NA Digest Sunday, December 3, 1995 Volume 95 : Issue 48

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.

URL for the World Wide Web: http://www.netlib.org/na-net/na_home.html -------------------------------------------------------

From: NA Digest <na.digest@na-net.ornl.gov>
Date: Sun Dec 3 12:40:34 EST 1995
Subject: NA Digest Calendar

The Netlib Conferences Database is on the Web at:

http://www.netlib.org/confdb/Conferences.html

NA Digest Calendar
Date Topic Place NA Digest #

Dec. 8 Runge-Kutta Centenial Amsterdam, Netherlands 45
Dec. 10-14 Global Optimization Szeged, Hungary 26
Dec. 11-13 Quasi-Monte Carlo Methods Hong Kong 36
Dec. 11-15 Latin American Congress Concepcion, Chile 44
Dec. 14-16 Dynamical Systems/Numerical Analysis Atlanta, GA 40
Dec. 14-20 Winter School on Iterative Methods Hong Kong 38
Dec. 16-19 Geophysical Inverse Problems Yosemite, CA 19

1996

Jan. 13-15 Course on Wavelets and Filter Banks Tampa, FL 42
Jan. 19-21 Boundary Elements Kiel, Germany 41
Jan. 28-30 Discrete Algorithms Atlanta, GA 25

Feb. 12-14 Network Optimization Problems Gainesville, FL 94:47
Feb. 12-15 Computational Differentiation Santa Fe, NM 29
Feb. 25-27 PVM User Group Meeting Sante Fe, NM 46
Feb. 29 Conference Honouring John Pollard Sydney, Australia 44

Mar. 4- 6 Numerical Combustion New Orleans, LA 37
Mar. 11-13 Structural Mechanics Dortmund, Germany 40
Mar. 21-24 Conference Honoring Ivo Babuska College Park, MD 45
Mar. 23 South Eastern Linear Algebra Meeting Williamsburg, VA 46

Apr. 1- 4 State of the Art in Numerical Analysis York, England 41
Apr. 9-11 Real Numbers and Computers Marseille, France 26
Apr. 9-13 Short Course in Optimization Hampton, VA 42
Apr. 9-13 Copper Mountain Conference Copper Mountain, CO 40
Apr. 11-13 Applied and Computational Mathematics Pittsburgh, PA 48
Apr. 13 Discrete Mathematics Day Ottawa, Canada 48
Apr. 14-16 Theoretical and Applied Mechanics Tuscaloosa, AL 35
Apr. 22-25 Linear Algebra in Optimization Albi, France 48
Apr. 25-26 Fortran Futures London, England 48

May 19-23 Computational Fluid Dynamics Freiburg, Germany 30
May 20-22 SIAM Conference on Optimization Victoria, BC, Canada 26
May 20-23 Parallel CFD Capri, Italy 34
May 21-24 Graphics Interface Conference Toronto, Canada 18
May 27-30 Volterra Centannial Tempe, AZ 42

June 3- 8 Domain Decomposition Bergen, Norway 28
June 10-13 Iterative Methods Toulouse, France 47
June 10-15 Honor Lax and Nirenberg Venice, Italy 29
June 11-14 Numerical Fluid Flow Breckenridge, CO 37
June 13-15 Algebraic Multilevel Iteration Methods Nijmegen, Netherlands 42
June 15-19 Hyperbolic Problems Hong Kong 36
June 17-20 SIAM Conference on Discrete Mathematics Baltimore, MD 29
June 17-20 Integral Methods in Science and Engin. Oulu, Finland 24
June 17-21 Householder XIII Symposium Pontresina,Switzerland 31
June 20-21 Mitrinovic Memorial Conference Belgrade, Serbia 26
June 24-26 CFD Short Course Ruston, LA 45
June 24-27 Numerical Analysis Russe, Bulgaria 45
June 24-28 Networks and Systems Saint Louis, MO 27
June 26-28 Time-Frequency Methods for Finance Geneva, Switzerland 47
June 26-28 Images, Wavelets and PDE's Paris, France 41

July 1- 4 Finite Element Methods Jyvaskyla, Finland 44
July 1 -5 Grid Adaptation in Computational PDEs Edinburgh, Scotland 40
July 7-11 ASME Fluids Engineering Division San Diego, CA 35
July 8-12 Prague Mathematical Conference Prague, Czech Rep. 03
July 8-12 Quality of Numerical Software Oxford, England 19
July 8-19 Numerical Analysis Summer School Leicester, England 41
July 9-12 Monte Carlo Methods Salzburg, Austria 45
July 15-19 Computational Mechanics Miskolc, Hungary 48
July 22-27 Canadian Mathematical Society Quebec, Canada 44
July 24-26 Symbolic and Algebraic Computation Zurich, Switzerland 42
July 27-30 Conference Honoring Mike Powell Cambridge, England 45

Aug. 7- 8 Workshop on Interval Arithmetic Recife, Brazil 48
Aug. 14-17 International Linear Algebra Society Chemnitz 48
Aug. 18-21 Parallel Computing Lyngby, Denmark 38
Aug. 19-21 Parallel Irregular Problems Santa Barbara, CA 44
Aug. 21-24 Total Least Squares Leuven, Belgium 38
Aug. 25-31 Congress Theor. & Appl. Mechanics Kyoto, Japan 94:46
Aug. 27-29 Parallel Processing Lyon, France 47

Sep. 2- 5 Nonlinear Programming Beijing, China 18
Sep. 9-13 "ECCOMAS 96" Paris, France 23
Sep. 9-14 Ill-Posed Problems Moscow, Russia 23
Sep. 15-17 Control System Design Dearborn, MI 43
Sep. 30 Interval Methods Wuerzburg, Germany 43

Oct. 9-11 SIAM Conference on Sparse Matrices Coeur d'Alene, ID 48
Nov. 6- 8 Innovative Time Integrators Amsterdam, Netherlands 40
Dec. 17-19 Mathematics in Signal Processing Warwick, England 48


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

From: Dianne O'Leary <oleary@cs.umd.edu>
Date: Fri, 1 Dec 1995 08:57:34 -0500
Subject: Householder Meeting Deadline Approaches

THE HOUSEHOLDER MEETING ON NUMERICAL ALGEBRA

Reminder: The deadline for submission of abstracts for
the Householder Meeting on Numerical Algebra is JANUARY 5, 1996.

The Householder Meeting on Numerical Algebra will be held
Monday June 17 - Friday June 21, 1996 in Pontresina, Switzerland,
at the Kronenhof. More complete information about the meeting and
the Householder Prize can be found at the Websites

http://www.cs.umd.edu/users/oleary
http://www.inf.ethz.ch/department/WR/html/householder/

and information is also available via ftp from

cs.umd.edu in directory pub/faculty/oleary.


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

From: Jack Dongarra <dongarra@cs.utk.edu>
Date: Sun, 26 Nov 1995 16:48:59 -0500
Subject: BLAS Birds-of-a-feather Session at SC'95

On November 13th and 14th, 52 people attended a Workshop on
Parallel BLAS and Related Interfaces, featuring 28 talks with
lively discussion throughout the meeting. The goal of the workshop
was to stimulate thought, discussion, and comment on the future
development of a set of standards for basic matrix data structures,
both dense and sparse, as well as calling sequences for a set of
low-level computational kernels for the parallel and sequential settings.

The existing BLAS have proven to be very effective in assisting portable,
efficient software for sequential and some of the current class
of high-performance computers. We would like to investigate the
possibility of extending the currently accepted standards to provide
greater coverage of sparse matrices and provide additional facilities
for parallel computing.

We propose to extend the Workshop ideas by holding a birds-of-a-feather
session at SC'95 in San Diego at 12:15 pm, Thursday, December 7th,
in Room 8 of the conference center. The purpose of this BOF is to
review the findings of the Workshop in November and to explore what
interest there is in the community to hold an on going forum to
develop calling sequences for a set of low-level computational kernels
for the parallel and sequential settings, as well as standards for
basic matrix data structures, both dense and sparse.

If you can, please join us at the BOF on December 7th, 12:15 in San Diego.

For additional information on the BOF see
http://wwwcs.hunter.cuny.edu/~wardc/bof_info/BLAS.html
http://www.netlib.org/utk/papers/sblas-meeting.html

Jack Dongarra


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

From: Hans Engler <ENGLER@guvax.acc.georgetown.edu>
Date: Tue, 28 Nov 1995 12:38:30 -0500 (EST)
Subject: Singular Values of Diagonal plus Circulant

Does anybody know a ``fast'' method for computing the singular values
of an nxn matrix A = C + D, where D is diagonal and C is
circulant? Note that there are only 2n-1 different coefficients in A.
``Fast'' here means anything faster than O(n^3) flops. Pointers
to theoretical results are also appreciated. The problem came up
in an application in medical imaging.

Hans Engler
Dept. of Mathematics
Georgetown University
Washington, DC 20057
U.S.A.
engler@guvax.georgetown.edu


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

From: Zhi Zhang <zhang@abratech.com>
Date: Wed, 29 Nov 1995 11:33:22 -0800
Subject: Isolated Problem Approach in Boundary Element Method

I am using the BEM (Boundary Element Method) to compute the potentials
on the scalp due to generators within the human brain.
To improve the accuracy, the so called IPA (Isolated Problem Approach) was
used. But the results were quite strange: compared with the analytic
solution, the whole BEM (with IPA) potential map (over all nodes) appeard to be
scaled by a non-unit factor. When the mesh was refined (by adding more
nodes), the non-unit scaling factor becomes closer to 1.

Does anyone have any experince with IPA, or knows any reference about IPA?
Please let me know, and I would like to ask some specific questions about
IPA. Any suggestions/comments will be highly appreciated.

Thanks in advance.

Zhi Zhang (John)
Tel. (415) 289-7458
email: zhang@abratech.com


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

From: Peter Bastian <peter@ica3.uni-stuttgart.de>
Date: Mon, 27 Nov 1995 13:05:47 +0100
Subject: Software for Unstructured Grids

Dear colleagues,

We are happy to announce the availability of version 3.1 of our software "UG"
(short hand for Unstructured Grids).

UG is a flexible software library for the development of adaptive multigrid
methods on unstructured meshes. The main idea of UG was to implement most parts
in a problem-independent way, so that the software can be reused for many
different applications.

The main features of the problem-independent part are:

* 2D (triangles, quadrilaterals) and 3D (tetrahedra) unstructured meshes
* tree-based local refinement and coarsening
* user-definable number of degrees of freedom in nodes, edges, faces and
elements (conforming, non-conforming and mixed finite-elements can be
implemented)
* BLAS and iterative methods (smoothers, multigrid, cg)
* powerful script language
* Multi-window graphical user interface for X11 and Macintosh. No special
graphics libraries required
* 1000 pages of documentation (man-pages, postscript)
* Parallel version with reduced functionality available (2D, degrees of
freedom only in nodes, dynamic load balancing, interfaces for PVM, NX,
PARIX, T3D, MPI)

Applications in the current distribution include:

* 2D scalar diffusion equation with linear conforming and nonconforming
finite-elements
* 2D linear elasticity with conforming and non-conforming finite-elements
* 2D incompressible, stationary Navier-Stokes equations
* 3D convection-diffusion equation

Many other applications are currently being developped on the basis of UG.
UG is available for free for all non-profit organizations. For further
information use:

* World Wide Web: http://www.ica3.uni-stuttgart.de/~ug
* FTP: ftp.ica3.uni-stuttgart.de/pub/ug/ug3.1
* email: ug@ica3.uni-stuttgart.de


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

From: Simon Streltsov <simon1@bu.edu>
Date: Wed, 29 Nov 1995 21:33:43 -0500 (EST)
Subject: Web Site: Updates in Global Optimization

I have started a web page that keeps track of

- upcoming events,
- recent publications ,
- web pages of individual researchers

in global optimization and relevent areas. Address is

http://cad.bu.edu/go

The intent of this page is to facilitate communications between researchers.

You are welcome to send in pointers to your resources via a simple online form
http://cad.bu.edu/go/#form or via e-mail (see the form below).

Simon Streltsov Dept of Manufacturing Engineering
simon1@bu.edu Boston University
phone 617-353-4209 44 Cummington St
fax 617-353-5548 Boston MA 02215


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

From: Diana Blackmore <blackmore@siam.org>
Date: Thu, 30 Nov 95 09:09:02 EST
Subject: Reid Prize: Call for Nominations

Second Reid Prize To Be Awarded
at SIAM Annual Meeting in Kansas City

SIAM is soliciting nominations for the 1996 W.T. and Idalia Reid
Prize in Mathematics. The prize, established in memory of
long-time University of Oklahoma mathematics professor W.T. Reid,
who died in 1977, recognizes outstanding work in the areas of
differential equations and control theory. The recipient will be
asked to present a lecture at the 1996 SIAM Annual Meeting in
Kansas City, where the prize will be awarded.

The prize was awarded for the first time in 1994 to Wendell
Fleming of Brown University, who was cited for his pioneering
research in geometric measure theory, the calculus of variations,
differential games, and stochastic control and filtering, as well
as for his generous nurturing of generations of applied
mathematicians and his loyal service to the mathematical sciences
community.

Letters of nomination for the prize should be sent to Reid Prize,
SIAM, 3600 University City Science Center, Philadelphia, PA
19104-2688; fax: (215) 386-7999. Additional information can be
obtained from Donna Blackmore at (215) 382-9800 or
blackmore@siam.org.

Nominations must be submitted to SIAM by January 1, 1996.


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

From: Houyuan Jiang <jiang@maths.unsw.edu.au>
Date: Tue, 28 Nov 1995 13:24:16 +1100
Subject: New Book: Recent Advances in Nonsmooth Optimization

RECENT ADVANCES IN NONSMOOTH OPTIMIZATION
Edited by
Ding-Zhu Du, Liqun Qi and Robert S. Womersley
World Scientific, 1995
ISBN: 981-02-2265-3

Nonsmooth optimization covers the minimization or maximization of functions
which do not have the the differentiability properties required by classical
methods. The field of nonsmooth optimization is significant, not only because
of the existence of nondifferentiable functions arising directly in
applications, but also because several important methods for solving difficult
smooth problems lead directly to the need to solve nonsmooth problems, which
are either smaller in dimension or simpler in structure.

This book contains twenty five papers written by forty six authors
from twenty countries in five continents. It includes papers on theory,
algorithms and applications for problems with first-order nondifferentiability
(the usual sense of nonsmooth optimization) second-order nondifferentiability,
nonsmooth equations, nonsmooth variational inequalities
and other problems related to nonsmooth optimization.


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

From: Roger Hockney <R.W.Hockney@ecs.soton.ac.uk>
Date: Fri, 1 Dec 95 14:28:08 GMT
Subject: New Book: The Science of Computer Benchmarking

I have recently published a book with SIAM, entitled:

"The Science of Computer Benchmarking"
Roger W. Hockney
ISBN 0-89871-363-3

Available at the SIAM stand, Supercomputing95, San Diego

Published November 1995, it consists of 129 pages and is
a softcover volume at US$ 21.25. Those of you interested
in computer benchmarking and performance analysis should
find the book valuable. It is a tutorial exposition of
the methodology and low-level benchmarks of the Parkbench
committee's report on parallel computer benchmarking,
together with the dimensionless theory of scaling and the
graphical presentation of results. It is suitable as a
teaching text for tutorials, advanced undergraduate and
MSc courses. The chapter headings are:

Chapter-1: "Introduction"
Chapter-2: "Methodology"
Chapter-3: "Low-level Parameters and Benchmarks"
Chapter-4: "Computational Similarity and Scaling"
Chapter-5: "Presentation of Results"

Please contact SIAM for further ordering information:
service@siam.org

Or the author regarding the book itself:
Roger W. Hockney
(Professor Emeritus, Reading University, UK)
(Visiting Professor, Southampton University,UK)
e-mail: rwh@ecs.soton.ac.uk
Ordinary mail: 4 Whitewalls Close, Compton, Newbury,
England, UK.
Telephone: +44 (1635) 578 679 (also fax after speaking).


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

From: Zahari Zlatev <luzz@sun2.dmu.dk>
Date: Tue, 28 Nov 95 08:47:13 +0100
Subject: New Book: Air Pollution Modelling

Dear colleagues,

A new book "Computer Treatment of Large Air Pollution Models"
has recently been published. It is assumed that the models are
described by systems of partial differential equations and also
that the models are considered on large space domain (in order
to be able to study the long-range transport of pollutants in
the atmosphere). The models are non-linear because advanced
chemical schemes are attached to them. This is why the discretization
of the partial differential equations leads to huge computaional
problems that are to be handled with fast numerical algorithms
on high-speed computers. The use of parallel and vector computers
is discussed in the book.

This book may be of interest not only to people that are
working with large air pollution models, but also to scientists
that are working with other large-scale computational problems
described by systems of partial differential equations.

The book reference is:

Zahari Zlatev
"Computational Treatment of Large Air Pollution Models"
Kluwer Academic Publishers,
Dordrecht-Toronto-London, 1995


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

From: Roger Ghanem <ghanem@jhu.edu>
Date: Tue, 28 Nov 1995 08:35:06 -0500
Subject: New Book: Stochastic Finite Elements

I like to announce the availability on the web of portions of the
book: "Stochastic Finite Elements: A Spectral Approach" by myself
(Roger Ghanem) and Pol Spanos. The book is published by Springer-Verlag,
and it can be found at: http://venus.ce.jhu.edu/book/book.html

Roger Ghanem, 201 Latrobe, Johns Hopkins University, Baltimore, MD 21218.
Tel: 410 516 7647 Fax: 410 516 7473 ghanem@jhu.edu


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

From: Rob Bisseling <Rob.Bisseling@math.ruu.nl>
Date: Mon, 27 Nov 1995 13:03:58 +0100
Subject: Netherlands Mathematical Research Institute

Master Class 1996-1997 of the Mathematical Research Institute

The Mathematical Research Institute (MRI) in The Netherlands has been set up
jointly by the mathematics departments of the universities of Groningen,
Nijmegen, Twente and Utrecht. The MRI organises a Master Class (MC).
Every year three parallel MC programmes are offered, each with its own topic.
The MC student follows one of these programmes. The language for
all MC activities is English.

Topics for the Master Class 1996-97
# Numerical Analysis and Supercomputing
# Stochastics and Operations Research
# Complex Geometry and Topology

Nine courses in Numerical Analysis and Supercomputing will be offered.
The teachers are:
O. Axelsson, R. Bisseling, J. Duistermaat, B. Polman,
G. Sleijpen, C. Traas, H. van der Vorst, P. Zandbergen

More information can be obtained from:

Ms. Jean Arthur (secretary MRI)
University of Utrecht, Mathematical Research Institute
P.O. Box 80.010
3508 TA Utrecht
The Netherlands

(tel: 31-30-2531472, fax: 31-30-2518394, e-mail: mri@math.ruu.nl).

The full text of this announcement can be obtained through:
http://www.math.ruu.nl/mri

Applicants seeking admission to the MC in September 1996 are
requested to contact the secretariat as soon as possible.
Deadline for applications: March 1st, 1996.


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

From: Trini Flores <flores@siam.org>
Date: Mon, 27 Nov 95 13:35:33 EST
Subject: SIAM Conference on Sparse Matrices

Second SIAM Conference on Sparse Matrices

October 9-11, 1996
The Coeur d'Alene Resort
Coeur d'Alene, Idaho

Organizers

Esmond G. Ng, Oak Ridge National Laboratory
Daniel J. Pierce, The Boeing Company

Program Committee

Ake Bjorck, Linkoping University, Sweden
Iain S. Duff, Rutherford Appleton Laboratory, United Kingdom and
CERFACS, France
Roland W. Freund, AT&T Bell Laboratories
J. Alan George, University of Waterloo, Canada
John R. Gilbert, Xerox Palo Alto Research Center
Gene H. Golub, Stanford University
Esmond G. Ng, Oak Ridge National Laboratory
Daniel J. Pierce, The Boeing Company
Horst D. Simon, Silicon Graphics Computer Systems

Conference Themes

o Applications
o Iterative Methods for Non-Hermitian Matrices
o Parallel Sparse Direct Methods
o Preconditioning Techniques
o Sparse Eigenvalue Computations
o Sparse Methods in Optimization
o Sparse Regularization and Rank-Deficient Methods
o Structured matrices

How to Participate

The program committee invites you to participate by submitting
an extended abstract not more than one page in length. Electronic
submissions are encouraged. If you need a LaTeX macro to format
your one-page, single-spaced extended abstract, send your request
for a macro to meetings@siam.org.

Completed abstracts should be sent to each of the following:
meetings@siam.org
esmond@msr.epm.ornl.gov
dpierce@espresso.rt.cs.boeing.com

and should arrive on or before APRIL 15, 1996 to be considered
for presentation.

Conference Location

The conference will be held in Coeur d'Alene, Idaho, at The Coeur
d'Alene Resort. The resort is located right on Lake Coeur
d'Alene among the Bitterroot Mountains of Northern Idaho. It is
45 minutes by car from Spokane International Airport in Spokane,
Washington.

World Wide Web

Information regarding the conference can be accessed in
electronic format through the World Wide Web:
http://www.siam.org/conf.htm


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

From: Ian Proudler <proudler@signal.dra.hmg.gb>
Date: Mon, 27 Nov 95 13:26:41 GMT
Subject: Mathematics in Signal Processing

THE (UK) INSTITUTE OF MATHEMATICS AND ITS APPLICATIONS
4TH INTERNATIONAL CONFERENCE ON MATHEMATICS IN SIGNAL PROCESSING

University of Warwick 17th - 19th December 1996

1ST CALL FOR PAPERS

Signal processing constitutes an important area for the application of
mathematical concepts and techniques fuelled, for example, by recent
developments in mobile communications and multimedia systems. The aim
of this conference is to bring together mathematicians and engineers
with a view to exploring recent developments and identifying fruitful
avenues for further research. It is hoped that the meeting will also
help to attract more mathematicians into this important and challenging
field.

Keynote Speaker: Prof. Bart DeMoor,
Katholieke Universiteit, Leuven, Belgium.

Paper are invited on topics such as: Adaptive filtering, Inverse
Problems, Pattern recognition, Numerical linear algebra, Array signal
processing, Subspace Tracking, Nonlinear Optimisation, Fuzzy
Techniques, Bayesian Estimation, Applications of finite mathematics,
Statistical signal / image analysis, Approximation techniques, High
resolution spectral analysis, Nonlinear signal processing and modelling,
Time-frequency and time-scale analysis, Blind deconvolution /
equalisation, Higher order statistics, Mathematical morphology.

Applications of interest include: physical modelling, communications,
financial modelling, medicine, meteorology, radar, seismology, sonar,
ocean science, multimedia, instrumentation and control, etc.

WWW Site: http://siwg.dra.hmg.gb/signal/ima96/

Email(for the attention of Pam Bye): imacrh@v-e.anglia.ac.uk

Organising committee Chairman: Prof. J G McWhirter (DRA, Malvern)


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

From: Matthias Bollhoefer <matthias.bollhoefer@mathematik.tu-chemnitz.de>
Date: Tue, 28 Nov 1995 11:09:05 MEZ
Subject: Website for ILAS Conference

Website for the
SIXTH CONFERENCE OF THE INTERNATIONAL LINEAR ALGEBRA SOCIETY
August 14-17, 1996, Chemnitz

A website has been installed for the ILAS conference. Information is
available at the URL

http://www.tu-chemnitz.de/ilas/

-- Matthias Bollh"ofer

Matthias Bollh"ofer |phone: (+49) 371 531-2142
Fakult"at f"ur Mathematik |FAX: (+49) 371 531-2657
Technische Universit"at Chemnitz|email: bolle@mathematik.tu-chemnitz.de
D-09107 Chemnitz, GERMANY |


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

From: Vladik Kreinovich <vladik@cs.utep.edu>
Date: Tue, 28 Nov 95 09:28:26 MST
Subject: Workshop on Interval Arithmetic in Brazil

WAI'96 - II WORKSHOP ON INTERVAL ARITHMETIC
August 7-8, 1996
Recife - Pernambuco - Brazil

Call for contributions
First Announcement

WORKSHOP. The second Workshop on Interval Arithmetic has
the purpose to join researhers interested in scientific computation
and related topics to present and discuss recent advances on this
branch of computer science and its applications.

The major topics of interest include, but are not limited to:
theoretical foundations of the computational arithmetic, interval
algorithms, interval/arithmetic co-processors, interval probability,
programming languages for scientific computation, tools for
scientific computation.

The workshop is a part of a larger annual event being held on the
campus of the Federal University of Pernambuco on August 4-9, 1996:
the XVI Meeting of the Brazilian Computing Society. This meeting
will integrate a variety of events: presentations of technical
papers and invited talks, panels, tutorials, research workshops
and tool demonstrations.

SUBMISSIONS. Ten pages in English or Portuguese, preferably
in TeX, must be received by MARCH 15th, 1996 by e-mail to
wai96@di.ufpe.br.

PROGRAMME COMMITTEE: B. de M. Acioly (UFPE), D. M. Claudio
(UFRGS), J. Dias dos Santos (UFPE), M. de B. Correia (UFPE),
M. A. Campos (UFPE), M. A. C. de Oliveira (FUEL), R. D. Lins (UFPE),
T. A. Diverio (UFRGS), and others.

SPONSORS
UFPE - Universidade Federal de Palambuco
SBC - Sociedade Brasileira de Computacao

ADDRESS:
WAI96
Departamento de Informatica
C.C.E.N. Av. Prof. Luiz Freire s/n
Cidade Universitaria
Recife - PE- BRAZIL
CEP 50732-970
E-mail: wai96@di.ufpe.br
Fax: (081) 27108430, (081) 2714925

The information about this workshop and other interval-related events is
placed on the Interval Computations homepage
URL http://cs.utep.edu/interval-comp/main.html


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

From: Irwin Pressman <ipress@math.carleton.ca>
Date: Tue, 28 Nov 1995 17:33:21 -0500 (EST)
Subject: Discrete Mathematics Day at Carleton University

DISCRETE MATHEMATICS DAY SATURDAY APRIL 13, 1996

Carleton University Department of Mathematics and Statistics
Ottawa, Ontario , Canada

Speakers and Topics:

Lieven Vandenberghe, Stanford University and K.U. Leuven, Belgium
Determinant maximization under linear matrix inequality constraints

Robin Thomas, Georgia Tech
Tutte's 3-edge-colouring conjecture

Phillipe Flajolet, Inria, France
Combinatorial Processes, Symbolic Computation and Analysis of Algorithms

Contact: Irwin Pressman ipress@math.carleton.ca
Bruce Richter brichter@math.carleton.ca


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

From: Nick Gould <nimg@letterbox.rl.ac.uk>
Date: Wed, 29 Nov 1995 08:24:48 GMT
Subject: Workshop on Linear Algebra in Optimization

ILAY WORKSHOP ON LINEAR ALGEBRA IN OPTIMIZATION, April 22-25, 1996

The third in a series of four workshops, which are a part of the CERFACS
International Linear Algebra Year, will be held in the picturesque town
of Albi, near Toulouse in South-West France, from the 22nd to the 25th
of April, 1996. This workshop is devoted to Linear Algebra in
Optimization.

The first day and a half of the workshop will be a tutorial introduction
to optimization methods, while the remaining two and a half days will be
dedicated to invited and contributed talks addressing research topics.

A preliminary list of invited speakers for the workshop includes:

A. Bjorck (Sweden), A. Conn (USA), Yu. Evtushenko (Russia), R. Fletcher
(UK), A. Griewank (Germany), C. Lemarechal (France), J. More (USA),
W. Murray (USA), J. Nocedal (USA), M. Powell (UK), R. Schnabel (USA),
D. Shanno (USA), and Ph. Toint (Belgium).

The local organizing committee comprises O. Burdakov (CERFACS), M. Dayde
(ENSEEIHT-IRIT), I. Duff (CERFACS and RAL) and N. Gould (RAL).

A limited number of contributed 20 minute talks will be selected for the
workshop, and suitable contributions are solicited. The deadline for
submissions is the 31st of January, 1996. All submissions should include
a title, together with a one to two page abstract, and should be sent to
the

Secretariat de l'Annee Internationale de l'Algebre Lineaire,
Parallel Algorithms Project, CERFACS,
42 Avenue Gustave Coriolis,
31057 Toulouse, CEDEX, France

Fax: +33 61 19 30 00
Email: rault@cerfacs.fr

Because of support, the cost of the workshop has been kept to 1500 FFR
(which includes registration, documentation, lunches, tea and coffee),
with a reduction to 1000 FFR for full-time students, and a fee of 3000
FFR for non-academics. There are many reasonably-priced hotels close to
the workshop site. A list will be available soon.

Up-to-date information on the workshop (including the abstracts of talks,
registration forms, etc.) can be found via the WWW page

http://www.cerfacs.fr/~wlay/LAY/lay.html

or directly from the Secretariat.

Nick Gould
RAL


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

From: Trini Flores <flores@siam.org>
Date: Wed, 29 Nov 95 15:56:38 EST
Subject: Deadlines for SIAM Conferences

DATES TO REMEMBER

December 1, 1995
Deadline for submission of contributed abstracts for
Eighth SIAM Conference on Discrete Mathematics
June 17-20, 1996
Johns Hopkins University
Baltimore, Maryland

December 4, 1995
Deadline for advance registration for
SIAM Symposium on Inverse Problems: Geophysical
Applications
December 16-19, 1995
Marriott Tenaya Lodge, Yosemite

December 8, 1995
Deadline for submission of minisymposium proposals
and short course proposals for
1996 SIAM Annual Meeting
July 22-26, 1996
Hyatt Regency Crown Center
Kansas City, Missouri

January 5, 1996
Deadline for hotel reservation for
ACM-SIAM Annual Symposium on Discrete Algorithms
January 28-30, 1996
Sheraton Colony Square Hotel
Atlanta, Georgia

January 9, 1996
Deadline for submission of contributed abstracts for
1996 SIAM Annual Meeting
July 22-26, 1996
Hyatt Regency Crown Center
Kansas City, Missouri

February 12-15, 1996
Deadline for hotel reservation for
Second International Workshop on Computational
Differentiation
La Fonda Hotel
Santa Fe, New Mexico

January 12, 1996
Deadline for advance registration for
ACM-SIAM Annual Symposium on Discrete Algorithms
January 28-30, 1996
Sheraton Colony Square Hotel
Atlanta, Georgia

To submit abstracts, register, or obtain additional information
on these meetings do browse or contact these sources:

World Wide Web: http://www.siam.org/conf.htm
E-Mail: meetings@siam.org
Telephone: 215-382-9800
Fax: 215-386-7999


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

From: Ling Ma <maling+@andrew.cmu.edu>
Date: Wed, 29 Nov 1995 17:41:35 -0500 (EST)
Subject: Conference on Applied and Computational Mathematics

The Department of Mathematics and Center for Nonlinear Analysis
announce an international onference on

Modern Computational and Applied Mathematics
Carnegie Mellon University
Pittsburgh, PA 15213
April 11-13, 1996

Organizing Committee:

Roy Nicolaides (Chairman), J. Chandra, David Kinderlehrer.

The invited speakers include:

Weinan E (Courant)
R. Glowinski (Houston)
D. Gottlieb (Brown)
L. Greengard (Courant)
J. Grove(Stony Brook)
K.H. Hoffman (Munich)
R. Lipton (Princeton)
O. Nevanlinna (Helsinki)
A. Quateroni (Cagliari)
O. Pironneau (Paris)
S. Osher (UCLA)

Please visit our web site:

http://www.cmu.edu/mcs/math/acm.html

or contact:

Ms. Francine Pugsley
(412) 268--2545
(412) 268--6380(FAX)
cn0s@andrew.cmu.edu

for more detailed information on hotels and registration.


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

From: Gyorgy Szeidl <mechszgy@gold.uni-miskolc.hu>
Date: Fri, 1 Dec 1995 11:13:58 GMT
Subject: Numerical Methods and Computational Mechanics in Hungary

FINAL ANNOUNCEMENT AND CALL FOR PAPERS
NUMERICAL METHODS AND COMPUTATIONAL MECHANICS
IN SCIENCE AND ENGINEERING
15-19th July 1996
Miskolc, Hungary
Organized by Central European Association for Computational Mechanics
Janos Bolyai Mathematical Sociaty Hungarian Acadamy of Sciences
University of Miskolc

URL (from which more information can be obtained):

http://silver.uni-miskolc.hu:8080/index.html
(English Home Page of the University of Miskolc; please click on the item
Conference on Numerical Methods and Computational Mechanics 96' )

or

http://silver.uni-miskolc.hu:8080/home/confer/matnum/conf96m.html

FIELDS OF PRIMARY INTEREST
Numerical algebra (sparse and dense linear systems, eigenvalue problems,
nonlinear systems, parallel algorithms, etc.)
Numerical solution of differential equations (FEM, BEM, multigrid,
difference methods, spectral methods, parallel algorithms, etc.)
Computational mechanics (FEM, BEM, parallel algorithms etc)

INTERNATONAL SCIENCE COMMITTEE
I. BABUSKA, chairman (USA) B. SZABO, co-chairman (USA)
I. S. DUFF, (UK) B. GUO, (Canada)
R. HABER, (USA) V.P. IL'IN, (Russia)
C. JOHNSON, (Sweden) M. KRIZEK, (Czech)
J.T. ODEN, (USA) J. PITKARANTA, (Finland)
L.S. XANTHIS, (UK) W.L. WENDLAND, (Germany)
M.F. WHEELER, (USA) T. CZIBERE, (Hungary)
Z. GASPAR, (Hungary) I. PACZELT, (Hungary)
P. ROZSA, (Hungary)

CALL FOR PAPERS
Authors who wish to present a 20-minute contributed lecture should submit a
one-page abstract in English and LATEX format with article 12pt style.
Page size is A4. Authors are kindly asked to submit their abstracts
via e-mail.

DEADLINES
Submission of abstarcts and registration: 30th of January, 1996.
Final program, notification of acceptance: 31st of May, 1996.
Submission of the whole paper is due on: 16th of July, 1996.
(i.e., in Miskolc on the second day of the Conference)
Information on formal requirements concerning the whole paper will be
provided from the above URLs.

CONTACT PERSONS
A.GALANTAI or Gy.SZEIDL
Institute of Mathematics or Department of Mechanics
University of Miskolc
3515 Miskolc-Egyetemvaros
Hungary

Phone: 36-46-365111 Fax: 36-46-365174
e-mail: matnum@gold.uni-miskolc.hu (preferred)
matgal@gold.uni-miskolc.hu mechszgy@gold.uni-miskolc.hu


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

From: Richard Brankin <richard@num-alg-grp.co.uk>
Date: Fri, 1 Dec 1995 10:57:48 +0000 (GMT)
Subject: Conference on Fortran Futures

Fortran Futures '96
25/26 April 1996
Fortran Conference and Exhibition

Cosponsored by: The British Computer Society Fortran Specialist Group
Visual Numerics (VNI) and The Numerical Algorithms Group (NAG)

There will be a two day Fortran conference and exhibition held near
London in 1996.

The event, Fortran Futures 96, will be held at the Ramada Hotel near
London's Heathrow Airport on the 25th and 26th April 1996 and will be
jointly organised by NAG Ltd and Visual Numerics with support from the
BCS Fortran Specialist Group.

The event will consist of talks, lectures, tutorials, product
presentations and an exhibition covering all the main topics, issues
and players in the Fortran arena. Topics covered will include Fortran
90, Fortran 95, HPF and Fortran 2000.

The aim of the conference and exhibition is to provide a forum at
which the use and future of Fortran can be discussed. It will also
allow Fortran practitioners to meet and discuss amongst themselves, and
with suppliers and system developers, their needs and interests.

Attendees will consist of developers, managers and maintainers of
Fortran codes plus those using Fortran in teaching and research
environments all over Europe.

The organising committee are seeking a number of papers on work
underway or recently completed in which Fortran played a major part.
Papers should be submitted in English to Robert Iles (NAG) or Peter
Anderton (VNI) by 31 January 1995. Succesful authors will be notified
by 14th February and will be registered as Day Delegates at the
meeting. Travel and overnight costs are the responsibility of the speaker.

All enquiries to

Robert Iles: The Numerical Algorithms Group
R.M.J.Iles@nag.co.uk
http://www.nag.co.uk/other/ff96.html
or
Peter Anderton: Visual Numerics
peter.anderton@vniuk.co.uk
http://www.vni.com

Dr. R.W.Brankin | NAG Ltd, Jordan Hill Rd | Tel: +44 (0)1865 511 245 |
richard@nag.co.uk | OXFORD, UK, OX2 8DR | Fax: +44 (0)1865 310 139 |


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

From: S. F. Keating <keating@ecsuc.ctstateu.edu>
Date: Mon, 27 Nov 1995 12:23:30 -0500 (EST)
Subject: Position at Eastern Connecticut State University

APPLIED MATHEMATICS/COMPUTATIONAL MATHEMATICS (Tenure Track)
Eastern Connecticut State University

Assistant Professor in the areas of applied or computational mathematics,
preference to those in chaotic systems, mathematical complexity, or numerical
analysis. Applicants must have their Ph.D. in mathematics by June 15, 1996.
The teaching load is typically 12 credit hours per semester. The successful
applicant must demonstrate a strong commitment to teaching with technology.
Desirable characteristics include the ability to do research, grant writing,
collaboration with department members and development of cooperative programs.
Interested applicants should send a resume, transcript and three letters
of recommendation, at least one of which should specifically address their
teaching ability, to S. F. Keating, Chair, Mathematics Search Committee,
Department of Mathematics and Computer Science, Eastern Connecticut State
University, Willimantic, Connecticut 06226-2295. Eastern Connecticut State
University is an AAO/EEO employer and is aggressively recruiting female and
minority applicants in an effort to bring greater diversity to its work force
and community.

(If you have any questions please feel free to either call me at 860-465-5274
or to send me email. My Internet address is : keating@ecsuc.ctstateu.edu)

S.F.Keating
Chair
Mathematics Search Committee
ECSU


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

From: Paul Farrell <farrell@mcs.kent.edu>
Date: Tue, 28 Nov 1995 03:50:21 -0500 (EST)
Subject: Chair Position at Kent State University

Kent State University
Chairperson - Department of Mathematics and Computer Science

Kent State University invites applications and nominations for the
position of Chairperson of the Department of Mathematics and Computer
Science. Kent is a spacious, residential campus serving more than
22,000 students, situated in a small university town within 30 miles of
the Cleveland metropolitan area. The Department of Mathematics and
Computer Science is situated in the College of Arts and Sciences and
houses programs through doctoral level in Applied Mathematics, Computer
Science, Pure Mathematics, and Statistics. It currently consists of 23
faculty in the Mathematical Sciences and 11 in Computer Science at the
Kent campus and 20 faculty in the Mathematical Sciences at the regional
campuses. The department recently moved to a new building and has an
extensive network connecting SIMD and MIMD parallel processors, servers
and over 140 workstations, and X-terminals for faculty and student
use.

Applicants for the position must have an earned doctorate, an
international research reputation as evidenced by publications, a
successful history of grant activity, and other academic and scholarly
achievements. In view of the composition of the department, applicants
should have a strong research reputation among both computer scientists
and mathematical scientists. They must have the ability and vision to
guide the department into the 21st century by developing and
maintaining, in both disciplines, a strong program of scholarship,
publications and grantsmanship, an effective advising system, and a
strong teaching program. In addition, applicants should have the
ability to work well within the university community and to foster
interdisciplinary research and cooperation with industry. The
successful applicant will be encouraged and supported in maintaining an
active research program. The salary will be competitive.

Screening of applicants will begin February 1, 1996, for a start date
of July 1, 1996, and will continue until the position is filled.
Please submit a full resume, including a list of publications, a
statement of interest regarding the post, and the names, addresses,
telephone numbers and e-mail addresses of at least five references, or
a letter of nomination to Chairperson Search Committee, Department of
Mathematics and Computer Science, Kent State University, Kent, Ohio
44242, USA, FAX (216)-672-7824. Further information about the
Department is available on the World Wide Web at URL
http://www.mcs.kent.edu/ . Questions and enquiries can be sent by
e-mail to: chair-search@mcs.kent.edu. Kent State University is an
Affirmative Action/Equal Opportunity Employer.

Paul A. Farrell farrell@mcs.kent.edu
Assoc. Professor, Computer Science Phone: (216) 672-4004 ext 258
Department of Mathematics & Fax: (216) 672-7824
Computer Science Dept: (216) 672-4004
Kent State University
Kent, OH 44242, U.S.A.


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

From: Philip Aston <P.Aston@surrey.ac.uk>
Date: Wed, 29 Nov 1995 12:25:48 +0000 (GMT)
Subject: Positions at the University of Surrey

UNIVERSITY OF SURREY
Department of Mathematical and Computing Sciences
TWO LECTURESHIPS IN MATHEMATICS

These posts arise from the University's intention to enhance the
Departmental research and teaching base in nonlinear mathematics.
Applications are invited from candidates with a strong research
background in any branch of nonlinear pure or applied mathematics.
It is anticipated that one appointment will be made in the area of
applied numerical analysis and that the second will encompass pure
mathematics relevant to nonlinearity, eg differential topology,
functional analysis, ergodic theory. The appointments will seek to
broaden the existing group of applied and numerical analysts working
in dynamical systems, chaos, turbulence, bifurcation theory,
mathematical biology and associated computational methods.
Candidates should possess or be about to obtain the degree of PhD.
The successful applicants will be required to undertake the normal
range of teaching and supervisory duties in appropriate areas of
mathematics. They will also be encouraged to develop strong links
with other Departments at the University and with Industry and
Government Research Establishments. The posts are tenable as
soon as possible.

Informal enquiries may be made to
Prof R Shail (01483-259196, email r.shail@mcs.surrey.ac.uk) or
Dr Tom Bridges (01483-300800, ext 2633, email t.bridges@mcs.surrey.ac.uk).

Salary is on the Lecturer Grade A/Grade B Scale (#15,154 - #26,430 per annum)
according to age, qualifications and experience. Superannuation is
available under USS conditions.

Further particulars may be obtained from the Personnel Office (CVC),
University of Surrey, Guildford, GU2 5XH, telephone Guildford (01483)
259279. The further particulars also appear on the Department's
World Wide Web pages at http://www.surrey.ac.uk/MCS/mcsdept.html.
Applications in the form of a curriculum vitae (2 copies) including
the names and addresses of three referees should be sent to the same
address by 5 January 1996 quoting Reference 430.

This University is committed to an Equal Opportunities Policy.


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

From: Esmond Ng <esmond@msr.EPM.ORNL.GOV>
Date: Thu, 30 Nov 1995 23:33:52 -0500
Subject: Postdoc Position at Oak Ridge Nat'l Lab

OAK RIDGE NATIONAL LABORATORY

ALSTON S. HOUSEHOLDER FELLOWSHIP IN SCIENTIFIC COMPUTING

The Mathematical Sciences Section of the Computer Science &
Mathematics Division at the Oak Ridge National Laboratory (ORNL)
invites outstanding candidates to apply for the 1996 Alston S.
Householder Fellowship in Scientific Computing.

The Fellowship honors Dr. Alston S. Householder, founding Director of
the Mathematics Division (now Computer Science & Mathematics
Division) at ORNL, and recognizes his seminal research contributions
to the fields of numerical analysis and scientific computing. The
Householder Fellowship is supported by the Office of Mathematical,
Information, and Computational Sciences of the U.S. Department of
Energy. It is a one-year appointment (potentially renewable for a
second year) with competitive salary, fringe benefits, travel
opportunities, access to state-of-the-art computational facilities
(high-performance workstations and parallel architectures), and
collaborative research opportunities in active research programs in
advanced scientific computing and computational sciences. The
purpose of the Householder Fellowship is to promote innovative
research in scientific computing on advanced computer architectures
and to facilitate technology transfer from the laboratory research
environment to industry and academia through advanced training of new
computational scientists.

ORNL's Mathematical Sciences Section has research programs in design
of numerical algorithms and development of mathematical software for
advanced computer architectures, and computational techniques and
tools for solving "grand challenge" problems. Included are
opportunities in heterogeneous distributed computing, performance
evaluation of parallel computers and programs, mathematical modeling
of environmental problems and their solution on advanced computer
architectures, and computational statistics and biostatistics. The
Householder Fellow will be expected to participate in existing
projects consistent with his or her research interests.

Applicants must have completed a doctorate in computer science,
mathematics, or statistics (no more than three years prior to the
appointment) and have a strong background and research interest in
large-scale scientific computing. For further information about the
Fellowship, contact Michael R. Leuze by phone at (423) 574-3125 or by
e-mail (leuze@msr.epm.ornl.gov).

To apply, send your resume, statement of research goals, and three
letters of recommendation by February 16, 1996, to Householder
Fellowship, c/o PhD Employment, Oak Ridge National Laboratory, Dept.
NAD-196, P.O. Box 2008, Oak Ridge, TN 37831-6216. Finalists for the
Fellowship will be invited to visit ORNL. The selection committee's
final decision will be made in April of 1996. The selected Fellow
must be available to begin the appointment during the 1996 calendar
year.

ORNL is an equal opportunity employer committed to building and
maintaining a diverse workforce. Competition for this position is
open to citizens and permanent residents of the United States.


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

From: Richard Lehoucq <lehoucq@mcs.anl.gov>
Date: Wed, 29 Nov 1995 11:03:03 -0600 (CST)
Subject: Report on the CERFACS Workshop on Eigenvalues

A workshop on eigenvalues was held October 17--20, 1995 in Toulouse,
France. The workshop was the second in a series held during the
International Linear Algebra Year hosted by CERFACS. The workshop had the
theme "Eigenvalues and beyond: towards a key for physical and numerical
stability." The last few years have seen an increasing amount of attention
focused on both nonsymmetric eigenvalue problems and their, often physical,
origin. The three and half days of talks were carefully planned to bring
together researchers and practitioners from all over the world in order to
discuss their work.

We briefly review the twenty-two invited talks referring the reader to the
excellent home page on the world wide web located at the URL
http://www.cerfacs.fr/~wlay/LAY/eigen.html. The home page has a listing of
the abstracts of all the talks. Many of the speakers have also included
links to their transparencies and papers.

The first day was an industrially oriented tutorial one. The first talk was
given by Jean-Claude Dunyach of Aerospatiale (France). He gave an
informative account of the large scale nonnormal eigenproblems solved in
the aeronautical industry. In particular, he reviewed the prediction of
flutter phenomenon and the structural optimization process and how they
lead to nonsymmetric eigenvalue problems. The talk ended with some
numerical results arising from software developed through a cooperation
with CERFACS. Professor Francoise Chaitin-Chatelin of the University Paris
IX and CERFACS gave the second talk on the important, and often neglected,
subject of the safety of computer simulations in practice. A theory
for computability in finite precision arithmetic was presented which
demonstrates the sensitivity of numerical procedures upon the arithmetic
precision used. An example involving iterations with highly nonnormal
matrices illustrated the theory. Professor James Demmel of the University
of California, Berkeley, followed with an overview of the recent progress
in fast and accurate eigenroutines for symmetric matrices in the ScaLAPACK
library. Recent work on algorithms for computing tiny eigenvalues and
singular values to high relative accuracy was also given. Dr. John Lewis of
Boeing Computer Services gave a presentation on the many issues that need
to be addressed in order to develop a black box code for the solution of
the symmetric generalized eigenvalue problems that arise in industrial
settings. The final talk of the day was given by Professor L. Nick
Trefethen of Cornell University. His talk discussed what occurs when non-
normality meets nonlinearity. A model problem was presented that served to
illustrate the many issues that arise when a nonlinear problem is
linearized resulting in a highly non-normal operator. In such a situation,
the eigenvalues may not be reliable indicators of system stability. The
afternoon also saw an excellent poster presentation. The workshop
participants then retired to a Cocktail at the Donjon du Capitole located
in the Toulouse town hall.

The second day of invited talks centered on nonsymmetric eigenvalue
problems. Professor Ilse Ipsen of North Carolina State University started
the morning session with a presentation on determining an eigenvector given
an approximate eigenvalue using inverse iteration. The talk first reviewed
the various attempts used and then focused on the additional difficulties
faced when the matrix is nonnormal. Professor Carl D. Meyer of North
Carolina State University gave the next talk on aggregation methods for
nearly uncoupled systems. The analysis and computation of the steady state
behaviour of large evolutionary systems with many discrete states was
explored. Professor Alan Edelman of MIT followed with a talk that unified
techniques from numerical linear algebra, differential geometry, and
physics applications to specify what it means to do conjugate gradient
optimization for functions of subspaces. One of the applications discussed
was an understanding of the conjugate gradient algorithm for the local
density approximation of Schrodinger's equation. Dr. Jennifer A. Scott of
the Rutherford Appleton Laboratory gave the next talk on evaluating
software for solving large sparse unsymmetric eigenvalue problems. A
comprehensive study, including substantial numerical experiments, on the
major software implementations for both Arnoldi's and subspace iteration
was given. Mr. Vincent Toumazou followed with a presentation on the
parallel computation of spectral portraits. Professor Francoise Chaitin-
Chatelin gave the presentation of Mr. Serge Gratton who had the unfortunate
coincidence of being drafted by the French Army in the week preceding the
conference. New results on condition numbers for linear algebra
computations based on the use of Kronecker products were given. To end the
busy day, Professor Gilbert Strang gave a talk on computing eigenvalues of
Toeplitz matrices with 1 x 2 blocks. The results were used to give a simple
proof of the convergence of the cascade algorithm used by wavelet
computations. The meeting adjourned to CERFACS where the workshop
participants were given a tour and introduction to the impressive
facilities.

Professor Beresford N. Parlett of the University of California, Berkeley
started Thursday's session by discussing a new approach to the eigenproblem
for symmetric tridiagonal matrices. The algorithm given removes the need
for reorthogonalizing approximate eigenvectors associated with clustered
eigenvalues that is required when inverse iteration is used. Thus, a
parallel order n^2 algorithm using inverse iteration would result.
Professor Yousef Saad of the University of Minnesota discussed the
computation of an eigenvalue problem arising from the modelling of
electronic structure. Hundreds and possibly thousands of eigenvectors are
typically required and Professor Saad discussed the many attempts and
choices made for this computation. He also described the initial modelling
of the physical problem and the tortuous road required to arrive at an
eigenvalue problem. Professor Gerard Sleijpen of the University of Utrecht,
The Netherlands, followed with a talk on joint work with Henk A. Van der
Vorst on the Jacobi--Davidson method for eigenvalue computations. As the
name implies, the algorithm combines ideas of both Jacobi and Davidson
resulting in an improved algorithm. The algorithm was first given and some
impressive numerical results followed. A particularly interesting feature
of the algorithm is that it may be viewed as a combination of both Rayleigh
quotient iteration and Arnoldi's method. The final talk of the morning was
given by Professor Olavi Nevanlinna of the Helsinki University of
Technology. He demonstrated how the Nevanlinna characteristic function
could be extended to matrix valued meromorphic functions.

The afternoon was given over to sightseeing beautiful Toulouse. A guided
tour of Toulouse was organized for the benefit of the workshop
participants. The workshop banquet was held Thursday night at the Chez
Fazoul restaurant. A gourmet meal featuring the many local plates of
Toulouse was celebrated.

The final day's talks all dealt with Krylov based methods. Professor Dan
Sorensen of Rice University, USA, gave a presentation on implicitly
restarted Arnoldi/Lanczos methods for large scale eigenvalue problems. The
talk ended with a discussion of the ARPACK software and its use on large
scale eigenvalue problems arising from various applications. Dr. Jane
Cullum of IBM gave a talk that contrasted the Arnoldi and (nonsymmetric)
Lanczos algorithms for matrix eigenvalue problems. An interesting result
was given that showed a relationship between the Arnoldi reduction to upper
Hessenberg form and tridiagonal form computed by the nonsymmetric Lanczos
algorithm. Dr. Cullum finished her presentation by discussing a method for
generating test matrices for eigenvalue computations and a nonsymmetric
Lanczos algorithm that performs no reorthogonalization. The next talk was
given by Professor Zdenek Strakos of the Institute of Computer Science,
Academy of Sciences of the Czech Republic. His talk surveyed much of the
recent results on the convergence of Krylov space methods obtained by him
and other researchers. The analysis includes a careful rounding error
analysis that examines how orthogonality is lost during the computation of
the Arnoldi basis vectors. Professor Bo Kagstrom of the University of Umea,
Sweden presented joint work with Alan Edelman and Erik Elmroth on a
geometric approach to perturbation theory of matrices and matrix pencils
based on versal deformations and stratifications. The geometry of the
matrix pencils was used by treating them as points in n^2 space, where n is
the order of the matrices. The motivation is to better understand numerical
algorithms used for computing the solutions to generalized eigenvalue
problems. Dr. Valeria Simoncini of IMGA-CNR followed with a presentation on
Ritz and pseudo-Ritz values using matrix polynomials. An analysis of two
Krylov subspace methods using block Arnoldi methods was given. The talk
ended with Dr. Simoncini discussing a restarting approach for the Arnoldi
algorithm based on the approximate solution of a Riccati equation arising
from a perturbation theorem on approximate invariant subspaces of G. W.
Stewart. To close the final day of the workshop, the organizers wisely
selected Dr. Nick Higham of the University of Manchester, England. Dr.
Higham gave a fascinating account of the effect of rounding errors in
eigenvalue computations. Though a technical subject, Nick managed to give
a clear and easy to understand presentation on many topics. Those discussed
included the accurate solution of small eigenvalue problems through the
role of backward error in terminating iterative methods.

The local organizing committee of F. Chaitin-Chatelin, Valerie Fraysse,
Osni Marques with the help and support of Iain Duff, Chiara Puglisi and
Dominique Rault and the whole Parallel Algorithms team are to be commended
for an excellent meeting. The UNESCO center in Toulouse proved more
than cooperative and the workshop effortlessly brought together many
world class researchers in a very spirited and lively meeting.

The author gratefully acknowledges the helpful commets of Francoise
Chaitin-Chatelin, Iain Duff, Alan Edelman and Valerie Fraysse.



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

End of NA Digest

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