NA Digest Sunday, November 26, 1995 Volume 95 : Issue 47

Today's Editor:
Cleve Moler
The MathWorks, Inc.

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to

URL for the World Wide Web: -------------------------------------------------------

From: Esmond Ng <esmond@msr.EPM.ORNL.GOV>
Date: Tue, 21 Nov 1995 09:47:45 -0500
Subject: Householder Fellow in Scientific Computing

Ren-Cang Li has joined the Mathematical Sciences Section at the Oak
Ridge National Laboratory as the 1995 Householder Fellow in Scientific
Computing. Ren-Cang will receive his Ph.D. in Mathematics from the
University of California at Berkeley. His research interests include
numerical linear algebra and numerical solution of ordinary
differential equations.

The Fellowship honors the late Alston S. Householder, founding Director
of the Mathematics Division (now Computer Science and 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


From: Kathy Brenan <>
Date: Wed, 22 Nov 1995 11:17:45 -0800
Subject: Change of Address for Kathy Brenan

I have transferred with the Aerospace Corporation to their
Engineering Group located in Herndon, VA. My new address is

Kathryn Brenan
The Aerospace Corporation
The Hallmark Building, Suite 187
13873 Park Center Road
Herndon, VA 22071



From: Francois Robert <>
Date: Mon, 20 Nov 1995 09:19:57 +0100
Subject: A Query about Block Matrices

I have below a nice result, and also a nice question for the na-neters:

B is an (n by n) nonnegative matrix decomposed into blocks Bij
with square diagonal blocks Bii:

B11... B1k

B = ..........

Bk1... Bkk

Moreover, we are given a p-uple p=(p1, p2, of positive integers pi,
and one defines the matrix Mp, with blocks Mij, as follows:

Mii = (Bii)^pi

Mij = (I+Bii+(Bii)^2+...(Bii)^(pi-1))*Bij

(this matrix Mp comes from parallel asynchronous iterations on a linear
fixed-point problem with matrix B)

1)a nice result (a Stein-Rosenberg-type theorem)

We have the same result between B and Mp than between the usual Jacobi and
Gauss-Seidel procedures for M-matrices (ro is the spectral radius):

(i) if ro(B)=0, then ro(Mp)=0
(ii) if ro(B) is less than one, then ro(Mp) is less or equal to ro(B),
and greater or equal to ( ro(B) )^(Max pi) (from which (i) follows)
(iii) if ro(B) is equal to 1, the same for ro(Mp)
(iv) if ro(B) is greater than one, then ro(Mp) is greater or equal to ro(B)
and less or equal to ( ro(B) )^(Max pi)

Moreover, the same result is valid for boolean matrices (with, of course,
boolean operations and spectral radii)
Please note that Mp is NOT a block-Gauss-Seidel matrix of B.

Did you have you already seen this kind of result in the litterature?

2)a nice question

In the case where ro(B)=0,
(equivalently, there exists a positive integer q, less or equal to n,
such that B^q = 0 (the null matrix), or B can be taken strictly triangular)
The conjecture is that we have also:

(Mp)^q = 0

(which gives again ro(Mp)=0)
The result is quite simple for constant pi, but I'd like to proove
it in the general case.
Thanks a lot for any help,

Francois ROBERT, ENSIMAG, BP 53 F-38041 Grenoble Cedex 9


From: Burt Simon <>
Date: Tue, 21 Nov 1995 10:56:20 -0700
Subject: Eigenvalues of An Integral Operator


Let A be the real, positive definite, symmetric operator given by

$ Af = \int_0^1 r(x,y)f(y) dy $

What can be said about the rate that the eigenvalues of A approach 0 ?

A related question concerns the finite approximation of A based on
the evenly spaced mesh on [0,1] with n points. Let $\lambda_n$ be
the smallest eigenvalue of the approximate operator, $A_n$. How
does $\lambda_n$ approach 0 ?

Our numerical experiments seem to imply that when

$ r(x,y) = exp(-k|x-y|) $

$\lambda_n$ is order $1/n$ for any k, and when

$ r(x,y) = exp(-k(x-y)^2) $

$\lambda_n$ decreases exponentially fast. Any help would be greatly
appreciated. Thanks,

Burt Simon
Dept. of Mathematics
University of Colorado at Denver


From: Praveen Sharma <sharma@pnge.PSU.EDU>
Date: Tue, 21 Nov 1995 13:19:04 -0500
Subject: Multisplitting Algorithm Needed

I am looking for parellel algorithm/code for multisplitting by
BSOR (or any other method). Can anyone give me directions?


From: Dan Sorensen <>
Date: Wed, 22 Nov 1995 16:52:59 -0600 (CST)
Subject: New Version of ARPACK Available for Large Eigenproblems


ARPACK is a collection of Fortran77 subroutines for solving
large scale eigenvalue problems.

The codes are available by anonymous ftp from

or by connecting directly to the URL

The package is designed to compute a few eigenvalues and corresponding
eigenvectors of a general n by n matrix A. It is most appropriate for large
sparse or structured matrices A where structured means that a matrix-vector
product w <- Av requires order n rather than the usual order n**2
floating point operations. This software is based upon an algorithmic variant
of the Arnoldi process called the Implicitly Restarted Arnoldi Method (IRAM).
When the matrix A is symmetric it reduces to a variant of the Lanczos process
called the Implicitly Restarted Lanczos Method (IRLM). These variants
may be viewed as a synthesis of the Arnoldi/Lanczos process with the
Implicitly Shifted QR technique that is suitable for large scale problems.
For many standard problems, a matrix factorization is not required. Only
the action of the matrix on a vector is needed.

ARPACK software is capable of solving large scale symmetric, nonsymmetric,
and generalized eigenproblems from significant application areas.
The software is designed to compute a few (k) eigenvalues with user specified
features such as those of largest real part or largest magnitude.
Storage requirements are on the order of n*k locations. No auxiliary storage
is required. A set of Schur basis vectors for the desired k-dimensional
eigen-space is computed which is numerically orthogonal to working precision.
Numerically accurate eigenvectors are available on request.

Important Features:

Reverse Communication Interface.

Single and Double Precision Real Arithmetic Versions for
Symmetric, Non-symmetric, Standard or Generalized Problems.

Single and Double Precision Complex Arithmetic Versions
for Standard or Generalized Problems.

Routines for Banded Matrices - Standard or Generalized Problems.

Routines for The Singular Value Decomposition.

Example driver routines that may be used as templates to
implement numerous Shift-Invert strategies for all problem types,
data types and precision.

TO GET THE SOFTWARE: Connect by ftp to and login
as anonymous. Then change directories to


or connect directly to the URL as described above
and follow the instructions in the README file in that directory.

Users of previous versions of ARPACK are urged to replace what they
have been using with this new version. This new software has improved
numerical performance and many new features.

Authors: R. Lehoucq, D. Sorensen, P. Vu, C. Yang

For further information send e-mail to


From: Jacko Koster <>
Date: Fri, 24 Nov 1995 16:01:50 +0100
Subject: Past and Forthcoming Workshops at CERFACS


The first two of a series of four workshops of the International Linear
Algebra Year were held September 26-29 (Direct Methods) and October 17-20,
1995 (Eigenvalues and Beyond). The workshops were organized by CERFACS and
took place in Toulouse, France. Each workshop was attended by about 90
academic and industrial researchers. From their reactions we conclude that
both meetings were a great success and that they attained their objectives
as well as the expectations of the participants.

For both workshops, URLs exist that provide links to abstracts, posters and
papers that have been made available by the speakers:

*** Direct Methods, September 26-29, 1995

*** Eigenvalues and Beyond, October 17-20, 1995

The two forthcoming workshops of the ILAY are scheduled for next year.
The first workshop is:

*** Linear Algebra in Optimization, April 22-25, 1996

This workshop will be held in Albi, a town not far from Toulouse.

The second workshop is:

*** Iterative Methods, June 10-13, 1996

More information about these workshops will appear in future issues of
the NA-Digest.

An interesting feature of the ILAY is the visitor programme that
enables participants and speakers to visit CERFACS before or after
the workshops or for longer periods during the year. More information
about this and other ILAY features can be obtained in any of the
following ways:

- ILAY home page,
- anonymous ftp to, directory
- email to


From: Yves Robert <>
Date: Mon, 20 Nov 1995 10:14:38 +0100
Subject: European Conference on Parallel Processing

Ecole Normale Superieure de Lyon, France
August 27-29, 1996

Euro-Par is THE European conference on parallel processing. It is the
merging of CONPAR-VAPP and PARLE, which were already two major
events in the field. The goal of Euro-Par'96 is to gather people
interested in any aspects of parallel computing and architectures.

Euro-Par'96 is split into workshops which focus on specific domains.
Each workshop is chaired by a major leading researcher in the field.
You are thereby ensured to have the best opportunity to meet
colleagues and to discuss on-going science. Furthermore, there are
invited talks to hear about hot topics, and, since many workshops
run concurrently, you can meet a large spectrum of people and
broaden your expertise. Workshops duration will be between 1/2
day up to 2 days depending on the number of selected papers.

All the accepted papers will be published in the collection Lecture Notes
in Computer Science by Springer-Verlag. The proceedings of Euro-Par'96
will contain full papers (12 pages), short papers (8 pages), and
poster papers (4 pages).


01 Programming environment and tools
Jack Dongarra, Knoxville, USA
02 Routing and communication in interconnection networks
Robert Cypher, Baltimore, USA
03 Automatic parallelization and high performance compilers
Chris Lengauer, Passau, D
04 Distributed systems and algorithms
Friedemann Mattern, Darmstadt, D
05 Parallel languages, programming and semantics
Ian Foster, Argonne, USA
06 Parallel non numerical algorithms
Burkhard Monien, Paderborn, D
07 Parallel numerical algorithms
Ian Duff, Didcot, UK
08 Parallel DSP and image processing
Larry Davis and Joseph Jaja, College Park, USA
09 VLSI design automation
Francky Catthoor, Leuven, B
10 Computer arithmetic
Jean-Marc Delosme, Yale, USA
11 High performance computing and applications
Wolfgang Gentzsch, Genias, D
12 Theory and models of parallel computing
William McColl, Oxford, UK
13 Parallel computer architecture
Chris Jesshope, Surrey, UK
14 Networks and ATM
Paul Kuehn, Stuttgart, D
15 Optics and other new technologies for parallel computation
Ahmed Louri, Tucson, USA
16 Neural networks
Michel Verleysen, Louvain, B
17 Scheduling and load balancing
Apostolos Gerasoulis, Rutgers, USA
18 Critical systems
Amir Pnueli, Weizman, Is
19 Performance evaluation
Francois Baccelli, Nice, F
20 Instruction level parallelism
Guang R. Gao, Montreal, CA
21 High-level and meta-level control in parallel symbolic programs
Jean-Pierre Briot, Paris, F
22 Parallel and distributed databases
Erhard Rahm, Leipzig, D


Submitted papers should be no longer than 15 pages.
The cover page must clearly indicate the names, addresses,
affiliations, and electronic addresses of the authors. The workshop
number and the workshop name related to each submitted paper
must also be clearly indicated on the cover page.

Authors are encouraged to submit PostScript versions of their papers
electronically at Authors not having
access to electronic facilities are invited to send 6 copies of their
papers and a cover letter to :

Ecole Normale Superieure de Lyon,
F-69364 Lyon Cedex 07, France
Phone: +33 72 72 80 37, Fax: +33 72 72 80 80


Electronic submission February 18, 1996
Non electronic submission February 4, 1996
Notification of acceptance/rejection May 10, 1996
Camera-ready of accepted papers June 10, 1996


Please email to for any
questions related to Euro-Par'96. Other information is available on
W3 at :

Organization: Euro-Par'96 is organized by LIP (Laboratoire de
l'Informatique du Parallelisme URA CNRS 1398) and will take place
at the Ecole Normale Superieure de Lyon.
General chair: Pierre Fraigniaud
Program co-chair: Luc Bouge' and Yves Robert
Organization chair: Anne Mignotte


From: Seth Greenblatt <>
Date: Tue, 21 Nov 1995 16:13:20 +0000 (GMT)
Subject: Time-Frequency and Time-Scale Methods for Economics and Finance

Session on

Society for Computational Economics
Second International Conference on
Computing in Economics and Finance
Geneva, Switzerland, 26-28 June 1996

Economic and financial data are often analysed in either the time
domain or the frequency domain. If the data is stationary, then these
are useful approaches. However, it is often the case that economic
and financial data are non-stationary, or non-homogeneous in some sense.
In these cases, it is instructive to look at the data either on the
time-frequency plane or on multiple scales over time. Time-frequency
and time-scale methods allow us to do this. We can observe the
changes in behaviour over time.

Papers are sought that address these issues, as they relate to
economics and finance, in areas including, but not limited to:
wavelets, wavelet packets, time-frequency distributions (Wigner-Ville,
etc.), atomic decompositions, matching pursuit, basis pursuit,
windowed Fourier methods, and Gabor decompositions.

Papers and abstracts for this session should be sent to the session

Dr. Seth A. Greenblatt
Director, Centre for Quantitative Economics and Computing
Department of Economics
Faculty of Letters and Social Sciences
University of Reading
PO Box 218
Reading RG6 6AA
FAX: +44(1734)750236
Phone: +44(1734)875123, Ext. 4059

Abstracts of up to two pages are due by 20th January 1996, and final
papers are due by 20th March 1996. Abstracts may be sent by post,
email (text, postscript, or LaTeX), or FAX. Final papers may be sent
by post or email (postscript or LaTeX).

For more information regarding the Society or the Conference, please
consult the Society for Computational Economics Homepage:

Dr. Seth A. Greenblatt
Director, Centre for Quantitative Economics
and Computing
Department of Economics
University of Reading


From: Valery P. Il'in <>
Date: Tue, 21 Nov 95 17:45:34 +0700
Subject: Proceedings of AMCA-95 Conference

For Researches, Universities, Libraries
Proceedings of AMCA-95 Conference

The International Conference "Advanced Mathematics, Computations and
Applications" has been hold in Novosibirsk, June 20-24, 1995.

The Proceedings will be published by Computing Center SD RAS
(Novosibirsk) in December, 1995, with the support of Russian Foundation
of Basic Research (RFBR) and Italien C.N.R (Consiglio Nazionale Delle

The Proceedings are edited by Academicians A.S.Alekseev (Novosibirsk)
and N.S.Bakhvalov (Moscow).

The contents of Proceedings consist of 85 selected artictles on the
following topics of the Conference:
- Approximations and Applications
- Circulation of Atmosphere and Ocean
- Computational Algebra
- Computer Architectures \& Technologies
- Domain Decomposition \& Multigrid Methods
- Incorrect and Inversed Problems
- Industrial Mathematics
- Mechanics of Continuous Media
- Statistical Simulation

The list of authors includes invited speakers (A.S.Alekseev, O.Axelsson,
N.S.Bakhvalov, V.P.Dymnikov, R.Ewing, H.Fujita, Yu.A.Kuznetsov,
M.M.Lavrentiev, E.Magenes, A.S.Sarkisyan, J.Sundermann, V.S.Vladimirov)
and contributors - wellknown and young scientists from Russia, France,
Germany, Italy, Japan, USA and other countries.

The price of the book (800 pages, hard cover) is 99$.
The Proceedings will be distributed free to the participants of AMCA-95

If you are inerested to have the Proceedings of AMCA-95 Conference,
please contact:

AMCA--95 Organizing Committee
Computing Center Russian Ac.\ Sci.,
Prospect Lavrentieva, 6
630090, Novosibirsk 90, RUSSIA
Tel.: (007)3832--351157,
Fax: (007)3832--357942


From: Robert van de Geijn <>
Date: Tue, 21 Nov 1995 09:58:36 -0600
Subject: Graduate Fellowships in CAM at UT-Austin

The Computational and Applied Mathematics (CAM) program is an
interdisciplinary graduate program at the University of Texas at
Austin. This exciting program draws from the strengths of the various
departments and colleges on campus. Associated with CAM is the Texas
Institute for Computational and Applied Mathematics (TICAM).

As part of this program, twelve fellowships were created at a level of
$25,000 per year for qualified students. While these fellowships are
restricted to US citizens, a number of other funding opportunities
also exist, including TAships and RAships in participating
departments, as well as a few other fellowships.

For more information, see our web page:

call (512) 471-7386, or contact our Graduate Coordinator, Nita Cooke


From: Victor Pereyra <>
Date: Wed, 22 Nov 95 11:15:42 PST
Subject: Position at Weidlinger Associates


Weidlinger Associates, Consulting Engineers, has an opening in its Los Altos,
California office, to support the development and maintenance of commercial
software in the areas of Non-Destructive Testing and Geophysical Oil
Prospecting and Reservoir Characterization.

Ideally, the candidate should have a MS degree in applied mathematics,
scientific computing, physics, geophysics, applied mechanics or engineering
with a strong applied mathematics and computing background, or several years
of practical experience in areas not covered by his/her formal education.
We value communications skills highly.

Any and all of the abilities listed below are of interest:

1. Knowledge of Ultrasound Testing methods for inspecting nozzles, safe ends
and other shapes; mathematical abilities to develop geometric models using
analytic functions, B-splines and other interpolation functions and to
implement ray tracing in such models.

2. Knowledge of elastic wave propagation simulation, geological modeling,
seismic data processing, and other related matters as they pertain to petroleum
exploration and reservoir characterization.

3. Development and maintenance under both UNIX and DOS/WINDOWS of large,
complex, computer modeling packages. This includes graphics, user interfaces,
and programming languages, specifically, FORTRAN, C and C++.

4. Interest in and ability to provide telephone or e-mail consultation and
support to clients and licensees of the various software products, and in
developing and performing demonstrations for clients and in trade shows.

Citizenship or permanent residency is required.

Please respond to:, with a short statement establishing your
credentials and current address. DO NOT SEND FULL RESUMES OVER e-mail.


End of NA Digest