NA Digest Sunday, July 7, 1991 Volume 91 : Issue 27

Today's Editor: Cleve Moler

Today's Topics:

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to


From: Paul Saylor <mathworks!apple!!saylor>
Date: Sat, 29 Jun 91 21:36:23 -0500
Subject: Gene Golub Honored at Midwest NA Day

Report on the Second Annual Midwest NA Day
Steve Lee and Paul Saylor

The 2nd Annual Midwest NA Day was held on Saturday, May 11, 1991 on the
University of Illinois, Urbana-Champaign campus. The conference guest
of honor this year was Gene Golub who had returned for the weekend to
receive an honorary doctorate from the University.

The speakers (in order) were: Bill Gear (NEC Research Institute), David
Young (University of Texas, Austin), Mike Berry (University of Tennessee),
Linda Petzold (Lawrence Livermore National Lab), Dan Boley (University of
Minnesota), Richard Varga (Kent State), Steven Ashby (Lawrence Livermore
National Lab), Tony Chronopoulos (University of Minnesota), Ching-Tsuan Pan
(Northern Illinois University), Haesun Park (University of Minnesota), and Ben
Leimkuhler (University of Kansas). The 30 minute talks covered various
topics in linear algebra and differential equations. The event drew a total
of around sixty attendees with research interests in applied numerical
analysis and scientific computing.

The morning session began with some background on Gene's Illinois days.
Gene is a native of Chicago and started advanced education at Wright Junior
College which he attended for two years. At first he was in chemistry but
will say that math was more fun and that he has always enjoyed manipulating
formulas. Following junior college, he wanted to continue his education
somewhere but also wanted to stay at home and so commuted to the University
of Chicago, an hour and a half each way. In his words, ``After two years
of a junior college and then the University of Chicago I didn't do too
well. I did well on the homework but when it came to the exams, I don't
know, I just sort of froze up. Some of my friends were in Urbana, so at the
end of the year that's where I wanted to go.''

Gene was an undergrad at the U of I for only one year and so did not know
anyone from his class of '53. He worked at the Digital Computer Lab as a
senior and then as a graduate student. His supervisor was the applied
mathematician Jack Nash whom he had taken a course from. Gene's first
assignment which he worked on for six months was to program Milne's method.

After Gene completed Milne's method, it was never used because
Runga-Kutta-Gill, which had come along in the meantime, was so convenient.
Nash then had Gene do statistical calculations. For a long time in his work
with least squares, he just formed the normal equations. But in 1959 ``...
there was the Householder transformation. A spotlight came on and there was
Householder on the stage.''

There is a story that Gene's friends like to tell from his days as a grad
student. Dave Muller, Jim Robertson, Ted Poppelbaum and Gary Metze chipped
in to buy a fifty dollar 1940 Chrysler --- with automatic transmission, no
less. Gene's birthday is February 29 and in those days was celebrated on
non-leap years during a ten microsecond interval after midnight February
28. But for his true sixth birthday a special occasion was called for.
Poppelbaum made a grand speech about the significance of birthdays, and at
the end produced a set of car keys.

The group led Gene out to the car, which started smoothly. He was
surprised and pleased, but, alas, did not have a driver's license! With
volunteers present offering to help and the automatic transmission to make
it easy, Gene soon began driving lessons and learned enough to pass his
test in about three weeks.

With a new license in hand and full of good will, Gene naturally invited his
benefactors to go for a ride. It seemed harmless enough. Friendship and the
spirit of celebration brought the group out to Gene's Chrysler. With hardly a
pause for doubt, everyone was quickly in the car. The naturally inviting
sound of the purring engine made them relax and overcame any trace of
hesitation. It was a happy, euphoric moment as Gene pulled away from the
curb. Soon enough there was a feeling of regret when he reached the open
road and, with only limited experience, began passing other cars in two-way
traffic. The repeated terror of an impending head-on collision changed the
mood of the party. Nevertheless, if it was terror it was just terror and
the ride ended without incident. Over time, Gene became a safe driver and
ordinary car owner, coping with the insurance on a free automobile that
turned it into something expensive.

Gene's thesis advisor was A. Taub, who at a point near completion of the
work wanted some outside opinion. Gene once said about this, ``Varga came
to Illinois --- Taub invited him to check me out. He said `I have some
similar results. Why don't we write a joint paper?' The next day Taub came
up to me and said `I understand you have results similar to Varga's. You
know a Ph.D. is supposed to be an original piece of research. If Varga's
results are published first, then no Ph.D.!' I wasn't worried because, of
course, Varga had told me we would write the paper together, but I then
thought `Who is this guy?' I was new at this game then at which everybody
is so honorable. So I worked on the paper with Varga, and that's when you
know everything very well. Varga did most of the actual writing. Varga is a
great mathematician and there I was saying to him `Well you see you take
this and it means that.' We were actually able to improve some of the
results of Sheldon. That was a very productive collaboration, even though
Varga was four, five thousand miles away. I've written papers with people
down the hall and not had a better relationship.'' (Gene refers to
Dick Varga as a great mathematician and on his part, behind Gene's back,
Dick refers to Gene as ``the world's most influential numerical analyst.'')

Besides the results in his thesis, Gene also discovered and developed
cyclic odd-even reduction when he was a grad student. But this method did
not get into his thesis due to stability problems. Cyclic reduction was not
exactly dead though as everyone knows. He developed the block idea later
and with contributions from Buzbee and Nielson together with the
stabilization ideas of Buneman, an elegant fast Poisson solver came about
in the late sixties.

After Illinois, Gene spent a year in England. From there he went to
Lawrence Berkeley Lab for a short time. Next he was at TRW for fifteen
months. Gene had been at TRW previously for a summer while in grad school
and it was his transportation needs in Los Angeles that partly suggested to
Gene's friends back in Urbana the idea of a car for his birthday.

Gene's work is noted both for its elegance and its wide usage. There is a
long list of his papers of supreme quality, at the highest order of
creativity, with ideas presented for the first time that are so familiar
now it is hard to think that they once were fresh. Elegance emerged early
in his career in his thesis and in his work on cyclic reduction. Some
admirers think that Golub is to algorithms as Mozart is to melody. (Mozart
recordings played over the sound system during breaks in the conference.)

On the personal side, Gene is known for his informal approach to people,
his cordiality and many small touches of kindness. His support of young
people shows that he remembers what it was like starting out, and how
important the nurturing process is.

The principal commencement ceremony at which Gene was given his honorary
degree occurred in the morning, Sunday, May 12. Nobel peace prize winner,
Oscar Arias Sanchez gave the commencement address, and an alumni
award was bestowed on Jim Brady. At the University of Illinois, the large
number of degree recipients scatter to smaller ceremonies later in the day
where they receive personal recognition. Gene spoke for a moment at one of
these to the degree candidates and the audience of family and friends. He
commented that when he started out computers were new. It was an exciting
time learning how to use those new devices and understand them. But it is
no different now, still an exciting time in mathematics, numerical analysis
and scientific computation.


From: Joe Grcar <>
Date: Fri, 05 Jul 91 14:12:24 PDT
Subject: Is Someone Taking This Down?

Concerning the historical tidbits that turn up in na.digest
(G W Stewart's recent history of "Gaussian" elimination and
the debate over Cholesky's nationality and pronunciation some
months ago, not to mention the same about Lanczos) I hope that
someone who writes textbooks (Gene and Charlie take note) is
writing this stuff down, so ten years from now it won't have
to be rediscovered.

Regards, Joe Grcar


Date: Mon, 1 Jul 91 10:15 EDT
Subject: Conference on Discrete Algorithms

Just a friendly reminder....The deadline for abstract submission
for the next SIAM-ACM Conference on Discrete Algorithms (SODA 92)
July 12, 1991

If you have not yet submitted your ten double-spaced typed pages
(12,000 bytes) of extended abstract, now is the time to do it.
Please mail eleven copies of the abstract to: Conference
Department, SIAM, 3600 University City Science Center,
Philadelphia, PA 19104-2688.

You will be notified of acceptance or rejection by September 30,

Thank you and we look forward to your participation.

The Program Committee


From: Peter P. M. de Rijk <>
Date: Tue, 2 Jul 91 12:27:13 METDST
Subject: EuroBen Benchmarking Workshop

Second EuroBen Benchmarking Workshop.

The EuroBen benchmarking group yearly organizes a workshop for
members and non-members of the group to discuss topics of
benchmarking high-performance computers.
The second EuroBen workshop will be held in Utrecht, The Netherlands
at 19-20 September this year. The subjects that will be discussed are:

- New results obtained with the EuroBen Benchmark.
- Evolution of the Benchmark.
- Presentation/interpretation issues.
- Other views on Benchmarking.

The attendance fee is Hfl. 90,-- ($ 45). Due to practical limitations
the maximum number of attendants is set to 40. Members of the EuroBen
Group take precedence over non-members.

Registration can be done by mail:

Aad van der Steen
Academic Computing Centre Utrecht
Budapestlaan 6
3584 CD Utrecht
The Netherlands

or by email:


From: J.C. Mason <>
Date: Tue, 2 Jul 91 14:21 BST
Subject: Conference on Algorithms for Approximation


3rd International Conference run by Cranfield Institute of Technology
at Lady Margaret Hall, Oxford. July 20 - 24, 1992

This will be the third in the series of conferences on Algorithms for
Approximation. As in the first two conferences at Shrivenham, the aim will be
to provide both good research and a friendly motivating atmosphere. The venue,
Lady Margaret Hall, is an Oxford college with ample parking nearby, set in an
attractive location adjacent to the University Parks and River Cherwell.

The conference will provide an opportunity for the exchange of ideas about
current research on the approximation of functions and data, including the
design and analysis of algorithms and the application of approximation theory
and methods to practical problem areas.

Keynote speakers who have already accepted invitations include:

I. Barrodale (Victoria, Canada) E. Grosse (AT & T Bell Labs, USA)
C. de Boor (Wisconsin, USA) W. Light (Leicester, England)
C. Brezinski (Lille, France) T. Lyche (Oslo, Norway)
M.G. Cox (NPL, England) J.C. Mason (Cranfield, England)
W. Dahmen (Berlin, Germany) M.J.D. Powell (Cambridge, England)
J. Gregory (Brunel, England) S. Seatzu (Cagliari, Itlay)

It is also anticipated that about 40 contributed papers will be presented.

The Programme Committee is led by J.C. Mason (Cranfield) and M.G. Cox (NPL).

It is planned that the proceedings will be published in hardback and all
contributed papers will be considered for inclusion.

Accommodation will be available at Lady Margaret Hall where many rooms have
en-suite facilities. The normal conference booking will start with Monday
lunch and end with Friday lunch, and we anticipate a full board charge
(including conference dinner) of about 180 plus a conference fee of about
100. We hope to offer a reduced conference fee to bona fide students.

A booking form will be sent to you with the 2nd Announcement, if you indicate
your interest in attending and/or submitting a paper by writing to

Mrs. Elizabeth Smith
Applied & Computational Mathematics Group,
RMCS (Cranfield),
Swindon, Wilts. SN6 8LA


From: Heinz W. Engl <K310773%ALIJKU11@pucc.PRINCETON.EDU>
Date: Thu, 04 Jul 91 17:07:27 SDT
Subject: Surveys on Mathematics for Industry

The first issue of "Surveys on Mathematics for Industry" (Springer)
has appeared and contains:

B.Simeon, C.Fuehrer, P.Rentrop, Differential-Algebraic Equations
in Vehicle Systems Dynamics

G.Farin, Splines in CAD/CAM

A.Quarteroni, Domain Decomposition and Parallel processing for the
Numerical Solution of Partial Differential Equations

Inspection copies available at ICIAM!

Subscription information:
P.O.B 89
A-1201 Wien

Heinz W. Engl (, Linz, Austria
(Managing Editor)


Date: Fri, 05 Jul 91 14:16:12 BST
Subject: Contents of IMA Journal Numerical Analysis


N I M Gould An algorithm for large-scale quadratic programming.

Y Yuan A modified BFGS algorithm for unconstrained
R H Chan Toeplitz preconditioners for Toeplitz systems
with nonnegative generating functions.
T E Simos Some new four-step exponential-fitting methods
for the numerical solution of the radial
Schrodinger equation.
B Bialecki Sinc-collocation methods for two-point boundary
value problems.
A K Pani and P C Das A priori error estimates for a single-phase
quasilinear Stefan problem on one space dimension.
P D Loach and A J Wathen On the best least squares approximation of
continuous functions using linear splines with
free knots.
I C Demetriou and Least squares smoothing of univariate data to
M J D Powell achieve piecewise monotonicity.
I C Demetriou and The minimum sum of squares change to univariate
M J D Powell data that gives convexity.


From: Murli Gupta <>
Date: Fri, 05 Jul 91 12:11:12 EDT
Subject: Report on Multigrid Conference

The NSF/CBMS Regional Research Conference on Multigrid and
Multilevel Adaptive Methods for Partial Differential Equations
was held at The George Wasghington University, Washington, D.C.
during June 24-28, 1991

The principal speaker, Professor Stephen F. McCormick of
The University of Colorado at Denver delivered 10 lectures:

1. Multigrid Tutorial I
2. Multigrid Tutorial II
3. Multigrid Computation: Performance Assessment, Programming
Strategies, and Folklore
4. Algebraic Multigrid (AMG): Automatic Solver for Ax=b
5. Multilevel Adaptive Methods\
6. Basic Theory for Variationally Formulated MG and FAC
7. Multilevel Projection Methods: Basic Concepts
8. Multilevel Projection Methods: Computation
9. Multilevel Methods for Image Reconstruction and Riccati Equations
10. Perspectives: Summary of Lectures and Some Open Research Directions

The following participants delivered invited lectures at the conference:

Steve Schaffer, New Mexico Tech
Semi-coarsening Multigrid Methods for Elliptic PDE's with
Highly Discontinuous Coefficients

Sam Yee, Air Force Geophysical Laboratory
A Dual-Resolution Semi-Implicit Method for NWP Problems

David M. Young, University of Texas at Austin
Parallel Multigrid

James W. Thomas, Colorado State University
Applications of the Fast Adaptive Composite Grid Method
Time Dependent Problems

Ulrich Ruede, Technische Universitaet Muenchen
Fully Adaptive Multilevel Methods

Daniel Quinlan, University of Colorado at Denver
Parallel Multilevel Adaptive Methods

Robert J. Plemmons, Wake Forest University
Extended SOR Iterations for Markov Chains

Jianping Zhu, Mississippi State University
Multilevel grid method for history matching multi-phase
multi-dimensional reservoir models

John Goodrich, NASA Lewis Research Center
Multigrid methods for driven cavity

Scott R. Fulton, Clarkson University
Applications of Multigrid Methods in Meteorology

J. Gordon Wade, University of Southern California
A Multigrid Application to Electrical Impedance Tomography

Van Emden Henson, Naval Postgraduate School
Fourier Transforms on Irregular Grids: The Anterpolated
Discrete Fourier Transform (ADFT)

Zhiqiang Cai, Courant Institute of Mathematical Sciences
Theory for Finite Volume Element Methods

Michael Mascagni, Supercomputing Research Center
In search of :"Multi-scale Monte Carlo methods for solving
elliptic PDEs on massively parallel computers"

Yi-ling F. Chiang, West Virginia University
On the use of adaptive Chebyshev iteration in Chebyshev problems

An additional feature of the conference was the Multigrid Lab
run by Chaoqun Liu and John Ruge of the University of Colorado at Denver
where software material was demonstrated and made available
for hands-on experimentation to the conference participants.

The conference was very succesful and generated a great deal
on interaction between the principal lecturer and the participants
and among the participants.

Further information on the conference may be obtained from the
Conference Director:

Murli Gupta 202/994-4857
Department of Mathematics mmg@gwuvm.bitnet
George Washington University


End of NA Digest