NA Digest Saturday, February 23, 1991 Volume 91 : Issue 8

Today's Editor: Cleve Moler

Today's Topics:


From: Jack Dongarra <>
Date: Wed, 20 Feb 91 08:10:58 -0500
Subject: NA-Net's Expansion

I happened to be looking at the NA-Net database and came up with
the following data on NA-Net growth since it moved to ORNL.

Individuals reachable thru
Dec 90 1592
Jan 91 1679
Feb 91 1851



From: Gene Golub <golub@Cholesky.Stanford.EDU>
Date: Fri, 22 Feb 91 16:26:46 PST
Subject: Congratulations to Joe Traub

Congratulations to Joe Traub for being awarded the Emanuel R.Piore
Award of the IEEE "for pioneering research on algorithms and
computational complexity , parallelism and optimal iteration theory,
and for leadership in computing education."

Gene Golub


Date: Mon, 18 Feb 91 08:27 PDT
Subject: New Book by David Watkins

My book ``Fundamentals of Matrix Computations'' is finally in print.
This book is intended for an upper division or (relatively easy)
graduate course. It attempts less than Golub and Van Loan but spends
more effort on motivation and explanation. If your students have
trouble reading Golub and Van Loan, they might like my book. The
publisher is John Wiley and Sons. The table of contents is listed
David Watkins
Washington State University

1. Gaussian Elimination and its Variants
2. Sensitivity of Linear Systems; Effects of Roundoff Errors
3. Orthogonal Matrices and the Least Squares Problem
4. Eigenvalues and Eigenvectors I
5. Eigenvalues and Eigenvectors II
6. Other Methods for the Symmetric Eigenvalue Problem
7. The Singular Value Decomposition (SVD)
A. Fast Rotators
B. Software for Matrix Computations


From: Dan Peterka <>
Date: 20 Feb 91 22:31:44 GMT
Subject: Adaptive Mesh Refinement Techniques

I need a crash course in Adaptive Mesh Refinement Techniques and would
like some pointers to the definitive works dealing with this subject.
I'd also like to know where I might beg, borrow or steal a Fortran
implementation of AMR to use as a starting point in bootstrapping
myself. I'm particularly interested in methods dealing with
rectangular meshes, but triangular meshes would be OK too.

Thanks for any help...

Dan Peterka S-CUBED 3398 Carmel Mtn Rd. (619) 587-8338 San Diego, CA 92121


From: Chuck Ferrara < adiron!>
Date: 21 Feb 91 14:17:39 GMT
Subject: Solution of Block Toeplitz Linear System

I need to implement, in real time, a solution for a real, symmetric, block
toeplitz, linear system. This is used to design optimal matched filters
for 2-D signal processing. I have scene solutions and source code for
op. count estimates or references would be very helpful.
Thanks in advance

Chuck Ferrara uunet!adiron!chuck
PAR Government Systems Corp.
220 Seneca Tpk
New Hartford, NY 13413
(315) 738-0600 X247


From: Roger Alexander <>
Date: 21 Feb 91 18:35:58 GMT
Subject: Seeking Recent Version of PITCON


The PITCON code of Rheinboldt and Burkardt, software for
numerical analysis of parametrized nonlinear equations, exists in
several versions. I have the 1981 TOMS version distributed
by IMSL, and am looking for a more recent one.

In Rheinboldt's 1983 lectures at University of Arkansas,
published by Wiley, an Appendix presents a completely rewritten
PITCON. Is this (or something more recent?) available anywhere
in machine-readable form?

Many thanks for any help.

Roger Alexander PING:
Department of Mathematics DING: (515) 294-7579
400 Carver Hall
Iowa State University
Ames, Iowa 50011 USA


From: Tim Monks <>
Date: Fri, 15 Feb 91 17:02:41 EST
Subject: Sparse Least Square Solvers


Are there any routines available that will solve least square problems
for rectangular sparse matrices ?

I am trying to solve a regularized least squares problem for Au = b,
where A is non-square, large (typically 16384 x 4096), and sparse.

I am thinking that methods based on QR or SVD factorizations would be
appropriate, as A is not well enough conditioned to solve the normal
equations with LU decomposition. The only routine I have found that
will do this is NAg (F04QAF) which is based on the TOMS Alg. 583 LSQR
by Paige & Saunders. This solves either the least square or a damped
least squares problem (ridge regression) using a Lanczos algorithm. It
uses reverse communication to calculate requisite matrix-vector
products. I haven't got LSQR out of netlib (yet) so I can't say how it
will work.

I have looked elsewhere for relevant routines. Generally I've found that
the routines exist, but require full storage and are generally for dense
matrices. The sparse libraries are usually restricted to square,
generally symmetric matrices.

So - are there any sparse routines available to solve least squares
problems ?

If not, what methods would you recommend to solve this type of problem ?
I have read Censor's row-action paper, but would prefer a column-
generation method as this suits my problem much better. I have
tried Gauss-Seidel & SOR on both the unregularised, and a regularized
objective, but have not had much success. Any pointers would be

Here is a summary of the libraries I have heard about, or looked at -
none appear to be satisfactory for the job. Those marked (?) are those
I am not sure about.


Linpack - (DQRDC, DQRSL) QR factor & LS solve.
IMSL - (LQRRR & LSQRR) same routine
Classpack - (KDQRDC & KDQRSL) same routine
Vanhuffel - Total least squares solution to Ax=b, A any size.


Harwell MA28 - LU factorisation only
ITPACK 2C - Symmetric PD. CG.
ITPACK 3A - CG and Orthomin variants. Overdetermined ?
MADPACK - A nxn. Multigrid Gaussian elimination, Gauss-Seidel, Orthomin.
NSWC library - (SPSLV) LU factorisation only.
PORT - LU factorisation + algebra ?
SGE - Gaussian elimination
SLAP - A nxn. CG
SPARSEPAK - Direct methods ?
Sparse1.3 - LU factorisation only
Yale - Direct methods ??
pcgpack - ?
y12m - ?

Dr. Tim Monks
Image Processing & Data Analysis Group | (direct) (+61-3)566-7448
BHP Research - Melbourne Laboratories | (switch) (+61-3)560-7066
245 Wellington Rd, Mulgrave, 3170, | (fax) (+61-3)561-6709


From: S. I. Hariharan <>
Date: Wed, 20 Feb 91 19:04:30 EST
Subject: Department Head Position at University of Akron


The Department of Mathematical Sciences invites applications and
nominations for the position of department head. A PhD in the mathematical
sciences (mathematics, applied mathematics, statistics, or computer science)
and a strong commitment to teaching and research are required. Some
administrative and/or professional experience in a mathematical sciences
PhD program is desirable. The Department consists of thirty-seven full-time
faculty members and offers BS degrees in mathematics, applied mathematics,
statistics, and computer science, and MS degrees in mathematics,
applied mathematics, and statistics. The department has proposed graduate
programs for the MS in Computer Science, and the PhD in Applied Mathematics.
The University of Akron is the third largest state university in
Ohio (30,000 day and evening students) and offers a multitude of associate,
bachelors, masters, and doctorate degree programs in the physical and
social sciences, engineering, and education.
Review of applications will begin April 15, 1991 and continue until
the position is filled. Tentative inquiries are desirable and will be treated
confidentially. Please send a curriculum vita and names of at least three
references to: Dr. Chand Midha, Chair, Search Committee, Department of
Mathematical Sciences, The University of Akron, Akron, OH 44325-4002
Phone: (216)-972-7128/7400).
The University of Akron is an equal opportunity/affirmative action employer.
Women and minorities are encouraged to apply.


From: Danny Hershkowitz <MAR23AA%TECHNION@TAUNIVM.TAU.AC.IL>
Date: Wed, 20 Feb 91 22:00:02 IST
Subject: International Linear Algebra Society

E-mail Address: MAR23AA @ TECHNION (bitnet)
Edited by Danny Hershkowitz
20 February 1991

ILAS-NET Message No. 104

FROM: Charlie Johnson


We are pleased to announce that we have again been selected by the National
Science Foundation (NSF) as one of a few nation-wide sites to offer funded
research experiences for undergraduates (REU) in the summer of 1991. The
theme of our summer research program is Matrix Analysis and Applications.
The predominant objective is to provide talented students with experience
at how mathematics is done in contrast to the deducive fashion in which it
is normally taught.

We are soliciting applications from undergraduate students who will not
graduate prior to June 1991 both within our geographical region and across
the nation. At least eight students will be selected on a competitive basis.
Each student will receive approximately $2400 ($2000 stipend plus $400
toward subsistence) and free housing in one of the College's residence

The NSF-REU program will be conducted from June 17 to August 9. Each
student will select a topic in matrix analysis and a primary advisor with
whom to work. The first week of the program will contain a series of
background lectures to establish a common language for the entire program
and to advance students to an appreciation of certain "open" (yet accessible)
problem areas. A computer laboratory consisting of tem 80386 class
personal computers and computer software such as MATLAB and MATHEMATICA
will be available to aid in mathematical exploration. In addition students
will gain a sense of the environments in which pure and applied mathematics
is carried out. Specifically, visits are planned to NASA/Langley Research
Center, ICASE, VIMS (Virgina Institute of Marine Science", CEBAF (Continuous
Electronic Beam Accelerator Facility), and NIST (formerly the National
Bureau of Standards).

To apply: Send a letter containing the following (a) a statement of your
interest in mathematics and (b) a description of the mathematics courses
(e.g., textbook titles) you have completed as well as the grades received.
Also, please arrange to have two letters of recommendation sent from
mathematics instructors and/or any other person(s) who are best qualified
to comment on your mathematical ability and experience. Send all application
material to Professor George T. Rublein at:

Department of Mathematics
The College of William & Mary
Williamsburg, VA 23187-8795

Please send all materials by March 16, 1991. If you have any questions or
need more information, please contact Professor Rublein by mail, telephone
(804)221-2028, or bitnet GTRUBL@WMVM1. You may also contact Charles
Johnson for questions.


From: Bob Grossman <>
Date: Thu, 21 Feb 91 17:43:55 CST
Subject: Workshop on Analytic, Intelligent and Database Computing

The NASA Ames - University of Illinois at Chicago
Symposia in Computation and Control 91

New Directions in Computing:
Analytic, Intelligent and Database Computing

University of Illinois at Chicago
April 7 and 8, 1991

Traditionally, computers have been used for numerical
simulations of physical processes, while the modeling
and analysis were done off line. Today computers are
used not only for numerical studies, but also to derive
better models, more accurate algorithms, and to automate
many of the computations that were formerly done by hand.
This conference will explore three areas: analytic
computing, or the application of symbolic and numeric
algorithms to nonlinear systems; intelligent computing,
or combining symbolic and numeric techniques with methods
incorporating knowledege of science and mathematics; and
database computing, or using databases as an integral
part of the computing environment. The speakers are:

Hal Abelson, MIT
"Intelligence in Scientific Computing"

Robert Grossman, University of Illinois at Chicago
"Analytic Computing"

John Guckenheimer, Cornell University
"Computing and Dynamical Systems"

Leo Joskowicz, IBM T. J. Watson Laboratory and
Elisha Sacks, Princeton University
"Automating Scientific and Engineering Reasoning"

Jack A. Orenstein, Object Design Inc.
"Object Oriented Databases"

Arie Shoshani, Lawrence Berkeley Laboratory
"Scientific Databases"

Richard Zippel, Cornell University
"Integrating Numeric and Symbolic Computing"

The workshop is sponsored by: NASA Ames; The Software Technologies
Program at the University of Illinois at Chicago; and the
Laboratory for Advanced Computing at the University of Illinois at Chicago.

To register, contact: Dr. Winifred Lyday, Software Technologies
Research Center, University of Illinois at Chicago, Mail Code 454,
Suite 226, 2201 West Campbell Park Drive, Chicago, IL 60612,
phone: (312) 996-5203, fax: (312) 413-3029,


End of NA Digest