NA Digest Sunday, February 9, 1992 Volume 92 : Issue 6

Today's Editor:

Cleve Moler
The MathWorks, Inc.
moler@mathworks.com

Submissions for NA Digest:

Mail to na.digest@na-net.ornl.gov.

Information about NA-NET:

Mail to na.help@na-net.ornl.gov.

-------------------------------------------------------


From: Gene Golub <golub@a31.ima.umn.edu>
Date: Tue, 4 Feb 92 22:42:29 CST
Subject: Happy New Year

We wish all our Chinese and Asian coleagues a very happy and healthy Year of
the Monkey.

Jack, Gene, Cleve and Bill -- the NA-net Gang of Four


------------------------------

From: Bill Rosener <rosener@cs.utk.edu>
Date: Mon, 3 Feb 92 13:37:44 -0500
Subject: Past Issues of the NA-Net Digest.

Past issues of the NA-NET Digest can be obtained through netlib.
For help on using netlib, send mail to: netlib@ornl.gov
Have the first line of your mail message be "send index"

To obtain the NA-NET Digest index send the following message

mail netlib@ornl.gov
Subject:
send index from na-digest


If you know the year and issue you are looking for, then send
a message similar to the one below. The following message would
return issue number 4 from the year 1992.

mail netlib@ornl.gov
Subject:
send v92n04 from na-digest

-- Bill Rosener


------------------------------

From: Jim Kirkpatrick <JIMKIRK@corral.uwyo.edu>
Date: Mon, 3 Feb 1992 14:18 MST
Subject: Is the CPC Program Library Online?

I recently obtained copies of some articles on random number generators,
published in Computer Physics Communications. Apparently, source code for
examples are available from the "CPC Program Library," Queen's University
in Belfast. Does anybody know if this library can be accessed via FTP or
some other form of on-line service? In a few cases, the examples were not
published in the article, so I can't even type them in!

Sounds like a worthy adjunct to Netlib.

Jim Kirkpatrick jimkirk@corral.uwyo.edu


------------------------------

From: Zahari Zlatev <neuzzz@alliant.uni-c.dk>
Date: Mon, 3 Feb 92 13:12:46 EST
Subject: Don't Supresss the Wiggles

Two weeks ago I sent a message to NA.DIGEST in which I asked
a question about an old paper. I got more than 75 answers. I
should like to thank everybody who answered or tried to answer my
question. Quite a few people wanted to know more about the paper.
There are two papers: one in proceedings and the other in Computers
and Fluids. The full reference to the journal paper is:

P. M. Gresho and R. L. Lee,
"Don't surpress the wiggles -- they're telling you something",
Computers and Fluids 9(1981), 223-253.

Many people think it is worthwhile (re)reading this paper. I
should add here that I do not know the authors and have not been asked
to advertize for the paper.

Best regards,
Zahari Zlatev


------------------------------

From: Tina Flores <flores@siam.org>
Date: Fri, 07 Feb 92 15:47:15 EST
Subject: SIAM Conference on Control and Its Applications

SIAM Conference on Control and Its Applications
September 17-19, 1992, Minneapolis, Minnesota

The deadline to submit contributed abstracts has been
extended to FEBRUARY 21, 1992.

Submit now your 100-word abstract and title either by
FAX to 215-386-7999, or by E-mail to meetings@siam.org
To help in formatting your submission by electronic mail,
plain TeX or LaTeX macros are available upon request. Call
215-382-9800 or send your request by e-mail to
meetings@siam.org

We look forward to your participation.

The Conference Organizing Committee


------------------------------

From: Alan Karp <KARP@PALOALTO.VNET.IBM.COM>
Date: Mon, 3 Feb 92 10:17:57 PST
Subject: Refereeing

I have been following the discussion about refereeing in the NA
digest with some interest. Like most of us, I write only fair,
perfectly correct referee reports and receive biased, error plagued
reviews of my work. I have no solution for the problem of people who
refuse to referee papers or are always late with their reviews.
However, I believe there is something we can do about poor reviews.

Poor reviews fall into two categories, those that are cursory and
those that are erroneous. The latter category includes both those
that miss the point of the paper or some key point in it and those
that are used as a forum for the reviewer to insult the authors.

Cursory reviews are a problem because they let poor papers get
published. The more poor papers that appear in the journals, the
more readers rely on the reputation of the authors to decide what to
read. This problem makes it hard for people not in the "club" to get
their work read. Erroneous reviews of both kinds are a problem
because they make it hard for good work to get published and clog
the system with resubmissions, secondary reviews, and acrimonious
correspondence.

I believe that both sets of problems are caused by anonymous
refereeing. (I have read all the letters touting the advantages of
anonymity in Physics Today and disagree with them.)

For the past 10 years I have been signing all my reviews. Revealing
my identity to the authors has had two effects. I am very careful
about making negative comments and pointing out errors, and I no
longer find myself making off-hand remarks that might be taken as
insulting. (I discovered this fact when rereading some anonymous
reviews I had done years ago.)

Does signing my name inhibit me? You bet it does. If I am not sure
something is wrong, I either check carefully or ask a carefully
worded question. I probably give more attention to criticisms in
reviews than I do to the facts in my own work. (Someone will referee
my papers and check them for me.)

What about a graduate student or assistant professor reviewing the
work of a famous person? Can we expect errors to be pointed out? Of
course we can if the errors are obvious. If there is only a
difference of opinion, the senior person's view will carry more
weight, as it would in any forum.

What about reprisals? I reject your paper so you refuse to talk to
me and instruct all your staff to reject my papers. First of all, if
reviews are signed, you are less likely to review my work unfairly.
Secondly, if I sign my name to a negative review, I am sure to have
my facts straight even if we disagree about them. With our current
system, I have heard of reprisals against innocent parties. ("It
must have been Dokes who reviewed this paper. I'll get even with him
some day.")

Signing my name also gives me the luxury of contacting the authors
directly if I have a question. Often, misunderstandings can be
resolved in a few minutes instead of months of letter writing
through an intermediary.

Will doing away with anonymous reviewing lead to overly kind
reviews? It may, but I have a solution to that problem and that of
cursory reviews; attach the reviewers name to the paper! The
reviewer will be forced to share the embarassment if any obvious
errors appear in print. On the plus side, the reviewer will be given
credit in print for the work done. (A list of referees for the year
doesn't do it.)

Often, there are honest differences of opinion. I have seen
publications that follow a controversial paper with a statement from
the referee and a rebuttal by the author. This discussion is the
most enlightening part of these papers. I heartily recommend that
all journals adopt this practice.

By the way, I have only had two papers rejected for publication out
of some 40 or so submitted. One was fully justified (I had made a
critical error); the other was not (I will submit the paper
elsewhere).


------------------------------

From: Mel Ciment <mciment@nsf.gov>
Date: Mon, 3 Feb 1992 17:29:18 -0500
Subject: NSF Grand Challenge Application Groups

******** ABSTRACT OF ANNOUNCEMENT NSF 92-7 ******************

NATIONAL SCIENCE FOUNDATION

OPPORTUNITIES IN HIGH PERFORMANCE
COMPUTING AND COMMUNICATIONS

a component of the U.S. High Performance Computing and Communications
Program

Fiscal Year 1992

Grand Challenge Applications Groups

The National Science Foundation (NSF) announces opportunities for
group oriented research for Fiscal Year 1992 in connection with the U.S.
High Performance Computing and Communications (HPCC) Program. Six to eight
proposals are expected to receive funding as Grand Challenge Application
Groups as a result of the opportunities described in this initial
announcement.

Activities supported under this announcement are expected to achieve
significant progress on Grand Challenge Applications - fundamental problems
in science and engineering, with broad economic and scientific impact,
whose solution could be advanced by applying high performance computing
techniques and resources. This HPCC activity will provide funding for
multidisciplinary groups of scientists, engineers, and mathematicians to
apply emerging high performance computing and communications systems to
advance the solution of diverse science and engineering problems. The
emphasis will be on support for groups requiring HPCC capabilities, where
such focused, cross disciplinary support is generally unavailable or
difficult to obtain. Any area of science and engineering supported by NSF
is eligible for funding under this solicitation.

Grand Challenge Applications Groups are expected to employ testbed systems
exploiting new and emerging computer and communications architectures, to
prepare the groundwork for the HPCC goal of sustained teraflop computing on
important application problems by the mid 1990's. Projects funded through
this effort will focus on the fusion of disciplinary research with emerging
high performance computing environments and architectures, within the
framework of the HPCC program goals. It is anticipated that projects will
include aspects of design of models, algorithms and software to fully
realize the potential of parallel, distributed and heterogeneous computing
systems on Grand Challenge Application problems.

Awards are planned to be in the range of $300,000 to $800,000 per year for
a period of three to five years, but there are no firm restrictions on size
or duration. The number and size of awards will be based on the quality
and potential impact of the proposals reviewed, and the availability of
funds.

Proposers interested in submitting a proposal must submit a letter of
intent to NSF by March 9, 1992. The dealine for submission of proposals is
April 30, 1992.

To receive the full announcement, NSF 92-7, address requests to
hpccgrps@nsf.gov, or utilize the NSF electronic dissemination Science and
Technology Information System, STIS.


------------------------------

From: Lloyd Fosdick <lloyd@wilkinson.cs.colorado.edu>
Date: Tue, 4 Feb 1992 21:53:03 -0700
Subject: IFIP WG2.5 Position Paper on LCAS

The following position paper has been sent to the X3 Secretariat of the
International Standards Organization (ISO) by Working Group 2.5 of the
International Federation of Information Processing Societies (IFIP).
John Reid is the principal author.

Comments on Version 3.1 of Draft ISO/IEC 10967:1991
Language Compatible Arithmetic
by
Members of IFIP Working Group 2.5 (Numerical Software)
31 Jan 1992

An extremely encouraging development of the past several years for
those concerned with constructing portable numerical software has been
the increasingly widespread use of hardware that conforms to the IEEE
standard for binary floating-point arithmetic (ISO/IEC 559:1989 and
ANSI/IEEE 754-1985). This was designed with great care and has many
features that assist the construction of software that is robust and
executes rapidly.

The proposed new standard specifies many desirable features for a
floating-point type in a loose enough manner to encompass most of
today's important hardware, though not Cray computers. Thus its
adoption would put some pressure upon Cray Research to offer an
alternative, but the pressure is there anyway since the workstation is
becoming more and more the preferred mode of access and it is clearly
desirable to be able to run small test cases on the workstation and see
the same results for these cases on the supercomputer.

The authors clearly have a problem in deciding how loose the
requirements should be, but it is really very hard to understand why
they regard as acceptable the VAX D-format with its very
restricted range (about 10^(-38) to 10^(+38)) for its precision (about 17
decimals) and the IBM double precision format whose range (about
10^(-76) to 10^(+76)) is little better.

The case that we wish to make is that the requirements should be very
tight. The effect of a loose requirement is that software written to be
robust in execution on all present and future machines that conform to
a loose standard will be very verbose and probably slow in execution.
An example is provided by Kahan (1991, Fig. 3). More likely is that
programmers will not bother to add all these extra tests and the code
will fail unexpectedly when moved to different hardware. An example of
this is provided by Tydeman (1991), who considers the computation
x / sqrt(x^2 + y^2)
which might reasonably be expected never to yield a value greater than
one. Such a tight specification is already provided by the IEEE
standard for binary floating-point arithmetic, or the IEEE
radix-independent standard for floating-point arithmetic (IEEE
854-1987), which has the great merit of permitting radix 10. Our
suggestion is that any new standard be firmly rooted on these
standards.

A further very serious defect of the proposed new standard lies in its
treatment of exceptions, for which 'notification' is required.
Notification may consist either of alteration of the control
flow of the program (the authors say this is their preferred choice)
or the output of a message in a 'hard-to-ignore' manner. Alteration of
the control flow is becoming an increasingly unrealistic choice on
today's hardware - the run-time penalty on vector or parallel hardware
may be prohibitive. The alternative of millions of lines of output
messages is even more unacceptable. What is needed is a mechanism that
records the event without interupting the execution flow. Corrective
action can then be taken when it is needed, perhaps by repeating the
calculation with a different algorithm or in a higher precision,
without penalizing the normal case. An excellent foundation has been
provided for such an approach in the IEEE standards, which reinforces
the value of basing any new standard on these standards.

Overall, one has the feeling that the proposed standard seeks to make
the best of a bad situation (except that Cray arithmentic is regarded
as beyond the pale). In fact, the situation has been improving steadily
over the last few years as more and more computers with IEEE arithmetic
come into use. The IEEE standard simplifies the problem of comparing
and analyzing the results of computations. Computations performed on
different machines can be compared and interpreted more easily,
numerical software can be expected to produce consistent results on
different systems, and so forth. The proposed new standard represents a
turn away from a strict arithmetic standard, making such comparisons
and interpretations far more difficult, and is thus detrimental to
progress in scientific computation.

References:

Kahan, W. [1991]. Analysis and refutation of the LCAS. ACM SIGNUM
Newsletter, 26, 2-15.

Tydeman, F. [1991]. Comments on LCAS draft 3.1. Presented to X3T2 in
July 1991.


------------------------------

From: Marcin Paprzycki <PBACAD::M_PAPRZYCKI"@utpb.pb.utexas.edu>
Date: Tue, 4 Feb 1992 9:12:54 GMT-0600
Subject: Permian Basin Supercomputing Conference 1992

PERMIAN BASIN SUPERCOMPUTING CONFERENCE 1992
March 13-15
The University of Texas of the Permian Basin
Odessa, Texas

Invited Speakers

Friday, March 13

I. Gladwell, Comparing Direct and Iterative Methods for Boundary Problems
C. Bischof, LAPACK: Portable Linear Algebra Software

Saturday, March 14

D.R. Kincaid, The ITPACK and NSPCG Software Packages
R. Plemmons, Large-Scale Bock Toeplitz Least Squares Computations
B.N. Datta, Large-Scale And Parallel Matrix Computations In Linear Control

Sunday, March 15

Robert A. van de Geijn, Scalable Dense Linear Algebra Libraries
D.C. Sorensen, An Implicitly Restarted Arnoldi Method

Marcin Paprzycki, Conference Chairman
Department of Mathematics and Computer Science
University of Texas of the Permian Basin
Odessa, TX 79762
Phone: (915) 367-2244


------------------------------

From: Paul Van Dooren <vdooren@maggie.csl.uiuc.edu>
Date: Wed, 5 Feb 92 08:53:26 CST
Subject: IMA Workshop on Linear Algebra for Control Theory

IMA WORKSHOP
on
LINEAR ALGEBRA FOR CONTROL THEORY
June 1--5
Minneapolis, Minnesota

The Institute for Mathematics and its Applications will sponsor a
workshop on linear algebra for control theory. This workshop is being
held in the context of the Applied Linear Algebra Year. The number of
talks has been limited in order to leave plenty of free time for
informal discussions. In addition facilities will be made available
for informal sessions in the late afternoons or evenings.

Talks are grouped according to the following themes,
with 4 or 5 speakers officially scheduled per theme :

NUMERICAL LINEAR ALGEBRA FOR CONTROL
(organized by A. Bunse-Gerstner and V. Mehrmann)
A. Bunse-Gerstner, R. Byers, A. Laub, P. Van Dooren

H-INFINITY CONTROL
(organized by A. Ran)
J. Ball, E. Jonckheere, D. Limebeer, H. Trentelman, M. Verma

CANONICAL FORMS AND INVARIANTS
(organized by D. Hinrichsen)
T. Antoulas, P. Fuhrmann, U. Helmke, R. Ober, Praetzel-Wolters

RING-THEORETIC METHODS IN LINEAR CONTROL
(organized by E. Sontag and B. Wyman)
J. Brewer, D. Cobb, G. Conte, E. Kamen

MATRIX THEORY IN CONTROL
(organized by L. Rodman)
I. Gohberg, L. Lerer, V. Mehrmann, I. Zaballa

Questions about the workshop can be directed to Paul Van Dooren,
University of Illinois, Urbana, IL 61801 (vdooren@uicsl.csl.uiuc.edu).
For details about local arrangements, please contact Willard Miller,
IMA, University of Minnesota, Minneapolis, MN 55455 (miller@ima.umn.edu).


------------------------------

From: Lloyd Fosdick <lloyd@wilkinson.cs.colorado.edu>
Date: Thu, 6 Feb 1992 09:55:07 -0700
Subject: Scientific Computing Workshop for Faculty

Summer Workshop for Faculty
High-Performance Scientific Computing
University of Colorado, June 8 - 19, 1992

This residential workshop is designed for college faculty
members interested in teaching scientific computing to
undergraduate students. Its objectives are to present
an early draft of course material we have developed to
potential instructors, to critique it, and to
discuss ways of incorporating it into a college curriculum.
This material was developed for an undergraduate course in
"High-Performance Scientific Computing" under a grant from the
National Science Foundation.

Students in our course learn to use high-performance
workstations and supercomputers through laboratory exercises
representative of scientific applications.
The exercises include numerical computation, scientific visualization,
and performance measurement. Our laboratory is equipped with
DEC 5000 and SGI Indigo workstations, and Xterminals; supercomputers
are accessed over a network.

Approximately 60% of the workshop time will be devoted to
laboratory sessions. The remainder will be devoted to tutorials on
the course material and to discussions of
how to teach this subject to undergraduates, including
practical matters of hardware and software
requirements, teaching methodology, and breadth and depth of the
subject matter.

Participants will reside in Kittredge Residence Halls on the campus.
Classes and laboratory sessions will be held in the Engineering Center
located nearby. Residence expenses (room and board) will be paid by
a grant from the National Science Foundation. Transportation expenses will
not be covered.

Attendance will be limited to sixteen participants, selected according to
these criteria:

Statement of interest. This statement describes the applicant's
interest in participating in this workshop and relevant academic
experience.

Satisfaction of prerequisites. The prerequisites are experience with
the Unix operating system and C or Fortran,
knowledge of undergraduate numerical analysis.

Faculty members from four-year colleges are especially encouraged to apply.

Inquiries and applications should be sent to: Professor Lloyd Fosdick,
Dept. of Computer Science, University of Colorado, Boulder, CO 80309-0430;
tel. (303) 492-7507; e-mail lloyd@cs.colorado.edu.

APPLICATION DEADLINE MARCH 30


------------------------------

From: Lloyd Fosdick <lloyd@wilkinson.cs.colorado.edu>
Date: Thu, 6 Feb 1992 10:01:50 -0700
Subject: Scientific Computing Workshop for Students

I would appreciate it if those of you who are in contact with
undergraduate students would bring the following announcement to
their attention:


Summer Workshop for Undergraduates
High-Performance Scientific Computing
University of Colorado, May 25 - June 5, 1992


This residential workshop is designed for upper division undergraduate
students who are interested in learning about the use of supercomputers,
high-performance workstations, and visualization in scientific
computing. Workshop participants will use a Connection Machine (CM 2),
a Cray Y-MP, DEC 5000 and SGI Indigo workstations, AVS and IDL (scientific
visualization tools) for solving problems in molecular dynamics,
wave motion, and visualization of data.

Participants will reside in Kittredge Residence Halls on the University campus.
Classes and laboratory sessions will be held in the Engineering Center
located nearby. Residence expenses (room and board) will be paid by
a grant from the National Science Foundation. Transportation expenses will
not be covered.

Attendance at the workshop will be limited to sixteen students. Selection
of participants will be based on the following criteria:

Statement of interest. This statement describes
why the applicant is interested in participating in the workshop,
relevant academic experience and interests, and long-range goals.

Letter of recommendation by faculty member. This letter should be
written by a faculty member who has served as the student's
academic advisor or teacher. It should describe the student's
strengths, academic performance, and potential for achievement.

Satisfaction of prerequisites. The prerequisites are: experience with
the Unix operating system, two semesters of calculus, one semester of
numerical analysis, two semesters of courses in physical or biological
science.


Inquiries and applications should be sent to: Professor Lloyd Fosdick,
Dept. of Computer Science, University of Colorado, Boulder, CO 80309-0430;
tel. (303) 492-7507; e-mail lloyd@cs.colorado.edu.

APPLICATION DEADLINE MARCH 30


------------------------------

From: J. C. Diaz <diaz@babieco.mcs.utulsa.edu>
Date: Sat, 8 Feb 1992 20:08:15 -0600
Subject: Announcement for SIAM Texas--Oklahoma Meeting

SIAM Texas--Oklahoma Meeting
April 3 to 4, 1992
Tulsa, Oklahoma

Invited Speakers

Dr. A. Yeremin
Department of Numerical Mathematics
Moscow, Russian Academy of Sciences

Dr. L. Kolotilina
St. Petersburg Branch of the
V.A. Steklov Mathematical Institute
Russian Academy of Sciences

Dr. M. Minkoff
Argonne National Laboratory


The SIAM Texas--Oklahoma Section Meeting is to be held
in Tulsa, Oklahoma, April 3 and 4, 1992. All talks will be
in room M1 of the Keplinger Hall on the campus of the
University of Tulsa, (E 5th Str. and S. Harvard Ave). The
conference will run from 2:00 PM to 6:00 PM on April 3, and
from 8:30 AM to 11:30AM and 1:00PM to 4:00 PM on April 4,
1992.

The main theme of the conference will be Scientific
Computing but contributed papers in all of areas of
mathematical sciences are invited. If you would like to
present a talk at the meeting send an abstract by March 15
to the organizers: J.C. Diaz, (diaz@babieco.mcs.utulsa.edu),
R. Redner, (redner@tusun2.mcs.utulsa.edu), Department of
Mathematical and Computer Sciences, The University of Tulsa,
Tulsa, OK 74104-3189.

Registration fee is $10 ($5 for students). Advanced
registration will continue until March 23, 1992. For more
information on travel, accommodations, etc., send e-mail to
the same address.

Lodging: A block of rooms have been reserved in the
Westin Hotel, (3rd Str. and S. Main Ave.). Rates are $59 per
night for double occupancy, or single occupancy. To obtain
these rates you must make reservations before March 19 by
telephoning 1-800-228-3000, and refer to SIAM.


------------------------------

From: Iain Duff <ISD%IBM-B.RUTHERFORD.AC.UK@VTVM2.CC.VT.EDU>
Date: Wed, 05 Feb 92 16:30:20 GMT
Subject: Team Leader position at CERFACS

CALL FOR APPLICATION

Position: Team leader of the Post-Processing and Visualisation
Team at CERFACS.

CERFACS ( Centre Europeen de Recherche et de Formation Avancee
en Calcul Scientifique ) is a "Groupement d'Interet Public" which
members are: Matra Marconi Space, Aerospatiale, CNRS, INRIA, ONERA,
GIE informatique CEA, Meteo France, Institut national polytechnique
de Toulouse, Universite Paul Sabatier, Region Midi-Pyrenees,
Technopolis-Csata, Centre Commun de Recherche des Communautes
Europeennes.

CERFACS is located in Toulouse (South of France).

CERFACS is a center of excellence whose expertise is the use of
high performance computers to solve problems in applied science
and in engineering and a focal point in Europe for reasearch
and training in these areas.

CERFACS is offering expertise in:

- Parallel Algorithms
- Computational Aerodynamics
- Instabilities and Turbulence
- Climate Modelling and Global Change

Candidates should submit:

1- a detailed C.V.
2- a proposal of scientific program for the
Visualization and Post-Processing Team

Final application must be received before February, 29th, 1992 at:


CERFACS Selection Commitee
42 Avenue Gustave Coriolis
31057 Toulouse Cedex
FRANCE
e-mail: adm@cerfacs.fr


------------------------------

From: University of Dortmund <UMA019%DDOHRZ11.bitnet@CUNYVM.CUNY.EDU>
Date: Thu, 06 Feb 92 15:05:43 SET
Subject: Position at University of Dortmund

Am Fachbereich Mathematik der Universitaet Dortmund
ist voraussichtlich eine

UNIVERSITAETSPROFESSUR (C4) fuer GEOMETRIE

zum 1. Maerz 1993 wiederzubesetzen.

Vorzugsweise ist an einen(e) Vertreter(in) gedacht, der(die) sich in der
Forschungsrichtung
Diskrete Geometrie oder Computer - Orientierte Geometrie
wissenschaftlich besonders ausgewiesen hat.

Habilitation fuer das Fach Mathematik oder habilitationsadaequate
Leistungen werden vorausgesetzt. Es wird die Beteiligung an der
Ausbildung der Studenten der Mathematik und anderer Fachrichtungen
erwartet. Im uebrigen richten sich die Einstellungsvoraussetzungen
nach Paragraph 49 WissHG des Landes NRW.

Schwerbehinderte Bewerber(innen) mit gleicher Eignung werden bevorzugt
behandelt.
Die Universitaet strebt eine Erhoehung des Anteils von Frauen in
Forschung und Lehre an und bittet deshalb Wissenschaftlerinnen
nachdruecklich um ihre Bewerbung.

Bewerbungen werden mit den ueblichen Unterlagen erbeten bis zum
1. Maerz 1992 an den Dekan des Fachbereichs Mathematik,
Universitaet Dortmund, Postfach 50 05 00, 4600 Dortmund 50,
Tel. (0231) 755(1)-3050.

Bitte benutzen Sie nur den brieflichen Weg fuer Ihre Bewerbung,
der Dekan.


------------------------------

From: Wei-Chang Shann <shann@math.psu.edu>
Date: Sat, 8 Feb 92 23:18:25 -0500
Subject: Position at National Central University of Taiwan

NATIONAL CENTRAL UNIVERSITY, TAIWAN, R.O.C.

Department of Mathematics

The Department of Mathematics at the National Central
University in Taiwan invites applications for tenure track
or visiting positions beginning in Fall, 1992. Numerical
analysts are specially welcome. Currently we have only two
levels of professorship: associated and full. Teaching load
is usually 7 hours per week, including 4 hours of calculus.
Minimum qualifications are a Ph.D. in Mathematical Sciences,
and a strong commitment to teaching and research. The
department now has 22 faculty members, around 200 under-
graduate and 30 graduate sutdents. We offer B.S., M.S. and
Ph.D. degrees in Mathematics.

Starting salary for Associated Professors is about NT$850,000
per year for the first two years. The currency exchange rate
today is US:NT = 1:24.74. For a middle class citizen like us
the tax rate is around 10%.

Math Department has a VAX 6510 and a micro VAS 3900, running
VMS; and an expanding network of DECstation 5000-series,
running Ultrix. IBM PC clones are common here. Every faculty
member has one. Virtually all computing facilities in Taiwan
are connected by telnet. You have a good chance to find the
the exact machine you like. But it may take you some efforts
to setup a software working environment.

Applications will be evaluated beginning early April. Send
vitae, transcripts and three letters of reference to the
following address. Since it is not easy to have an interview,
a copy of the draft of your Ph.D. thesis and/or other
publications will be much helpful. For further information,
contact me (Shann) by e-mail: t210001@sparc20.ncu.edu.tw

Prof. Hua Yang, Chairman
Department of Mathematics
National Central University
Chung Li, Taiwan
R. O. C.

email: NCUT010@TWNMOE10.BITNET
tel: 011-886-3-426-7209, 011-886-3-425-6704
FAX: 011-886-3-425-7379


------------------------------

From: George Labahn <glabahn@daisy.waterloo.edu>
Date: Fri, 7 Feb 92 13:22:57 -0500
Subject: Report on Extrapolation and Rational Approximation meeting

A Report on the International Congress on
Extrapolation and Rational Interpolation

held on Jan 13-17, 1992 Tenerife, Canary Islands, Spain


The International Congress on Extrapolation and Rational Interpolation was
held this year on the island of Tenerife in the Canary Islands of
Spain. The four day conference was the latest in a series of such congress's,
the previous one being in Luminy, France in 1989. The conference was organized
by C. Brezinski of University of Lille, France and N. Hayek, P. Gonzalez-Vera,
F. Perez-Acosta, C. Gonzalez-Concepcion, M. Camacho, J. Betancor and
M. Jimenez from the University of La Laguna, Spain. Participants included
researchers from around the world - Australia, Belgium, Canada, England,
Finland, France, Germany, Israel, Italy, Japan, Norway, Poland, Portugal,
Romania, Russia, Scotland, South Africa, Spain, Sweden, Switerland, the
Netherlands and the US. The conference was great - good talks, great location
and fine treatment for all the participants.

The conference was organized into five plenary talks along with approximately
70 short (~30 min.) talks given in two parallel sessions. The conference began
bright and early on Monday the 13-th with the first of the plenary talks -
this one by Gene Golub entitled 'Generating unknown orthogonal polynomials from
known orthogonal polynomials'. His talk (parts of which were based on some
recent joint work with Bernd Fischer of Hamburg) centered around some problems
in least squares approximation. In particular a description was given of the
problem of efficiently generating new orthogonal polynomials from old when
additional points are to be included with the previous set of approximation
points. The technique used, that of modified moments, was also shown to lead
to solutions of other problems, including for example Gaussian quadrature rules
for integrals having singularities near the endpoints of the interval of
integration. The other plenary talk on the first day was given in the afternoon
by Philip Rabinowitz from the Weizmann Institute of Science in Israel. Entitled
Extrapolation methods in Numerical Integration, the talk gave a survey of the
progress made since 1971 in using various extrapolation methods in solving
numerical integration problems. Included were methods dealing with
singular integrands and evaluation of highly oscillatory infinite integrals

Besides the plenary talks, the highlight of the first days talks
came immediately after Gene Golubs talk. Claude Brezinski presented a light
talk detailing a little of the history of some of the leaders in the fields
of extrapolation and rational approximation . The talk included a mini-album
of photos of such past luminaries as Hermite and Pade. Since some of the photos
were on the fun side, Brezinski felt compelled to include one or two photos of
himself in equally embarrassing poses - including one that probably caught him
viewing his first UFO. He also mentioned that, when he first approached the
descendants of Pade during the research for his book on Pade, they were most
surprised by the interest - they had no idea that their grandfather was anybody
famous.

Additional plenary talks were given on the Tuesday and Thursday. Martin
Gutknecht from ETH in Zurich gave an interesting talk discussing a new Pade
algorithm discovered by two Canadians, Stan Cabay and Ron Meleshko. The main
feature of their algorithm is that it is the first known fast numerically
(weakly) stable Pade algorithm. The algorithm also provides a fast stable
algorithm for inverting Hankel and Toeplitz matrices. Gutknecht surveyed the
various approaches used in both Pade computation and stable methods for
Toeplitz inversion and gave some idea of how Cabay and Meleshko
accomplished their results. Finally Gutknecht gave some conjectures and
details on how their approach can be used to establish stable algorithms in
other related fields such as rational interpolation and alternate Pade
algorithms. He also highlighted the relation of their stable algorithm to the
non-symmetric Lanczos process. The second plenary talk of this day was a
talk by Daniel Bessis from France. Titled the ''Moment Problem Formulation
of the Schrodinger Equation'', Bessis showed how to determine in an exact
manner the lowest eigenstate of Schrodinger equations. In particular he
showed how one can replace the full Schrodinger equation by a moment
problem and obtain the desired information about this lowest eigenstate.
The talk was well received and generated a number of interesting discussions
during the rest of the conference.

The next day was an off day for the conference. The organizers were kind
enough to provide the conference attendees a bus tour of Teide, the famous
volcano of the island. Lots of conversation and scenic sights - the terrain
of the island seemed to change every ten kilometers - sometimes looking like
the forests of Canada while at other times like a magazine pictorial of New
Mexico.

The fourth day of the conference featured the final plenary talk of the
conference given by Avram Sidi of the Technion in Israel -
''Rational Approximation from power series of vector-valued meromorphic
functions and their applications of the matrix eigenvalue problem". The
talk developed vector-valued rational approximation procedures for a
vector-valued function via applying vector extrapolation methods to the
sequence of partial sums of the corresponding Taylor series. The talk
included discussions of convergence and results on poles. Generalizations
of the work were given that impacted on the finding of largest distinct
eigenvalues and eigenvectors of certain matrices. The end of the talk
featured a lively discussion between the speaker and Gene Golub - a discussion
that continued into the next day at lunch. Finally a special thanks from
the speaker was given to Wolfgang Throns daughter, a new MD of approximately
14 days standing, who helped the speaker recover from loosing his voice just
24 hours previously.

The rest of the conference was filled by the short talks and lots of social
events. Examples included an energetic and easily understood talk by
Bernhard Beckermann discussing a crossrule for rational interpolation; a
clear presentation by Ana Cristina Matos on some acceleration properties for the
vector epsilon algorithm for a pair of families of vector sequences; some
interesting results from Peter Graves-Morris on the use of functional Pade
approximants for solving integral equations (main memory - doesn't always work
well, unfortunately); a talk by Jeanette Van Iseghem on best choice of poles for
certain Pade approximants ( main memory - no fun to study this problem - even
the easy case is real hard ); a description of a general way to construct lots
of Pade-like approximants by Marc Van Barel and Ademar Bultheel; and lots and
lots more that should also be mentioned (no ranking is implied above - just
some that I remember for one reason or other). Some attendees reminded me of
the talk by Jose Javier Martinez, who appears to have a reputation for lots of
humour in his talks - enough to worry his supervisor and coauthor about what he
would say during his 30 minutes of prime time. Still others mentioned the talks
from the large contingent of researchers (+ 1 new medical doctor) that came from
Colorado.

Of course the conference was much more than just the talks. There was a three
hour break every afternoon. This enabled some participants to catch up with
their work (well okay I admit I only saw one person do this) - others swam or
worked on their tans beside the pool. The participants were treated to
wonderful meals every evening. Lots of people continued to compare ideas well
into the late evening (I learned that my new Spanish and Portugese friends
could give very lucid talks in English, translating very complex ideas from
their language into mine. Yet somehow the phrase ''please, its after 2am and I
need to get some sleep before tomorrows first talk at 9'' could not be
translated into their home languages). North Americans also had to get used to
late dining. This was especially evident on the first night - participants
boarded a bus for the restaurant at 7pm - only to find that the bus ride was
to last two hours. After one hour one was barely able to hear the roar of
the bus engine over the stomach rumbling noises coming from the assembled
gathering. Lots of participants had chocolate bars or apples at the ready
for a 5 pm snack on the next day.

The highlight of the social activities occurred on the Thursday evening when
the participants were wined and dined at the Casino Taoro. A great meal
including lots of little elegant touches - menus specially engraved for
the conference, cigars and flowers handed out after the meal, a local folk band
invited to entertain at the end of the evening. Memories of the meal
included: Claude Brezinski and Michela Redivo Zaglia dancing a wild and
impressive flamengo; about half the crowd joining hands into long lines to
dancing some sort of a snake-like dance throughout the tables (watching 40
academics dance as one is usually not anyones idea of fun - but this was
something truly different); Gene Golub bouncing up and down during another
dance; Carsten Carstensen deciding that sipping wine was a good technique
for countering any nearby odor of cigarette smoke (which he disliked) - not a
good decision considering the cigar giveaway; the folk group loving
serenading Ana Cristina Matos with a Portugese love song. Afterwards
many of the conference attendees went up to do a little gambling in the
casino itself. Memories here include: George Labahn loosing money (it's my
memory); Zelia Da Roche winning at the slot machines (main technique -
only use one slot machine until one wins big - then move to another one
immediately. Made no sense but it did work for her); Marc Van Barel doubling
his salary when one of his numbers came in at the roulette wheel; most
conference people watching not gambling (these were, after all, smart people).

I do not know if the above description sounds appealing or just silly - I
do know that for me it was a wonderful conference and I would love to go again
for both people reasons and academic reasons. I look forward to attending the
next such conference.

George Labahn (with help from Gene Golub and Ana Matos)
Symbolic Computation Group
University of Waterloo,
Waterloo, Canada
glabahn@daisy.waterloo.edu



------------------------------

End of NA Digest

**************************
-------