NA Digest Sunday, April 4, 1993 Volume 93 : Issue 14

Today's Editor:

Cleve Moler
The MathWorks, Inc.

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to


From: Samir Matar <>
Date: Wed, 31 Mar 93 15:59:27 BST
Subject: John A. Gregory

John A. Gregory

Numerical Analysts everywhere will be shocked and saddened
to learn of the sudden death on Friday 26 March 1993 of
Dr. John A. Gregory. John had just returned from jogging when
he collapsed and died of a heart attack.
John had been a successful teacher for 23 years at
Brunel University where he was a Reader. His research interests
included approximation theory and CAGD. He supervised many
successful Ph.D. students and postdoctoral Research Fellows in
these fields.
He will be missed at Brunel University. The wider
numerical analysis community has lost a talented and dedicated
All E-mail letters of condolence to John's wife, Pat, c/o
the writer(, will be forward to her.

E.H. Twizell, S.A. Matar
Mike K Warby

Dept. of Maths & Stats
Brunel University
Uxbridge, Middx., UB8 3PH, U.K.


From: F. Brezzi <>
Date: Thu, 01 Apr 93 12:38:06 ITA
Subject: Ricardo Nochetto Receives Sacchi Landriani Prize

On March 11, 1993 the Istituto Lombardo, Accademia di Scienze e Lettere
awarded the International Giovanni Sacchi Landriani Prize for 1993 to
Professor Ricardo H. Nochetto of the Department of Mathematics of the
University of Maryland, College Park. The prize recognizes outstanding
contributions to the field of numerical methods for partial
differential equations made during the preceeding five years and
carries a monetary award of ten million Italian lire. Professor
Nochetto was selected by a distinguished jury from among fifteen
candidates of the highest international caliber. The jury cited his
results in several areas of partial differential equations and
numerical analysis, giving particular emphasis to his work on free
boundary problems, problems involving change of phase, and maximum norm
estimates. Luigi Amerio, President of the Istituto Lombardo, conferred
the award on Nochetto during the inauguration ceremonies for the
190th academic year of the institute held in the Palazzo Brera
in Milan.

The biennial prize was established by family, friends, and colleagues
of Giovanni Sacchi Landriani, a brilliant numerical analyst who died
tragically in 1989 at the age of 31. Nochetto is the second recipient
of the prize, the first having been awarded to Professor Douglas
N. Arnold of the Pennsylvania State University in 1991.


From: Steve Campbell <>
Date: Fri, 2 Apr 1993 12:46:08 +1100
Subject: High Index DAEs

High Index DAEs

I am in the process of compiling a survey of linear or nonlinear
differential algebraic equations (DAEs) (also known as singular,
descriptor, or semi-state systems) of index four or higher which
occur in applications. One way that these systems occur is in path
or other constrained control problems where the relative degree is
three or higher. I would be interested in any papers, preprints, or
reports dealing with DAEs which are index four or higher.

Both general discussions and specific models are being sought.
Hardcopy or email can be sent directly to the addresses below.
The report will hopefully be ready for public distribution by sometime
this summer. However, information on this topic is always welcome.

Incidently, most of the high index applications found so far have had
index five or six, although I think I have found one index ten problem.

Thank you for your assistance.
Steve Campbell

Department of Mathematics
Box 8205
North Carolina State University
Raleigh, NC 27695-8205 USA

Phone: 1-919-515-3300
FAX: 1-919-515-3798
email: (Fastest)


From: Rob Schreiber <>
Date: Mon, 29 Mar 93 09:50:16 PST
Subject: Parallel L D**-1 U versus L (D**(-1),U)

Ned Anderson points out that it speeds up backsolves if you store the
reciprocals of the diagonal elements of the upper triangular factor U
in the LU decomposition. This replaces N divides in the backsolve with
N multiplies, which are usually faster.

The tridiagonal LU decomposition and backsolve loops go like this:

D( 1 ) = ONE / D( 1 )
DO K = 1, N - 1
MULT = DL( K ) * D( K )
D( K+1 ) = D( K+1 ) - MULT*DU( K )
B( K+1 ) = B( K+1 ) - MULT*B( K )
D( K+1 ) = ONE / D( K+1 )
* D**(-1),U and (L**(-1))B calculations complete.
* Back solve:
B( N ) = B( N ) * D( N )
DO K = N - 1, 1, -1
B( K ) = ( B( K )-DU( K )*B( K+1 ) ) * D( K )

Since some new RISCs cannot divide worth a damn, this has real value,
especially if you have to do a lot of backsolves.

On really parallel machines, there is something else badly wrong with
backsolve, for full matrices, that Anderson's variant doesn't address.
Here is full backsolve, using this form of storage for U, annotated to
point out the parallelism:

* Back solve:
DO K = N, 1, -1
* This statement is a sequential bottleneck
B( K ) = B( K ) * U( K, K )
* This loop is a parallel DAXPY
DO I = 1, K-1
B( I ) = B( I ) - U( I, K ) * B( K )

Assume we have N processors. The DAXPYs take (roughly) unit time, and the
multiplications take unit time, so we finish in time 2N. If instead we
store an LDU factorization, with unit triangular L and U, and also
store D**(-1) rather than D, the sequential multiplies and divides are
completely removed from the backsolve. Now, full backsolve becomes

* D**(-1) B via parallel vector multiply
DO K = 1, N
B( K ) = B( K ) * U( K, K )
* Back solve:
DO K = N, 1, -1
* This loop is a parallel DAXPY
DO I = 1, K-1
B( I ) = B( I ) - U( I, K ) * B( K )

Now, with N processors, we finish in time N, better by a factor of two.

How stable? Using Matlab, I tried this on the upper triangular factors
of 100 matrices generated with rand(100,100). I measured the infinity-norm
of the residual (of the upper triangular system) with a random right-hand side.
The L D U residuals were 60 percent larger, on average.



Date: Fri, 2 Apr 93 12:12:10 -0500
Subject: LAPACK, Version 1.1, New Routines

Revisions Since the First Public Release of LAPACK

Since its first public release in February, 1992, LAPACK has had three
updates -- two minor updates and one major update. Minor updates consist
of code and documentation corrections, whereas major updates introduce
new routines as well as corrections. The two minor updates that occurred
were on June 30, 1992 and October 31, 1992, referenced as version 1.0a and
version 1.0b, respectively. The major update was denoted by version
1.1 and occurred on March 31, 1993. All LAPACK routines reflect the
current version number with the date on the routine indicating when it was
last modified. For more information on the revisions detailed below please
refer to the LAPACK release_notes file on netlib.

Minor updates will be eliminated as they were only needed to catch the
inevitable flux of bug fixes that accompany any initial release. We plan
to have only one or two updates a year, and provide a PRERELEASE
directory on netlib to contain new software that is being considered for
inclusion. Users can then provide input and experimentation with these
prerelease routines.

At the time of each update, there is a tar file available via xnetlib
that contains all revisions for the update. You can then simply tar
this file on top of your existing LAPACK directory. The entire package
available in lapack.tar.z will always be the most up-to-date and include
all revisions.

Bug Fixes and Code Modifications:

CHEGST, CHSEQR, SSTEBZ (and their double precision equivalents)
had errors corrected;

CLANST --> CLANHT name change to avoid notation confusion;

Special version of ILAENV was missing from the TESTING and TIMING

All -EVX drivers and xGESVX had incorrect workspace requirements
(WORK for the real routines, and RWORK in the complex routines)
specified in their Argument sections;

xGTCON also had incorrect workspace requirements specified in
its Argument section;

Added functionality has been incorporated into the expert driver
routines that involve equilibration (xGESVX, xGBSVX, xPOSVX, xPPSVX,
xPBSVX). The option FACT = 'F' now permits the user to input a
prefactored, pre-equilibrated matrix. The testing routines in
TESTING/LIN that involve these routines have also been modified;

LIN Timing Suite modifications to allow the input of the number of
right hand sides, and timing code added to time the GT and PT
routines including comparisons with LINPACK;

xGESVD bug fix and added functionality;

Comments added to SLAEBZ and SSTEBZ to enforce SCALAR mode in
certain loops; in addition, in front of each such "DO" statement
a series of compiler directives has been added which should
enforce SCALAR mode on most popular vector computers;

STIM21, code correction to fix variable ITCNT in COMMON misdeclared as
INTEGER instead of REAL.

New Routines:

xGEGS and xGEGV, driver routines for the Generalized Nonsymmetric
Eigenvalue Problem with accompanying computational and auxiliary
routines xGGBAK, xGGBAL, xGGHRD, xHGEQZ, xTGEVC, and SLAG2;

xGGQRF and xGGRQF, computational routines for the generalized QR
and RQ factorizations;

xGGSVD, driver routine for the generalized singular value
decomposition with accompanying computational and auxiliary routines

xGGGLM, driver routine for the solution of the generalized linear
regression model;

xGGLSE, driver routine for the solution of the constrained linear
least squares problem.

Additional Features:

On-line manpages (troff files) for LAPACK driver and computational
routines, as well as most of the BLAS routines, available via xnetlib.

As always, all questions/comments should be sent to




From: Dean Schulze <>
Date: 03 Apr 1993 13:04:37 -0700 (MST)
Subject: Problems with "High Quality" Random Number Generators

A recent Physical Review Letter [1] points out that serious problems
can arise in Monte Carlo computations due to subtle correlations in "high
quality" random number generators. The quality of these number generators
was determined to be "good" because they passed a battery of tests for
randomness. However, they produced erroneous results when used together
with the Wolff algorithm for cluster-flipping in a simulation of a 2
dimensional Ising model for which the results are known. The author of
this Letter, Alan M. Ferrenburg of the University of Georgia, says that
an algorithm must be tested together with the random number generator
being used regardless of which tests the random number generator has
passed on its own.

In another development, Shu Tezuka of IBM, Tokyo and Pierre L'Ecuyer
of the University of Montreal have proven that the Marsaglia-Zaman random
number generators are "essentially equivalent" to linear congruential
methods [2]. (Linear congruential number generators produced better results
in Ferrenburg's simulations than random number generation algorithms
that are of higher quality, however.)

[1] Alan M. Ferrenburg, D.P. Landau, and Y. Joanna Wong,
"Monte Carlo simulations: Hidden errors from 'good'
random number generators", Phys. Rev. Lett., 69, pp.
3382-4, 1992.

[2] Science News, v142, pg. 422, 1992.


From: Kristjan Jonasson <>
Date: Tue, 30 Mar 93 16:31:55 CET
Subject: CUTE on PC's

I have now installed CUTE (Constrained and Unconstrained Testing
Environment, see NA-Digest #11 1993) on an IBM PC compatible under
Salford Fortran 77. Anyone interested may obtain the installation
.bat-files (scripts) from me (for the time being).

Kristjan Jonasson
Institute for Numerical Analysis
Technical University of Denmark


From: Philippe Toint <>
Date: Mon, 29 Mar 93 14:33:39 +0200
Subject: LSNNO is Now Available

LSNNO is now available

LSNNO (Large Scale Nonlinear Network Optimization) is a set of Fortran
subroutines, whose purpose is to solve optimization problems whose objective
function is nonlinear (possibly nonconvex) and whose constraints represent
linear flow conservation in a network.

LSNNO was written by D. Tuyttens (FUNDP, Namur, Belgium). It provides a
variety of algorithmic strategies to handle derivative information, including
several approximation schemes. A full description of the package and a
discussion of its performances can be found in

Ph. L. Toint and D. Tuyttens,
"LSNNO: a Fortran subroutine for solving large-scale nonlinear network
optimization problems",
ACM Transcations on Mathematical Software,
vol. 18(3), pp. 308-328, 1992.

LSNNO is written in standard ANSI Fortran 77. It uses double precision
arithemetic. A fully detailed User's Guide is included as a LaTeX file in the
software distribution.

The package may be obtained in one of two ways. Firstly, the reader can
obtain LSNNO electronically (and free of charge) via an anonymous ftp call to
the account (Internet number, in the
directory lsnno) at Facultes Universitaires Notre-Dame de la Paix (Namur).
We request that the userid is given as the password. This will serve to
identify those who have obtained a copy via ftp.

Secondly, the package can be obtained on a floppy disk or magnetic cartridge
at a minimal price, intended to recoup the costs of media, packaging,
preparation and courier delivery. Potentially interested parties should
contact Ph. Toint for obtaining a suitable order form.

It is also intended that LSNNO will be available via NETLIB soon.

Ph. L. Toint
D. Tuyttens


From: Lester Ingber <>
Date: Sun, 4 Apr 1993 13:28:57 -0700
Subject: Very Fast Simulated Reannealing (VFSR) on Netlib

Very Fast Simulated Reannealing (VFSR) Netlib & Statlib updates

To get on or off blind-copy VFSR e-mailings, just send
an e-mail to with your request.

VFSR is a robust algorithm for nonlinear optimization. Major changes
since it was made publicly available under the GNU Copyleft in Nov
92 include support for independent setting of scales of annealing
of parameters, extended documentation, and a new option novel to
optimization algorithms permitting VFSR to recursively optimize
its own Program Options for a given system. The unique latter
capability required a major rewriting of the code to pass only local
pointer-parameters in the vfsr module.

It should be noted that some people have generally criticized simulated
annealing algorithms as _not_ containing enough free parameters to
tune different complex systems. While VFSR has demonstrated that
it is surprisingly robust given its default parameters, it also is
capable of quite a bit of "tuning"; I put these parameters outside
the code, in a separate structure, to enable people so motivated to
be able to tune their optimizations of their systems.

Codes in Netlib and Statlib have been updated, replacing version 6.38.

NETLIB (compressed share file) version 9.3
[login as netlib, your_login_name as password]
cd opt
get vfsr.Z
mail [AT&T Bell Labs, NJ, USA]
mail [Oak Ridge Natl Lab, TN, USA]
mail [U Kent, UK]
mail [Oslo, Norway]
mail [U Wollongong, NSW, Australia]
and send the one-line message
send vfsr from opt
(It may take a week or so for the code in to propagate
to the other netlib sites.)

STATLIB (uncompressed share file) version 9.3
[login as statlib, your_login_name as password]
cd general
get vfsr
and send the one-line message
send vfsr from general
(It may take a few days for the new version to be installed here.)

After getting the file, on your local machine:
ingber% uncompress vfsr.Z [If a compressed file was obtained]
ingber% sh vfsr
ingber% cd VFSR.DIR
The directory VFSR.DIR contains the code.

If you do not have ftp access, get information on the FTPmail service
by sending the word "help" as a message to
If you receive VFSR via e-mail, then first `uudecode mailfile',
(where mailfile may be a synthesis of several files) to get vfsr.Z,
and then follow the previous directions.

If this is not convenient, and if your mailer can handle large files,
I directly can send you the code or papers you require via e-mail.
Sorry, I cannot assume the task of mailing out hardcopies of code
or papers.


|| Prof. Lester Ingber [10ATT]0-700-L-INGBER ||
|| Lester Ingber Research Fax: 0-700-4-INGBER ||
|| P.O. Box 857 Voice Mail: 1-800-VMAIL-LI ||
|| McLean, VA 22101 EMail: ||


Date: Mon, 29 Mar 93 16:17:11 AST
Subject: Call for Papers

Final Call for Participation
The 5th UNB AI Symposium
August 11-14, 1993
Sheraton Inn, Fredericton
New Brunswick

We invite researchers from the various areas of Artificial Intelligence,
Cognitive Science and Pattern Recognition, including Vision, Learning,
Knowledge Representation and Foundations, to submit articles which
assess or review the progress made so far in their respective areas, as
well as the relevance of that progress to the whole enterprise of AI.
Other papers which do not address the theme are also invited.

Four 70 minute invited talks and five panel discussions are devoted to
the chosen topic: "Are we moving ahead: Lessons from Computer Vision."
The speakers include (in alphabetical order)
* Lev Goldfarb
* Stephen Grossberg
* Robert Haralick
* Tomaso Poggio
Such a concentrated analysis of the area will be undertaken for the
first time. We feel that the "Lessons from Computer Vision" are of
relevance to the entire AI community.

Information for Authors
April 10, 1993:
Four copies of an extended abstract (maximum of 4 pages including
references) should be sent to the conference chair.

May 15, 1993:
Notification of acceptance will be mailed.

July 1, 1993:
Camera-ready copy of paper is due.

Conference Chair: Lev Goldfarb
Mailing address:
Faculty of Computer Science
University of New Brunswick
P. O. Box 4400
Fredericton, New Brunswick
Canada E3B 5A3

Phone: (506) 453-4566
FAX: (506) 453-3566

Symposium location
The symposium will be held in the Sheraton Inn, Fredericton which
overlooks the beautiful Saint John River.


From: L. Ridgway Scott <Scott@UH.EDU>
Date: Tue, 30 Mar 1993 22:27:55 -0600
Subject: Computer Design: A New Grand Challenge


A minitrack of The Architecture Track of the 27th Hawaii
International Conference on System Sciences is being organized
on the theme of High Performance Computation in the field of
Computer Design. Computer manufacturing is a major industrial
consumer of computation. The design of VLSI chips is highly
dependent on software tools that perform various modeling,
optimization and testing functions. Enhancement of these tools
to utilize techniques of High Performance Computation can en-
hance dramatically the design of new chips and computer systems.

The primary object of the Minitrack will be to bring together
people who specialize in the use of modern parallel super-
computers and computer architects who are using extensive
amounts of computing in computer design.

Three types of papers are solicited for the Minitrack. One type
describes large computations related to computer design and how
a significant increase in computational power would dramatically
enhance the way computers are designed. Also, papers devoted to
improved algorithms will be encouraged provided they are
directed to practical computer design issues. And papers on the
modification of existing algorithms to achieve reasonable speed-
ups on a large number of processors are also solicited. Papers
with a mixture of all three aspects especially encouraged.

Areas of interests include

* Device, Circuit, Logic & Register models of chip function.
* Relations between different modeling levels, mixtures of levels.
* Optimizing the placement of components and routing of wires.
* Testing a final design for correctness.
* Supercomputer computation in the design of commercial systems.

1993 Deadlines

o Eight copies of the manuscript by June 4
o Notification of accepted papers by August 31
o Camera-ready copies of accepted manuscripts due by October 1

The paper should be sent to
Ridgway Scott
Department of Mathematics
Houston, TX 77204--3476
phone: 713-743-3445 FAX: 713-743-3505 e-mail:

If you intend to submit a paper, or are willing to serve as a
referee, please respond via e-mail to Papers chosen
for the Minitrack will be considered for publication in a special
issue of the International Journal of Supercomputer Applications.


From: Dan Zwick <zwick@uvm-gen.EMBA.UVM.EDU>
Date: Thu, 1 Apr 1993 19:24:51 -0500 (EST)
Subject: Workshop on Constrained Approximation and Optimization

Workshop on
Algorithms for Constrained Approximation and Optimization

May 23--26, 1993
Stowe, Vermont

The Organizers:
D. Zwick, University of Vermont
R.A. Zalik, Auburn University

We are pleased to announce the "Workshop on Algorithms for Constrained
Approximation and Optimization" to be held at the Stowehof Inn, Stowe,
Vermont, May 23-26, 1993. This workshop will be funded by the National
Science Foundation, with partial support from the University of Vermont.
If you would like to attend this exciting workshop, please notify me by
email with the topic of your talk and your mailing address, insofar as it
differs from the one in the CML. Regretfully, space and time limitations
require that we restrict the number of speakers, so attendance is by
invitation only.
The "Workshop on Algorithms for Constrained Approximation and Optimization"
will bring together leading researchers in numerical analysis and
approximation theory to discuss new and innovative approaches to solving
difficult problems in the area of constrained approximation and optimization.
The participants, who will come from academia, private and government
research institutes, and industry, are mostly experts in either constrained
approximation or constrained optimization. While there is a nonempty
intersection in the problems and techniques familiar to these two groups,
there has not been much interaction and cross-fertilization between them.
The purpose of the proposed workshop is to provide a pleasant and productive
setting for participants to engage in an interchange of ideas, to foster
collaborative research among the participants, and to expose young researchers
and graduate students to the exciting mathematics being done at the forefront
of these fields.
The conference proceedings will be published either in the "Annals of
Numerical Mathematics," a new journal devoted to conference proceedings of
this type, or in "Numerical Algorithms," depending on the number of
contributed papers.
We strongly encourage participants who have sufficient funds to support
graduate students, postdocs, and other new researchers at their institutions
who might be interested in attending and being exposed to the exciting
mathematics being done at the forefront of these fields. Members of this
group who do not have their own research funds will be given top priority
in the use of the NSF funds.
We also wish to make a special effort to reach mathematicians traditionally
underrepresented at workshops of this nature---women, members of minorities,
and disabled persons---and have been assured that all conference facilities
and a number of the rooms are accessible to persons with disabilities. So
please help spread the word!
Dan Zwick

Dan Zwick Tel: (802) 656-4314
Math Department, Univ of Vermont (802) 658-1978 (home)
16 Colchester Ave. e-mail:
Burlington VT 05401-1455 fax: (802) 656-8802


From: Daniel B. Szyld <>
Date: Thu, 1 Apr 93 23:07:57 EST
Subject: AWM WORKSHOP before SIAM meeting

Association for Women in Mathematics (AWM) WORKSHOP

AWM will sponsor a Workshop in conjunction with the annual
SIAM meeting for women graduate students and postdocs in
applied mathematics. It will be similar to the workshops
held at major mathematics meetings over the past three
years and will be funded by NSF and ONR. The workshop will
take place on July 11, 1993, before the SIAM Meeting (July
12 to 16) at the Wyndham Franklin Plaza Hotel in
Philadelphia, PA.
Funding for travel and subsistence will be offered for up to
ten women graduate students and ten women postdocs to attend
the AWM Workshop. Those attending are encouraged to stay for
the rest of the SIAM meeting. Departments are urged to
provide some institutional support.
To be eligible for funding, graduate students must have
begun work on a thesis problem. Postdocs must have received
their Ph.D. within approximately the last five years. All
applications should include a curriculum vita and a concise
description of research; graduate students should include a
letter of recommendation from their advisor.
Nominations by other mathematicians (accompanied by the
information described above) are welcome.
The deadline for applications is April 10, 1993. They should
be sent to: Ginny Reinhart, Executive Director, AWM, 4114
Computer & Space Sciences Building, University of Maryland,
College Park, MD 20742-2461; (301) 405-7892.


From: Sanzheng Qiao <>
Date: Sun, 4 Apr 93 09:25 EDT
Subject: Southern Ontario NA Day


The Eleventh Annual Southern Ontario Numerical Analysis Day
is to be hosted at McMaster University, Hamilton, Ontario,
on Saturday, May 1, 1993. Gene Golub is our distinguished
speaker this year. There are a total of 14 presentations. The
program is availabe via anonymous ftp on
( under the directory public/seminars/na.

Hamilton is located about 60 miles from the Niagara Falls
and about 40 miles from Toronto. McMaster University is on
the west side of the city. Directions to the city and hotel
information can also be found in the directory public/seminars/na.

We are planning on having a wine and cheese party on the
Friday (April 30) evening. If you are interested in joining
us please advise.

Sanzheng Qiao
Dept. Comp. Sci. & Syst.
McMaster University
Hamilton, ON L8S 4K1
(416)525-9140 ext. 7234


From: John Coleman <>
Date: Thu, 01 Apr 93 15:47:54 BST
Subject: Lectureships at University of Durham

Department of Mathematical Sciences
Lectureships in Mathematical Sciences

Two or more temporary posts in Mathematical Sciences are available for two
years from 1 October 1993, or as soon as possible thereafter. The Department
of Mathematical Sciences has active research groups in Pure Mathematics,
Mathematical Physics, Numerical Analysis and Statistics, and is committed to
excellence in both research and teaching. One post will be in Numerical

Salaries are likely to be on the Lecturer A scale (13,400 - 18,576 pounds per

Further details may be obtained from the Director of Personnel, Old Shire Hall,
Durham, DH1 3HP (tel 091 374 3158) to whom applications (three copies, including
curriculum vitae and the names of three referees) should be sent, to arrive no
later than Tuesday 4 May 1993. Please quote reference A189.

(The personnel office will be closed for Easter for the period 8-13 April.)

The Numerical Analysis group is part of the Applied Mathematics section of the
department which obtained Grade 5 in the recent Research Assessment Exercise.

E-mail queries may be addressed to


From: Richard Brualdi <>
Date: Mon, 29 Mar 93 09:35:03 CST
Subject: Contents, Linear Algebra and its Applications

Linear Algebra and its Applications
Contents Volume 185

J. William Helton (La Jolla, California)
and Hugo J. Woerdeman (Williamsburg, Virginia)
Symmetric Hankel Operators: Minimal Norm Extensions
and Eigenstructures 1

Adam W. Bojanczyk, Ruth Onn, and
Allan O. Steinhardt (Ithaca, New York)
Existence of the Hyperbolic Singular Value Decomposition 21

Mario Taboada (Los Angeles, California)
A Theorem in Linear Algebra With Applications to the
Geometry of Quadrics 31

Jean-Claude Evard (Laramie, Wyoming) and
Frank Uhlig (Auburn, Alabama)
Comparison of the Sets of Polynomials p and q
Such That X=p(A) and Y=q(B) Are Solutions of the Matrix
Equations f(X)=A and f(Y)=B 41

Dragoslav Herceg and
Natasa Krejic (Novi Sad, Federal Republic of Yugoslavia)
On the Convergence of the Unsymmetric Successive
Overrelaxation (USSOR) Method 49

Elizabeth A. Kalinina and
Alexei Yu. Uteshev (St. Petersburg, Russia)
Determination of the Number of Roots of a Polynomial
Lying in a Given Algebraic Domain 1

Leonid Lerer (Haifa, Israel) and Leiba Rodman (Williamsburg, Virginia)
Sylvester and Lyapunov Equations and Some Interpolation
Problems for Rational Matrix Functions 83

Moody T. Chu (Raleigh, North Carolina)
The Stability Group of Symmetric Toeplitz Matrices 119

K. Zietak (Wroclaw, Poland)
Subdifferentials, Faces, and Dual Matrices 125

H. Valiaho (Helsinki, Finland)
The Boolean Pivot Operation, M-matrices, and Reducible Matrices 143

H. Miranda and Robert C. Thompson (Santa Barbara, California)
A Trace Inequality With a Subtracted Term 165

Peter Lancaster (Calgary, Alberta, Canada) and
Qiang Ye (Winnipeg, Manitoba, Canada)
Rayleigh-Ritz and Lanczos Methods for Symmetric Matrix Pencils 173

James W. Demmel (Berkeley, California) and
William Gragg (Monterey, California)
On Computing Accurate Singular Values and
Eigenvalues of Matrices With Acyclic Graphs 203

Gilbert Helmberg and Peter Wagner (Innsbruck, Austria) and
Gerhard Veltkamp (Eindhoven, the Netherlands)
On Faddeev-Leverrier's Method for the Computation of the
Characteristic Polynomial of a Matrix and of Eigenvectors 219

Wang Xinmin (Beijing, People's Republic of China)
Generalized Extrapolation Principle and Convergence
of Some Generalized Iterative Methods 235

Hongxiang Li (Shanghai, People's Republic of China) and
Gangsong Leng (Changsha, People's Republic of China)
Matrix Inequality With Weights and Its Applications 273

Author Index 279


From: Daniel Baltzer <>
Date: Wed, 31 Mar 1993 10:47:33 -0700
Subject: Contents, Numerical Algorithms

Contents NUMERICAL ALGORITHMS, Editor-in-Chief: Claude Brezinski,
Laboratoire d'Analyse Numerique et d'Optimisation, UFR IEEA - M3,
Universite de Lille 1, France, fax: +33 - 20 43 49 95, e-mail:

Numerical Algorithms, Volume 4 (1993), issues 1,2:

Computing the real roots of a polynomial by the exclusion algorithm, J.-P.
Dedieu and J.-C. Yakoubsohn

On the computation of a versal family of matrices, L. Stolovitch

Model reduction techniques for sampled-data systems, F.D. Barb and M.Weiss

Automatic solution of regular and singular vector Sturm-Liouville problems,

An efficient Total Least Squares algorithm based on a rank-revealing
two-sided orthogonal decomposition, S. van Huffel and H. Zha.

On the design of an upwind scheme for compressible flow on general
triangulations, Th. Sonar.

Fast orthogonal decomposition of rank deficient Toeplitz matrices, P.C. Hansen.

A homotopy algorithm for a symmetric generalized eigenproblem, K.Li and
T.-Y. Li.

A note on the implementation of the successive overrelaxation for linear
complementarity problems, W. Niethammer.

Requests for free sample copies and orders are to be sent to: J.C. Baltzer
AG, fax: +41-61-692 42 62,


From: Carlos Moura <>
Date: Sat, 3 Apr 93 14:23:02 EST
Subject: Contents, Computational and Applied Mathematics

(Matematica Aplicada e Computacional)
Vol. 11, Issue 2, 1992

* Special Issue on Conservation Laws *

Aparecido J. de SOUZA
Stability of Singular Fundamental Solutions under Perturbation
for Flow in Porous Media

Jesus C. da MOTA
The Riemann Problem for a Simple Thermal Model
for 2-Phase Flow in Porous Media

Multiphase Flow Models with Singular Riemann Problems

Hyperbolic Systems of Conservation Laws with
Rotationally Equivariant Flux Function

The Multi-Fractal Hypothesis and Anomalous Diffusion


End of NA Digest