NA Digest Saturday, October 30, 1999 Volume 99 : Issue 43

Today's Editor:
Cleve Moler
The MathWorks, Inc.

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to

URL for the World Wide Web:

From: G. W. Stewart <>
Date: Mon, 25 Oct 1999 14:55:57 -0400 (EDT)
Subject: Rounding Error

On October 11, I posted the following query in the NA Digest.

I'm giving a talk later this month on rounding error to the
Baltimore-Washington Section of SIAM. I would appreciate any
interesting anecdotes on real-life embarrassments due to rounding
error. I recall (though I don't know the details) stories about an
index on the Canadian stock market drifting off true and something
about a missile in the Iraq war. If you know the details of these
or any other stories, please pass them on to me.

Pete Stewart

The response was certainly gratifying. Within two days I received 64
messages. They were a great help in preparing my talk, which I gave
on October 20 at Johns Hopkins. My thanks to everyone. (For those
who are interested, a postscript file of my transparencies can be found

I was surprised to find how few incidents were reported. Apparently,
with some notable exceptions, the problems due to roundoff are
themselves at the level of roundoff.

Many responders asked me to post a summary. Here is a list of the
incidents I used in my talk with references and a brief description.

1. The Patriot and the Scud.
1. General Accounting Office Report GAO/IMTEC-92-26.
2. Robert Skeel, "Roundoff Error Cripples Patriot
Missile," SIAM News, July 1992.

On February 25, 1991, during the Gulf War, a Patriot missile
defense system let a Scud get through. It hit a barracks, killing
28 people. The problem was in the differencing of floating point
numbers obtained by converting and scaling an integer timing
register. The GAO report has less than the full story. For that
see Skeel's excellent article.

2. The short flight of the Ariane 5.
1. Report by the Inquiry Board.

On June 4, 1996, the first Ariane 5 was launched. All went well
for 36 seconds. Then the Ariane veered off course and
self-destructed. The problem was in the Inertial Reference System,
which produced an operation exception trying to convert a 64-bit
floating-point number to a 12-bit integer. It sent a diagnostic
word to the On-Board Computer, which interpreted it as flight data.

Ironically, the computation was done by legacy software from the
Ariane 4, and its results were not needed after lift-off.

3. The Vancouver Stock Exchange
1. The Wall Street Journal November 8, 1983, p.37.
2. The Toronto Star, November 19, 1983.
3. B.D. McCullough and H.D. Vinod
Journal of Economic Literature
Vol XXXVII (June 1999), pp. 633-665.
(References communicated by Valerie Fraysse)

In 1982 (I figure) the Vancouver Stock Exchange instituted a new
index initialized to a value of 1000.000. The index was updated
after each transaction. Twenty two months later it had fallen to
520. The cause was that the updated value was truncated rather
than rounded. The rounded calculation gave a value of 1098.892.

4. Parliamentary elections in Schleswig-Holstein.
1. Rounding error changes Parliament makeup
Debora Weber-Wulff
The Risks Digest
Volume 13, Issue 37, 1992

In German parliamentary elections, a party with less than 5.0% of
the vote cannot be seated. The Greens appeared to have a
cliff-hanging 5.0%, until it was discovered (after the results had
been announced) that they really had only 4.97%. The printout was
to two figures, and the actual percentage was rounded to 5.0%.

5. For a few more examples try searching the RISK Digest
( with the key word "rounding"
(David Goldberg's idea).

Once again, thanks to all of you for your help.

Pete Stewart


From: Mike Hosea <>
Ross Lippert <>
Miguel Sousa Lobo <>
Date: Mon, 25 Oct 1999 00:18:15 -0500
Subject: Re: Difference Between 1x1 Matrices and Scalars

Ilse Ipsen wrote:
- Consider the matrix product A*B*C where
- A = [1 1] B = [1] C=[1]
- [1] [2]
- [3]
- The product A*B*C is not defined because B*C is not defined.
- Nevertheless MATLAB 5.3 computes the answer [2]
- [4]
- [6]
- presumably because it associates (A*B)*C. In this case A*B
- happens to be a scalar but should be treated as a 1x1 matrix.
- Is this something to worry about?

Mike Hosea responds:

I agree mostly with what other respondents have been saying, but they have
left me somewhat perplexed by leaping to conclusions of what should be done
about this problem. Unless I missed it, nobody as yet has mentioned what it
would cost to impose formal correctness in this matter, nor has much been
said about the ultimate consequences of not doing so. So, let's consider it.

First, what happens to us if it is not fixed. Well, it has been established
that Matlab will allow some constructs that aren't formally correct. Okay,
but is this really a problem? Matlab 5.3 behavior in this regard is not
new, and the software has long proven its usefulness. If the alternative is
to issue an error, I must wonder how this error message would add to the
usefulness of the software. Matlab is not an environment for validating
symbolic mathematics, and the lack of errors from any number-crunching
environment is hardly convincing evidence of the validity of the mathematics
behind what is being computed. If the manifestations of the problem are
limited to instances where Matlab should have issued an error but didn't,
and hence returned some results in accordance with a certain interpretation
instead of issuing no results at all, then I do not see that the current
behavior has any significant negative impact, at least of a practical

Second, what happens to us if it is fixed. Well, it is common enough in
mathematical papers to equivocate n x 1 (and sometimes 1 x n) matrices with
n-vectors, and it is likewise common to regard the matrix product of x' and
y, where x and y are n x 1 matrices, as a handy way of representing the dot
product. This avoids the need to define a formal transformation from the
set of n x 1 matrices to the set of n-vectors and to use it every time one
wants to do a dot product (or, perhaps, to use its inverse when one wants to
do a matrix product). For example, almost any treatise on numeric
eigenvalue computation or nonlinear optimization will contain equations like
"alpha = (x'Ay)/(x'y)". This is not done because the writers are unaware of
the formal distinction between a 1x1 matrix and a scalar. It is done
because it is convenient, and everybody knows what the writers mean. To do
otherwise would rightly be regarded as pedantic. The equivalence of scalars
and 1x1 matrices in Matlab certainly helps here, and I am certain that
Matlab users have taken advantage of it for a long time, often eschewing
dot(x,y) in favor of x'*y because it more closely approximates the form of
the expression they are trying to represent in Matlab's language. I would
not be surprised to hear that the current behavior was designed to support
this inclination, and I shudder to think of all the programs that would
break if Matlab were to start issuing error messages in this case.

From a pedagogical perspective, we should indeed worry about such things.
Students are often confused enough as it is, and for teaching purposes in
elementary linear algebra, I might prefer an environment that is quite a bit
more rigid and pedantic than Matlab. From a practical perspective, however,
changing Matlab behavior in this matter is likely to cause a lot of problems
for existing Matlab programs in exchange for next to no benefit.


Ross Lippert responds:

OK, here are 2 cents from a fan of lex and yacc.

This all comes down to the parsing of a*b*c. Treating MATLAB as
a formal grammar following rules like
expr--> expr * expr | variable | literal | ....
we find that a*b*c is ambiguous, in the sense that is can be generated
by the two different sequences
expr --> expr*expr --> a*expr --> a*expr*expr --> a*b*c = a*(b*c)
expr --> expr*expr --> expr*c --> expr*expr*c --> a*b*c = (a*b)*c
This is not new to MATLAB or any other language that has a comfortable
artihmetic syntax. We like and use ambiguous grammars likes this in C,
fortran, whathaveyou.

The way the ambiguity gets resolved in such languages is by the use of
associativity rules in the parser. If * is right associative we do the
first sequence, always. If left, we take the second.

My first cent: This "bug" is nothing new
The associativity of an operator can have observable consequences
without the use of matrices. For example, let
>> x = -1; y = 1+eps; z = 1;
>> x+y+z == 1
returns 'true' is + is parsed as left associative and 'false' is + is
right associative.

Thus, this sort of trouble has been with us before we had fancy Matrix
language interpreters. It has to do with operators more than with
matrices. Just as one can generate statements which will raise error
messages if parsed with a different associativity with matrices, so can
one do the same with mere scalars.

My second cent: This "bug" requires documentation
Assuming that this discussion would not be taking place, if the
associativity of the MATLAB operators was not explicitly described
somewhere in official MATLAB documentation, here is why I think that it
should be documented clearly.

Some languages, (C, but not Fortran) do not have specified
associativities with operators and, under "-O" conditions, sometimes
will reassociate expressions. This can have some unfortunate
consequences for scalar expressions, but even more unfortunate
consequences for matrix expressions.

So, someone writing a MATLAB parser (syntax checker, clone, MATLAB to C
converter, MATLAB compiler, etc), who does not know what the
associativity of each operator is, might guess left associative (and
most people do) and get it right, or guess right associative, or
reassociate in an optimization step, and thereby produce bizarre errors
in their result.

If the associativity of the operators is left undocumented, then
statements such as A*B*C should be considered machine/version dependent,
and, as the version/dependent syntaxes of MATLAB do, should generate a


Miguel Sousa Lobo responds:

Yet one more comment about matrix multiplication in Matlab. All this
discussion would become pointless if The Mathworks added a line to its
documentation saying:

"By default, multiplication in Matlab is associated left to right (and
parsed accordingly.)"

This is Matlab's current (and perfectly defensible) behavior.

Best regards,


From: Wen Chen <>
Date: Mon, 25 Oct 1999 10:54:55 +0900
Subject: New Journal, Nonlinear Sciences and Numerical Simulation

Dear Colleagues:

A new "international journal of Nonlinear Sciences and Numerical Simulation"
will be launched in the coming 2000 year. The sponsorships of institutes
and universities worldwide are invited to join in us to activate the
nonlinear sciences. The nomination of member of editorial board is also
welcome. Aims and Scope of this journal is given below

The International Journal of Nonlinear Sciences and Numerical Simulation
publishes original papers on all subjects relevant to nonlinear sciences and
numerical simulation. The highest priority will be given to those
contributions concerned with a discussion of the background of a practical
problem, the establishment of an appropriate nonlinear model, the
determination of a solution, approximate or exact, analytical or numerical,
and a discussion of the relevance of the results when applied to the
real-life problem.
The following types of manuscript are encouraged:
1) New nonlinear model for a real-life problem with possible exciting
2) New analytical techniques for new nonlinear problems with physical
3) Numerical simulation revealing possible hidden pearls in nonlinear
Although the journal concentrates mainly on the applications side, review
articles dealing with establishment of nonlinear models, new numerical or
analytical techniques, with potential for wider application to real-life
problems, are invited by the Editor, and will be published from time to

For more details please contact

Editor-in-Chief Dr. Ji-Huan He
149 Yanchang Road, Shanghai University, Shanghai Institute of Applied
Mathematics and Mechanics, Shanghai 200072, China.

Yours sincerely,



From: A.J.W.v.d. Boom <>
Date: Mon, 25 Oct 1999 14:33:25 +0200
Subject: NICONET E-letter

The fifth issue of the NICONET E-letter is now available and can be
accessed from the World Wide Web URL:

or can be downloaded as compressed postscript file from the WGS ftp site:
( filename: E-issue-5.Z)

1. Welcome to the NICONET E-letter number 5
2. New additions to SLICOT since July 1999
3. New NICONET Reports since July 1999
4. Niconet events
5. (Forthcoming) Meetings and symposia attended by NICONET partners


From: Kamal Abdali <>
Date: Mon, 25 Oct 1999 09:56:59 -0400
Subject: NSF's Information Technology Research Program

The National Science Foundations's newly announced Information Technology
Research (ITR) program should be of interest to the scientific
computing/computational science research community. There is a web site that includes the program announcement, various
information items, and a FAQ.

The ITR program has been developed in response to the President's
Information Technology Advisory Committee (PITAC) Report to the
President. The report, entitled "Information Technology Research:
Investing in Our Future" , contains a number
of findings and recommendations. High end computing is one of the Technical
Research priorities in the report. The high-end computing research
recommended by PITAC is covered in, among other places, the section
entitled "Advanced Computational Science" in the ITR announcement. This
section invites proposals for leading-edge research in algorithms, software
and systems that is applicable to scientific and engineering computation.
The list of research topics explicitly mentioned in this section include:
numerical methods, optimization, symbolic and algebraic computation,
computational geometry, software libraries, problem-solving environments,
computer graphics and scientific visualization.


From: T. Terlaky <terlaky@mail.cas.McMaster.CA>
Date: Fri, 29 Oct 1999 14:57:03 -0400
Subject: Change of Address for T. Terlaky

My new address is:

Prof. Tam=E1s Terlaky
Office: JHE/214
Department of Computing and Software
McMaster University
Hamilton, Ontario, Canada, L8S 4L7
Phone: (905) 525-9140 ext. 27780
Fax: (905) 524-0340


From: Mikhail Shashkov <>
Date: Tue, 26 Oct 1999 11:10:40 -0600 (MDT)
Subject: Conference in Honor of Burton Wendroff


To honor BURTON WENDROFF on the occasion of his 70-th Birthday.

March 8-10, 2000
Los Alamos National Laboratory
Los Alamos, New Mexico, USA

The main theme of this conference will be the numerical
solution of systems of conservation laws but other areas
of importance for scientific computing will also be


March 8, 2000, Wed.

Stan Osher
"From Lax-Wendroff to TVD, ENO, WENO, Level Set Methods,
Ghost Fluids & Back Again"

Pete Miller
"Computational Transport Theory
in the Early Days: Wendroff's Contributions"

Herb Keller
A Rank Theory of DAE's.
W.J. Rider and L.G. Margolin
"Entropy Consistency through Nonlinear Dispersion"

Seymour Parter
"Preconditioning Spectral Collocation Methods for
Elliptic Problems"

Vidar Thomee
"Maximum-norm Estimates for Parabolic Finite Element Equations"
"Stable Schemes for Evolving Half-Space with General Boundary

Reuben Hersh
"Stable Schemes for Evolving Half-Space with General Boundary

March 9, 2000, Thursday.

P. Lax - TBA

Herb Kranzer
"Nonstrictly Hyperbolic Conservation Laws and Singular Shocks"

Joel Smoller
"Cosmology with a shock-wave"

Mac Hyman
"High-Order Predictor-Corrector Methods for Conservation Laws"
Blair Swartz
"Swiftly slicing specific fractions from n bodies in n-space"

Richard Liska
"Composite Schemes for Conservation Laws"

Randy LeVeque
"Three-dimensional wave-propagation algorithms for hyperbolic systems:
some upwinded and limited variants of Lax-Wendroff"

Jerry Brackbill
"The Immersed boundary Method: Some successes and Challenges"

March 10, 2000, Friday.

Eitan Tadmor
"Central Schemes: Convergence and Error Estimate for
High Resolution Methods"

H. Bruce Stewart
"The Virtues of a Computational Toolkit for Dynamical Systems"

Heinz Kreiss
"The numerical solution of nonlinear time dependent
partial differential equations"

Norman Bleistein
"High Frequency Imaging: from String Contruction to
Micro-Local Analysis"

Bill Morton
"Vorticity-preserving Lax-Wendroff Schemes on Arbitrary Meshes"

Mikhail Shashkov
"Discrete Vector Analysis"

Len Margolin and P.K. Smolarkiewicz
"Turbulence Modeling Using Nonoscillatory Lax-Wendroff Methods"

Sessions for contributed papers are not planned at this time,
but all who are interested are welcome to attend.

For additional information about this conference,
please visit the conference web page

or contact

Dr. Mikhail Shashkov
MS-B284, Group T-7, Theoretical Division,
Los Alamos National Laboratory,
Los Alamos, New Mexico, USA, 87545


From: A.J.W.v.d. Boom <>
Date: Wed, 27 Oct 1999 11:04:51 +0200
Subject: NICONET Workshop

Second NICONET workshop December 3, 1999

The Second NICONET workshop will be held at INRIA Rocquencourt, France.
The workshop is sponsored by the EU. The workshop covers all aspects of
numerical software in systems and control and their implementation and use
in industrial practice. See the NICONET website at:

for more details.

Workshop secretariat:
Mrs. Christine Bren
Fax: + 33-1-39635638
Mail: INRIA, Relations Exterieures
Domaine de Voluceau Rocquencourt B.P. 105


From: Trini Flores <>
Date: Thu, 28 Oct 1999 09:41:08 -0500
Subject: SIAM Conference on Materials Science

Third SIAM Conference on Mathematical Aspects of MATERIALS SCIENCE
May 21-24, 2000
Crowne Plaza Hotel
Philadelphia, Pennsylvania

NOVEMBER 15, 1999 -- Deadline for submission of minisymposium

For information on how to submit, please visit

or send e-mail to


From: Workshop <>
Date: Thu, 28 Oct 1999 18:04:46 +0200 (MET DST)
Subject: Workshop on Numerical Methods for Kinetic and Hyperbolic Equations

Workshop on
Ferrara, Italy, December 17-18, 1999


Department of Mathematics, University of Ferrara
Via Machiavelli, 35 I-44100 Ferrara, Italy


L. Pareschi (University of Ferrara),
V. Ruggiero (University of Ferrara),
G. Toscani (University of Pavia),
R. Natalini (IAC, Rome).

The workshop is organized jointly with the TMR projects "Asymptotic
methods in kinetic theory" and "Hyperbolic systems of conservation
laws" and is partially supported by MURST, Progetto Nazionale di Ricerca
"Analisi numerica: metodi e software scientifico".

Topics will include:

- Deterministic and Monte Carlo methods for kinetic equations
- Multiscale hyperbolic and kinetic problems
- Hyperbolic relaxation approximations
- High order methods for conservation laws

List of speakers:

S. Jin (Atlanta), R. Natalini (Rome), G. Russo (L'Aquila),
D. Aregba Driollet (Bordeaux), C. Buet (Paris),
S. Cordier (Paris), M. Falcone (Rome), F. Filbet (Nancy),
A. Frezzotti (Milano), I. Ibragimov (Saarbrucken),
A. Klar (Berlin), M. Lemou (Toulose), L. Mieussens (Bordeaux),
G. Naldi (Milano), G. Puppo (Torino), S. Rjasanow (Saarbrucken),
V. Romano (Bari), C. Schmeiser (Vienna), B. Wennberg (Goteborg).


For further information consult the web page

For registration write a message with your data (name, university or
institute, address etc.) to


From: Michael Philippsen <>
Date: Thu, 28 Oct 1999 12:08:41 +0200
Subject: ACM 2000 Java Grande Conference

ACM 2000 Java Grande Conference
Sponsored by ACM SIGPLAN
San Francisco, California, June 3-4, 2000

The Java Grande Conference focuses on the use of Java in the broad area of
high-performance computing; including engineering and scientific
applications, simulations, data-intensive applications, and other
emerging application areas that exploit parallel and distributed computing or
combine communication and computing. A day of tutorials will be held on the
day following the conference. The conference precedes the JavaOne
2000 conference, which would enable the Java Grande attendees to expose
themselves to the latest in basic Java Technology.

Authors are invited to submit manuscripts that demonstrate timely results,
technologies, or experiences that are most likely to have impact on the
use of Java in high performance computing systems.

Topics of interest include but are not restricted to:

Java use for scientific and engineering applications
Java frameworks and libraries for high-performance computing
Implementation techniques for Java on high-performance systems
Java numerics and Java extensions for high-performance computing
Java compilation and optimization for high-performance computing
Java development tools and environments for high-performance computing
Java performance and benchmarking


Dennis Gannon
Department of Computer Science
Indiana University
Bloomington, IN 47401
NAS Division
NASA Ames Research Center
MS 258-5
Moffet Field, CA 94035
Phone: (650) 604 1934


Piyush Mehrotra
MS 132C
3 West Reid Street - Building 1152
NASA Langley Research Center
Hampton, VA 23681
Phone: (757)-864-2188
Fax: (757)-864-6134


Henri Bal, Vrije University
Sandra Baylor, IBM
Aart Bik, Intel Corporation
Siddartha Chatterjee, University of North Carolina
Ken Kennedy, Rice University
Scott Kohn, Lawrence Livermore National Laboratory
Arvind Krishnamurthy, Yale University
Timothy Lindohm, Sun Microsystems
Satoshi Matsuoka, Tokyo Institute of Technology
Roldan Pozo, NIST
Vivek Sarkar, IBM
Suresh Srinivas, SGI
Vaidy Sunderam, Emory University
Gregor von Laszewzki, Argonne National Laboratory
Martin Westhead, EPCC
Kathy Yelick, University of California at Berkeley


Michael Philippsen
Computer Science Department
University of Karlsruhe
Am Fasanengarten 5
76128 Karlsruhe Germany
Phone: 49-721-608-4067
Fax: 49-721-608-7343


From: Andy Philpott <>
Date: Fri, 29 Oct 1999 13:20:28 +1300
Subject: Australian and New Zealand Industrial and Applied Mathematics Society


ANZIAM is the Australian and New Zealand Industrial and Applied Mathematics
Society. The 36th Annual Applied Mathematics Conference and Annual Meeting
of ANZIAM will be held at the Copthorne Resort in Waitangi on the shores
of the Bay of Islands, New Zealand, from the evening of Tuesday 8 February 2000
to lunchtime on the following Saturday 12 February.

The annual conference of ANZIAM is an established gathering of
applied mathematicians, scientists and engineers with wide-ranging interests.
It provides an interactive forum for presentation of results and discussions
by students, academics and other researchers on applied and industrial problems
derived in many scientific fields and amenable to quantitative description and

The invited speakers are:

Professor Jerzy Filar, University of South Australia
Dr Derek Goring, National Institute of Water and Atmospheric Research, NZ
Professor Peter Hunter, Auckland University
Professor Peter Jackson, Auckland University
Professor Nancy Kopell, Boston University
Dr Margaret Wright, AT&T Bell Laboratories

Further information and registration forms are available from the web page:

or e-mail


From: Luc Wuytack <>
Date: Fri, 29 Oct 1999 09:20:55 +0200 (MET DST)
Subject: Congress on Computational and Applied Mathematics


An International congress on Computational and Applied Mathematics
will be organized on July 17-July 21, 2000 at the University of Leuven,
Belgium. The congress will bring together people working on the analysis and
application of computational techniques for solving real scientific problems.
The list of invited speakers includes:
Baker C.T.H. (U.K.), van Dooren P. (Belgium), Pryce J.D. (U.K.),
Watson L.T. (USA), Sloan D. (U.K.), Brezinski C. (France),
Yamamoto T. (Japan), Meek D.S. (Canada).
Short communications will be accepted for presentation. An abstract of
at most 1 page should be sent before March 30, 2000 to:
Prof. M. Goovaerts, K.U. Leuven
CRIR, Minderbroederstraat 5, B-3000 Leuven (Belgium)

ICCAM-2000 will also concentrate on the project "Numerical Analysis
in the 20th century". Several survey talks will be given covering different
fields of numerical analysis. The Proceedings of ICCAM-2000 will be published
in the Journal on Computational and Applied Mathematics.
More information can found at:


From: Paulina Chin <>
Date: Fri, 29 Oct 1999 13:21:13 -0400 (EDT)
Subject: Symposium on Symbolic and Algebraic Computation

ISSAC 2000
International Symposium on Symbolic and Algebraic Computation
St Andrews University, Scotland, August 7-9, 2000

ACM sponsorship pending

ISSAC provides an opportunity to learn of new developments and to present
original research in all areas of symbolic and algebraic computation.
ISSAC 2000 will be held at St Andrews University, Scotland's oldest
university. Planned activities include invited presentations, research and
survey papers, poster sessions, tutorial courses, vendor exhibits and
software demonstrations.


Topics of the meeting include, but are not limited to:
Algorithmic mathematics: Algebraic, symbolic, and symbolic-numeric
algorithms including: simplification, polynomial and rational function
manipulations, algebraic equations, summation and recurrence equations,
integration and differential equations, linear algebra, number theory,
group computations, and geometric computing;
Computer sciences: Theoretical and practical problems in symbolic
mathematical manipulation including: computer algebra systems, data
structures, computational complexity, problem solving environments,
programming languages and libraries for symbolic-numeric-geometric
computation, user interfaces, visualization, software architectures,
parallel or distributed computing, mapping algorithms to architectures,
analysis and benchmarking, automatic differentiation and code generation,
automatic theorem proving, mathematical data exchange protocols;
Applications: Problem treatments incorporating algebraic, symbolic,
symbolic-numeric and geometric computation in an essential or novel way,
including engineering, economics and finance, architecture, physical and
biological sciences, computer sciences, logic, mathematics, statistics,
and uses in education.


Research results and insightful analyses of current concerns are the primary
focus. Papers will be reviewed by the Program Committee and additional
referees. Survey articles may be suitable for submission, if clearly
identified as such, and will be considered in a separate category from the
research papers. Simultaneous submission for publication elsewhere is not

Authors are invited to submit papers to the Program Committee Chair before
December 17, 1999. Submission instructions can be obtained from the
conference Web page or by sending e-mail to

A poster session will be organized separately. Information is available
on the conference Web site.


For more information, please refer to the conference Web page, E-mail inquiries may be sent to or to one of the conference committee members
listed below:

General Chair: Tomas Recio,
Local Arrangements Chair: Steve Linton,
Program Committee Chair: Chandrajit Bajaj,
Tutorial Chair: Josef Schicho,
Exhibitor Chair: Marc Moreno Maza, Numerical Algorithms Group,
Poster Session Chair: Arjeh Cohen,
Editor: Carlo Traverso,
Treasurer: Colin Campbell,
Publicity Chair: Paulina Chin,


From: Marcin Paprzycki <>
Date: Fri, 29 Oct 1999 23:56:25 -0500 (CDT)
Subject: Special Issue on Parallel and Distributed Simulation


this message is to remind everyone whois interested in contributing to
the Special Issue of the PDCP Journal (

edited by: Carl Tropper, McGill University

that the submission deadline is DECEMBER 15th, 1999.

More details about the special issue and the journal can be found on the
PDCP WWW site.

All the best,

Marcin Paprzycki phone: 601-266-6639
Department of Computer Science and Statistics or: 601-266-4949
University of Southern Mississippi
Hattiesburg, MS 39406-5106, USA FAX: 601-266-6452


From: Marcin Paprzycki <>
Date: Sat, 30 Oct 1999 00:00:13 -0500 (CDT)
Subject: Volume Dedicated to Prof. Galligani


This messge is to remind everyone interested in contributing to the

(volume intended to celebrate the 65-th birthday of Ilio Galligani)
Edited by L. Brugnano and D. Trigiante, University of Florence, Italy

that the submission deadline is NOVEMBER 15th.

More information about the volume and the book series that it will be
published in can be found no the www site.


Marcin Paprzycki phone: 601-266-6639
Department of Computer Science and Statistics or: 601-266-4949
University of Southern Mississippi
Hattiesburg, MS 39406-5106, USA FAX: 601-266-6452


From: Floyd Hanson <>
Date: Mon, 25 Oct 1999 00:47:48 -0500
Subject: Faculty Position at University of Illinois at Chicago

The Department has active research programs in all areas of pure
mathematics, computational and applied mathematics, combinatorics and
computer science, statistics, and mathematics education. See for more information.

Applications are invited for a tenure track or tenured position,
effective August 21, 2000, in Computer Science, broadly defined.
Current areas of interest include algorithms, coding theory,
combinatorial optimization, combinatorics, complexity, computational
mathematics, computational statistics, cryptography, data mining,
graph theory, language design, learning theory, logic, numerical
analysis, and universal algebra. The position is initially budgeted
at the Assistant Professor level, but candidates with a sufficiently
outstanding research record may be considered at higher levels.
Applicants must have a Ph.D. or equivalent degree in mathematics,
computer science, or a related field, an outstanding research record,
and evidence of strong teaching ability, with particular interest in
programming and algorithms. Salary negotiable.

Send vita and direct 3 letters of recommendation, indicating the
position being applied for, to Henri Gillet, Head; Dept. of
Mathematics, Statistics, and Computer Science; University of Illinois
at Chicago; 851 S. Morgan (M/C 249); Chicago, IL 60607. No e-mail
applications will be accepted. To ensure full consideration,
materials must be received by December 21, 1999. Minorities, persons
with disabilities, and women are particularly encouraged to apply. UIC
is an AA/EOE.

For more information about the position or institution/company:

Deadline for Applications: December 21, 1999


From: Axel Voigt <>
Date: Mon, 25 Oct 1999 14:51:28 +0200
Subject: Research Positions Stiftung Caesar, Bonn


The Modeling group of caesar is searching for candidates for a research
project in industrial mathematics. The project is carried out in cooperation
with industrial partners. Its subject is modeling, numerical simulation and
optimal control of crystal growth processes. The models are nonlinear partial
differential equations (heat transfer by conduction, convection and radiation).
The position is limited to two years, a prolongation is planned.

The ideal candidate has a master / Ph.D. in mathematics, physics, engineering,
computer sciences or in a similar subject. He / She should be familiar with at
least one of the subjects: scientific computing / mathematical modeling /
partial differential equations / optimal control / crystal growth. Knowledge
of the German language is not expected and only advantageous as far as
everyday life is concerned. Salary is according to the German BAT IIa tariff.
It depends on age and family status.

Interested candidates should contact us as soon as possible and provide their
CV, testimonies of their academic record (if possible in English or German
translations) and a description of their scientific background and interests.
One or two references of university teachers would be appreciated.

Further questions? Please feel free to contact Axel Voigt, Stiftung caesar,
Friedensplatz 16, 53111 Bonn, Germany, phone: +49(0)228-9656-236,

Please send your application to Ulrike Wermescher via e-mail: or to the address below.

Stiftung caesar
Center of Advanced European Studies and Research
Postfach 7025
53070 Bonn, Germany

Tel +49 (0)2 28 9656 - 122
Fax +49 (0)2 28 9656 - 111


From: Homer Walker <walker@WPI.EDU>
Date: Mon, 25 Oct 1999 19:17:23 -0400 (EDT)
Subject: Visiting Faculty Positions at Worcester Polytechnic Institute


The WPI Mathematical Sciences Department invites applications for
two visiting assistant professorships to begin in the fall of 2000.
An earned Ph.D. or equivalent degree is required. Successful candidates
must demonstrate strong research potential and evidence of quality
teaching, and will be expected to contribute to the department's
research activities and to its innovative, project-based educational
programs. Preferred research interests are areas of applied and
computational mathematics compatible with those represented in the
department: partial differential equations with applications in fluid
and continuum mechanics, composite materials, computational modeling
and simulation, numerical analysis, optimization, control theory,
applied probability, and discrete mathematics.

Appointments will be made for two years, contingent on satisfactory
teaching, with possible renewal for a third year at the discretion
of WPI and the appointee. WPI's academic year is divided into
four seven-week terms. The teaching load for these positions is
five courses over the four terms, i.e., one course in each of
three terms and two courses in the remaining term.

WPI is a private and highly selective technological university
with an enrollment of 2700 undergraduates and about 1000 full-
and part-time graduate students. Worcester, New England's second
largest city, offers ready access to the diverse cultural and
recreational resources of the region together with opportunities
for urban, suburban or rural lifestyles.

The Mathematical Sciences Department has 24 full-time faculty and
supports a PhD program and MS programs in applied mathematics and
applied statistics, as well as a full undergraduate program. For
additional information, see

Qualified applicants should send a detailed curriculum vitae, a
brief statement of their specific teaching and research objectives,
and three letters of recommendation, at least one of which addresses
teaching potential, to Visitor Search Committee, Mathematical Sciences
Department, WPI, 100 Institute Road, Worcester, MA 01609-2280, USA.

Applicants will be considered on a continuing basis beginning January 1,
2000 until the positions are filled.

WPI offers a smoke-free environment. To enrich education through diversity,
WPI is an affirmative action, equal opportunity employer.


From: Christine Cunningham <>
Date: Mon, 25 Oct 1999 11:11:32 -0500
Subject: Research and Development Position at Wolfram Research

Wolfram Research, makers of Mathematica, is seeking an outstanding
candidate to add to its research and development team in the area of
numerical computation. A successful applicant should have a Ph.D. in
numerical analysis, applied mathematics, or a related discipline in which a
significant component consists of practical implementation of mathematical
algorithms. We are most interested in people who have a broad and general
view of how their work fits into the overall framework of a technical
computational system. Consequently we will consider applicants with
expertise in a broad range of topics such as differential equations, linear
algebra, optimization, root finding, and special function evaluation.
Implementation experience in C/C++ and Mathematica skills would be a strong

Wolfram Reseach is located in Champaign, Illinois -- home of the University
of Illinois.

You can contact us:

by filling out a web employment application
at (the best way)
by sending email to
by sending a fax to 217-398-0747
by sending regular mail to:
Human Resources Department, Wolfram Research, 100 Trade Center Drive,
Champaign, IL 61820-7237, USA


From: P. Wesseling <>
Date: Tue, 26 Oct 1999 11:42:42 +0200 (METDST)
Subject: Student Position at Delft University of Technology

Department of Applied Mathematical Analysis
Faculty of Information Technology and Systems
Delft University of Technology

Applications are invited for a funded PhD studentship in the project
"Unstructured staggered schemes for the Navier-Stokes equations".
Further information may be found on the Internet at


From: Richard Braun <>
Date: Tue, 26 Oct 1999 09:44:40 -0400 (EDT)
Subject: Postdoctoral Positions at University of Delaware

Postdoctoral Positions
Department of Mathematical Sciences
University of Delaware

The Department of Mathematical Sciences at the University
of Delaware has funding for two postdoctoral positions
in scientific computing or computational and applied mathematics.
These positions will be an important part of an active program
which involves graduate students in industrial applied mathematics problems;
the program is supported with an NSF Group Infrastructure Grant.
Current projects involving industry include theory for
electromagnetic scattering, micromagnetics, fluid mechanics, materials
processing and free boundary problems. These positions will be
part of a group of professors, graduate students and postdocs.
The projects will involve working with both faculty and industrial
scientists or engineers on research problems of mutual interest.

The research part of position A will involve computation in fluid
dynamics and/or materials science problems with free boundaries.
The research for position B will involve scientific computation
and/or modeling in fluid dynamics, electromagnetic scattering
or computational finance. Both two-year positions have a
salary of $42,500 per year with additional travel and expense
money totaling $6,000; one course per semester of teaching will
also be required. The duration of both positions is two years
beginning in January 2000 or later. Applicants
should have received the PhD before beginning at the University of
Delaware, and the degree should be in the mathematical, computational
or physical sciences.

Please send a vita, research summary, a cover letter,
and have letters of recommendation (including one addressing teaching)
sent to:
GIG Postdoctoral Positions
Department of Mathematical Sciences
University of Delaware
Newark, DE 19716.

Applications will be considered until the positions are filled.

The University of Delaware is an equal opportunity employer.


From: Lyle Long <>
Date: Tue, 26 Oct 1999 19:36:52 -0400
Subject: Faculty Positions at Penn State University

Aerospace Engineering
Penn State University
Faculty Positions

The Department of Aerospace Engineering invites applications from, and
nominations of, highly-qualified individuals committed to teaching and
research in all areas relevant to aerospace engineering. Rank, tenure
status, and salary will be commensurate with experience and accomplishments.
Candidates with inter-disciplinary interests who can work effectively as
part of teams are encouraged to apply. Areas of particular relevance at
this time include (but are not limited to): computational and/or experimental
aerodynamics, composite structures, guidance and control, rotorcraft
aerodynamics or rotorcraft flight dynamics, and propulsion. Potential
candidates and nominators are encouraged to view our department website: Applicants must have an earned doctorate
in aerospace engineering or a related field, and at least one degree in
aerospace engineering is desirable. Applications should include a cover
letter specifying desired position, 2 copies of a resume, and a list of at
least 3 references. Responses received at the address below before
November 15, 1999 are assured full consideration, but the search will
remain open until the positions are filled.

Professor Cengiz Camci
Chair, Faculty Search Committee
Department of Aerospace Engineering
233A Hammond Building
The Pennsylvania State University
University Park, PA 16802


From: Esmond Ng <>
Date: Wed, 27 Oct 1999 22:52:28 -0700
Subject: Postdoctoral Positions at Lawrence Berkeley National Laboratory

The Scientific Computing Group at the Lawrence Berkeley National
Laboratory has three immediate visiting postdoctoral openings in various
computational science areas. Detailed descriptions of the openings can
be found at the following URL's.

Climate and earth sciences applications on massively parallel high
performance computers. [An announcement of this opening appeared in NA
Digest V.99, #41]

Parallel algorithms for image analysis in single-particle reconstruction
from electron micrographs.

Large-scale electronic structure calculations and materials simulations.

Applications/inquiries should be sent to

The Lawrence Berkeley National Laboratory (LBNL) is a multidisciplinary
DOE laboratory located on the hills of the campus of the University of
California at Berkeley, and is an equal opportunity/affirmative action


From: C. R. Traas <>
Date: Fri, 29 Oct 1999 17:47:09 +0200
Subject: Postdoctoral Positions in The Netherlands


This 4-year project, funded by the Dutch Science Foundation (NWO), consists
of three research projects dealing with different aspects of wavelet
research, and the establishment of a national seminar on wavelets, including
an international visitor's programme.

The projects are:

(project leader: dr. H.J.A.M. Heijmans, Centre for Mathematics and Computer
Science, Amsterdam).

(project leader: dr. J.B.T.M. Roerdink, Institute for Mathematics and
Computing Science, University of Groningen).

(project leaders: prof. dr. C.R. Traas, University of Twente, and
dr. H.G. ter Morsche, Technical University of Eindhoven).

Below you find a more detailed description of projects (ii) and (iii), for
which we have two open positions (one PhD, one postdoc).
Applications should be addressed to the respective project leaders,
as mentioned below under the different sub-projects.


TITLE: Wavelet Methods for Numerical Analysis

Wavelets and multiscale type methods are becoming increasingly important
with respect to numerical simulation and scientific computing, in particular
where large and complex problems have to be solved. Specifically in the
following tasks a significant contribution of these methods is envisaged:
- Preconditioning of large systems which may arise in the discretization
of elliptic PDE's or integral equations.
- Sparsening of dense matrices.
- Adaptive approximation.
The fundamental property of wavelets to be capable to zoom in onto local
properties of a function, is a key to the expectations mentioned above.

The research project concerns the investigation of the impact of wavelets
onto the solution methodology in and around the Boundary Element Method, in
particular the possibility of sparsening the systems' matrix. This
impact is certainly dependent on the class of kernels considered.

The project will be carried out in the research group "Numerical
Analysis" of the Mathematics Department of the University of Twente.
The research interest of this group is in Splines and Wavelets
and their applications in PDE's, integral equations and filtering.
There will be close cooperation with the Numerical Analysis group at
the university of Eindhoven, where related research is done within
the same covering 4-year project.

1: PhD in Applied Mathematics
2: Affinity with wavelets and PDE's/Integral equations.

Applications for the position, including a Curriculum Vitae (name,
address, degree, date, etc.) and name and email address of two professional
references, should be addressed (by surface mail) to

University of Twente
Faculty of Applied Mathematics
P.O.Box 217
7500 AE Enschede
The Netherlands
Tel: +31 53 4893408


Institute for Mathematics and Computing Science

TITLE: Wavelet-Based Signal Detection in Functional Neuroimaging

Many techniques currently exist to produce images of the human brain, such
as CT (Computerized Tomography), MRI (Magnetic Resonance Imaging) or PET
(Positron Emission Tomography). These imaging techniques are not only useful
to obtain images of brain structure, but also of brain function. The
detection of significant changes in neurological activity is not an easy
task, since in general these changes are small and distributed over the
whole brain, although not in the same amount. Statistical analysis of image
differences corresponding to different experimental conditions can be
subdivided in two classes: techniques which first partition the brain into
regions of interest, based upon anatomical structure, and techniques in
which one creates images of a statistical parameter: Statistical Parametric
Mapping, by applying statistical hypothesis testing to each element of the
difference between the images with and without activation, based upon
modelling the difference image as a realization of a Gaussian random field.
The aim of the research is a study of wavelet decompositions of brain images
to improve signal detection, and to establish a connection between the
current detection methods in neuroimaging.

This project is connected to GNIP: Groningen NeuroImaging Project, which
involves many research groups from the Academic Hospital (AZG) and the local
research school BCN (Behavioural, Cognitive and Neurosciences).

The project will be carried out in the research group `Scientific Computing
and Imaging' of the Department. The research interests of the group are
particularly concentrated in high performance computing for visual
perception and neural-net modelling, morphological image processing,
computerized tomography, scientific visualization and geometric algorithms.

1. MsC in Mathematics with strong affinity to Computer Science, or vice versa.
2. Experience in image processing and/or spatial statistics.
3. Demonstrably good computing skills.

Applications for the position, including a Curriculum Vitae (name, address,
degree, date, etc.) and name and email address of two professional
references, should be addressed (by surface mail) to:

Dr. Jos Roerdink
University of Groningen
Department of Mathematics and Computing Science
P.O. Box 800
9700 AV Groningen
The Netherlands
Tel: +31-50-3633931


From: Vladik Kreinovich <>
Date: Sun, 24 Oct 1999 11:11:30 -0600 (MDT)
Subject: Contents, Reliable Computing

Reliable Computing
Volume 6, Issue 1, 2000

Special issue on Reliable Geometric Computations
Guest editors: Helmut Ratschek, Jon G. Rokne

Introduction to the Special Issue: What Can One Learn from Box-Plane
Helmut Ratschek, Jon G. Rokne

Robust and Efficient Ray Intersection of Implicit Surfaces
Ole Caprani, Lars Hvidegaard, Mikkel Mortensen, Thomas Schneider

Correct Delaunay Triangulation in the Presence of Inexact Inputs and
Jeffrey S. Ely, Anthony P. Leclerc

Exact Computation of Delaunay and Power Triangulations
Marina Gavrilova, Helmut Ratschek, Jon G. Rokne

Removing Degeneracies by Perturbing the Problem or Perturbing the World
Pierre Alliez, Olivier Devillers, Jack Snoeyink

Interval Constraint Plotting for Interactive Visual Exploration of
Implicitly Defined Relations
Timothy J. Hickey, Zhe Qiu, Maarten H. van Emden

Reminiscences "Back in the Good Old Days..."
column edited by George F. Corliss

Interval Integration
L. B. Rall


End of NA Digest