NA Digest Sunday, December 24, 2006 Volume 06 : Issue 52

Today's Editor:
Tamara G. Kolda
Sandia National Labs

Submissions for NA Digest:

Mail to

Information via email about NA-NET:

Mail to


From: "David McIlhagger" <>
Date: Mon, 18 Dec 2006 16:59:58 -0000
Subject: Banded or Regularly Structured Systems

I have a method to form a banded approximation to a system of equations and
was wondering if there were any available cheap methods to solve for the
eigenvalues that avail of the structure of the problem.

David McIlhagger
School of Electrical & Electronic Engineering
The Queen's University of Belfast
Phone: +44 (0)28 90974085


Date: Tue, 19 Dec 2006 10:11:19 +0200
Subject: Whereabouts of Juan Rivero?

Does anyone have a current email address for Juan Rivero? He was with the IBM
Scientific Centers in Los Angeles and then Caracas (Venezuela). Subsequently he
taught Computer Science in Alaska until about 2003.

Thank you very much
Al Inselberg


From: Ying Wai Fan <>
Date: Mon, 18 Dec 2006 20:57:24 -0500 (EST)
Subject: Re: Fortran vs Matlab

In response to the recent discussion on Fortran vs Matlab, I would like to
draw people's attention to a third language Python. Python is kind of in
between Fortran and Matlab, it is very expressive and yet can call
low-level libraries. It comes with an interactive shell, so you can play
around with it as in Matlab. Python packages Numpy, Scipy and Matplotlib
provide functions similar to those in Matlab. Python can read a lot of
different data files including Fortran data files and Matlab mat files.

What I really like about Python is that it is a "glue" language. You can
call C, Fortran libraries and even Matlab functions in Python with almost
no effort at all using a wrapper. So no more mex-files as in Matlab.

Building user interface is also pretty easy in Python. I just built a web
interface for my course project on image deblurring. You can also use
Tcl/Tk, GTK, WxWindows toolkits with Python.

Ying Wai (Daniel) FAN
PhD Candidate
Dept of Mathematics and Computer Science
Emory University
(404)727-5611 fax


From: "Pereyra, Victor" <>
Date: Mon, 18 Dec 2006 10:21:59 -0800
Subject: MATLAB

I have been following this lively discussion and I would like to make a
distinction that does not come out on the previous discussions:

a) Teaching Numerical Analysis to the crowds (Engineers, Mathematicians, etc.)
for problem solving, as a requirement, etc.

b) Teaching Numerical Analysis to specialists (Computational Mathematicians).

For (a) I believe that MATLAB (of which I am not an user) or the likes are
probably adequate in order to concentrate on the NA, rather that in the
implementation details. I believe also that these students, if eventually they
go to a real working environment that is not MATLAB geared and are required to
interact with real programming languages would regret not having learned the
details while they were young.

For (b), including and specially for Graduate students writting theses in
Scientific Computing programs, I feel strongly that a MATLAB (or the likes)
centered approach is not appropriate. After all, who develops MATLAB and the
likes? At least one large project using a real programming laguage is
essential for any people that expect to do practical work in Scientific
Computing. It is not a waste of time and it is not demeaning to program and
debug, it is an essential part of our work. NA people were delivering open
software long before the term was dreamt off: it is an important part of our
contribution to society.

Victor Pereyra
Weidlinger Associates Inc.
Mountain View, CA 94040
Voice (650) 230-0210
FAX (650) 230--0209


From: David Stewart <>
Date: Mon, 18 Dec 2006 14:33:11 -0600
Subject: Re: MATLAB vs Fortran

Many people responded to Mike Gander's question about programming
languages for teaching numerical analysis. Most of them, and we,
think that MATLAB is great teaching language for numerical
analysis, as it leaves time to concentrate on the numerical issues.

However, Louis Komzsik says that using "canned" routines
is not good enough for training tomorrow's scientific
programmers. We agree, but perhaps the best place for teaching
these issues is not in a numerical analysis course. Teaching
one language (e.g., Fortran 90+, or C/C++)
is also not adequate given that most programmers
have to live in a multi-language software world.

Recently we wrote a book published by Cambridge University Press:
Writing Scientific Software: a guide to good style
ISBN: 0521675952 (paperback), 0521858968 (hardcover)
This book is about all the other stuff that is important
to actually programming numerical algorithms including
software engineering, blocked algorithms, efficient use of memory
hierarchies, efficient object-oriented programming, BLAS
& LAPACK, software tools, modularization, etc.
Regarding language, we treat C, C++, Fortran 90, and Java.

We encourage courses treating these issues --- our graduate
students need them.

Suely Oliveira and David Stewart
University of Iowa


From: Dana Proctor <>
Date: Mon, 18 Dec 2006 21:49:19 -0700
Subject: Re: Fortran versus Matlab in teaching numerical analysis

With regard to Mr. Gander's question:
Should one use Fortran, one of the top 10 algorithms of the 20th
century, a language in which most large scale existing codes are
written ? Or should one use Matlab, or even C++ or Java ?

First I like to state that I'm a layman in this area, but did take several
Numerical Analysis classes in college and created a tool which I call
NA_WorkSheet: I subscribe to the
NA_Digest, I guess just as after taking a year of economics, I like to read
the Wall Street Journal. Anyway from the web site Britannica they give this
definition (

"Branch of applied mathematics that studies methods for solving complicated
equations using arithmetic operations, often so complex that they require a
computer, to approximate the processes of analysis (i.e., calculus).

The arithmetic model for such an approximation is called an algorithm, the set
of procedures the computer executes is called a program, and the commands that
carry out the procedures are called code. An example is an algorithm for
deriving p by calculating the perimeter of a regular polygon as its number of
sides becomes very large. Numerical analysis is concerned not just with the
numerical result of such a process but with determining whether the error at
any stage is within acceptable bounds."

I feel the education community does a dis-service to it students in not
approaching the subject from the perspective of a programming language. I
realize many of the students coming to you have been raised in a graphical
user environment, but the necessity of learning to understand, and create
mathematical algorithms in code I feel will benefit them in their future
careers. Surely the engineers among them will more likely be using a tool like
Maple, both those that are mathematical majors, need to be able to understand
and evaluate coded algorithms to better insure that results given to them are
worthy of consideration as solutions. The question of using Fortran or some
other programming language such as C or Java should not be the question, but
the requirement.


From: "Kirsten Wilden" <>
Date: Tue, 19 Dec 2006 14:13:10 -0500
Subject: CFP, SIAM Conference on Mathematics for Industry, Oct 9-11, 2007

SIAM Conference on Mathematics for Industry: Challenges and Frontiers
Hyatt Regency Philadelphia, Philadelphia, Pennsylvania
October 9-11, 2007

The Call for Presentations for this conference is available at:


March 9, 2007: Minisymposium proposals
April 9, 2007: Abstracts for contributed and minisymposium speakers

For additional information, contact SIAM Conference Department at


From: Sekretariat Logik Heidelberg <>
Date: Wed, 20 Dec 2006 10:55:59 -0500
Subject: Professorship in Applied Computer Science, University of Heidelberg

Professorship (W3) in Applied Computer Science at the Faculty of Mathematics
and Computer Science, University of Heidelberg

We encourage applications for the position of "Professor for Applied Computer
Science". The research activities of the ideal candidate should complement the
existing research in scientific computing including its applications in
natural and life sciences. Preferred areas are data management and data
exploration. The position includes teaching of computer science, in particular
Bachelor and Master classes in the field of data base and information systems.

The position is permanent but the first contract is temporary.
The University of Heidelberg intends to increase the number of women in
teaching and research; women are therefore explicitly encouraged to apply.
According to German law, disabled applicants with an equivalent high
qualification will be given preference.

Applications with the usual documents (curriculum vitae, description of
scientific interests, list of publications (no reprints), record of teaching
activities) should be submitted until January 14, 2007, to: Dekan der
Fakultät für Mathematik und Informatik, Universität Heidelberg, Im Neuenheimer
Feld 288, D-69120 Heidelberg, Germany.


From: Angela Kunoth <>
Date: Wed, 20 Dec 2006 11:34:40 +0100
Subject: Advertisement Bonn Junior Fellows

The Excellence Cluster "Mathematics: Foundations, Models, Applications"
at the University of Bonn brings together researchers in Theoretical
and Applied Mathematics, and Mathematical Economics. The cluster looks
forward to fill

eight W2-Professors (``Bonn Junior Fellows'')

within the next few years. These temporary positions are limited to a
period of five years.

The professorships. The professorships intend to encourage independent
research by outstanding young researchers, in analogy to the research
groups of the "Max-Planck-Society" or the "Emmy-Noether-Program" of the
German Research Foundation. The professorships are endowed with a budget
for travel and for inviting guests, and with a post-doc position.
Professors are encouraged to supervise doctoral students. Teaching
obligations amount to four hours per week during semesters. If desired,
teaching can be limited to the graduate program. Depending on the
respective research focus, candidates become members of the "Faculty of
Mathematics and Natural Sciences", or of the "Faculty of Law and
Economics" at the University of Bonn.

The candidate. We are looking for young researchers who are about to
develop their own research agenda, and who have already demonstrated
their scientific independence by relevant publications. Candidates who
received their Ph.D in Germany should have some international experience.
German language skills are not necessary. All legal provisions about
equal opportunities apply.

Research focus. The candidates' research focus should be in one of the
Cluster's fields of activity, see

As not all positions will be assigned at the same time, there will be
several separate selection rounds. The application deadline for the
first round is February 28th, 2007. Deadlines for the following rounds,
and further details will be published at Applications should be sent to
Prof. Dr. Herbert Koch, Chairman of the "Junior Search Committee"
c.o. Frau Anke Thiedemann, Poppelsdorfer Allee 82
53115 Bonn, Germany

For the complete job advertisement, see


Date: Tue, 19 Dec 2006 13:40:46 +0100
Subject: PhD and post doc positions in Numerical Analysis of X-ray data

Four Ph.D. and post doc positions are available at the Centre of Excellency:
"Metal Structures in four Dimensions" at Risoe National Laboratory in
Denmark. The center has been the pioneer behind a new x-ray imaging technique
for in situ 3D visualisation of materials, known as 3DXRD. This technique is
based on tomographic reconstruction principles and very large data sets,
emerging from 3D detectors. We have a close collaboration with groups in
applied mathematics, e,g. CUNY.

The candidates will be part of an international network aiming at the design
and implementation of new algorithms for 3DXRD. We seek candidates with a
strong background in scientific computing. An interest in tomographic
reconstruction or image analysis or the implementation of large sets of linear
equations would be favourable.

The full text of the announcement can be found at Applications must include a letter of
motivation, a CV, and the listing of at least two referees. Applications and
inquires should be send by e-mail to Prof. Henning Friis Poulsen, The submission deadline is January 15, 2007.


From: Sebastian Reich <>
Date: Wed, 20 Dec 2006 12:01:44 +0100
Subject: Postdoctoral position at University of Potsdam

Postdoctoral Position in Computational Climate and Weather Prediction
at University of Potsdam

The position is available at the Department of Mathematics at the University
of Potsdam (Germany) and is part of collaborative work with the Freie
Universit\"at Berlin, Imperial College London, and the CWI Amsterdam.

The group has developed novel numerical methods based on Lagrangian and
semi-Lagrangian formulations of atmospheric fluid dynamics. These methods
display excellent conservation properties, which are of fundamental importance
to long term predictability. The multi-scale aspect of atmospheric dynamics is
treated by a regularization approach jointly developed with the UK Met Office.

The successful applicant will be part of the theoretical and practical
development of these methods into a fully three dimensional simulation
methodology for atmospheric circulation. Desirable prerequisites are

-a PhD in mathematics, physics or meteorology
-experience with computational fluid dynamics/PDEs
-knowledge of mathematical concepts in fluid dynamics

The position (BAT IIa-O equivalent) is initially for 2 year period with the
possibility of extension for an additional 2 years.

Please send your complete application (cover letter, CV, summary of thesis, 2
references) preferably in PDF format to:

Prof. Dr. Sebastian Reich
Department of Mathematics
University of Potsdam
Am Neuen Palais 10
D-14469 Potsdam
Phone: +49 331 977 1859
home page:

Closing date for applications: 1st March 2007


From: "Jose E. Castillo" <>
Date: Thu, 21 Dec 2006 11:40:25 -0800
Subject: Ph.D. Studies in San Diego

The Computational Science Research Center at San Diego State
University is looking for qualified applicants for its
interdisciplinary Ph.D. program (joint with Claremont Graduate
University) in Computational Science.

We have financial support in the form of Teaching, Graduate and
Research Assistantships and Fellowships. Areas of interest include
Nonlinear Dynamics, Biomathematics, Soft Condensed Matter Physics,
Relativistic Astrophysics, General Relativity,
Material Sciences, Geophysics, Nuclear Physics and Physical Oceanography.

Please see our web page ( for details about our
program including application process, deadlines and faculty research

Best Regards,
Jose E. Castillo PhD
Director / Professor
Computational Science Research Center
San Diego State University
619-5947205/3430, Fax 619-594-2459


From: "Prof. T. Tang" <>
Date: Thu, 21 Dec 2006 16:58:11 +0800 (HKT)
Subject: Contents, Commun. Comput. Phys. Vol. 2 No. 1 (2007)

Communications in Computational Physics

Volume 2, Number 1, February 2007

Review Article:
X. S. Liu, Y. Y. Qi, J. F. He and P. Z. Ding
Recent progress in symplectic algorithms for use in quantum systems.
Commun. Comput. Phys., 2 (2007), pp. 1-53.

Regular Articles:
D. Xiu and J. Shen
An efficient spectral method for acoustic scattering from rough surfaces.
Commun. Comput. Phys., 2 (2007), pp. 54-72.

X. Li and Q. Nie
Surface diffusion on stressed solid surface.
Commun. Comput. Phys., 2 (2007), pp. 73-86.

J. Fuite, R. E. Marsh and J. A. Tuszynski
An application of Prony's sum of exponentials method to pharmacokinetic data
Commun. Comput. Phys., 2 (2007), pp. 87-98.

T. S. Lee, X. Liu, G. C. Li and H. T. Low
Numerical study on sinusoidal fluctuated pulsatile laminar flow through various
Commun. Comput. Phys., 2 (2007), pp. 99-122.

W. Bao and C. Zheng
A time-splitting spectral method for three-wave interactions in media with
competing quadratic and cubic nonlinearities.
Commun. Comput. Phys., 2 (2007), pp. 123-140.

A. Kurganov and C. T. Lin
On the reduction of numerical dissipation in central-upwind schemes.
Commun. Comput. Phys., 2 (2007), pp. 141-163.

W. Guo, T. P. Schulze and W. E
Simulation of impurity diffusion in a strained nanowire using off-lattice KMC.
Commun. Comput. Phys., 2 (2007), pp. 164-176.


From: Lothar Reichel <>
Date: Fri, 22 Dec 2006 10:16:09 -0500
Subject: Contents, ETNA, vol 23

Table of Contents, Electronic Transactions on Numerical Analysis (ETNA),
vol. 23, 2006. ETNA is available at and at several
mirror sites. ETNA is in the extended Science Citation Index and the
CompuMath Citation Index.

i-viii Table of contents and abstracts

1-4 David S. Watkins
A case where balancing is harmful

5-14 Alexander Malyshev and Miloud Sadkane
Condition numbers of the Krylov bases and spaces associated with the
truncated QZ iteration

15-37 Antonija Duvnjak and Hermann J. Eberl
Time-discretization of a degenerate reaction-diffusion equation
arising in biofilm modeling

38-62 Serge Nicaise and Emmanuel Creusé
Isotropic and anisotropic a posteriori error estimation of the mixed
finite element method for second order operators in divergence form

63-75 Toshihiro Yamamoto
Toward the Sinc-Galerkin method for the Poisson problem in one type
of curvilinear coordinate domain

76-87 L. Bergamaschi, R. Bru, A. Martínez, and M. Putti
Quasi-Newton preconditioners for the inexact Newton method

88-104 Michael H. Hohn
Solution of singular elliptic PDEs on a union of rectangles using
sinc methods

105-128 Wayne Joubert and Jane Cullum
Scalable algebraic multigrid on 3500 processors

129-140 Mohamed El-Gamel
The Sinc-Galerkin method for solving singularly-perturbed
reaction-diffusion problem

141-157 David S. Watkins
On the reduction of a Hamiltonian matrix to Hamiltonian Schur form

158-179 Olaf Schenk and Klaus Gartner
On fast factorization pivoting methods for sparse symmetric
indefinite systems

180-201 M. K. Kadalbajoo and K. K. Sharma
Parameter-uniform fitted mesh method for singularly perturbed delay
differential equations with layer behavior

202-218 Kenshu Shimomukai and Hidesada Kanda
Numerical study of normal pressure distribution in entrance flow
between parallel plates, I. Finite difference calculations

219-250 Kendall Atkinson and David Chien
A study of the fast solution of the occluded radiosity equation

251-262 Ahmad Reza Naghsh-Nilchi and Shahram Daroee
Iterative sinc-convolution method for solving radiosity equation
in computer graphics

263-287 Oren E. Livne and Grady B. Wright
Fast multilevel evaluation of smooth radial basis function

288-303 Iliya A. Brayanov
Uniformly convergent difference scheme for singularly perturbed
problem of mixed type

304-319 Mario Gotz
On extremal problems related to inverse balayage

320-328 Toshihiro Yamamoto
Approximation of the Hilbert transform via use of Sinc convolution

329-338 Yoshinori Miyazaki, Nobuyoshi Asai, Dongsheng Cai, and Yasuhiko Ikebe
Numerical computation of the eigenvalues for the spheroidal wave
equation with accurate error estimation by matrix method

In 2006 ETNA also is publishing special volumes on

"Saddle Point Problems: Numerical Solution and Applications" edited by
Michele Benzi, Richard B. Lehoucq, and Eric de Sturler (vol 22),

"Orthogonalnt Polynomials and Mathematical Physics" edited by
R. A'lvarez-Nodarse, J. Arvesu', and F. Marcella'n (vol. 24),

"Constructive Function Theory" edited by Wolfgang Dahmen, Jeff Geronimo,
Xin Li, Doron Lubinsky, Igor Pritsker, and Ian Sloan (vol. 25, in progress)


From: "Gladwell, Ian" <>
Date: Tue, 19 Dec 2006 13:36:38 -0600
Subject: Contents, ACM Transactions on Mathematical Software

Table of Contents
ACM Transactions on Mathematical Software
December 2006, Volume 32, Number 4

For more information, including abstracts and access
to full text, see

Inverting the symmetrical beta distribution
Pierre L'ecuyer, Richard Simard
Pages: 509 - 520

Block algorithms for reordering standard and generalized Schur forms
Daniel Kressner
Pages: 521 - 532

The design and implementation of the MRRR algorithm
Inderjit S. Dhillon, Beresford N. Parlett, Christof Vömel
Pages: 533 - 560

Algorithm 857: POLSYS&lowbar;GLP---a parallel general linear product homotopy
code for solving polynomial systems of equations
Hai-Jun Su, J. Michael McCarthy, Masha Sosonkina, Layne T. Watson
Pages: 561 - 579

Algorithm 859: BABDCR---a Fortran 90 package for the solution of bordered ABD
linear systems
Pierluigi Amodio, Giuseppe Romanazzi
Pages: 597 - 608

Algorithm 860: SimpleS---an extension of Freudenthal's simplex subdivision
Eduardo N. Gonçalves, Reinaldo M. Palhares, Ricardo H. C. Takahashi, Renato
C. Mesquita
Pages: 609 - 621

Algorithm 861: Fortran 90 subroutines for computing the expansion coefficients
of Mathieu functions using Blanch's algorithm
Danilo Erricolo
Pages: 622 - 634

Algorithm 862: MATLAB Tensor Classes for Fast Algorithm Prototyping
Brett W. Bader, Tamara G. Kolda
Pages: 635 - 653

End of NA Digest