NA Digest Sunday, September 17, 1995 Volume 95 : Issue 37

Today's Editor:
Cleve Moler
The MathWorks, Inc.

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to

URL for the World Wide Web: -------------------------------------------------------

From: Cleve Moler <>
Date: Sun, Sept. 17, 1995, 11:51 EST
Subject: MATLAB Conference

A quick reminder that the 1995 MATLAB will be held October 16-18,
at the Hyatt Regency Cambridge, overlooking the Charles River,
halfway between Harvard and MIT. We will have invited speakers,
expository and tutorial talks by MathWorks developers and Toolbox
authors, user contributions, minicourses, the MATLAB Lab, and a
party. Further information is available off our home page

Or, if you aren't wired to the Web, send me email and we will send
you a conference annoucement.

-- Cleve Moler


From: Zhongxiao Jia <jia@Mathematik.Uni-Bielefeld.DE>
Date: Mon, 11 Sep 1995 16:55:53 +0200 (MET DST)
Subject: Change of Address for Zhongxiao Jia

Dear colleagues,

I will begin to work at a new institution next month. The following
address is valid from now on:

Zhongxiao Jia
Department of Applied Mathematics
Dalian University of Technology
Dalian 116023
P.R. China

Thanks for your attention!

Sincerely yours,

Zhongxiao Jia


From: Art Owen <art@playfair.Stanford.EDU>
Date: Mon, 11 Sep 1995 16:04:29 -0700
Subject: Quadrature for Optimization and Search

I would like to get some pointers to the literature on the following
problem area:

Given the function f(x,t), integrate out x to obtain the function I(t).
Suppose interest centers on minimizing I over t, or on finding t to solve
I(t) = 0. In the second problem t and I may be vectors of the same dimension.

An obvious approach to this is to pick a good quadrature rule in x, and
approximate I(t) by \sum w_i f( x_i, t ) and then optimize or solve using
the approximation. Another approach is to use points x_i(t) that depend
on t. Under the former method, the x_i must be good "uniformly in t".
Under the second rule there could be unpleasant numerical artifacts if
x_i is not a smooth function of t.

Where might I find a good discussion of how to solve these problems,
and of the resulting error analysis?

[For random x_i and w_i==1, the problem looks like statistical asymptotics.
It may be possible to modify statistical analysis, but I'd hate to reinvent
a wheel.]

-Art Owen, Dept Statistics, Stanford University


From: Robert Nachbar <robert_nachbar@Merck.Com
Date: Fri, 15 Sep 1995 16:51:44 -0400
Subject: N-dimensional Rotation Matrices

I'm working with n-dimensional vector spaces that are generated from
N-dimensional (n<<N) distance matrices by embedding (diagonalization involved).
Because of the symmetries in the various examples, one obtains degenerate
subspaces. These subspaces are not optimally oriented with respect to special
rows/columns of the distance matrix due to the nature of the diagonalizer. What
I would like to do is reorient the the subspace so that the projections of the
special points fall on the axes or in coordinate hyper-planes. For example,
with n=4, I want {w1,x1,y1,z2}.R1 -> {w1',0,0,0} (R1 is 4*4). then, once I've
fixed the first point, then I want to fix the second, say {w2,x2,y2,z2}.R2 ->
{w2',x2',0,0} (R2is 4*4). In fact, the second step could be simplified to
{x2,y2,z2}.R2' -> {x2',0,0} (R2' is 3*3).

Can anyone point me in the right direction for an algorithm or cite a
pertainant reference?

Thanks in advance,


Dr. Robert B. Nachbar Merck Research Laboratories 908/594-7795 RY50S-100 908/594-4224 FAX
PO Box 2000
Rahway, NJ 07065


From: Karl Sabelfeld <>
Date: Sun, 10 Sep 1995 06:20:35 GMT
Subject: New Journal, Monte Carlo Methods and Applications

This year a new international journal "Monte Carlo
Methods and Applications" started. Three first issues are
available. The publisher is VSP, Utrecht, The Netherlands.
To submit a paper, please send a Latex file to the Editor,
Sabelfeld Karl ( or a hard copy
Karl Sabelfeld,
630090, Novosibirsk,
Akad. Lavrentieva, 6
Computing Centre.

or to any member of the Editorial Board.

All the information about subscription is available
via E-mail

Karl Sabelfeld,
The Editor.


From: John Mathews <>
Date: Sun, 10 Sep 1995 08:29:25 -0800 (PST)
Subject: Software for Mathews Numerical Methods Book

I am glad to announce the availability of:

NUMERICAL METHODS: Mathematica Notebooks

These files are located at:

Also available are:

NUMERICAL METHODS: Matlab; C; FORTRAN; Pascal Programs

The files are located at:

Best Regards,

John Mathews
Dept. of Mathematics
California State University Fullerton
Fullerton, CA 92634 USA


From: Tim Davis <>
Date: Thu, 14 Sep 1995 12:28:24 -0400
Subject: UMFPACK Version 2.0: General Unsymmetric Sparse Matrix Solver

I would like to announce the release of UMFPACK Version 2.0, the unsymmetric-
pattern multifrontal package.

UMFPACK Version 2.0 is a package for solving systems of sparse linear systems,
Ax=b, where A is sparse and can be unsymmetric. It is written in ANSI Fortran
77. There are options for choosing a good pivot order, factorizing a
subsequent matrix with the same pivot order and nonzero pattern as a
previously factorized matrix, and solving systems of linear equations with
the factors (with A, L, or U, or their transposes). Iterative refinement,
with sparse backward error estimates, can be performed. Both single and
double precision routines are available. UMFPACK Version 2.0 is based on a
combined unifrontal/multifrontal algorithm that enables a general fill-in
reduction ordering to be applied but avoiding some of the data movement of
previous multifrontal approaches. Version 2.0 is up to twice as fast as
Version 1.1, and uses at little as half the memory. For some matrices,
Version 2.0 has about the same performance as Version 1.1. The improvement
obtained depends on the matrix, and how much can be gained from exploiting
unifrontal-style data movement.

UMFPACK Version 2.0 is available via anonymous ftp to, and in NETLIB
( Some licensing restrictions
apply. For related technical reports, see
(you can also obtain UMFPACK Version 2.0 from my home page).

Tim Davis
Joint work with Iain Duff, Rutherford Appleton Laboratory.


From: Steve Hammond <hammond@niwot.scd.ucar.EDU>
Date: Mon, 11 Sep 1995 16:56:06 -0600
Subject: Workshop on Numerical Fluid Flow In Spherical Geometry


(PDEs on the Sphere)

June 11-14, 1996
Breckenridge, Colorado


As part of the DOE CHAMMP program a workshop on 'Numerical solutions
of fluid flow in spherical geometry' will be held in Breckenridge,
Colorado, June 11-14, 1996. This is the fifth in the series of meetings
which have been hosted on a rotating basis by Argonne National Laboratory,
the National Center for Atmospheric Research, and Oak Ridge National


The primary emphasis of the workshop is:

The development of new methods for and application of new and existing
methods for computing fluid flow in spherical geometry; in particular,
how new methods eliminate shortcomings of current methods and how
they address the problems unique to spherical geometry. Application
to the shallow water equations and to complete baroclinic models
are of interest as well as pure transport for chemistry models.

The workshop will also cover:

1. Topics of interest include monotone advection, geodesic grid systems,
TVD methods, flux correction methods, vector and scalar spectral methods,
alternative discretizations, and Eulerian and semi-Lagrangian methods.
These can be applied to three dimensional baroclinic equations, two
dimensional shallow water equations, or pure transport in spherical

2. The presentation of results from current or new methods is encouraged.
The comparison of these methods is facilitated by application to
standard test sets.

3. Performance of current methods of choice and new methods on parallel
computers, including limitations associated with algorithm and computer
design are appropriate. Implementation details for distributed
and shared memory architectures as well as discussion of attendant
difficulties are of interest. Characteristics of parallel computers that
may affect computational algorithms, in particular, future design
aspects could be considered as they might affect algorithm


The organizing committee for the workshop consists of:

James Hack (, (303) 497-1387)
Steven Hammond (, (303) 497-1811)
Paul Swarztrauber (, (303) 497-1220)
David Williamson (, (303) 497-1372)

Please contact the organizing committee if you have any questions
about the conference program.


The workshop will be held at the Beaver Run Resort in the victorian
town of Breckenridge, Colorado. Breckenridge is located approximately
2 hours west of Denver International Airport. Hourly shuttle service
is available between DIA and Breckenridge.

For questions concerning the conference logistics please contact:

Karen Friedman
Scientific Computing Division
P.O. Box 3000
Boulder, CO 80307
(303) 497-1276
FAX: (303) 497-1804


From: Trini Flores <>
Date: Tue, 12 Sep 95 16:23:19 EST
Subject: SIAM Conference on Numerical Combustion

SIAM Conference on Numerical Combustion
March 4-6, 1996
Le Meridien Hotel, New Orleans, Louisiana

The Call for Papers is accessible through the World Wide Web. Point
your browser to:

The deadline for submission of an extended abstract is SEPTEMBER 25.
Send submissions to:

The Preliminary Programs, hotel and registration information, for the
1995 SIAM Annual Meeting and the 1995 SIAM Conference on Geometric
Design are now available through the World Wide Web. Point your
browser to the following URLs, respectively:

The Call for Participation for the following conferences are likewise
on the World Wide Web:

1996 SIAM Conference on Optimization

1996 SIAM Workshop on Computational Differentiation

1996 SIAM Conference on Discrete Mathematics


From: Hirofumi Amano <>
Date: Sat, 16 Sep 95 15:08:47 +0900
Subject: Parallel Supercomputing Conference in Japan

Dear Colleagues,

Though we have very little time left, we are going to have:

International Symposium on Parallel and Distributed
Supercomputing (PDSC'95)

on September 26-28 in Fukuoka, Japan. The WWW page for the symposium
is now available at:

Some of the highlights of this symposium are:


Practical Parallel Processing
David Kuck (Kuck & Associates)

Parallel Computation: Views from End-User Pragmatism
Takao Tsuda (Kyoto Univ.)


The Effect of Optimizing Compilers on Architecture and Programs
Michael Wolfe (Oregon Graduate Institute)

Application of Program Dependence Analysis in Distributed Software
Jingde Cheng and Kazuo Ushijima (Kyushu Univ.)

A Multicomputer HPC Platform: Experiments and Outcomes
Bob Gingold (Australian National Univ.)

Emerging Supercomputer Applications in Information Infrastructure
George Cybenko, Carl Beckmann, Robert Gray, Katsuhiro Moizumi
(Dartmouth College)

Massively Parallel Processing System Jump-1
Hidehiko Tanaka (Univ. Tokyo)

Vienna Fortran and the Path Towards a Standard Parallel Language
Barbara Chapman (Univ. Vienna), Piyush Mehrotra (NASA Langley Res.
Center), Hans P. Zima (Univ. Vienna)

Hirofumi Amano

Computer Center
Kyushu University
Hakozaki, Higashi, Fukuoka 812-81
Phone: (+81)-92-641-1101 Ext. 2510
FAX: (+81)-92-631-3196


From: Jerzy Wasniewski <>
Date: Fri, 15 Sep 1995 16:26:07 +0200 (METDST)
Subject: Report on the Second Workshop on Applied Parallel Computing

Report on the Second Workshop on Applied Parallel Computing, PARA95

The Second International Workshop on Applied Parallel
Computing in Physics, Chemistry and Engineering Science
(PARA95), and the Tutorial on ScaLAPACK and Parallel NAG
Library on IBM SP2 and SGI Power Challenge were held in Lyngby,
Denmark August 21-24, 1995. The Conference was organized and
sponsored by the Danish Computing Centre for Research and
Education (UNI*C) and the Institute of Mathematical Modelling
(IMM) of the Technical University of Denmark (DTU). Support was
also received from the Danish Natural Science Research Council
through a grant for the EPOS project (Efficient Parallel
Algorithms for Optimization and Simulation) and from the IBM
and SGI computing organizations.
The Workshop was preceded by a one-day Tutorial on ScaLAPACK
and the Parallel NAG Library. The ScaLAPACK (Linear Algebra
Package for Parallel Distributed Memory Computers) and the
Parallel NAG (Numerical Algorithms Group) Library were
presented and discussed. During the Tutorial time was allocated
for practical exercises on the IBM SP2 and SGI Power Challenge
computers. More than 60 people attended the Tutorial.
The Workshp attracted more than 90 participants from around
the world. Authors from 17 countries submitted 60 papers,
of which 12 were invited and 48 were contributed. There were
three parallel sessions.
The conference was opened by remarks from Professor Kaj
Madsen (Technical University of Denmark) and Professor Dorte
Olesen (The Danish Computing Centre for Research an Education).

Tuesday 22 August, 1995:
Paul Woodward (University of Minessota), "Using clusters of
shared memory multiprocessors to simulate turbulent fluid
Aake Nordlund (Copenhagen University), "Visualizing
astrophysical MHD turbulence".
Ahmed Sameh (University of Minessota), "Parallel algorithms
for block tridiagonal systems".
Zahari Zlatev (Danish Environmental Research Institute),
"Parallel computations with large atmospheric models"

Wednesday 23 August, 1995:
Rob Bisseling (University of Utrecht), "Sparse matrix
computations on bulk synchronous parallel computers".
Peter Lomdahl (Los Alamos National Laboratory),
"Multi-million particle molecular dynamics on MPPs".
Jerzy Wasniewski (the Danish Computing Centre for Research
and Education, UNI*C), "Portable parallel numerical
Jack Dongarra (Tennessee University and Oak Ridge National
Laboratory), "Templates and client server computing for
linear algebra".
Fred Gustavson (IBM Research Centre), "An efficient 3d
parallel algorithm for matrix multiplication".

Thursday 24 August, 1995
James Glimm (University at Stony Brook), "Front tracking: A
parallelized CFD method for internal boundaries and
Alex Krasnitz (University of Copenhagen), "Lattice field
theory in a parallel environment".
Antonio Lagana (University of Perugia), "Parallel approaches
to the solution of reactive scattering equations".
Martyn Guest (CCL, Daresbury Laboratory), "High performance
computational chemistry: NWCHEM and fully distributed
parallel algorithms".

Due to the excellent work of the organizing committee the
Workshop went very smoothly. The informality of the meeting gave
the participants much time to converse with new and old colleagues.
Tuesday afternoon saw the Inauguration of the new supercomputer
IBM SP2, followed by the PARA95 reception. IBM contributed 120
bottles of champagne.
On Wednesday night conference participants met at the "Peter
Liep" restaurant, Dyrehaven, where Professor Knut Conradsen
(Technical University of Denmark) gave a very entertaining
speech after an excellent dinner.
The PARA95 papers, both invited and submitted, will be
published by Springer-Verlag in their Lecture Notes in Computer
Science series.

Jack Dongarra & Jerzy Wasniewski


From: P. R Graves-Morris <>
Date: Wed, 13 Sep 1995 14:40:05 +0100 (BST)
Subject: Position at University of Bradford, England


This new post of professor (or reader) in applied mathematics has
been created as part of the university's policy of excellence in
research. Aimed at strenghthening and reinforcing the existing
wide-ranging research carried out by the department, it should
also raise the department's profile in a number of new areas,
including the university's science and engineering departments
and industry.

Appropriately qualified applicants from either an industrial or
an engineering background must have an established international
reputation in original research, including recent quality
publications. Their research interests should ideally complement
or reinforce existing strengths in heat/mass transfer, computational
PDEs, or solid and fluid mechanics.

Further particulars and an application form (quoting ref CRM/NANET)
are available from the Personnel Office at Bradford University,
Bradford, West Yorkshire BD7 1DP, England, or by phoning the
office on 0044-1274-383105, or by FAX on 0044-1274-385009.

Closing Date for applications: 6 October 1995.



From: Paul Matthews <>
Date: Wed, 13 Sep 1995 11:57:54 +0100
Subject: Postdoctoral Position at University of Nottingham.

Postdoctoral Research Opportunity

Applications are invited for a 3-year postdoctoral research assistant
in the Department of Theoretical Mechanics at the University of
Nottingham. The project title is
`The effect of rotation on the formation of buoyant flux tubes in the Sun'
and funding is from the Particle Physics and Astronomy Research Council.
Applicants should hold, or expect to obtain in the near future, a PhD
in an applied mathematical discipline. Some knowledge of numerical
methods and fluid mechanics will be required.

The project will be based in Nottingham under the joint direction
of Dr.P.C.Matthews (University of Nottingham), Prof.D.W.Hughes
(University of Leeds) and Dr.M.R.E.Proctor (University of Cambridge).
The proposed start date is 1 January 1996.

Applications, including a full CV and the names and addresses of two
academic referees, should be made in writing to
Dr.P.C.Matthews, Department of Theoretical Mechanics, University of
Nottingham, Nottingham NG7 2RD, UK.
There is no application form.
The closing date for applications is 31 October 1995.

Informal enquiries may be made to:
Dr.P.C.Matthews (Tel: 0115 951 3840, email:


From: Heinz W. Engl <>
Date: Fri, 15 Sep 1995 11:07:37 EDT
Subject: Postdoctoral Position at Johannes Kepler University

Open Postdoc Positions in Industrial Mathematics

At the Chair for Industrial Mathematics at the Johannes Kepler
Universitdt in Linz (Austria), a full-time research position is to be
filled immediately (now for 18 months, renewable for up to 4 years).
The position is financed by industry and involves research in inverse
problems, especially in connection with parameter identification and
inverse heat conduction problems that arise in steel industry. Good
knowledge about the numerical solution of pdes is essential,
knowledge about inverse problems is desirable. The gross annual
salary will be (depending on prior experience and age) between
360.000 and 420.000 Austrian Schilling. Citizens of E.C. countries
and of Switzerland and Norway do not need visa or work permit, others

In a few months, a second position of the same type will probably be

Applications should be sent to Prof.Heinz W. Engl, Chair for
Industrial Mathematics, Johannes Kepler Universitdt, A-4040 Linz,

Information about the Chair for Industrial Mathematics can be
obtained in the WWW at


From: Francisco de A.M.Gomes Neto <>
Date: Wed, 13 Sep 95 16:30:27 EST
Subject: Position at UNICAMP, Brazil


The Department of Applied Mathematics of the State University
of Campinas (UNICAMP) invites applications for faculty positions in
the area of operations research/industrial engineering. A Ph.D. or
equivalent qualification is required.
Appointments are for three years initially with possible
Candidates are expected to contribute to research and teaching
activities in a system oriented program on production research.
Applications to arrive by *October 31, 1995* should consist of
a full c. v. including a list of publications, and names and addresses
of at least two referees.
Applications and requests for information should be addressed to:
Prof. Jose Plinio de Oliveira Santos
Caixa Postal 6065
13083-970 Campinas, SP
Tel.: + 55 192 397900
Fax: + 55 192 395808


From: Richard Brualdi <>
Date: Mon, 11 Sep 1995 15:45:04 -0500 (CDT)
Subject: Contents, Linear Algebra and its Applications

Contents Volume 229, November 1, 1995

Krzysztof C. Kiwiel (Warsaw, Poland)
Iterative Schemes for the Least 2-Norm Solution of Piecewise
Linear Programs 1

Changqin Xu (Heifei, Anhui, People's Republic of China)
Bellman's Inequality 9

Isabel Faria (Lisboa, Portugal)
Multiplicity of Integer Roots of Polynomials of Graphs 15

F. A. Arlinghaus (Youngstown, Ohio), L. N. Vaserstein
(University Park, Pennsylvania), and Hong You
(Changchun, People's Republic of China)
Products of Involutory Matrices Over Rings 37

Carmine Di Fiore and Paolo Zellini (Roma, Italy)
Matrix Decompositions Using Displacement Rank and
Classes of Commutative Matrix Algebras 49

Khalide Jbilou (Villeneuve d'Ascq, France)
Projection-Minimization Methods for Nonsymmetric Linear Systems 101

Jin-Hsien Wang (Kaohsiung, Taiwan, Republic of China)
Sums and Products of Two Quadratic Matrices 127

John A. Holbrook (Guelph, Ontario, Canada)
Interpenetration of Ellipsoids and the Polynomial Bound of a
Matrix 151

H. Hu (DeKalb, Illinois)
Positive Definite Constrained Least-Squares Estimation of Matrices 167

Yixun Shi (Bloomsburg, Pennsylvania)
Solving Linear Systems Involved in Constrained Optimization 175

Author Index 191


From: George Anastassiou <>
Date: Fri, 15 Sep 1995 12:31:03 CDT
Subject: Contents, Computers and Mathematics

An International Journal
with Applications

Editor in Chief: Ervin Y. Rodin
Special issue on "Concrete Analysis"
Guest Editor
George A. Anastassiou

Contents Volume 30, Number 3-6, Sept. 1995


G. A. Anastassiou XV Preface

J. A. Adell and J. de la Cal 1 Bernstein-Durrmey Operators

G. A. Anastassiou 15 Comparison Theorems on Moduli
of Continuity

C. Badea 23 K-Functionals and Moduli of
Smoothness of Functions

G. Baszenski, F. J. Delvos 33 A United Approach to
Accelerating Trigonometric

L. Chong and G. A. Watson 51 Characterization of a Best and
Unique Best Approximation
from Constrained Rationals

H. Esser, St. J. Goebbels, 59 Sharp Error Bounds for the G.
Luttgens and R. J. Nessel Crank-Nicolson and Saulyev
Difference Scheme in
Connections with an Initial
Boundary Value Problem for the
Inhomogeneous Heat Equation

B. Germano, P. Natalini 69 Computing the Moments of the
and P. E. Ricci Density of Zeros for
Orthogonal Polynomials

H. H. Gonska and 83 Local Smoothness of Functions
D. X. Zhou and Bernstein-Durrmeyer

H. H. Gonska and 103 The Strong Converse Inequality
X. I. Zhou for Berstein-Kantorovich

C. W. Groetsch 129 Regularized Product
Integration for Hadamard
Finite Part Integrals

Y. K. Hu 137 Positive and Copositive Spline
Approximation in Lp {0,1}

R. K. Kovacheva 147 A Note on the Limiting
Distribution of the Zeros of
Pade' Approximants
for Some Classes of Functions

G. Mastroianni and 155 Interlacing Properties of the
D. Occorsio Zeros of the Orthogonal
Polynomials and Approximation
of the Hilbert Transform

J. Prestin and F. Sprengel 169 Multivariate Trigonometric
Wavelet Decompositions

J. B. Prolla 183 Semilattices and Lattices of
Bounded Continous Functions

B. Shekhtman 191 Interpolation of Individual

J. Shi and R. Huotari 197 Simultaneous Approximation
From Convex Sets

O. Shisha 207 The Genesis of the Generalized
Riemann Integral

A. K. Varma and E. Landau 213 New Quadrature Formulas Based
on the Zeros of Jacobi

W. R. Wade 221 A Walsh System for Polar

M. Weba 229 Cubature of Random Fields by
Product-Type Integration Rules

R. A. Zalik 235 Some Remarks on Spectral

M. Zhou and G. A. 243 Some Saturation Properties of
Anastassiou the Representation Formulate
for {C_o} m-Parameter Operator

M. Bartelt and W. Li 255 Error Estimates and Lipschitz
Constants for Best
Approximation in Continuous
Function Spaces

M. Campiti, G. Metafune 269 Uniformly Convergent Lagrange-
and D. Pallara Type Approximation

B. L. Chalmers and 277 The Geometry of Minimal Shape-
M. P. Prophet Preserving Projections Onto

Z. Ciesielski 283 Fractal Functions and Schauder

S. Dahlke and P. Maass 293 The Affine Uncertainty
Principle in One and Two

M. He 307 The Faber Polynomials for
Circular Lunes

T. X. He 317 Shape Criteria of Bernstein-
Bezier Polynomials Over

O. Kounchev 335 Optimal Recovery of Linear
Functionals of Peano Type
through Data on Manifolds

E. B. Lin 353 Multi-Dimensional Wavelet
Stieltjes Transforms

K. Marti 361 Differentiation of Probability
Functions: The Transformation

R. N. Mohapatra, 383 Norm Inequalities which Yield
F. Salzmann and D. Ross Inclusion for Euler Sequence

T. Nishishiraho 389 Approximation Processes with
Respect to Positive
Multiplication Operators

P. E. Ricci 409 Improving the Asympototics for
the Greatest Zeros of Hermite

F. Feher and M. J. Strauss 417 Weak-Type Interpolation and

M. Genuit and 433 A Problem of L. Campbell on
G. Schottler the Equivalence of the
Kramer and Shannon Sampling


End of NA Digest