popular method for solving such systems is called the "method of
least squares". This refers to solving linear systems Ax=b in which the matrix A is of
order mxn, usually with m much larger than n. Then a "solution" is found by
attempting to minimize the Euclidean size of the vector Ax-b. This is a difficult and
important problem that occurs regularly in nonlinear regression analysis in statistics.
For an up-to-date accounting of this problem and its solution, see the following text.
A Bjorck. Numerical Methods for Least Squares Problems, SIAM Pub., 1996.
B. Numerical Solution of Nonlinear Systems
Solving nonlinear equations is often treated numerically by reducing the solution process
to that of solving a sequence of linear problems. As a simple but important example,
consider the problem of solving a nonlinear equation f(x)=0. Given an estimate of the
root, approximate the graph of y=f(x) by the tangent line at ; and then use the root of the
tangent line to obtain an improved estimate of the root of the original nonlinear function
f(x). This is called Newton's method for rootfinding.
This procedure generalizes to handling systems of nonlinear equations. Let f(x)=0 denote
a system of n nonlinear equations in the n unknown components of x. In this, the role of
the derivative is played by f'(x), the Jacobian matrix of f(x). To find the root of the
approximating linear approximation, we must solve a linear system f'(
)
=f(
), a linear
system of order n. There is a large literature on Newton's method for nonlinear systems
and on ways to increase its efficiency. There are numerous other approaches to solving
nonlinear systems, most based on using some type of approximation by linear functions.
1. Single equations
For solving a single equation of one variable, f(x)=0, see the following books.
J Traub. Iterative Methods for the Solution of Equations, Prentice-Hall, 1964. This is
a classic text that examines this problem from almost every imaginable perspective.
A Householder. The Numerical Treatment of a Single Nonlinear Equation, McGrawHill, New York, 1970. This is a classic treatment of the quite old, but still interesting,
problem of finding the roots of a nonlinear equation. Included are special methods
for polynomial rootfinding.
2. Multivariate problems
For introductions to the numerical solution of nonlinear systems, see the following
texts.
J Ortega & W Rheinboldt. Iterative Solution of Nonlinear Equations in Several
Variables, Academic Press, 1970. This is a classic text on the subject. It covers
many types of methods in great detail, and it also presents the mathematical tools
needed to work in this area.
C Kelley, Iterative Methods for Linear and Nonlinear Equations, SIAM Pub., 1995.
This is an excellent introduction to the general area of solving nonlinear systems.
In addition, the solution of nonlinear systems for particular types of problems is often
discussed in the intended area of application. For example, discretizations of nonlinear
partial differential equations leads to special types of nonlinear systems, and special types
of methods have been developed for solving these systems. The solution of such
nonlinear systems is discussed at length in the literature for solving partial differential
equations.
C. Optimization
An important related class of problems occurs under the heading of optimization,
sometimes considered as a sub-area of "operations research". Given a real-valued
function f(x) with x a vector of unknowns, we wish to find a value of x which minimizes
f(x). In some cases x is allowed to vary freely, and in other cases there are constraints on
the values of x that can be considered. Such problems occur frequently in business and
engineering applications. This is an enormously popular area of research, with many new
methods having been developed recently, both for classic problems such as that of linear
programming and for previously unsolvable problems. In many ways, this area needs a
chapter of its own; and in the classification scheme for Mathematical Reviews, it has a
category (MR90) separate from that of numerical analysis (MR65).
For introductions with a strongly mathematical flavor, see the following, listed
alphabetically. Most of these books also address the practical problems of
implementation.
We begin our list of resources and references with a specialist journal; otherwise, the list
is alphabetical.
SIAM Journal on Optimization (ISSN 1052-6234), SIAM Pub. An important journal for
articles on optimization theory, especially from a numerical analysis perspective.
D Bertsekas. Nonlinear Programming, Athena Scientific Pub., 1995.
3. Special topics
G Baker and P Graves-Morris. Pade Approximants, 2nd ed., Cambridge University
Press, 1996.
T Rivlin. Chebyshev Polynomials, John Wiley, 1974.
G Szego. Orthogonal Polynomials, 3rd ed., American Mathematical Society, 1967.
A Zygmund. Trigonometric Series, Vols. I and II, Cambridge University Press, 1959.
4. Multivariate approximation theory
In the not too distant past, most approximation theory dealt with functions of one
variable, whereas now there is a greater interest in functions of several variables. For
example, see the following references to the current literature on multivariate
approximation theory.
C de Boor. Multivariate Piecewise Polynomials, Acta Numerica (1993), pp. 65-110.
C Chui. Multivariate Splines, SIAM Pub., 1988.
M Sabin. Numerical Geometry of Surfaces, Acta Numerica (1994), pp. 411-466.
5. Wavelets
"Wavelets" furnish a means to combine the separate advantages of Fourier analysis
and piecewise polynomial approximation. Although the first example of wavelets,
the Haar function, goes back to 1910, much of the research on wavelets and their
application is from 1980 onwards. Connected to wavelets is the idea of
"multiresolution analysis", a decomposition of a function or a process into different
levels of precision. Wavelets and multiresolution analysis is a very active area at
present, and we give only some of the better known references on it.
C Chui. An Introduction to Wavelets, Academic Press, 1992.
I Daubechies. Ten Lectures on Wavelets, SIAM Pub., 1992.
R DeVore and B Lucier. Wavelets, Acta Numerica (1992), pp. 1-56. This gives an
approximation theoretic perspective of wavelets and multiresolution. It is a tightly
written article, and it has a good bibliography.
S Mallat. Multiresolution Approximation and Wavelet Orthonormal Bases of L2(R),
Transactions of the Amer. Math. Soc. 315 (1989), pp. 69-87.
good introduction to the theory of numerical methods for solving ordinary differential
equations.
J Sanz-Serna and M Calvo. Numerical Hamiltonian Problems, Chapman & Hall, 1994.
L Shampine. Numerical Solution of Ordinary Differential Equations, Chapman & Hall,
1994. An excellent text with good coverage of questions of software implementation, by
one of the foremost designers of such software.
B. Partial Differential Equations
This is a truly enormous subject, on the same scale as the entire general area of algebra.
It encompasses many types of linear and nonlinear partial differential equations (PDE);
and the various types of applications lead to other areas of specialization in studying the
mathematics, e.g. the mathematics of fluid flow and the study of the Navier-Stokes
equation. Among the most important linear PDE are those of order two; they are
classified as elliptic, parabolic, and hyperbolic, and these correspond to different types of
physical phenomena. There are important equations of orders other than two, especially
of order one; and many of the most important PDE are nonlinear.
Numerical methods are similarly quite varied, and we can merely skim the surface with
our references. With hyperbolic problems, finite difference methods are the dominant
form of discretization, with some aspects of the treatment of elliptic problems used as
tools. Elliptic problems are solved principally using finite difference and finite element
methods, with boundary element methods also important. Parabolic problems (linear and
nonlinear) were solved traditionally by finite difference methods, but the "method of
lines" has become a favored treatment in the past 30 years. We give a sampling of books
and apologize for what the reader may consider to be major omissions.
We begin our list of resources and references with a specialist journal; otherwise, the list
is alphabetical.
Numerical Methods for Partial Differential Equations (ISSN 0749-159X), John Wiley.
This is an important journal for research on the numerical solution of partial differential
equations.
D Braess. Finite Elements, Cambridge University Press, 1997. A good introduction to
finite element methods.
F Brezzi and M Fortin. Mixed and Hybrid Finite Element Methods, Springer-Verlag,
1991.
W Briggs, V-E Henson, S McCormick. A Multigrid Tutorial, 2nd ed., SIAM Pub., 2000.
The multigrid method is the most powerful iteration method for solving discretizations of
elliptic PDE, and this is an up-to-date and well-written introduction to that method.
C. Integral equations
Integral equations arise directly and as reformulations of ordinary and partial differential
equations. Boundary value problems for differential equations can be reformulated as
Fredholm integral equations, and initial value problems can be reformulated as Volterra
integral equations. Both of these types of integral equations are more general than the
reformulations, and most Fredholm and Volterra integral equations cannot, in turn, be
returned to a differential equation formulation. There are also integral equations of many
other types, including Cauchy singular integral equations and hypersingular integral
equations. All of these have important applications in the natural sciences and
engineering.
We begin our list of resources and references with a specialist journal; otherwise, the list
is alphabetical.
Journal of Integral Equations and Applications (ISSN 0897-3962), Rocky Mountain
Mathematics Consortium. This journal is devoted to the study of integral equations,
especially to the numerical analysis for them and to applications of them.
K Atkinson. The Numerical Solution of Integral Equations of the Second Kind,
Cambridge University Press. This is a comprehensive look at linear integral equations of
Fredholm type. Chapters 7-9 give an introduction to boundary integral equations and
their numerical solution.
C Baker. The Numerical Treatment of Integral Equations, Oxford University Press,
1977. This is a classic text that ranges widely in discussing most types of integral
equations.
H Brunner and P van der Houwen. The Numerical Solution of Volterra Equations,
North-Holland Pub., 1986. This summarizes well the literature on solving Volterra
integral equations. In more recent years, researchers have extended these results to
Volterra integro-differential equations.
D Colton and R Kress. Inverse Acoustic and Electromagnetic Scattering Theory, 2nd ed.,
Springer-Verlag, 1998. An excellent presentation of the subject in the title.
C Groetsch. Inverse Problems in the Mathematical Sciences, Friedr. Vieweg & Sohn,
Braunschweig, 1993. This is a very good introduction to ill-posed inverse problems and
their numerical solution. Many such problems are posed as integral equations, and that is
the author's focus in this book.
W Hackbusch. Integral Equations: Theory and Numerical Treatment, Birkhaeuser
Verlag, 1995. This gives a general look at integral equations and their numerical
analysis.
R Kress. Linear Integral Equations, 2nd ed., Springer-Verlag, 1999. This is a masterfully
written introduction to the theory of integral equations. In addition, it has several wellwritten chapters on the numerical solution of integral equations.
P Linz. Analytical and Numerical Methods for Volterra Equations, SIAM Pub., 1985.
S Proessdorf and B Silbermann. Numerical Analysis for Integral and Related Operator
Equations, Birkhaeuser Verlag, 1991. This book gives a quite abstract, and important,
framework for the singular integral equations associated with many boundary integral
equation reformulations of PDE.
V. Miscellaneous Important References
There are a number of important references that do not fit easily into the above schemata.
We present some of those here.
P Henrici. Applied and Computational Complex Analysis, Vols. I, II, and III, John Wiley,
1974-1986. These volumes contain a wealth of information on complex analysis and its
application to a wide variety of mathematical problems.
M Overton. Numerical Computing with IEEE Floating Point Arithmetic, SIAM Pub., 2001.
This is an excellent discussion of floating-point arithmetic for numerical computations,
carried out for the format of such arithmetic used on most digital computers.
F Stenger. Numerical Methods Based on Sinc and Analytic Functions, Springer-Verlag,
1993. This develops the theory of the sinc function and applies it to a wide variety of
problems.
VI. History of Numerical Analysis
Numerical analysis is both quite old and quite young. It is old in that most people doing
mathematics in the past, including the quite distant past, did numerical calculations and
developed numerical algorithms. It is young in that much of what we study today has been a
consequence of the use of digital computers, and computers continue to shape the area today.
H Goldstine. A History of Numerical Analysis: From the 16th Through the 19th Century,
Springer-Verlag, 1977. This is a very good presentation of the development of numerical
algorithms, beginning with logarithms, looking in depth at the contributions of Newton, and
going onto the contributions of Euler and others up through the end of the 1800s.
J-L Chabert, et al.. A History of Algorithms: From the Pebble to the Microchip, SpringerVerlag, 1999. This is a history of computing from the perspective of developing algorithms,
but it has a significant overlap with numerical computing.
S Nash (editor). A History of Scientific Computing, ACM Press, 1990. This is a collection of
articles discussing numerical analysis, especially in association with the use of digital
computers. It fills in some of the history of numerical analysis in the 20th Century.