NA Digest Sunday, July 30, 1989 Volume 89 : Issue 29
Today's Editor: Cleve Moler
Today's Topics:
-------------------------------------------------------
From: Philippe Caussignac <caussi%masg12.epfl.ch@Forsythe.Stanford.EDU>
Date: Mon, 24 Jul 89 13:01:11 GMT
Subject: Positions in Lausanne
RESEARCH POSITIONS AT THE SWISS FEDERAL INSTITUTE OF TECHNOLOGY
The numerical analysis group of the Math. Dept of the Swiss Federal
Institute of Technology in Lausanne (EPFL) is looking for two young
researchers to work on projects partly supported by grants.
One of these projects is in fluid dynamics (study of parabolized
Navier-Stokes equations applied to hypersonic viscous flow around
blunt bodies) and the other in MHD (numerical analysis of eddy
currents problems) . Both positions will be available at the beginning
of April 1990 and in principle for two years.
Successful candidates should have a thorough background in numerical
analysis and scientific computing and possibly a Ph.D . The research
will take place in small teams of three to four persons . There are
about 15 researchers in the numerical analysis group in the Math.
Dept. of the EPFL ; they have acccess to facilities like Silicon
Graphics workstations and Cray 2.
Lausanne is a very nice little town by the Lake of Geneva and the
mountains Alps are at one hour by car.
For application or questions, please contact :
Ph. Caussignac or R.Touzani
EPFL/DMA EPFL/DMA
CH-1015 LAUSANNE CH-1015 LAUSANNE
Switzerland Switzerland
Phone : 41 21 693 25 78 Phone : 41 21 693 42 47
Email : caussi@masg12.epfl.ch Email : touzani@masg15.epfl.ch
------------------------------
From: George Byrne <GDBYRNE%ERENJ.BITNET@Forsythe.Stanford.EDU>
Date: Mon, 24 Jul 89 14:55:54 EDT
Subject: New Yorker Articles on Radiation Recommended
The New Yorker Magazine had an article in three parts on electromagnetic
radiation, which may be of some interest to NA-NET readers/users. The
articles appeared June 12, 19, and 26, 1989 beginning on pp. 51, 47, & 39,
respectively. Of these, the third is perhaps of most interest to us,
since it suggests the possibility of cataracts (and other eye problems)
being induced by using VDTs (video display terminals). As one who is
having serious eye problems, including cataracts, the article is of
substantial interest to me. Elsewhere in this long series, it is suggested
that electric blankets can lead to problems, especially for pregnant
women.
The articles are not lightly written; but they may be important to us for
health reasons.
George Byrne
------------------------------
From: Wlodek Proskurowski <proskuro%castor.usc.edu@usc.edu>
Date: Mon, 24 Jul 1989 12:53:44 PDT
Subject: Multigrid Routine in NAG
Since I was approached last week at the SIAM Conference in San Diego to give
answers to my recent quiz, here they are for all to share:
Answers to the quiz on the NAG multigrid routine:
1. Accuracy
The exact solution has been chosen as u(x,y)=x^2+y^2. As a consequence, the
central differences (second order accurate) give rise to a zero truncation
error in approximating the first order terms, while the one sided differences
(first order accurate) produce a substantial trucation error for large c,
here c=100). Of course, the 5-point stencil to approximate the Laplacian
also is second order accurate and does not contribute to the truncation
error for this choice of u.
2. Convergence rate (simplified to 1-D case)
a. central differences.
Denote k=ch/2, where h is the step size h=1/n. Then the difference operator,
matrix A has the form: A=tridiag(-1-k,2,-1+k) and its eigenvalues are
l_i(A)=4-2sqr(1-k^2)cos(iPh). This spectrum shifts to the left and becomes
complex with growing value of k, to the point where the relaxation scheme used
in the multigrid altogether diverges. With fixed c and h approaching to zero,
the performance of the multigrid aproaches that for the Laplacian alone.
b. one-sided differences.
Denote k=ch.
Then A=tri(-1-k,2+k,-1) and A=tri(-1,2-k,-1+k) for back- and forward schemes.
Eigenvalues are l_i=2+k+2sqr(1+k)cos(iPh) and l_i=2-k+2sqr(1-k)cos(iPh).
One of the schemes is converging nicely depending of the sign of c.
------------------------------
From: John Connolly <CONNOLLY%UKCC.BITNET@Forsythe.Stanford.EDU>
Date: Sun, 23 Jul 89 19:54:07 EDT
Subject: IBM Supercomputing Contest
CALL FOR PAPERS
1989 IBM 3090 SUPERCOMPUTING COMPETITION
Cash Awards for Outstanding Work
in
Large Scale Computer Analysis and Modeling
The IBM Corporation and IBM Canada Ltd.
are sponsoring the
1989 IBM 3090 Supercomputing Competition
IBM invites authors from industry, research and academia to submit
papers in competition for major cash awards. The competition will be
for First, Second and Third Prizes in each of four divisions:
The divisions are:
1) Physical Science and Mathematics
2) Engineering
3) Life and Health Sciences
4) Social Sciences, Humanities and Fine Arts
The prizes in each division are:
First Prize - $25,000 US
Second Prize - 15,000 US
Third Prize - 10,000 US
An additional $10,000 will be awarded to universities that provide
substantial assistance to 1st Prize papers.
All papers must describe analysis and modeling work done using an IBM
3090 Supercomputer as the primary computational system. Judging will be
by panels of noted non-IBM experts in each division. Winning and other
selected papers will be published in IBM's PROCEEDINGS: 1989 IBM 3090
Supercomputing Competition
To enter the competition, authors must submit an abstract by October 2,
1989. All necessary information is provided in the General Information
Brochure which may be obtained from your local IBM Branch Office or by
contacting one of the Competition Administrators:
In the United States:
IBM Corporation Dept 72/BNG
44 S. Broadway
White Plains, NY 10601-4495
(914) 686-6318
In Canada:
IBM Canada Ltd. Dept 2/645
245 Consumers Road
North York, Ontario M2J1S2
(416) 758-4136
A preliminary abstract and registration must be postmarked by October 2,
1989. Final papers must be received by January 15, 1990. Results of
the competition will be announced by March 1, 1990.
------------------------------
From: Ken Jackson <krj@na.toronto.edu>
Date: Tue, 25 Jul 89 19:41:32 EDT
Subject: Human Rights Violations -- China
As chairman of the SIAM Committee on the Human Rights of Mathematical
Scientists, I just received a letter from Kari Hannibal, the Acting
Director of the Science and Human Rights Program of the Office of
Scientific Freedom and Responsibility (OFSR) of the American Association
for the Advancement of Science (AAAS), addressed to "Affiliates of the
OFSR Human Rights Program" asking for our help in documenting recent
human rights violations in China. Let me quote from her letter:
"As you are all keenly aware, the suppression of public dissent in
China in June 1989 following weeks of peaceful demonstration for
reform of government policies has resulted in hundreds of deaths,
arrests of over 1,600 people, trials, and executions. The OFSR
Science and Human Rights Program is coordinating with several
human rights organizations to obtain and confirm information on
the situation of professionals and students within the scientific,
medical and engineering communities who have experienced
violations of their human rights."
"We would like to ask your help in this documentation effort.
If you or your colleagues become aware of actions by the Chinese
government which compromise the human rights of scientists,
engineers, health professionals, or students in these fields,
would you please forward the information to us. We, in turn, will
attempt to keep you as up to date as possible with new information
as it comes to us ..."
"For those of you working in education institutions, we would urge
you to ask your respective universities to be flexible in their
dealings with students who wish to stay for an additional year.
We have been told that some students may now be arriving in North
America with an incomplete set of certificates and test scores,
having left China at the time of the demonstrations which
coincided with the period of exams."
"Finally, there have been some reports of harassment or
surveillance of Chinese students studying in North America. If
you learn of such actions, would you please let us know? Several
human rights organizations are concerned about such actions and
are collecting information on it."
Please forward any relevant information you have either to me or to
Kari Hannibal directly.
Ms. Kari Hannibal,
Acting Director of the Science and Human Rights Program,
Office of Scientific Freedom and Responsibility,
American Association for the Advancement of Science,
1333 H Street, NW,
Washington, DC, 20005.
(Phone: 202-326-6792)
(Cable: ADVANCESCI)
Prof. Kenneth R. Jackson, krj@na.toronto.edu (on Internet, CSNet,
Computer Science Dept., ARPAnet, BITNET)
University of Toronto, krj@na.utoronto.ca (on CDNnet and other
Toronto, Ontario, X.400 nets (Europe))
Canada M5S 1A4 ...!{uunet,pyramid,watmath,ubc-cs}!utai!krj
(Phone: 416-978-7075) (on UUCP)
(FAX: 416-978-4765)
------------------------------
From: Robert Skeel <skeel@cs.uiuc.edu>
Date: Wed, 26 Jul 89 11:32:04 -0500
Subject: Courseware for Numerical Analysis
In the July 23rd NA Digest Patrick Gaffney asks the question
"Why would anyone wish to use Mathematica for presenting a Numerical
Analysis course in preference say to MATLAB or to the new Kahaner,
Moler, Nash book?"
I do not know enough about either Mathematica or MATLAB to offer
advice to others. However, I am qualified to answer the question
that is posed above because I have decided to try Mathematica
for the computational part of an intro course in numerical methods.
It gets tiring to have to learn and re-learn a multitude of different
languages for performing various tasks on the computer, so a single
system like Mathematica that offers numerics, symbolics, and graphics
is attractive. I think it would appeal to the students to learn a
powerful system that is capable of doing a lot of things. Certainly
most of our students do not like to use FORTRAN (and most of our faculty
do not like to teach FORTRAN). Moreover, Mathematica may well
represent the future of mathematical computing (for most users).
One of the big unanswered questions is the extent to which symbolic math
would be useful in a n a course. There are two ways in which
it might be useful:
(1)numerical computation can profit from greater use of analytical
knowledge, and (2) written homework in n a involves a lot of tedious algebra.
Bob Skeel, DCS, U of Ill at U-C
------------------------------
From: Iain Duff <duff@antares.mcs.anl.gov>
Date: Wed, 26 Jul 89 12:09:34 CDT
Subject: CERFACS Enlarges Its Computing Base
I thought readers of the digest might be interested in a recent
development at CERFACS, a European Centre for Research and Advanced Training
in Supercomputing. The following announcement was issued last week by BBN,
who have recently announced an upgrade to their Butterfly multiprocessor
based on the new Motorola chip. Although this is essentially a shared memory
machine (single address space), it can benefit one to use it as a local
memory one since memory is distributed and there are noticeable delays if
a node is accessing data in the memory of another node. Although CERFACS
does have a small project on an INTEL hypercube, this marks the Centre's
first main move towards distributed memory computing.
If you require any further information on CERFACS please contact me.
Iain Duff
na.duff@na-net.stanford.edu
BBN Sells New High-Performance RISC Computer
to European CERFACS
CAMBRIDGE, Mass., July 18, 1989 -- BBN Advanced Computers Inc. today
announced the sale of a 30-processor TC2000(TM) computer to the European
Centre for Research and Advanced Training in Scientific Computation
(CERFACS).
CERFACS, based in Toulouse, France, intends to use the TC2000 system for
such tasks as simulating and modeling complex aerospace design factors
(e.g., turbulence and wind sheer).
Incorporating Motorola's 88000 microprocessor, the BBN TC2000 system is
the first large-scale implementation of RISC (reduced instruction set
computing) technology. The multiprocessing architecture allows
field-expansion from eight to 504 processors, with corresponding
increases in memory, memory-access bandwidth and I/O capabilities.
Four additional characteristics of the TC2000 system support rapid
program development and high-performance execution: 1) two concurrent
multiprocessor operating systems are available: the nX(TM) operating
system supports general-purpose software development, at the same time
as multi-tasking execution takes place under the pSOS+m(TM) real-time
kernel. (The nX system is based on UNIX(R)4.3BSD(TM), and the pSOS+m
kernel has been adapted from Software Components Group, Inc.); 2) all
processors can share memory over the third-generation Butterfly(R)
switch, at 38 MHz per processor; 3) a unique software-controlled
clustering option lets users designate processor groups to run programs
under either OS simultaneously; and 4) the TC2000 system offers the
Xtra(TM) programming environment, featuring the industries' only
integrated, graphical (X Window System(TM)), multiprocessing tools for
development and performance analysis. "The TC2000 system is an
excellent platform for our applied research on multiprocessor computer
architectures," said Jean-Claude Ippolito, Director of CERFACS. "It
incorporates high-performance RISC processors and also features advanced
software development tools that will improve our utilization of existing
software programs."
"This sale will be the first installation of the TC2000 system in
Europe," said Dave Micciche, BBN Advanced Computer's vice president of
sales, marketing and customer service. "This is a significant market
opportunity for our new system. The CERFACS installation will allow
some of Europe's leading computational scientists and engineers to work
with the TC2000 computer, which incorporates several technological
innovations and represents the latest development in time-critical
computing."
CERFACS was created in October 1987 to combine research and training
activities in scientific computation for European researchers, engineers
and scientists. The centre, comprised of industries, laboratories, and
universities, studies fundamental applications in such fields as
aeronautics, meteorology, the petroleum industry and hydrodynamics.
------------------------------
From: Gene Golub <golub@na-net.stanford.edu>
Date: Wed, 26 Jul 1989 11:00:15 PDT
Subject: Golub/Van Loan, 2nd Edition
The new edition of Golub/Van Loan, "Matrix Computations" is now out. It
contains about 20% new material, mainly on parallel computation. We hope you
find the book of use and of interest. Hopefully, most of the bugs have been
eliminated but, of course, a few new ones have probably been introduced.
We thank those of you who helped us with the corrections of the first edition.
The cost of the paperback is about $US 30. (For some reason, there is
a 15% surcharge when it is sent to Europe.) If you cannot easily
obtain it, you can send an order via FAX to Johns Hopkins Press (FAX
number: 301/338-6998). Please include your credit card number and its
expiration date. You should also include your postal address since the
cost of sending a copy of the book by FAX would be rather high.
Good reading!
Gene
------------------------------
From: George Byrne & Alan Hindmarsh <GDBYRNE%ERENJ.BITNET@Forsythe.Stanford.EDU>
Date: Wed, 26 Jul 89 13:51:43 EDT
Subject: Comments on Dundee N.A. Meeting
The 13th Biennial Conference
on
Numerical Analysis, Dundee
George D. Byrne
Computing and Telecommunications Systems Division
Exxon Research and Engineering Company
Annandale, NJ 08801
and
Alan C. Hindmarsh
Computing and Mathematics Research Division, L-316
Lawrence Livermore National Laboratory
P.O. Box 808
Livermore, CA 94550
The 13th Biennial Conference on Numerical Analysis was held
at the University of Dundee, Dundee Scotland, 27 - 30 June 1989.
To our knowledge this is the only regularly scheduled,
international, numerical analysis conference in the world. There
were about 200 participants, many of whom took advantage of the
scheduling of this conference and the IMA Conference on
Computational Ordinary Differential Equations (London) during
successive weeks. (Please see our article on that conference
elsewhere in this issue of SIAM News.)
Dundee is a colorful blend of the old and the new. For
example, in 1689 the Viscount Dundee raised the standard of James
VII on Dundee Law, the high hill just behind the headquarters
dormitory, overlooking the Firth of Tay. More recently, Dundee
was a whaling town. It is now a bustling city with (by American
standards) old, stone buildings, modern shopping malls, old
churches, a large university, the famous Firth of Tay bridge, and
a very friendly, helpful populace. We can think of few other
places where a lone woman would guide an obviously bewildered man
to a shop!
A lasting memory will be a walk up Dundee Law with a group
of very amicable Canadians to watch the beautiful sunset up the
Firth of Tay. The lights of the villages sparkled below in the
cool, clean air. The dark, Scotch hills formed a sharp, artistic
contrast to the sunset and the serene Firth. And this was at 11
PM (2300)!
A few of us took advantage of the long-lighted evenings by
visiting some of the local pubs to savor the local color and
products. The friendly patrons seemed amused by our array of
accents: English, Australian, North American, New Zealander, and
1
2
several varieties of Continental European.
It is worth noting that there were several entire families
who stayed in the headquarters dormitory. It was a
pleasant change to see children ranging in age from toddlers,
through pre-teens, teenagers, and university ages. There seemed
to be plenty to see and to do in and around Dundee. The local
tourist information office and the meeting organizers were most
helpful.
As for the dormitory facilities, they included community
pantries for cooking or making tea and coffee. The common bath
facilities were separated for females and males, and, like the
rest of the dorm, neat and clean. The Scotch breakfasts brought
back memories of the table groaning breakfasts of the rural
midwestern U.S. The service was outstanding, courteous, and
carried out with good humor.
The Dundee Numerical Analysis meetings are among the best
organized of any conference series, and this last one was fully
up to previous standards. The organizers had rather thoroughly
planned the registration procedures, lodging arrangements, meal
service, bus transportation between residence and conference
halls, lecture room setups, and numerous other details. There
was even a cash bar set up in West Park Hall each evening. Dave
Griffiths and Alistair Watson seemed to be ubiquitous in carrying
out the many necessary organizational tasks, and the staff people
were very efficient in all aspects of the meeting. The result
was a very smoothly operating meeting with ample opportunities
for both formal and informal discussions. About the only
obstacle to good interactions that was not under the control of
the organizers was some intermittent jack hammering just outside
the second floor lecture room.
The conference was organized around 16 invited talks, and 96
submitted talks in groups of either three or four parallel
sessions. The invited speakers were:
Mike Baines (U. Reading)
Jim Bramble (Cornell U.)
Kevin Burrage (U. Auckland)
Tom Coleman (Cornell U.)
Peter Deuflhard (Konrad-Zuse-Zentrum)
Roger Fletcher (U. Dundee)
Bengt Fornberg (Exxon Research & Engineering Co.)
Ivan Graham (U. Bath)
Nicholas Higham (Cornell U.)
Claes Johnson (Chalmers Institute of Technology)
Tom Lyche (U. Oslo)
Mike Powell (U. Cambridge)
Tom Russell (U. Colorado)
Dan Sorensen (Argonne National Lab.)
Alistair Spence (U. Bath), and
Alistair Watson (U. Dundee)
3
As might be expected from this roster alone, the topics presented
were wide-ranging. Overall, the quality of the talks was quite
high, and the invited speakers did an excellent job of conveying
their material to an audience with highly varied backgrounds.
Several areas seem to be more active than others, including
moving finite elements, finite element methods for hyperbolic
problems, Runge-Kutta methods, iterative methods for linear
algebraic systems, and analysis of special PDEs. Classical
methods in PDEs, optimization, linear algebra, and other areas
were also well represented. There were several talks on parallel
methods, and a few on applications. Two talks included
entertaining poetry as well as enlightenment: Burrage's "The
Hunting of the Snark" (with the Lewis Carroll poem in the
abstract), and Butcher's slightly less authentic "By the Silvery
Tay." Of course, a meeting like this is as valuable for the
opportunities that it gives to make informal contacts as it is
for the formal talks, perhaps even more so. And private
discussions seemed to be going on throughout the conference.
One unfortunate event during the conference was Professor
A. (Ron) Mitchell's suffering a heart attack. Ron has success-
fully guided the confernces and helped to make them what they are
today. We are pleased to report that he is recovering nicely.
We all wish him a speedy recovery and return to his low handicap
golf game.
The conference banquet was "just lovely," as the Brits might
say. John Mason the after dinner speaker, gave a speech laced
with the sharp humor that some of us associate with the British
Isles. For example, several conference speakers had plugged
their forthcoming books, planned conferences, and new journals.
John deftly retaliated by announcing the new "Pacific Journal of
Slightly Applied Numerical Analysis." One conference speaker had
also mentioned his paper "Stiff or Not Stiff, That is the
Question." The mythical co-author was H. Amlet. John pointed
out other possibilities: F. Letcher, B Ramble,.... It was great
fun, aided by a copious supply of good wine that accompanied a
dinner of smoked Scotch Salmon, hearty Scotch beef, marvelous
fellowship, and a beautiful setting.
On the last day the waitress captain asked, "Will ye be
comin' back?" Aye.
------------------------------
From: George Byrne & Alan Hindmarsh <GDBYRNE%ERENJ.BITNET@Forsythe.Stanford.EDU>
Date: Wed, 26 Jul 89 13:55:55 EDT
Subject: IMA Computational ODE Meeting
Here is a rewrite of the London SIAM article. It includes revisions
suggested by Ian Gladwell, representing IMA.
GDB
The IMA Conference
on
Computational Ordinary Differential Equations, London
George D. Byrne
Computing and Telecommunications Systems Division
Exxon Research and Engineering Company
Annandale, NJ 08801
and
Alan C. Hindmarsh
Computing and Mathematics Research Division, L-316
Lawrence Livermore National Laboratory
P.O. Box 808
Livermore, CA 94550
The IMA (Institute of Mathematics and its Applications)
Conference on Computational Ordinary Differential Equations was
held at the Imperial College of Science and Technology, London,
3-7 July 1989. There were about 150 participants from around the
world. In fact, this meeting seemed to have an even stronger
international flavor than the Dundee Conference. (Please see the
article on that conference elsewhere in this issue of SIAM News.)
Imperial College was established in 1907 and is located in
the South Kensington section of London. Hyde Park and a dazzling
array of museums are nearby. The excellent tube (underground or
subway) and bus service provided easy access to Harrod's, the
theater district, a gathering of tall ships on the Thames, and
the famous sights of London. Many of the participants, spouses,
and children took advantage of these opportunities. The
Conference tour of Hampton Court was a tour de force in
architecture, art, and English gardens - and a welcome break from
the lectures. Ironically, the 800th anniversary of the Lord
Mayoralty of London was celebrated with a fireworks display over
the tall ships -- on 4 July, American Independence Day.
London has much of the flavor of New York City - busy,
noisy, and crowded. It also has stunning sights, quiet pubs, and
a rich history. There really are red (and other colored) double
decker buses, the distinctive black taxis, blizzards of helmeted
motorcycle couriers, and the pervasive smell of Diesel exhaust.
The people were warm and kind (unlike residents of some other
major cities). In what other great city would a taxi driver show
his passenger the route to the airport on a map, tell of his
visit to America, and say "The meter says twenty-three sixty.
Pay no mind to it. I know you are short. Just pay me what you
can." Or hear a barmaid speak with the accent of the American
plains in a pub called the Plough? London is an intrinsic
1
1
contradiction. Or as a logic professor used to say, "People,
people, all, all universal statements, including this one, are
false."
The Conference was ably organized by Jeff Cash, Roland
England, Ian Gladwell, and Arieh Iserles. In addition, a great
deal of effort was spent by two IMA staff people--Yvonne May and
Pamela Irving, who were extremely pleasant and helpful throughout
the meeting. For a while, all of the organizers were subjected
to some grumbling over the level of the registration fee, which
was then reduced by 22%. There was also some confusion arising
from communication problems among the organizing committee, the
IMA staff, and the university residence staff; but all problems
seem to have resolved amicably and behind the scenes.
The conference included papers on both initial value
problems (IVPs) and boundary value problems (BVPs). It was
organized around 8 invited talks, 9 highlighted talks, and 64
contributed papers (including 7 poster papers), with the
contributed talks given in
groups of three parallel sessions. The invited speakers were:
John Butcher (U. Auckland)
Herb Keller (Caltech)
Bob O'Malley (Rensselaer)
Linda Petzold (Lawrence Livermore)
Rob Russell (Imperial College)
Chuz Sanz-Serna (U. Valladolid)
Larry Shampine (S.M.U.), and
Marc Spijker (U. Leiden).
And the highlighted speakers were:
U. Ascher (U. British Columbia)
G. Bader (U. Heidelberg)
G. D. Byrne (Exxon Research & Engg. Co.)
G. F. Corliss (Marquette U.)
U. Kirchgraber (ETH Zurich)
J. F. B. M. Kraaijevangaer (U. Leiden)
J. Lawson & M. Berzins (U. Leeds)
B. J. Leimkuhler & O. Nevalinna (Helsinki U. of Technology)
H. W. Tam (U. Illinois)
By comparison to earlier meetings on numerical ODEs, the most
striking technical aspect of this meeting was strong showing of
Runge-Kutta methods, which were featured in 16 of the papers. A
resurgence of interest in RK methods, of both implicit and
explicit types and both for IVPs and BVPs, is related to a
combination of new results on nonlinear stability, opportunities
for parallel implementations, and new results on their use for
differential-algebraic equations (DAEs). Moreover, modern
software for symbolic manipulation and calculation makes much of
the research possible.
In numbers of papers, the topic of applications, mainly to
PDE (partial differential equations) problems by the method of
lines, was a close second to RK, with 14 papers. (Of course,
these categories overlap considerably.) But it has been true for
a long time that a major motivation for work in methods for ODE
initial value problems (at least) is the solution of
time-dependent PDEs. Applications to circuit models and
mechanics problems were addressed more than once, but probably
the most exciting application talk was Zadunaisky's search for a
tenth planet by way of inverse problems for orbital ODEs. A
fairly new popular topic was that of DAE methods, which were
featured in 9 papers, covering both new theoretical results, and
1
various challenging application problems. The search for good
parallel algorithms for ODEs was the subject of 8 papers. Other
recurring themes included general linear (or hybrid) methods,
dynamic system issues (invariants, periodicity, etc.),
continuation methods, defect correction schemes, new types of
iterations for initial value problems, asymptotic analysis, and
interval methods.
The particular mix of topics addressed at this conference,
together with the expository skills of the invited and
highlighted speakers, had an especially strong effect in
encouraging cross-topic interaction. Thus, for example, IVP
method researchers were able to get some good exposure to BVP
work and vice versa; people doing traditional numerical analysis
were comfortably exposed to some non-traditional topics like
interval methods and Hamiltonian systems; people whose work has
been restricted to ODEs learned a lot about the very different
properties of DAEs and methods for their solution; and so on. A
conference with a narrower scope does not have that advantage.
Oddly, larger conferences often discourage such
cross-fertilization, simply because the size and breadth of the
list of subject areas is too formidable. As the conference came
to a close, everyone seemed to come away with a broader
familiarity with the active topics in computational ODEs, and a
greater sense of personal familiarity with the people working in
this area, both through the formal presentations and through many
informal discussions and interactions.
The conference banquet followed a sherry reception, a fine
custom to our way of thinking. The food was excellent and
included an entree of steak pie. By the way, the rumors that the
food in the United Kingdom is poor and the lager is warm are
false! The after dinner speaker was Hans Stetter, whose speech
was at once imaginative, humorous, thought provoking, and a
beacon for younger researchers in ODEs. Hans began by suggesting
that observers on the tenth planet might wonder why
numerical analysts cluster and scatter quasi-periodically. This
caused us all to realize that our world is indeed small and that
the world of practicing numerical analysts is much smaller still.
He used these thoughts to suggest that numerical analysts really
form a family and that as such we should help one another. (He
said nothing of sibling rivalry, which is just as well.) Hans
suggested that many of us were working in ODEs because it was
comfortable for us to do so. He proposed that we should move on
to the more challenging field of PDEs . Hans then talked about
liberty and truth and how important they are to us all; and the
persecution of those who seek these ideals. We paused to
remember those who suffered and are suffering during the recent
Chinese uprising and its suppression.
Hans modestly suggested that we gather again next June in
Helsinki. How about your birthday meeting next year in Wien,
Hans?
1
Yes, the weather was hot for much of the conference, the
dormitory facilities were somewhat spartan, but we would strongly
recommend London and an IMA conference to our colleagues. We
also strongly recommend the intellectual stimulation and the
better understanding of mankind and friendships that are fostered
by international conferences and travel The age of the lone
mathematician working in his attic, encrypting his work is long
past. Some of us need to realize this truth.
------------------------------
From: Homer Walker <walker@math.usu.edu>
Date: Fri, 28 Jul 89 08:39:02 MDT
Subject: Utah State Miniconference on Nonlinear Systems
MINICONFERENCE ON NEWTON-LIKE METHODS FOR LARGE-SCALE NONLINEAR SYSTEMS
The Mathematics and Statistics Department at Utah State University, with
support from the U.S. Department of Energy, is hosting a Miniconference
on Newton-Like Methods for Large-Scale Nonlinear Systems.
Dates: August 28-29, 1989.
Location: Sherwood Hills Resort, in the Wellsville Mountains about
20 miles from Logan, Utah.
Confirmed invited participants at this time are P. Brown, A. Griewank,
C. T. Kelley, J. Meza, H. Mittelmann, S. Nash, A. Poore, and F. Potra.
We are encouraging arrivals on Saturday, August 26, which will allow
significant savings on airfare for many people. Sunday, August 27,
will be a day of informal discussions and various recreational and
social activities.
For more information, contact Homer Walker, Mathematics and
Statistics Department, Utah State University, Logan, UT 84322-3900,
e-mail: na.walker@na-net.stanford.edu, walker@math.usu.edu, or
uf7099@usu.bitnet.
------------------------------
End of NA Digest
**************************
-------