NA Digest Saturday, April 2, 1988 Volume 88 : Issue 14

Today's Editor: Cleve Moler

Today's Topics:


From: Heinz Engl <>
Date: Tue, 29 Mar 88 15:33:59 EDT
Subject: Positions at University of Linz, Austria

Several positions as 'Universitaetsassistent' (roughly equivalent
to U.S. Assistant Prof., both temporary and tenure track) will be
available from July 1988 or some months later at the newly created
Chair for Industrial Mathematics at the Johannes-Kepler-Universitaet
in Linz, Austria.
Consideration will be given to applicants holding a Ph.D. (or at least
a diploma or master's degree) in any area of applied and computational
mathematics. Preference will be given to those with some experience
in solving industrial problems, since duties include (besides teaching
and research) work on research projects from industry. Some knowledge
of German is desirable, but not absolutely necessary. However, it
would be necessary to learn enough German for teaching as soon as
Yearly salary is around $US 20000 -- at current exchange rate.
This is not yet a formal advertisement, but a preliminary information.
Those interested should send relevant information (including e-mail
address, if applicable) as soon as possible to
Prof. Heinz W. Engl
Institut fuer Mathematik
A-4040 Linz, AUSTRIA
E-mail: K310773 at AEARN.BITNET or na.engl at


From: Andy Wathen <>
Date: Wed, 30 Mar 88 14:36+0100
Subject: Jobs at Bristol



Applications are invited for:
(a) a temporary lectureship in NUMERICAL ANALYSIS
(b) a temporary lectureship in STATISTICS
both tenable for one year, and
(c) a two-year post of COMPUTER OFFICER to assist with software
and hardware management for courses and projects within the
department. This is an opportunity for a graduate to participate
in a wide range of computer-related activities involving PCs
and mainframes

The above posts are tenable from a mutually agreed date before October 1988
on a salary within the range UK pounds 8675-14500 depending on age,
qualifications and experience.
Further particulars can be obtained from the Registrar and Secretary,
University of Bristol, Senate House, Bristol BS8 1TH, England to whom
applications, containing a c.v. and names of three referees should be
sent, to be received by 16 May 1988, quoting ref JC1.


From: Iain Duff <duff@anl-mcs.ARPA>
Date: Wed, 30 Mar 88 12:48:36 cst
Subject: Leslie Fox Prize 1988

The high standard of the previous Leslie Fox Prize Competitions was
maintained this year in the presentations given by the finalists at a meeting
at Imperial College, London on Monday March 28th.
In a very hotly contested competition, the first prize was awarded to
Nick Higham of the University of Manchester for his presentation on Analysis
of the Cholesky decomposition of a semi-definite matrix.
The standard was so high that the Committee decided to present second prizes
to the other finalists, who were (in alphabetical order) ...
T Hagstom (SUNY, Stony Brook). Asymptotic boundary conditions for
computational studies of wave propagation.
P T Harker (Univ of Pennsylvania) Accelerating the convergence of the
diagonalization and projection algorithm for finite-dimensional variational
R H Jackson (Univ of Cambridge) An order of convergence for radial basis
T Tang (Univ Leeds) On the spline collocation methods for non-standard
Volterra Integro-Differential equations.

Congratulations are due to all candidates for providing a memorable and
stimulating occasion providing ample evidence of the continued health of
numerical analysis research amongst the younger ( < 31) members of our

The next Fox Prize will be awarded in 1989, either in June or September.
An announcement and call for entries will appear shortly.


From: Ken Jackson <>
Date: Wed, 30 Mar 88 14:35:27 EST
Subject: Constrained Approximation Problem

Constrained Approximation Problem

An economist here posed the following question:

Given a set of interpolation points (x(i),y(i)), i=1,..,n, and two
values b1 and b2, does there exist a twice continuously differentiable
function y which interpolates the points (x(i),y(i)) and satisfies
|y'(x)| <= b1 and |y''(x)| <= b2?

An alternate form of the question is:

Given the interpolation points (x(i),y(i)), i=1,...,n, what are the
minimum values of b1 and b2 such that there does exist a function y
satisfying the problem above?

He's also interested in other generalizations of these problems.

Clearly, b1 has to be at least as large as the maximum magnitude of the
first divided difference of each pair of adjacent data points and b2
has to be at least as large as twice the magnitude of the second
divided difference of each adjacent triple of data points. If b1 or
b2 is smaller than this, then the problem has no solution. But, if b1
and b2 are equal to these bounds, then it is easy to construct examples
of problems that have no solution as well. So this simple first try
at a solution is not sufficient.

I would guess that problems of this sort are quite common in science
and engineering, but I have not seen any discussion of them in the
literature. Do you know a reference? Or do you know a good solution
technique for problems of this sort?

Ken Jackson, (csnet)
Department of Computer Science, uunet!!krj (uucp)
University of Toronto, krj@csri.toronto.cdn (ean x.400)
Toronto, Canada M5S 1A4 (arpa)
(416) 978-7075 krj@csri.utoronto (bitnet)


From: Matt Crawford <>
Date: 29 Mar 88 20:41:46 GMT
Subject: Using Netlib

Given the writeup in this month's Unix Review, I think I had better
send this reminder and correction. To access the mathematical
software in the Argonne "netlib", write to only
if you are using the internet. (There is no underscore after
"netlib", although the Unix Review "Off the Shelf" column had one.)

If you are using UUCP, write to research!netlib instead.

Someone in uucp-land just asked for five complete libraries and the
requests had bad return addresses, so oddjob no longer forwards uucp
requests to anl-mcs.

Matt Crawford
University of Chicago


From: Mary Medeiros <MFH@MATH.AMS.COM>
Date: Wed 30 Mar 88 9:46:41-EST
Subject: AMS Reviews in Numerical Analysis

The American Mathematical Society has just published a five-volume set "Reviews
in Numerical Analysis, 1980-86." The books contain the complete reviews as
they appeared in Mathematical Reviews during 1980-86 of all of the over 17,600
reviews of articles, books, and conference proceedings in numerical analysis.
Author and key indexes appear in Volume 5. Prices: List $250, Individual AMS
members $150, MR Reviewers $125. Order code: REVNAN/86EM. To order, call
1-800-556-7774 in the continental U.S. or write to American Mathematical
Society, P.O. Box 6248, Providence, RI 02940. Visa and Master Card accepted.
Prepayment required.

End of NA Digest