NA Digest Sunday, May 28, 1989 Volume 89 : Issue 21

Today's Editor: Cleve Moler

Today's Topics:


From: Iain Duff <>
Date: Fri, 19 May 89 10:20:41 CDT
Subject: Hotels for the ODE meeting in London

Ordinary Differential Equation Conference.. London ... July 1989.

Several people have asked for information on hotels in the
vicinity of Imperial College. Limited accommodation is available
for spouses in the college although no children under the age of 16
will be accommodated.
The conference office referenced below is that of Imperial College
and the telephone number for further information, bookings, etc. is
at the end of the message.


The hotels listed below, all within walking distance of Imperial
College, are bookable through the Conference Office. The rates
indicated are rates offered to the Conference Office and should
you book direct with the hotel you will probably be charged the
normal tariff rate, which will be higher.

All rates were current on 1st April and so might possibly be slightly
higher. Although VAT is always included in the price, for your
information it still stands at 15%.

All hotels have bars and restaurants.

All prices are in pounds.


This is a purpose built hotel situated on the Cromwell Road - all rates
are inclusive of Service, VAT and Continental Breakfast.

Single room with bath - 34.00
Double room with bath - 50.00


Situated on Hogart Road, this hotel is a Victorian terrace conversion -
all rates are inclusive of Service, VAT and Continental Breakfast.

Single room with bath - 36.00
Double room with bath - 49.00


I have not visited so I do not know what it is like - all prices are
inclusive of Service, VAT and Continental Breakfast.

Single room with bath - 40.00
Double room with bath - 53.00


Situated on the Cromwell Road, this hotel is a Victorian terrace
conversion - all prices are inclusive of Service, VAT and Continental

Single room with bath - 44.00
Double room with bath - 60.00


Situated on Sumner Place, this hotel is a Victorian terrace conversion -
all prices are inclusive of Service, VAT and English Breakfast.

Single room with shower - 45.00
Double room with bath - 60.00


This hotel is situated on the Cromwell Road and is closest to Imperial
College - all prices are inclusive of Service, VAT and English

Single room with bath - 47.00
Double room with bath - 54.00


Situated on Cromwell Road, this hotel is a Victorian terrace conversion
- all prices are inclusive of Service, VAT and Continental Breakfast.

Single room with bath - 57.00
Double room with bath - 76.00


This hotel situated on Queen`s Gate is a Victorian terrace conversion -
all prices are inclusive of Service and VAT.

Single room with bath - 64.75
Double room with bath - 84.15


Situated on Sumner Place, this hotel is a Victorian terrace conversion -
all rates are inclusive of Service, VAT and English Breakfast.

Single room with bath - 65.00
Double room with bath - 80 00


Situated on Queen`s Gate, this hotel is a Victorian terrace conversion -
all rates are inclusive of Service and VAT.

Single room with bath - 71.10
Double room with bath - 84.60

To make a reservation please contact Miss Di Pullar-MacMillan on
+44-1-589-5111 extension 3182.


Date: 21 May 1989 16:49:53 EST
Subject: Seeking Block Toeplitz Solver

I am looking for the BLOCK-Toeplitz solver Tx=b, where
T - BLOCK Toeplitz; with many b-vectors.
I would prefere FORTRAN, MATLAB, or C implementation.
I am aware about the TOEPLITZ package available from NETLIB
(routine TGSLC), but it doesn't factor T !

Also, but this is rather secondary, I would like to find the inverse
of T (of course NOT to solve the linear equations system but for some
other reason --- having T^{-1} allows to find some physical quantities,
which, in effect, correspond to the solution of infinitely many R.H.S).
In fact my T is complex, symmetric (but non-Hermitian),
and has block-block-Toeplitz structure;
but you may ignore this --- I will be happy with the general, real,
block-Toeplitz solver.

It seems there is a lot of theoretical work
on this subject (see e.g. journal Lin. Alg.and Appl.) related to many
diverse fields (multivariate analysis, image processing, etc.)
but very little actual implementations ?!

P. Flatau


From: George Byrne <GDBYRNE%ERENJ.BITNET@Forsythe.Stanford.EDU>
Date: Mon, 22 May 89 10:21:39 EDT
Subject: Help With NAMELIST Under Microsoft Fortran

I am interested in locating a public domain Fortran routine, which will
allow the use of NAMELIST under a Microsoft Fortran Compiler, version
4.01. (The request is for someone else.) Yes, I know that Lahey F77L
supports NAMELIST.

George Byrne


From: Art Werschulz <>
Date: Mon, 22 May 89 14:53:48 EDT
Subject: Stability of Negative Norm Projectors


I need pointers to proofs of the following (if they're known). Please
give a counterexample if they're false.

They are described in (plain) TeX. In what follows, $s\ge 0$,
$p\in(1,\infty]$, and $n$ is a nonnegative integer. We use
more-or-less standard notation for Sobolev spaces, norms, inner
products, and the like.

Let $P_{-s,n}$ denote the orthogonal projector of the Sobolev space
$H^{-s}(\Omega)$ onto a finite element subspace of dimension~$n$ and
degree~$k$ (i.e., the subspace consists of piecewise-polynomials of
degree at most $k$). Show that there is a constant $C$, independent
of $n$, such that if $v\in W^{-s,p}(\Omega)$, then
$$\|P_{-s,n} v\|_{W^{-s,p}(\Omega)} \le C \|v\|_{W^{-s,p}(\Omega)} .$$

Show that
$$\|v\|_{W^{-s,p}(\Omega)} =
\sup_{w\in L_{p'}(\Omega)}
{|\langle v,w \rangle_{H^{-s/2}(\Omega)|} \over \|w\|_{L_{p'}(\Omega)}}


Art Werschulz

USEnet: ...!columbia!!agw

ATTnet: Columbia University (212) 854-8642 854-2736
Fordham University (212) 841-5323 841-5396


From: David Womble <>
Date: 24 May 89 08:08:00 MST
Subject: Fellowship at Sandia National Laboratories


The Applied Mathematical Sciences Research Program at the
Department of Energy is sponsoring a research fellowship in
applied mathematics at Sandia National Laboratories. The
fellowship will provide an outstanding research opportunity for
young scientists who are actively pursuing leading-edge research
in computational mathematics or numerical analysis. The
successful candidate will have recently earned a doctorate in
mathematics and will have made strong contributions to numerical
computation or computer science.

The fellowship appointment will be made in the Mathematics
and Computational Science Department at Sandia, which has a
strong research staff and provides an exceptional environment
for the research and application of advanced parallel computing
methods. The department maintains strong programs in analytical
and computational mathematics, physics and engineering, as well
as in advanced computational approaches for parallel computers,
computer graphics, and computer architectures and languages. We
also provide a unique parallel computing environment, including
a 1024-processor NCUBE/ten, a Connection Machine and several
large Cray supercomputers.

Applications from qualified candidates should include a
resume, a statement of research goals, and the names of three
references. The fellowship appointment includes a highly
competitive salary, moving expenses, and a generous professional
travel allowance.

If you are interested in this unique opportunity, please
send your application to

Robert H. Banks, Division 3531-86A
Sandia National Laboratories
P. O. Box 5800
New Mexico 87185

The closing date for applications is July 31, 1989, and a
successful candidate will be announced by September 15, 1989.

Further inquiries can be made by calling R. C. Allen at
(505) 844-2248.

Equal Opportunity Employer M/F/V/H
U.S. Citizenship is Required


From: David Whitley <>
Date: Thu, 25 May 89 16:14:49 CDT
Subject: Parallelizing the Simplex Method

Bob Bixby and I are interested in finding out what work has been done on
parallelizing the simplex method for linear programming. There is no work
listed in the Science Citation Index from 1984 through Jan-Feb 1989 whose
title contains both the word 'simplex' and any form of the word 'parallel';
this strongly suggests that no work on this subject has been published. If
anyone has any information concerning this subject, we would appreciate it
if you would let us know by writing to either or
Thank you.

David Whitley
Rice University


From: Bill Venables <munnari!!wvenable@uunet.UU.NET>
Date: Tue, 23 May 89 13:18:30 CST
Subject: Eigenvalues of Diagonally Perturbed Matrices

Suppose S is a positive definite symmetric matrix and S(eps) is a scale
perturbation of it, that is it has the form

S(eps) = (I + eps)S(I + eps)

where eps is a *diagonal* matrix of small entries. I am interested in
expansions of the eigenvalues and eigenvectors of S(eps) about those of S near
eps=diag(0). Does anyone know of work done in this area?

Thanks in advance,

Bill Venables (


From: Ronald Thisted <>
Date: Fri, 26 May 89 14:15:11 cdt
Subject: Average-case Rounding Analysis

Most numerical analysis books give error bounds for the accumulation of
an inner product in terms of a worst case scenario, that is, errors at
each step are assumed to have the same sign (no cancellation) and
magnitude close to machine precision. [For example, Golub and Van Loan,
p. 35]

Can anyone supply pointers to the literature for analysis based on
the average case? I am interested in three kinds of work: (a) assuming
that errors are of the form tu, where u=machine eps, and t is uniform on
(-0.5, 0.5), (b) empirical or theoretical investigations of the validity
of the uniform assumption, and (c) any specific work on error bounds
associated specifically with the inner-product computation.


Ronald Thisted
Dept of Statistics/The University of Chicago


From: Yin Zhang <!>
Date: 26 May 89 21:17:17 GMT
Subject: Sparse QR Factorization

Does anybody know of the existence and availability of programs for
sparse QR decomposition (in Fortran or C)? Pointers are appreciated.
Yin Zhang
Rice University, Houston


From: Howard Wilson <>
Date: Sat, 27 May 89 12:03:42 CDT
Subject: Forming a MATLAB Users Group

I am an engineer interested in efficiently using computers to perform
mathematical computations. After working with Fortran for many years, I
recently became an entusiastic MATLAB user. There is a rapidly growing
community of people who use MATLAB for algorithm development and other
scientific computing tasks. This software gives a very friendly inter-
active environment while providing powerful mathematical tools such as
the major capabilities of LINPACK as well as functions for differential
equation solution, Fourier transformation, numerical integration, spline
interpolation, and graph plotting. Most of the functions and operators
handle scalar, vector, or matrix arguments, and work with either real or
complex quantities. The programming features of flexible function format
and dynamic memory allocation make code development very easy. In fact,
MATLAB combines most of the best features of Fortran and APL.
The community of MATLAB is now large enough that a User Group would be
helpful to communicate user experiences, interchange software, and
cultivate wider usage of MATLAB. An inquiry made to original licensees of
the software generated more than a hundred responses favoring formation
of a user group. The current notice is made to contact other users
who may not have been reached by the first inquiry. If you would like to
join and participate in development of a MATLAB user group, please notify
me so that we can include your name for receipt of future information.
Send any replies to


From: Gene Golub <>
Date: Sun, 28 May 1989 11:50:38 PDT
Subject: FAX Numbers

I attach a list of FAX numbers which I have collected over the past several
months. If you want to add your name or make a change, please send me a msg.


FAX numbers as of May 28, 1989. A "+" indicates an international number.

Arizona State University : Hans Mittelmann

ATT : Numerical Analysis : Eric Grosse

CINECA, Bologna, Italy : Valeria Frontini

Cornell University : Electrical Engineering : Franklin Luk

Courant Institute : Michael Overton

Duke University : Computer Science : Daniel B. Szyld

ETH-Zurich : Martin Gutknecht

George Washington University : Mathematics : Murli Gupta

Harwell : Iain Duff
+44-235-832591 and +44-235-432375

Helsinki University of Technology : Mathematics : Olavi Nevanlinna

IMA, UK : Catherine Richards

Flinders University : Mathematics : Jerry Kautsky

Minnesota : Mathematics : Hans Weinberger

MIT : Mathematics : Nick Trefethen

NBS : Francis Sullivan

NSF : Kamal Abdali

ORNL : Mathematics : Bob Ward

Oxford Computing Lab : Bill Morton

McGill University, Canada : Computer Science : Chris Paige

Penn State University : Computer Science : Jesse Barlow

Philips Lab, Belgium : Paul Van Dooren

Princeton : School of Engineering : Bradley Dickinson

Purdue University : Computer Science : John Rice
Walter Gautschi

Royal Institute of Technology : NADA : Germund Dahlquist

Sandia : Juan Meza
Joe Grcar

SIAM : Administration

Southern Methodist University : Mathematics : Ian Gladwell

SRI : Julia Olkin : 415-859-4171

Stanford University :
Computer Science: Gene Golub

Electrical Engineering : Thomas Kailath

University of Adelaide : Computer Science : Sylvan Elhay

UBC : Computer Science : Jim Varah

UCLA : Computer Science : Tony Chan

UCSD : Mathematics : Jim Bunch

United Technologies Research Center : Edward Smith

University of Bergen : Bengt Aspfall and Petter Bjorstad

University of California : John DePillis

University of Fribourg, Switzerland : Mathematics : Jean-Paul Berrut

University of Illinois : CSRD : Bill Harrod and Ahmed Sameh
Computer Science : Bill Gear

University of Leuven, Belgium : Marc Moonen

University of Liverpool, England : R. Wait

University of Maryland : UMIACS: Howard Elman

University of Minnesota : IMA : Willard Miller

University of New South Wales: School of Mathematics: Ian Sloan
+61- 2- 662 6445

University of Technology, The Netherlands : Piet Wesseling

University of Toronto : Computer Science : Ken Jackson

University of Waterloo : Alan George
Henry Wolkowicz
Andrew R. Conn

Vancouver, Computer Science : Uris Ascher

Weidlinger Assoc. : Victor Pereyra

Weizmann Institute, Israel : Uri Ascher

Xerox Park : John Gilbert


End of NA Digest