NA Digest Sunday, August 21, 1988 Volume 88 : Issue 33

Today's Editor: Cleve Moler

Today's Topics:


From: Bracy Elton <!>
Date: 18 Aug 88 00:51:49 GMT
Subject: Cellular Automata,PDEs,Boundary Conditions

Hi, I'm wondering whether anyone is doing or knows of any work in using
cellular automata for modeling fluid flow problems. I'm particularly
interested in any analysis into the realization of Dirichlet and Neumann
boundary conditions in cellular automata for fluid flow problems.

Please respond via e-mail.

Bracy Elton
Lawrence Livermore National Laboratory or


From: Richard Franke <>
Date: Thu, 18 Aug 88 8:42:06 BST
Subject: Franke Joins ONR, London

I am now at the Office of Naval Research in London, where my mailing address

Office of Naval Rsearch OR Office of Naval Research
Branch Office London, Box 39 223 Old Marylebone Road
FPO New York 09510-0700 London NW1 5TH, England
(telephone: 01-409-4471)
Richard Franke


From: Walter Gander <>
Date: 18 Aug 88 10:47 +0100
Subject: FFT for IBM PC Wanted

A friend of mine (Luciano Molinari) is looking for a fast FFT programm
for the IBM PC. Preferably written in C or Microsoft Assembler.

Please let us know where we could get such a software. Many thanks.

- Walter Gander

postal address: ETH Zuerich, Inst. f. Informatik, CH-8092 Zuerich,


From: Per Christian Hansen <mcvax!cuobs!pch@uunet.UU.NET>
Date: Wed, 17 Aug 88 18:13:03 +0200
Subject: Perturbation of singular vectors

I am interested in the following problem:

Let R be an upper triangular matrix with 1's on the diagonal and
elements R(i,j) that decrease in magnitude with the 'distance' from
the diagonal. Example: R may be the inverse of the Cholesky factor
of the matrix
| 1 .25 |
|.25 1 .25 |
| .25 1 .25 |
| . . . |
| . . |
which occurs in connection with spline approximations.

Let A = U*Sigma*V' be an ill-conditioned matrix with ill-determined
numerical rank derived from an ill-posed problem. Its singular values
decay gradually towards zero, and the number of oscillations in the
singular vectors U(:,i) and V(:,i) increase with i .

Finally, let A~ = A*R = (U~)*(Sigma~)*(V~)' .

What can then be said about the relations between the singular vectors
or singular subspaces of A and A~ ?

Experiments with MATLAB suggests that both U'*(U~) and V'*(V~) are
close to identity matrices.

Per Christian Hansen
Copenhagen University Observatory, Denmark
email to


From: Piet Wesseling <mcvax!dutinfh!piet@uunet.UU.NET>
Date: 18 Aug 88 14:33:54 EST (Thu)
Subject: GAMM Conference on CFD

First announcement and call for papers

The GAMM Committee for Numerical Methods in Fluid Mechanics organizes the


September 27-29, 1989

University of Technology, Delft, The Netherlands

The subjects of the conference are:

1. Theory of numerical methods in fluid mechanics: finite difference methods,
finite element methods, spectral methods, etc. The emphasis will be on
innovations in methods.

2. Application of numerical methods to fluid mechanical problems in
aerodynamics,hydrodynamics, propulsion, fluid machinery, nuclear reactor
technology, meteorology, biomechanics, etc.

The duration of each presentation will be 20 minutes plus discussion. Talks
on work in progress are welcome. Prospective contributors are invited to
submit sufficiently detailed abstracts of 1 to 2 pages of text plus figures
and tables by MARCH 7, 1989. The abstracts should include: field of study,
definition of problem, approach, results and conclusions. The abstracts should
concern unpublished work.
A selection of papers will be made on the basis of these abstracts. Authors
will be notified of acceptance by May 20, 1989. The proceedings will be
published by Vieweg in the series Notes on Numerical Fluid Mechanics. A book
of abstracts will be available at the time of the meeting.

The conference language is English.

Please write to the correspondence address below if you want to be put on the
mailing list of the organizing committee.

Conference chairman: P.Wesseling, Dept. of Technical Mathematics and

Conference secretary: W.J.Bannink, Dept. of Aerospace Engineering

Address all correspondence to: Mrs.R.Komen-Zimmerman
Congresbureau TU Delft
Stevinweg 1, 2628 DELFT, THE NETHERLANDS
Telephone (015)781340 Telex 38151 butud nl
Telefax (015)781855

CONFERENCE SPONSORS: Royal Netherlands Academy of Sciences
Shell International Petroleum Company Ltd.
N.V. Nederlandse Gasunie
Philips Research Laboratories
Gesellschaft fuer Angewandte Mathematik und Mechanik
IBM Nederland N.V.
Delft University of Technology

Please post


From: David Hough <dgh@Sun.COM>
Date: Thu, 18 Aug 88 20:34:48 PDT
Subject: ANSI C - Third Public Review

The third public review of X3J11's Draft ANSI Standard C
is nearing its close on 1 September 1988. This third review
is based upon a draft dated 13 May 1988 which is little
changed from earlier drafts except that the controversial
"noalias" keyword was removed.

Consequently the Draft still leaves a good deal to be
desired from the numerical point of view.

I have two documents available for electronic distribution.
I will be glad to send you tbl/troff -ms source for these;
please specify which if you only want the second one described below.
Unfortunately the Draft ANSI Standard itself is not publicly
available in electronic form.

The first available document is my 29 March 1988 commentary prepared
for the second public review period (30 pages), with X3J11's
formal responses of 22 April interspersed. The following were

Greg Astfalk Larry Breed D. Burton
W. J. Cody Iain Johnstone W. Kahan
Zhishun Alex Liu David Mendel Jim Meyering
K-C Ng Gene Spafford Philippe Toint
Stein Wallace

The second available document is a draft, subject to revision
until about 25 August, of my commentary for the third public
review. It's only about 10 pages since I generally avoided
directly repeating what was in the earlier document.
I'm looking for additional reviewers and conspirators
on this one. The abstract follows:

The proposed C standard suffers numerical
shortcomings - many inherited from its precursors
- in areas of interest to providers of portable
mathematical software. I comment in detail upon
the following aspects of the proposed standard:

Comment #1, Section 3.9: encourage sound practices
Comment #2, Section 3.9: disparage hazardous practices
Comment #3, Section 1.1: emphasize surprises in rationale
Comment #4, Section 1.1: anticipate supplemental standards
Comment #5, Section use "significand"
Comment #6, Section <float.h> has too many names, not enough information
Comment #7, Section round conversions between floating types
Comment #8, Section fix arrays
Comment #9, Section 4.5: exceptions in mathematical functions
Comment #10, Section 4.5: tell more in the rationale
Comment #11, Section 4.5: standardize hypot
Comment #12, Section delete modf
Comment #13, Section 4.7: specify which signals can arise


From: Gene Golub <golub%kulesat.uucp%blekul60.bitnet@Forsythe.Stanford.EDU>
(Gene Golub in Leuven)
Date: Fri, 19 Aug 88 10:43:36 GMT
Subject: NATO Advanced Study Institute in Leuven

REPORT of the NATO Advanced Study Institute

The NATO Advanced Study Institute on "Linear Algebra, Digital
Signal Processing and Parallel Algorithms" took place in Leuven
from August 1 through August 12. There were over 90 participants
(including the 15 invited speakers) from 16 countries and there
was a significant industrial participation.

The invited speakers included : M. Bellanger (TRT, France), B.
Bitmead (ANU, Australia), A. Bjorck (Linkoping, Sweden), R. Brent
(ANU, Australia), Y. Genin (PRLB, Belgium), S. Hammarling (NAG,
England), I. Ipsen (Yale, USA), T. Kailath (Stanford, USA), F. Luk
(Cornell, USA), J. McWhirter (RSRE, England), G. Meurant (CEL,
France), D. Sorensen (Argonne, USA), J. Vandewalle (KUL, Belgium),
and the codirectors G. Golub (Stanford, USA) and P. Van Dooren
(PRLB, Belgium).

The goal of this meeting was to synthesize the three topics
mentioned in the title as there is currently a great deal of
activity in each one of these areas. It was felt that there were
many interactions at the meeting, not only between participants
that were familiar with each one of these areas but also
between people working in different areas of interest.

The Proceedings will include the invited presentations, some
of the contributed talks and all the abstracts. They will be
published by Springer-Verlag in the NATO ASI Series and should
appear early next year.

The following areas emerged as major themes at the meeting :

1) Singular value and eigenvalue decompositions, including

The current techniques used for the computation of the singular
values and eigenvalues of matrices were addressed by several
speakers. Modified singular value and eigenvalue problems were
also discussed, especially in view of their application in various
signal processing problems (e.g. total least squares, generalized
SVD, time varying eigenvalue problems etc.)

2) Toeplitz matrices, including special algorithms and

Several speakers focussed on so-called fast or O(n**2) algorithms
for computing L.L' and Q.R decompositions of a Toeplitz matrix.
Special attention was given here to Levinson and Schur type
algorithms as well as to their split version. The fast algorithms
based on displacement ranks and on updating and downdating were
also addressed.

3) Recursive least squares in linear algebra, digital signal
processing and control

The recursive least squares algorithms occurring in signal
processing and control often have a specific structure, hence
allowing for fast solutions. Various of these fast algorithms were
explained and their potential numerical deficiencies were pointed
out. Issues as error build-up, error feedback, exponential
windowing, sliding windowing and so on, were addressed.

4) Updating and downdating techniques in linear algebra and
signal processing

The two main techniques of updating and downdating are low rank
corrections and low norm corrections. Low rank corrections are
used in divide and conquer methods for eigenvalue and singular
value computations and were shown to yield powerful parallel
computational methods. They can also efficiently be used for
computing new least squares solutions for modified data fitting
problems using the theory of orthogonal polynomials and modified
moments. Low norm corrections are used in slowly time varying
problems as encountered in various signal processing problems.

5) Error analysis and stability of algorithms and sensitivity
analysis of special recursive least squares problems

Error propagation in linear algebra is a well established
discipline but insufficiently known to the signal processing
community. Their basic principles were explained and their
application to specific problems in signal processing were
addressed (fast recursive least squares, Toeplitz solvers, Kalman
filtering, and so on). The relevance of special forms of stability
(weak, strong, mixed stability) was also emphasized in problems of
signal processing and linear algebra with special structure.

6) Special architectures (including supercomputers and distributed
processor arrays) for linear algebra and signal processing

Exploiting the parallelism of present (and future) computer
architectures is an area of significant interest the last few years.
The state of the art in computational algorithms for supercomputers
and distributed arrays of processors (such as systolic arrays) were
widely covered. Also special problems in linear algebra and signal
processing were given full attention.

Although most of the presentations can easily be classified in one
of the above "themes", it became apparent during the meeting that
there was a strong interconnection between several of the themes.
This lead to lively discussions both during lectures and breaks.
Several of the contributed talks focussed on specific applications
of these topics. In particular we mention here radar technology,
medical applications and robotics.

For the organizing committee,

Gene Golub Paul Van Dooren


From: Melvyn Ciment <>
Date: Fri, 19 Aug 88 17:44:53 -0400
Subject: Program Director Needed at NSF

FOUNDATION, invites applications for the position of Program
Director, Centers. This program is responsible for the management
and development of the five NSF National Supercomputer Centers.

The current Program Director, Paul Rotar, will be returning to NCAR
as of December 1, 1988.

Interested persons should contact;

Dr. Melvyn Ciment,
Acting Director
1800 G Street, N. W.
Washington D.C. 20550
Phone 202-357-7558.


From: Gene Golub <prlb2!kulcs!kulesat!golub@uunet.UU.NET>
(Gene Golub in Leuven)
Date: Thu, 11 Aug 88 14:31:42 GMT
Subject: Congratulations Joe Keller

Joe Keller, Professor at Stanford and a former VP of SIAM, was awarded the
National Medal of Science on July 15. Keller was cited for "his outstanding
contribution to the geometrical theory of diffraction....."

Congratulations, Joe, for this award; it's further recognition of your
great contributions.

Gene Golub


From: George Corliss <>
Date: Fri, 19 Aug 88 10:59:43 CDT
Subject: Differentiation Arithmetic: Examples & applications

We are currently compiling a bibliography on automatic differentiation
(differentiation arithmetic) and its applications. If you know of

1. work on computing numerical values of derivatives or partial derivatives
from recurrence relations, or
2. applications which would benefit from the ability to compute derivatives
accurately and efficiently,

please e-mail or surface mail citations to either of us.

For those people who are unfamiliar with differentiation arithmetic, it
is the calculation of the VALUES of derivatives of a function using
recurrence relations derived from the rules of calculus. It is neither
symbolic (no formula for derivatives is formed) nor numeric (no finite
differences are computed). An N-term Taylor series can be computed accurately
in O(N^2) time. Two basic references are: Rall, "Automatic Differentiation:
Techniques and Applications", and Kagiwada, et al., "Numerical Derivatives
and Nonlinear Analysis".

Applications include optimization, solving nonlinear systems, Taylor series
solutions of ordinary differential equations, and interval techniques for
bounding remainder terms.

Thank you in advance.

George Corliss
Department of Mathematics, Statistics and Computer Science
Marquette University
Milwaukee, WI 53233 USA

UUCP: ...!uwvax!marque!georgec
BITNET: 6591CORL@mucsd
NANET: na.corliss
phone: (414) 224-6599

Paul Davis
School of Mathematics
University of Bristol
University Walk, Bristol
BS8 1TW, England

BITNET: ukacrl!!DavisPH


End of NA Digest