NA Digest Sunday, March 15, 1992 Volume 92 : Issue 11

Today's Editor:

Cleve Moler
The MathWorks, Inc.

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to


From: Cleve Moler <>
Date: Sun Mar 15 06:21:03 PST 1992
Subject: Change of Address for Cleve Moler

For the past two and a half years, I've been commuting between my
home in California and the MathWorks main office outside Boston.
This week, I cut 2700 miles off that commute by moving to Massachusetts.
So, my business mailing address is now:

Cleve Moler
The MathWorks, Inc.
24 Prime Park Way
Natick, MA 01760

The telephone number for the MathWorks switchboard is: 508-653-1415.
(That's 508-65-pi). You can reach my office phone directly by
dialing 508-653-2452 and then keying in extension 325.

My e-mail address remains

Our family's new home is in Sherborn, which is just south of Natick and
Framingham and a few miles west of Wellesey. The mailing address is:
62 Russett Hill Road, Sherborn, MA 01770.

If you're in the Boston area, and would like to visit the MathWorks,
let me know -- we'd be happy to see you.

-- Cleve


From: Siamak Hassanzadeh <>
Date: Mon, 9 Mar 92 10:04:04-1795
Subject: New Address for Siamak Hassanzadeh

Effective immediately my address is

Siamak Hassanzadeh
Fujitsu America, Inc.
3055 Orchard Dr.
San Jose, CA 95134-2022

Phone: 408-456-7308

I will be working in the Computational Research Division of Fujitsu.
I look forward to continuing interaction with computational community.


From: Carlos de Moura <>
Date: Thu, 12 Mar 92 16:49:17 GMT
Subject: Address Change for C. A. de Moura

For March to August 1992 I will be working with the

Group de Calcul Parallele
Laboratoire de Physique Corpusculaire
College de France
11, Place M. Berthelot
75231 Paris 05 Cedex


Best Regards
Carlos A de Moura


From: Gene Golub <>
Date: Thu, 12 Mar 92 10:59:08 GMT
Subject: Bracewell on the Hartley Transform

The Hartley Transform
R.N. Bracewell

(Stanford EE370 Seminar for March 12, 1992)

Spectral analysis is usually done with the aid of the Fourier transform, which
is based on ideas inherited from the time of Fourier and makes full use of the
mathematical theory of complex analysis which, to a large extent, developed in
the last century in response to the need to resolve tricky problems of
integration and convergence thrown up by exploration of Fourier's theorem. As
students, we become familiar with the algebraic manipulation of complex
quantities and equations such as V=ZI become second nature to us; it is true
that when we come to evaluate ZI four multiplies are involved. In the days of
slide rules this factor four was very noticeable and although it is less
noticeable today, still, somewhere deep inside your calculator, four
multiplies are performed whenever two complex numbers are multiplied together.

When numerical Fourier analysis is performed complex operations are required
because the Fourier spectrum of a real waveform is necessarily complex.
Conversely, when we invert the Fourier transform we must have an algorithm
that accepts complex input. Thus, in the first case (real data) a lot of
unnecessary operations are carried out on an imaginary part that is
zero-valued, while in the second case (complex input) a lot of output values
are computed which are zero (or should be, to the limits of machine
precision). These elementary observations tell us that there should be a way
of performing spectral analysis of real data that avoids the wastefulness of a
transform that must be prepared for complex input, whereas most of the time
either our data are real, or our output is real.

This transform is the Hartley transform; it represents a waveform with
N sample values by N transform values which are real (see "Assessing the
Hartley Transform," IEEE Trans ASSP, vol. 38, 2174-2176, Dec 1990 and
references provided). Of course the Hartley transform graph is not the same
as the Fourier transform graph, which means that some of our intuition is
lost. On the other hand when we graph Fourier transforms we do not always
plot both the real and imaginary parts, which are not particularly clear to
grasp and which in fact change drastically for the same waveform if one only
changes the choice of origin of time. Usually we graph the power spectrum,
which is real and does not call on us to visualize the full complex transform.
The Hartley power spectrum is the same as the Fourier power spectrum;
likewise the Hartley phase is the same as the Fourier phase (with a 45 degree
shift). Also it is easy to learn how to see the real and imaginary parts of
the Fourier transform, given the Hartley transform, if you really want them.
In summary, it seems that we will continue to retain the advantage that we get
from fluency in complex algebra by using Fourier transforms in theoretical
work but that when we talk to computers which prefer real numbers we will
shift to the Hartley transform. By computing with real numbers we gain a
factor of two in speed in the inner loops of the algorithms. This can be very
important in new programming; however, if you have canned code that runs it is
usually not advised to tamper with it.

Interesting questions have been raised about the physical significance of the
Hartley transform relative to the Fourier transform. After all, Lord Kelvin
told us that "Fourier's theorem is not only one of the most beautiful results
of modern analysis, but it may be said to furnish an indispensable instrument
in the treatment of nearly every recondite question in modern physics." If
you have the opportunity to think about the Hartley transform in advance you
might like to consider whether the Fourier transform is more fundamental


From: John Strikwerda <>
Date: Fri, 13 Mar 92 13:53:31 -0600
Subject: Errata for Strikwerda's Textbook

I have made an errata list for my textbook available
by anonymous ftp. The text is:
Finite Difference Schemes and
Partial Differential Equations
published by Wadsworth & Brooks/Cole.

The errata are on the file
on the computer

John Strikwerda


From: Tina Flores <>
Date: Tue, 10 Mar 92 16:55:51 EST
Subject: Conference on Applications of Dynamical Systems

SIAM Conference on Applications of Dynamical
Systems, October 15-19, 1992, Salt Lake City, UT

The organizers for the conference are pleased to inform you
that the DEADLINE for submitting contributed abstracts has
been EXTENDED to MARCH 20, 1992. For those of you who
have not yet submitted your 100-word abstract, send it
NOW -- by e-mail to:
by fax to: 215-386-7999
or call the SIAM office at 215-382-9800 if you have any
questions. SIAM encourages electronic submission of
abstracts. To help in formatting your submission, plain TeX
or LaTeX macros are available upon request.


From: Tina Flores <>
Date: Thu, 12 Mar 92 09:52:09 EST
Subject: Short Course on Numerical Optimization and Software


Date: Sunday, May 10, 1992
Location: Hyatt Regency Hotel, Chicago, Illinois
Lecturers and organizers: Jorge J. More'
Stephen J. Wright

Both lecturers are with the Mathematics and Computer Science
Division, Argonne National Laboratory, Argonne, Illinois

The course will cover four main problem areas. These are
nonlinear equations and nonlinear least squares,
unconstrained optimization, constrained optimization, and
global optimization.

Registration Fees: SIAM Non-
Member Member Student

Advance $120 $135 $55
On-Site 135 155 75

Preprints, coffee and lunch are included in the registration
fees. Attendees are advised to preregister for the short
course. On-site registration cannot be guaranteed. Preprints
of the lecture materials will be distributed upon check-in
at the SIAM registration desk.

The short course will precede the Fourth SIAM Conference on
Optimization which will be held on Monday through
Wednesday, May 11-13, 1992, Hyatt Regency Hotel, Chicago,

Deadline for Advance Registration: May 4, 1992

Register NOW! By phone: 215-382-9800. By FAX: 215-386-7999.
by E-mail:


From: Joel Saltz <>
Date: Wed, 11 Mar 92 17:16:50 -0500
Subject: Scalable High Performance Computing Conference


Sponsored by
IEEE Computer Society

April 27-29, 1992

Williamsburg Hilton and
National Conference Center
Williamsburg, Virginia


SUNDAY, April 26, 1992


Computational Fluid Dynamics on Parallel Machines - Algorithms and
Applications, Ramesh Agarwal, McDonnell Douglas Research Laboratories

Using Massively Parallel Supercomputers: An Applications Perspective
Horst Simon, Computer Sciences Corporation, NASA Ames Research Center

Compilers for Scalable Architectures, Ken Kennedy, Rice University
and Joel Saltz, ICASE

SUNDAY, 6:00-9:30, Registration and Reception


Scalability of Data Transport, Harry Jordan, Univ. of Colorado

Parallel sessions on: Applications and Performance

Parallel sessions on: Applications and Languages

Interleaved Mass Storage: Parallel Processing in Secondary and Tertiary
Memory, Randy Katz, UC-Berkely

Parallel sessions on: Algorithms and Systems Support for Languages

Parallel sessions on: Molecular Dynamics and Tools

Evening: Technical Vendor Presentations


Parallel Methods and Applications for Macromolecular
Simulations, Bernard Brooks, National Institutes of Health

Parallel sessions on: Molecular Dynamics and Support for Irregular Problems

Parallel sessions on: Computational Fluid Dynamics and Systems Issues

Computational Aerodynamics with Unstructured Meshes, Dimitri Mavriplis, ICASE

Parallel sessions on: Communication and Tools

Parallel sessions on: Computational Fluid Dynamics and Compilers

Evening: Panel on Educational Issues in Parallel Computing


Performance Animation, Dennis Gannon, Indiana Univ.

Parallel sessions on: Applications and Languages

Parallel sessions on: Applications and Performance

For a complete elecronic or hard copy of the Advance Program
including a hotel reservation card, contact


From: Willard Miller <>
Date: Fri, 13 Mar 92 11:11:14 CST
Subject: IMA Workshop on Linear Algebra for Signal Processing

IMA Workshop on


April 6 -- April 10, 1992
Organizers: A. Bojanczyk and G. Cybenko

Signal processing is making increasingly sophisticated use of linear algebra
on both theoretical and algorithmic fronts. The purpose of this workshop is
to bring signal processing engineers, computer engineers, and applied linear
algebraists together for an exchange of problems, theories and techniques.
Particular emphasis will be given to exposing broader contexts of the signal
processing problems so that the impact of algorithms and hardware will be
better understood.

The workshop will explore five areas by having a sequence of talks devoted to
the underlying signal processing problem, the algorithmic and analytic
techniques and, finally, implementation issues for each area. The five areas
1) updating SVD and eigendecompositions;
2) adaptive filtering;
3) structured matrix problems;
4) wavelets and multirate signal processing;
5) linear algebra architectures (parallel/vector and other high
performance machines/designs).

Most of the workshop talks will be held in Conference Hall 3-180 on the entry
floor of the Electrical Engineering/Computer Science Building. This building
is located on the corner of Washington Avenue and Union Street, a block from
the IMA Main Office. The conference hall is on the Ethernet and has a
projection system for display of computer output.


Monday, April 6

Gene Golub, Stanford/IMA
The canonical correlations of matrix pairs and their numerical computation

Franklin T. Luk, Rensselaer Polytechnic University
Adaptive parameter estimation in signal processing

J. Cadzow, Vanderbilt University

Marc Moonen, Katholieke U. Leuven
Systolic algorithms for adaptive signal processing

Wenyuan Xu, University of Minnesota
New ideas in the design of signal-subspace detectors and estimators

Robert J. Plemmons, Wake Forest University/IMA
Preconditioned iterative least squares FIR system identification

Tuesday, April 7

Gilbert Strang, MIT
Wavelet transforms versus Fourier transforms

A. H. Tewfik, University of Minnesota
Wavelets in signal and image processing

Martin Vetterli, Columbia University
Wavelets, filter banks, and applications in compression

Jo Ward, Murdoch University
Stability of DARMA filters on spaces

M. Stewart, University of Illinois
A general linear algebraic framework for perfect reconstruction filters
Lothar Reichel, Kent State University

Wednesday, April 8

Simon Haykin, McMaster University
Fast implementation of the RLS algorithm

A. Steinhardt, Cornell University
Adaptive detection using sensor arrays

John Proakis, Northeastern University
Blind equalization

Thursday, April 9

Pierre Comon, Thomson Sintra, France
Looking for fast algorithms solving structured linear systems

Tom Kailath Stanford University

Israel Koltracht, University of Connecticut
Structured condition numbers

Georg Heinig, Universitaet Leipzig/IMA
Fast algorithms for generalized Cauchy matrices and rational interpolation

Gregory Ammar, Northern Illinois University
Updating and downdating Szego polynomials

Friday, April 10

Robert Schreiber, NASA Ames Research Center/IMA
Matrix computation on SIMD data-parallel architectures

E. Deprettere


University of Minnesota
514 Vincent Hall
206 Church Street S.E.
Minneapolis, Minnesota 55455
FAX (612) 626-7370 telephone (612) 624-6066


From: Joseph Oliger <>
Date: Fri, 13 Mar 92 14:34:51 PST
Subject: Visiting Positions at RIACS


NASA Ames Research Center
Moffett Field, CA

The Research Institute for Advanced Computer Science (RIACS) at the
NASA Ames Research Center, located in the San Francisco Bay Area
adjacent to Silicon Valley, is inviting applications for visiting
research positions for graduate students for the summer of '92 and for
post-doctoral appointments of up to two years begining in the Fall of '92.

RIACS carries out a basic research program in the computational
sciences to support the needs of NASA Ames scientific missions.
Specific areas of interest are: algorithms and software for parallel
scientific computation with applications to computational fluid
dynamics, adaptive and composite mesh methods for solving partial
differential equations, the design and implementation of compilers and
tools for parallel computers, and the analysis of high performance

The computing environment at NASA Ames Research Center includes a
Connection Machine (CM-2), an Intel iPSC/860, a Cray Y-MP and a
Cray-2. High performance graphics workstations are also available.

Visitors to RIACS are expected to collaborate with NASA Ames scientists.
Additional opportunities for collaboration abound with the many local
research universities and institutions.

Applicants should send resumes and descriptions of research interests
with references to:

Joseph Oliger, Director
NASA Ames Research Center
Mail Stop T041-5
Moffett Field, CA 94035-1000

Applications and inquiries may also be made via e-mail to:

RIACS is an Equal Opportunity Employer.



From: Richard A. Brualdi <>
Date: Tue, 10 Mar 92 16:37:35 CST
Subject: Contents: Linear Algebra and its Applications

Contents of LAA Volume 167, April 1, 1992

Preface 1

Daniel Hershkowitz (Haifa, Israel)
The Height Characteristic of Block Triangular Matrices 3

Naomi Shaked-Monderer and Abraham Berman (Haifa, Israel)
More on Extremal Positive Semidefinite Doubly Stochastic Matrices 17

Pal Rozsa (Budapest, Hungary) and Francesco Romani (Pisa, Italy)
On Periodic Block-Tridiagonal Matrices 35

S. W. Hadley (Waterloo, Canada), F. Rendl (Graz, Austria), and H. Wolkowicz
(Waterloo, Canada)
Symmetrization of Nonsymmetric Quadratic Assignment Problems and the
Hoffman-Wielandt Inequality 53

Bit-Shun Tam (Tamsui, Taiwan)
On the Structure of the Cone of Positive Operators 65

Adi Ben-Israel (New Brunswick, New Jersey)
A Volume Associated With mxn Matrices 87

Mark Krupnik (Haifa, Israel)
Changing the Spectrum of an Operator by Perturbation 113

Michael Gorodetsky (Haifa, Israel)
Inversion of Quasitriangular Block Toeplitz Matrices 119

J. C. Willems (Groningen, The Netherlands) and P. A. Fuhrmann (Beer Sheva,
Stability Theory for High Order Equations 131

Moshe Roitman and Zalman Rubinstein (Haifa, Israel)
On Linear Recursions With Nonnegative Coefficients 151

Yoav Yaacoby and Peter R. Cappello (Santa Barbara, California)
Decoupling the Dimensions of a System of Affine Recurrence Equations 157

Avram Sidi (Haifa, Israel)
Development of Iterative Techniques and Extrapolation Methods for Drazin
Inverse Solution of Consistent or Inconsistent Singular Linear Systems 171

Abraham Berman, Moshe Goldberg, and Daniel Hershkowitz (Haifa, Israel)
REPORT: Haifa 1990 Conference on Matrix Theory 205

Author Index 273


End of NA Digest