NA Digest Sunday, May 28, 1995 Volume 95 : Issue 22

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: John Pryce <>
Date: Thu, 25 May 1995 14:46:01 +0100 (BST)
Subject: Simeon Fatunla

Grieve, colleagues.

On 24th May I received by fax this brief message from Prof James
Ezeilo, director of the Nigerian Mathematical Centre in Abuja:

" I regret very much to have to report to you the tragic death of
our friend and colleague Prof Simeon Fatunla. He died on the
spot as a result of a head-on collision of his vehicle with
another vehicle at a point on the Auchi-Okene Road on Friday 19
May 1995. His driver also died on the spot.

Funeral arrangements are being organized by the University of
Benin and we will keep you posted.

May Prof Simeon Fatunla's soul rest in peace.
James O C Ezeilo "

Many people round the world will join with his family and his
associates, in mourning the death at age 51 of one of the
outstanding men of his generation and a leader of the academic
community in his country.

I will be in touch with colleagues both from Nigeria and
overseas, who have taken part in Simeon's Benin Conferences, to
decide what should be done to leave an enduring memorial of his
work and to help make permanent what he has begun.

John Pryce
Software Engineering Group
Royal Military College of Science
Shrivenham, Swindon, UK 25 May 95


From: Laurent Bartholdi <>
Date: Mon, 22 May 1995 14:48:24 +0200
Subject: A Continuous Choice of an Eigenvalue

I asked myself the following problem, without being able to answer it:
``Does there exist a continous function e from the space of 2x2 complex
matrices to the space of complex numbers, such that to each matrix
corresponds one of its eigenvalues?''
My first guess was that such a function does not exist. Indeed, take a matrix
with two different eigenvalues, for instance A= [0 1]
[1 0]
(eignenvals +1 and -1), and contruct a 'path in the matrix space', that is
a parametrization M(t) = [a(t) b(t)]
[c(t) d(t)]
such that M(-1) = M(1) = A, a, b, c, d continuous in -1<=t<=1,
and e(M(t)) = t, where e is our hypothetical eigenvalue function. This would
then show such a function cannot exist, as it would have two different values
on A.
Unfortunately all my attempts to build one such function have failed, so I
have come to doubt my conjecture.

Thanks to all, Larry


From: Ira Hammerman <>
Date: Wed, 24 May 95 12:30:29 +0300
Subject: Spatial Cluster Analysis

I am looking for efficient algorithms for clustering two
points on a two-dimensional plane. Can some recommend
1. algorithms
2. reports
3. books.

Much thanks,
Ira Hammerman


From: K. B. Williams <>
Date: Fri, 26 May 1995 18:22:31 -0400
Subject: Tests for Kalman Filter

I have written a program that implements the Kalman Filter with
Driving Noise. I need a test case that is more complicated than
my simple one, yet not overly complex either.

Program or literature references would be much appreciated.

K. B. Williams


From: Pat Worley <worley@haven.EPM.ORNL.GOV>
Date: Wed, 24 May 1995 12:48:20 -0400
Subject: PICL and PSTSWM software updates

PICL and PSTSWM software updates:


PICL v2.1 is now available from .
The primary differences between versions 2.0 and 2.1 are new ports for the
IBM SP multiprocessors and for the Cray Research T3D. The SP port is built on
top of MPL, while the T3D port is built on top of the Cray Research
multiprocessor implementation of PVM. README files are provided with the
source code distribution that describe how to compile and use PICL in the new

The Portable Instrumented Communication Library (PICL) is an instrumented,
low overhead message-passing compatibility library, supporting
both portable message-passing and portable performance data collection
without incurring unnecessary overhead. PICL is a research tool developed for
and used in projects in performance evaluation and high performance
computing, and is made available to the community as a research tool.
(It is not a product.)

In my own work, PICL is used for portability when the underlying
native communication library is a good match to the PICL semantics.
When the most efficient native library is not a good match, like SHMEM on the
T3D, I still use PICL for collecting performance data on the user-level
events. An example of this usage can be found in PSTSWM, described below.


PSTSWM v4.0 is now available from .
(It has been available for awhile, but the user guide has just been updated.)

PSTSWM is a message-passing benchmark code and parallel algorithm testbed
that solves the nonlinear shallow water equations on a rotating sphere using
the spectral transform method. PSTSWM was developed by to evaluate parallel
algorithms for the spectral transform method as it is used in global
atmospheric circulation models. Multiple parallel algorithms are embedded in
the code and can be selected at run-time, as can the problem size, number of
processors, and data decomposition. This flexibility allows the code to be
tuned on a parallel platform before benchmarking, thus evaluating the
multiprocessor on its ability to solve the numerical problem rather than
it ability to execute a given fixed parallel implementation. Six different
problem test cases are also supported, each with associated reference
solutions and solution and error analysis options.

As of 4/1/95, PSTSWM runs on the Cray Research T3D, the IBM SP-1 and SP-2,
the Intel iPSC/2, iPSC/860, DELTA, and Paragon (on both GP and MP nodes
and using either the NX or SUNMOS operating systems), the nCUBE/2 and
nCUBE/2S, across a network of SUN and IBM workstations, and on a Cray vector
machine (as a serial application). Message passing is implemented using MPI,
PICL, PVM, and/or native message passing libraries, with the choice being
made at compile time. In principle, it should also run on any other platform
on which MPI, PICL, or PVM is available.

To aid in tuning and in understanding the parallel performance, PSTSWM has
been instrumented for the collection of performance data using the PICL trace
and profile collection interface. The PICL implementation of the code must be
used in order to collect performance data on interprocessor communication
but a mixed PICL/native implementation is also provided that can be used to
collect data on events not related to message passing. In the mixed
implementation, the performance sensitive message passing uses native
commands and PICL is only used in the collection of the performance data.

For more information or for alternative ways of acquiring source code,

Pat Worley


From: Richard Lehoucq <>
Date: Thu, 25 May 1995 09:24:58 -0500 (CDT)
Subject: Report on TICAM Symposium

The Symposium on Advances and Trends in Computational and Applied
Mathematics, sponsored by the Texas Institute for Computational
and Applied Mathematics (TICAM), was held April 20-22, 1995 in
Austin, Texas. The symposium focused on the many interdisciplinary
topics that bring together applied mathematics, numerical analysis,
computational sciences, physics, and engineering.

The symposium also helped to introduce the recently established
academic and research program in Computational and Applied
Mathematics (CAM) at The University of Texas at Austin. Drawing
faculty together from the Natural Sciences and Engineering, it
provides an important new opportunity for instruction and research
in this rapidly growing area. The program is interdisciplinary and
consists of faculty representatives from six engineering departments,
as well as the departments of Computer Sciences, Mathematics and
Physics with contributions from Chemistry and Astronomy. The program
offers study leading to both the Ph.D. and Master of Science degrees.
An important component in the CAM program is the initiation of a new
research institute, the Texas Institute for Computational and Applied
Mathematics (TICAM). TICAM is an organized research unit drawing
research faculties and students from the participating departments.
Interdisclipinary research projects include large-scale parallel
computing, structural acoustics, computational fluid dynamics,
computer simulation of micro-structure of composite materials, and
ongoing projects in computational physics.

The two and half day symposium was structured around a number of
thirty minute keynote presentations. Each day also included parallel
sets of contributed sessions. The contributing speakers were given
fifteen minutes to discuss their research. The organizers assembled
an impressive group of computational scientists for the symposium.
Although the backgrounds of all the speakers varied, the title of
computational scientist could aptly be used for all since the use
of computing was of central importance to the presentations. Indeed,
there was such diversity among the talks that only a few could claim
to not have learned something new about areas outside their own. The
symposium organizers are to be commended for a well coordinated event.
The overall quality and content of the presentations was high. To give
the reader an idea of the breath of the talks, the following
quickly reviews the symposium and subjects of each of the keynote

Starting off the first day was Dr. Marye Anne Fox of the University of
Texas at Austin with a warm welcome for all the attendees. Professor
Tinsley Oden, the director of TICAM, followed with a brief introduction
to the activities and mission of the institute.

The first keynote talk was given by Jerry Bona of the University of
Texas at Austin. He considered the modeling of some of the aspects of
near-shore coastal dynamics. The goal was understanding how seashores
erode over time. In the second talk, Gene Golub of Stanford University
discussed the many applications of the theory of moments in numerical
analysis. One application is the efficient and stable computation of
the elements of the inverse of a symmetric matrix. Professor Golub
mentioned current research is underway that uses these estimates for
determining preconditioners for linear systems. Professor T.J.R. Hughes,
also of Stanford University followed with a talk concerning multi-scale
phenomena. The presentation examined a methodology for capturing the
various scales of phenomena in the finite element modeling of physical
processes. Jack Dongarra of the University of Tennessee and Oak Ridge
National Laboratory next discussed the SCALAPACK project. SCALAPACK is
the porting of the LAPACK software package onto parallel computers, in
particular, massively parallel computers. Professor Dongarra discussed
the many issues that arise in such an undertaking.

Starting off the afternoon session, Professor Bjorn Engquist of the
University of California, Los Angeles talked about multi-phase
computations in geometrical optics. Attempting to capture as much of
the frequency component of various equations describing geometrical
optics through different numerical schemes was discussed. Professor P.
Netaanmaki then followed with an overview of the work of the Laboratory
of Scientific computing he directs at the University of Jyvaskyla,
Finland. One particular research area he addressed was that of problems
in nonlinear optimal control. The final keynote talk of the day was given
by John Whiteman of Brunel University, United Kingdom, on discrete
schemes for hereditary problems of viscoelasticity. Professor Whiteman's
talk typified the synthesis of engineering, science and computational
mathematics of all the talks. The day concluded with a reception at the
Lila B. Etter Alumni center.

The second day of the symposium was started with a talk by Mary Wheeler
of Rice University. Professor Wheeler gave an overview of the research
being undertaken by her group on flow through porous media. Applications
include remediation of soils and aquifers and oil recovery. Simulating
the manufacturing process in microlithography was the subject of the next
talk given by Steven Orszag of Princeton University. Professor James
Bramble of Texas A&M University then followed with a mathematical talk on
the subject of least squares methods for Stokes equations. The fourth talk
of the day was given by David Young of the University of Texas at Austin.
The parallel implementation of alternating-type iterative solution methods
were discussed. The last keynote presentation before lunch was given by
Richard Ewing of Texas A&M University. Professor Ewing discussed an
interesting categorization of the multidisciplinary approach being used
to solve problems in environmental modeling. Following a break for lunch,
Randy Bank of the University of San Diego, La Jolla, examined the many
issues involved in the design of an algorithm for coarsening unstructured
meshes. Professor Tinsley Oden's talk followed and considered parallel
hp-adaptive methods for viscous flow. The methods described provide
efficient and robust criteria for deciding when to refine the mesh in an
adaptive fashion and increase the degree of the finite elements employed.
The goal was to develop a local error estimator to obtain near optimal
convergence rates. Professor Owe Axelsson of the University of Nijmegen gave
the last keynote presentation of the day. A detailed study was given on
developing scalable parallel multilevel algorithms for problems arising from
fluid flow. The day was concluded with presentations by the contributing
speakers. The symposium then retired to a reception at the Faculty center
followed by a banquet.

Ivo Babuska of the University of Maryland gave the first keynote
presentation of Saturday morning. Professor Babuska considered the many
complexities when using the finite element method on domains with unsmooth
boundaries and interfaces as well as the impact of rough coefficients.
The next hour was devoted to sessions by the contributing speakers.
Advances in element free and particle methods was the subject of the talk
given by Professor Ted Belytschko of Northwestern University. Professor G.F.
Carey of the University of Texas at Austin presented a survey and review of
the many numerical schemes for non-Newtonian flows. The final keynote
presentation of the symposium was given by Peter Pinsky of Stanford
University. He addressed the use of Galerkin least squares methods for
Hydrodynamic semiconductor device equations.

Professor Oden then concluded the half day session with some closing
remarks. In particular, he asked all the symposium attendees to thank
Ms. Ruth Hengst for a successful coordination of the event.
Finally, Professor Oden mentioned that a future issue of the Journal of
Computational and Applied Mathematics will contain papers by the keynote
speakers on their presentations.

(The author gratefully acknowledges Gene Golub and Ruth Hengst for their
comments on this report.)


From: Jiming Liu <>
Date: Wed, 24 May 95 11:06:40 EDT
Subject: Positions at Bell Labs

The Advanced Decision Support System (ADSS) organization at AT&T Bell Labs
is interested in outstanding candidates who are considering consulting
and development careers in Operations Research and Computer Science.

AT&T Bell Labs ADSS has a long record of developing large-scale decision
support systems and providing optimization-based consulting, object
oriented design and software development, and implementation of
client/server architectures. We have 3-4 openings for regular employees
as well as for post-doctoral positions with a possibility of converting
to regular employee positions.

What AT&T Bell Labs ADSS requires:

Education: Advanced degree (M.S. or Ph.D.) in Operations Research,
Computer Science or related field required. In addition, an
undergraduate degree in Engineering is desirable.

OR Expertise: Practical experience and a theoretical background in
general optimization and/or stochastic processes required. Knowledge
of network analysis, system design and analysis, decision support
systems, and basic statistical analysis desirable.

Work Experience: A minimum of 2 years of non-academic professional
experience required (may be waived for an applicant with a Ph.D.).
Experience in the telecommunication area is desired. Previous
consulting experience a plus.

Computer Skills: Proficiency in programming C++ or C required.
Knowledge of UNIX, object-oriented methodology, use of state-of-the-art
optimization and statistical packages desirable. Experience with
graphics and spreadsheets a plus.

General: Highly developed oral and written communication skills
as well as excellent interpersonal skills. Willingness to learn,
self-motivation, and self-management,

What AT&T Bell Labs ADSS can offer:

A challenging and informal work environment. Work on leading-edge
problems, develop innovative decision technologies, using the latest
in high-performance commercial software technology (object oriented
development, etc.). An opportunity to make an impact in a leading
industrial laborotory environment.

Competitive salaries, excellent benifits, and exciting career growth.

If your background and interests match these expectations, please send
your resume via fax, e-mail or regular mail to:

Jiming Liu
AT&T Bell Laboratories
Room 2L-320
101 Crawfords Corner Road
Holmdel, NJ 07733
fax. 908-949-4001


From: Chris Bischof <>
Date: Thu, 25 May 1995 15:47:39 -0500
Subject: Postdoctoral Position at Argonne Laboratory


Argonne National Laboratory has openings for postdoctoral research
positions in the Mathematics and Computer Science Division and invites
outstanding candidates to apply. One position is expected to be on
assignment at the Supercomputing Research Center in Bowie, Maryland.
Candidates should have a Ph.D. in applied mathematics, computer
science, computational chemistry or a related discipline and
comprehensive knowledge in numerical linear algebra and parallel

The successful candidate will participate in the development of
algorithms and software tools for large-scale dense eigenvalue and
orthogonal reduction problems under the umbrella of the PRISM (Parallel
Research in Invariant Subspace Methods) project. Information on the
PRISM project can be found in pub/prism on This project
is interdisciplinary in nature and interfaces with efforts in
computational chemistry.

The Mathematics and Computer Science Division supports an excellent
computational environment that includes access to high-performance
scientific workstations, a scientific visualization and virtually
reality laboratory, and state-of-the-art parallel computers.

Argonne is located in the southwestern Chicago suburbs, offering the
advantages of affordable housing and good schools, as well as easy
access to the cultural attractions of the city.

Applicants must have received their Ph.D. not more than three years
prior to the beginning of the appointment. The appointment is available
immediately and for a one-year term (renewable). Applications should
be addressed to Walter McFall, Box mcs-postdoc9, Employment and
Placement, Argonne National Laboratory, 9700 S. Cass Avenue, Argonne,
IL 60439, and must include a resume' and the names and addresses of
three references. For additional information, contact Chris Bischof

Argonne is an affirmative action/equal opportunity employer.


From: Antonio Messina <>
Date: Fri, 26 May 1995 08:35:30 -0600
Subject: Postdoctoral Fellowship in Bologna, Italy

Postdoctoral Research Fellowships
Numerical Magnetohydrodynamics

The EEC ``Human Capital and Mobility'' program has established
a numerical MHD grant, which will finance one-year fellowships
in different laboratories. The grant supports collaborations
in the following principal areas:

MHD Flows and Turbulence
Thermal Convection
Interstellar Medium; Stellar Winds; Solar Wind and Corona
Numerical Techniques for MHD

The fellowships are intended for citizens of the member states of
the European Community and associated states. The salary and
associated research costs will be at the standard rate of each state.

Candidates may not apply in their own country; they should write
directly to the group providing their curriculum vitae, a list of
recent publications, two letters of recommendation and a statement
of research interests.
This call is for a one-year position at the University of Bologna,
Dip. Fisica, the deadline for receipt of applications is June 20th 1995.

Contact names:
P. Londrillo, Osserv. astron. di Bologna, via Zamboni 33,
40126 Bologna, Italy. Email: 37907::londrillo,;
tel. (39) 51 6305720.
A. Messina, Dip. Fisica, via Irnerio 46,
40126 Bologna, Italy. Email:;
tel. (39) 51 6305212, fax: (39) 51 6305212.

Cray T3D MCA 64-8, Cray C92/2128 SSD, cluster of graphic ws


From: Rod Smith <>
Date: Fri, 26 May 1995 14:59:44 +0100 (BST)
Subject: Research Studentship at RMCS, Shrivenham



Research studentship in numerical mathematics/scientific computing

Applications are invited for the above studentship, which is for fees
and a maintenance bursary of approximately seven thousand pounds
per annum.

The Applied Mathematics and Operational Research Group is part of
the Royal Military College of Science (RMCS) situated at Shrivenham
in the Vale of the White Horse district of Oxfordshire.
The College is a faculty of Cranfield University and the research
student will be registered for a Cranfield degree.

Techniques of particular interest to the group are the numerical
solution of integral equations, including boundary element methods;
finite element and finite difference methods and approximation and
data fitting including neural networks.
Work on boundary elements has been continuing since 1980 with the
focus on numerical techniques, applications in fracture mechanics,
free surface problems in fluid mechanics and electrodeposition.

It is envisaged that the student will embark on research in the
application of modern computing techniques to the development of
more efficient boundary element methods. The research may
encompass numerical integration techniques, parallel processing
and adaptive meshing and will require extensive programming in
Fortran 90.

Although previous experience of boundary element methods is not required,
a strong mathematics/numerical analysis background with a good grounding
in computer programming is essential. Applicants should have
(or expect to obtain this summer) at least an upper second class
honours degree or an equivalent postgraduate qualification.
For further information or to apply for the position please contact

Dr Rod Smith/Dr.Venkat V S S Sastry, AMOR group,
RMCS Shrivenham, Swindon, Wilts SN6 8LA, UK
Tel:044 01793 785312
Fax:044 01793 782179


From: Marilyn Radcliff <>
Date: Tue, 23 May 1995 10:56:57 -0400 (EDT)
Subject: Contents, Approximation Theory

Table of Contents: J. Approx. Theory, Volume 81, Number 1, April 1995

J. Bustamante and G. Lopez Lagomasino
Hermite-Pade approximation to a Nikishin type system of meromorphic

Vladimir A. Operstein
A characterization of smoothness in terms of approximation by algebraic
polynomials in $L_p$

H.-B. Knoop and Xin-Long Zhou
On convergence of Hermite-Fej\'er interpolation polynomials

Amos Ron
Approximation orders of and approximation maps from local principal
shift-invariant spaces

Peter Borwein and Tam\'as Erd\'elyi
Dense Markov spaces and unbounded Bernstein inequalities

A. L. Brown
Metric projections in spaces of integrable functions

Knut Petras
Duality and lower bounds for relative projection constants

Timur Oikhberg
Absolute widths of some embeddings

L. Brutman and E. Passow
On the divergence of Lagrange interpolation to $|x|$


Gancho T. Tachev
A note on two moduli of smoothness

Margarita Nikoltjeva-Hedberg and Vladimir Operstein
A note on convex approximation in $L_p$

Ryszard Szwarc
A lower bound for orthogonal polynomials with an application to
polynomial hypergroups


Table of Contents: J. Approx. Theory, Volume 81, Number 2, May 1995

Grzegorz Lewicki
Best approximation in finite dimensional subspaces of $L(W,V)$

Kang Zhao
Simultaneous approximation from PSI spaces

S.P. Zhou
On rational lacunary approximation on the interval $[-1,1]$

Herman Bavinck and Roelof Koekoek
On a difference equation for generalizations of Charlier polynomials

F. M\'oricz
A quantitative version of the young test for the convergence of
conjugate series

Ying Guang Shi
Bounds and inequalities for arbitrary orthogonal polynomials on finite

Jutta Faldey and Wolfgang Gawronski
On the limit distributions of the zeros of Jonqui\'ere polynomials and
generalized classical orthogonal polynomials

Oleg V. Davydov
A class of weak Chebyshev spaces and characterization of best

H. T. Koelink and R. F. Swarttouw
A $q$-analogue of Graf's addition formula for the Hahn-Exton $q$-Bessel

J. M. Carnicer and J. M. Pe\~na
On transforming a Tchebycheff system into a strictly totally positive


Ryszard Szwarc
Uniform subexponential growth of orthogonal polynomials


From: Panos Pardalos <>
Date: Wed, 24 May 1995 20:41:13 -0400
Subject: Contents, Global Optimization

Table of Contents
Vol. 6 No. 3, 1995

/ A Relaxation Method for Nonconvex Quadratically Constrained
Quadratic Programs

HAROLD P. BENSON / A Geometrical Analysis of the Efficient
Outcome Set in Multiple Objective Convex Programs with Linear
Criterion Functions

SANTOS / Solution of Linear Complementarity Problems Using
Minimization with Simple Bounds

Optimization of Arborescent Multilevel Inventory Systems

Minimization Algorithm for One-Variable Rational Functions

Multiplicative Programming Problems via C-Programming

Book Reviews
P.D. PANAGIOTOPOULOS/ Hemivariational Inequalities: Applications
in Mechanics

Contents of all other issues can be found using


End of NA Digest