NA Digest Sunday, April 21, 1996 Volume 96 : Issue 16

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: Samir Elhedhli <samir@Bilkent.EDU.TR>
Date: Sun, 14 Apr 1996 16:40:53 +0400 (EET DST)
Subject: l(infinity) Algorithm


I am currently interested in designing an algorithm for the l_infinity
problem (Chebychev, minimax) and I need some algorithm to compare with.
If possibe, I want copies of the famous Barrodale-Phillips and
Conn-Bartels and Charalambous algorithm. Also if there is any other
efficient algorithm, I would like to have information about it.

Yours sincerely

Samir Elhedhli
Department of Industrial Engineering
Bilkent University
06533 Ankara Turkey.


From: Fred Kus <>
Date: Tue, 16 Apr 1996 13:08:26 -0400
Subject: Elliptic PDE Solver


I am looking for a solver for the 3D self-adjoint elliptic PDE :

div[E(x,y,z)*grad(U(x,y,z))] = R(x,y,z)

where E & R are user-specified functions. I am interested in
Cartesian as well as cylindrical coordinates.
Any suggestions or source code will be appreciated.

Thanks in advance.


Fred W. Kus INTERNET: fred@McMaster.CA
Computing & Information PHONE: (905) 525-9140 ext. 24160
Services FAX (905) 528-3773
McMaster University A. N. Bourns Bldg. Rm 131C
Hamilton, Canada L8S 4M1


From: Raghavendra K. Madyastha <>
Date: Thu, 18 Apr 1996 10:31:24 -0500 (CDT)
Subject: Question on Kronecker Products

I have a question related to minimization of a function involving kronecker
products and I can't seem to be able to crack it. I've looked in lots of
matrix analysis books for similar problems with little luck. I was
wondering if anyone had seen something similar to the problem I will pose

lower case letters -- vectors (or scalars, if specified)
upper case letters -- matrices

a^H -- conjugate transpose of 'a'
.kron. -- symbol for kronecker product of 2 matrices

Problem formulation:

minimize f(p,u) = (p^H .kron. u^H)*(E*E^H)*(p .kron. u)
subject to p^H*p = 1
and u^H*u = 1

'p' -- m x 1
'u' -- n x 1
'E' -- mn x k

Any insight or suggestions on this problem would be greatly appreciated.
Thank you.

Raghu Madyastha email:
Dept. of Electrical and Computer Engg, Ph: 713-527-8750 x3577
Rice University, P.O.Box 1892, Houston, TX 77251


From: Jeff Squyres <>
Date: Sun, 21 Apr 1996 17:54:52 -0500 (EST)
Subject: Object Oriented MPI, OOMPI

The Laboratory for Scientific Computing of the Department of Computer
Science and Engineering at the University of Notre Dame is proud to
[finally] announce Object Oriented MPI (OOMPI): A full-featured class
library for MPI. Some of OOMPI's features include:

Full MPI-1.1 functionality
Implemented as a thin layer on top of the C MPI bindings
Offers convienent and intuitive object-oriented abstractions
for message passing
Uses many of the powerful semantic features of the
C++ language, such as data typing, polymorphism, etc.

Full information, documentation, and the source code distribution for
OOMPI Version 1.0 can be obtained from:

Any questions, comments, and most especially feedback can be directed to

Jeff Squyres


From: (Mariano Gasca)
Date: Mon, 15 Apr 1996 17:53:56 +0100
Subject: New Book on Total Positivity

New book on the applications of Total Positivity.

The book Total Positivity and its Applications, (ISBN
0-7923-3924-X) has just been published by Kluwer Academic
Pub., Dordrecht, The Netherlands, in the series Mathematics and
its Applications. Edited by Mariano Gasca and Charles A.
Micchelli, the volume (about 500 pages) contains articles that
document the advances in the subject of Total Positivity during
the last two decades.

The material is divided into ten chapters: Spline
Functions, Matrix Theory, Geometric Modeling, Probability and
Mathematical Biology, Approximation Theory, Complex
Analysis, Statistics, Real Analysis, Combinatorics and Integral
Equations. Some of the articles are of a survey nature and
others present new results.

The book, dedicated by the editors to Prof. S. Karlin, author of
the influential treatise Total Positivity (Stanford U. Press) in
1968, is the result of an international meeting held in Jaca
(Spain) in September 1994. It is expected that the book will
be of value to mathematicians, engineers and computer
scientists whose work involves applications of Total Positivity.


From: Vasile Sima <>
Date: Tue, 16 Apr 1996 07:32:45 +0200
Subject: New Book on Linear-Quadratic Optimization


Algorithms for Linear-Quadratic Optimization
by Vasile Sima
Vol. 200 in "Monographs and Textbooks in Pure and Applied Mathematics"
Earl J. Taft and Zuhair Nashed (Series Editors).

Publisher: Marcel Dekker, Inc., New York
March 1996 / 384 pages, illustrated / Hardcover / ISBN 0-8247-9612-8

This book offers valuable theoretical, algorithmic, and computational
guidelines for solving the most frequently encountered linear-quadratic
optimization problems, providing an overview of recent advances in
control and systems theory, numerical linear algebra, and optimization.
It covers fundamental issues in control theory and applications, such
as regulator and estimator design, state estimation, and robust control.
Topics of special interest, including H-infinity and H-2 optimization,
defect correction, stability-enhancing scaling procedure for Schur
Riccati solvers, Hammarling's algorithms for non-negative definite
Lyapunov equations, ordering generalized real Schur forms, structure-
preserving techniques, finding eigenvalues of Hamiltonian matrices,
and multishift algorithms, are highlighted. Theoretical treatment
benefically combines with an insight into algorithmic techniques,
with relevant numerical issues, advanced software tools, implementation
details improving performances, practical recommendations, and solution
of specific problems.

The chapters have a similar organization: basic theory (with short
proofs); mathematical and computational problem(s); and associated
numerical solvers. Each particular solver has the following referred:
algorithmic templates; computational issues; applicability and
limitations. To be as self-contained as possible, the book includes
and examines state-of-the-art numerical linear algebra algorithms and
associated software tools, like LAPACK. All algorithms are written in
a suggestive and concise informal language, that facilitates computer

A summary of extensive numerical comparisons of various algorithms
is included. Many worked numerical examples, based on published
industrial models, are used for illustrative purposes. The book also
contains 20 figures, 11 tables, an index of algorithm definitions
(with over 80 entries), a subject index, and about 300 references.

By its emphasis on the latest algorithmic advances and areas of current
research interest, as well as by its broad coverage, methodological
guidances and practical guidelines, the book could be useful to applied
and industrial mathematicians, control engineers, electrical and
electronics engineers, computer programmers, researchers in automatic
control and dynamical optimization, and graduate students in these


1 Linear-Quadratic Optimization Problems
2 Newton Algorithms
3 Schur and Generalized Schur Algorithms
4 Structure-Preserving Algorithms
A Comparison of Riccati Solvers
B Notation and Abbreviations

Additional information can be obtained from:

Promotion Department
Marcel Dekker, Inc. or Hutgasse 4, Postfach 812,
270 Madison Avenue, CH-4001 Basel
New York, N.Y. 10016 Switzerland
(212) 696-9000 Phone 061-261-8482


From: Lutz Tobiska <Lutz.Tobiska@Mathematik.Uni-Magdeburg.DE>
Date: Wed, 17 Apr 1996 18:07:42 +0200 (MESZ)
Subject: New Book on Singularly Perturbed Diffferential Equations

Springer-Verlag have just published the book

Convection-Diffusion and Flow Problems"
by Hans-Goerg Roos, Martin Stynes & Lutz Tobiska.

Springer Series in Computational Mathematics No. 24
Published March 1996 in Heidelberg, Germany, ISBN 3-540-60718-8,
350 pages, cost 148 DM.

This book collects, explains and analyses basic methods and recent
results for the successful numerical solution of singularly perturbed
differential equations. Such equations model many physical
phenomena and their solutions are characterized by the presence of layers.
The book is a wide-ranging introduction to the exciting current literature
in this area. It concentrates on linear convection-diffusion equations and
related nonlinear flow problems, encompassing both ordinary and
partial differential equations. While many numerical methods are considered,
particular attention is paid to those with realistic error estimates.
The book provides a solid and thorough foundation for the numerical analysis
and solution of singular perturbation problems.

The full Table of Contents is available as a postscript file on the
WorldWideWeb: open Martin Stynes's home page at
then follow the link to Publications.

Prof. Dr. Lutz Tobiska |
Otto-von-Guericke Universitaet Magdeburg | Tel: +49-391-671 8650
Institut fuer Analysis und Numerik |
Postfach 4120 , Magdeburg, D-39016 | Fax: +49-391-671 8073


From: Bruce A Wade <>
Date: Tue, 16 Apr 1996 16:40:35 -0500 (CDT)
Subject: Midwest Numerical Analysis Day

The Midwest Numerical Analysis Day of 1996 will be held on
Saturday, April 27
at the University of Wisconsin- Milwaukee.
Information can be obtained on the World Wide Web at

The featured speakers are:

Kendall Atkinson, University of Iowa
A survey of recent work in the numerical solution of boundary
integral equations

Suhrit K. Dey, Eastern Illinois University
Solving Nonlinear Systems by perturbed Functional Iterations

Donald Greenspan, University of Texas-Arlington
Discrete Modeling

A complete list of speakers is available on the Web page.


From: R. Beauwens <>
Date: Wed, 17 Apr 1996 15:57:09 +0200 (DST)
Subject: IMACS Seminar on Monte Carlo Methods

APRIL 1-3, 1997

An IMACS Seminar on Monte Carlo methods will be held at Brussels Free
University (ULB) from April 1 to 3, 1997, jointly organised by IMACS
and ULB.

The purpose of the seminar is to provide a forum for the presentation
of recent advances in the analysis, implementation and applications of
Monte Carlo simulation techniques and, in particular to stimulate the
exchange of informations between specialists in these areas.

The topics should cover both theoretical developments:
- random numbers
- statistical analysis
- variance reduction
- perturbation techniques
- MC error analysis
and application fields:
- particle transport
- reliability analysis
- quantum mechanics
- statistical physics

We are now calling for original contributions describing new results.
A few tutorial review works can also be accepted.

International Program Committee:

Prof. K. Binder, Johannes Gutenberg-Universit&aumlt Mainz, Germany.
Dr. T. Booth, Lawrence Livermore National Laboratory, USA.
Prof. J.M. Caillol, Universit&eacute Paris XI Orsay, France.
Prof. D. Ceperley, University of Illinois at Urbana-Champaign, USA.
Prof. I. Dimov, Bulgarian Academy of Sciences, Bulgaria.
Prof. A. Dubi, Ben Gurion University of the Neguev, Israel.
Dr. W. Eckstein, Max-Planck-Institut f&uumlr Plasmaphysik, Garching, Germany.
Dr. A. Gandini, ENEA-CRE Casaccia, Roma, Italy.
Prof. J.H. Halton, The University of North Carolina at Chapel Hill, USA.
Prof. M.H. Kalos, Cornell University, USA.
Prof. S.E. Koonin, The California Institute of Technology, USA.
Prof. D. Levesque, Universit&eacute Paris XI Orsay, France.
Prof. J.D. Lewins, University of Cambridge, United Kingdom.
Prof. E.E. Lewis, Northwestern University, USA.
Prof. M. Marseguerra, Politecnico di Milano, Italy.
Prof. W.R. Martin, University of Michigan, USA.
Prof. J.W. Negele, Massachusetts Institute of Technology, USA.
Dr. F. Salvat, Universitat de Barcelona, Spain.
Dr. S. Seltzer, National Institute of Standards and Technology, USA.
Prof. J. Spanier, Claremont Graduate School, USA.
Dr. I.M. Sobol, Russian Academy of Sciences, Russia.
Prof. S. Tougaard, University of Odense, Denmark.
Prof. P.A. Whitlock, Brooklyn College, USA.

To be on our mailing list and get further informations, please see
our server:

Seminar address:
Alain Dubus
IMACS Monte Carlo Seminar
ULB, CP165
50, av. F.D. Roosevelt
B-1050 Brussels


From: Trini Flores <>
Date: Sat, 20 Apr 96 11:31:54 EST
Subject: SIAM Conference on Sparse Matrices

Second SIAM Conference on SPARSE MATRICES
October 9-11, 1996
The Coeur d'Alene Resort
Coeur d'Alene, Idaho

Deadline for submission of one-page extended abstracts: MAY 15, 1996.

Electronic submissions in LaTeX or ASCII files are encouraged. Send
completed abstracts to:

Information regarding the conference can be accessed in electronic
format via SIAM's World Wide Web:


From: Jerome Kreuser <>
Date: Sun, 21 Apr 1996 20:10:03 -0400 (EDT)
Subject: Modeling Workshop Using GAMS

Announcing a Workshop on Mathematical Programming Modeling in
Economics, Finance, and Projects Using the
General Algebraic Modeling System (GAMS)

Sponsored by the George Washington University
Research and Applications on Modeling Laboratory,
Department of Operations Research, and the
SEAS Computing Facility.

May 20-23, 1996 in Washington, D.C.

For more information see

Or email to


From: Jean Utke <>
Date: Mon, 15 Apr 1996 17:08:13 +0200
Subject: Faculty Position at Technische Univ. Dresden

Technical University Dresden

C3-Professorship for High Performance Computing

to be filled at the Faculty of Mathematics and Science, Dept. of
Mathematics, Institute of Scientific Computing.

The Technical University established this position in the context of the
acquisition of a high performance computer. The professor will be the
founding director of the interdisciplinary center for high performance

The prospective professor should follow the latest developments in high
performance computing and be involved in the development of new algorithms
and implementations. He/she will support other university departments with
respect to teaching and research that requires the use of high performance
computers. He/she is also expected to advise the partners inside and outside
the university regarding the development and use of centralized and
distributed computing resources.

In addition to publications and lectures, supervision of interdisciplinary
Diploma and PhD theses are expected. Habilitation or comparable experience
in the area of High Performance Computing are required. The applicant must
satisfy the employment conditions according to o51 SHG (Tertiary Education
Law of the State of Saxony) from 8/4/93. Women are especially encouraged to
apply. Applications from handicapped persons will be given particular

The application should consist of a Curriculum Vitae, an outline of
scientific background, a photo, lists of scientific publications and a copy
of the highest academic degree obtained. Applications must be received no
later than 4/30/96 by :
TU Dresden, Dekan der Fakultaet Mathematik und Naturwissenschaften,
Herr Prof. Dr. R. Kuehne, 01062 Dresden, GERMANY.


From: Christine Copello <>
Date: Tue, 16 Apr 96 07:52:00 PDT
Subject: Positions at Auto-trol Technology

Auto-trol Technology Corporation located in Denver, CO is a systems
integrator on the forefront of graphics and information technology.
Auto-trol implements distributed computer systems in industrial companies
and government agencies worldwide. Presently, we have opportunitites for
the following positions:

Software Development Engineer (All levels)

C++ required, some experience with CAD packages a plus, MOTIF a plus.

Software Test Engineers

Experience using automated test development preferably X-runner, data base
experience helpful and experience developing test plans/procedures.

Technical Writer

Technical writing experience in a high technology environment required.

Network Communication Software Engineer

Strong analytical and problems skills, UNIX experience, including shell
scripts; Oracle database experience, C programming. Experience in the
following areas are a plus: C++ programming; knowledge of network
management, including LAN, WAN and telecommunications.

Interested candidates should send a resume and salary history to:

Auto-trol Technology Corporation
12500 N. Washington Street
Denver, CO 80241

or E-mail your resume and salary history to:

If you have questions on the positions and would like to discuss these,
contact Chris Copello at (303) 252-2247.


From: Francoise Chatelin <>
Date: Tue, 16 Apr 1996 15:43:08 +0200
Subject: European Community Research Training Grant

European Community Research Training Grant (Post-Graduate level)

CERFACS is looking for a post-graduate student to jointly submit
an application for a European Community Research Grant (TMR programme).

We are looking for a graduate student with experience, interest
and background in mathematics and computers, to work on a
Ph.D. thesis on eigensolvers (in connection with evolution equations
and computation of exponentials). The work will be carried out in
the Parallel Algorithms Project (Qualitative Computing Group) at CERFACS,
in cooperation with the Aladin project at INRIA/IRISA, under the
supervision of Francoise Chatelin and Bernard Philippe.

Eligible candidates are non-French citizens from the European
Community (+ Iceland, Liechtenstein and Norway). Students interested
by this joint application should send relevant information (CV +
name of referees + motivation letter) to Valerie Fraysse
( by May 15th at the latest. More information
is available upon request.

For detailed information about CERFACS and the Parallel Algorithms Project
URL address:
For detailed information about the Qualitative Computing Group
URL address:
For detailed information about INRIA/IRISA
URL address:
For more information about Toulouse, the city where CERFACS is located
URL address:


From: Mike Osborne <>
Date: Wed, 17 Apr 1996 12:44:15 +1100
Subject: Positions at Australian National University



The Australian National University and Fujitsu Limited, the world's second
largest computer manufacturer, are collaborating in developing numerical
algorithms and software in a parallel vector environment for new generation
parallel supercomputers such as the VPP700 with a peak speed of 560 Gflops,
and the VPP300.

Mathematicians or computer scientists with experience in algorithm and
software development are sought to join this project (two positions
are anticipated). This is an exciting opportunity to contribute to the
development of software for some of the most powerful supercomputers
of the decade. The appointee will be under the direction of staff
from the School of Mathematical Sciences and the Computer Sciences
Laboratory, RSISE. Appointees will be expected to interact closely
with these groups which are also engaged in the development of
algorithms and software for the experimental multi-processor AP1000

We are seeking candidates with a strong background in computational
mathematics, experience in parallel and/or vector processing, and
demonstrated skills in developing numerical algorithms and
implementing them in practice. The successful candidates will be
required to develop software and undertake research in some of the
following areas: numerical linear algebra, numerical solution of
ordinary and partial differential equations, mathematical programming,
numerical approximations, signal processing. Current projects include
methods for large scale eigenvalue problems, direct and iterative
methods for sparse linear systems, development of multigrid
preconditioners, and extensions to Fourier transforms and wavelet
libraries. Experience in software development for parallel and/or
vector computers is desirable. The ability to complete projects on
schedule is essential. It is important that candidates be able to
take up the post at an early date.

An initial appointment of between one and three years will be negotiated.
Extension of the appointment may be possible, but will depend on continuing
collaborative arrangements with Fujitsu. Persons wishing to take the post
on secondment from other positions are encouraged to apply.

The ANU has assembled one of the most powerful and broadly-based
advanced computing environments to be found in a university.
Resources at the ANU include a 28 Gflop Fujitsu VPP300, a 128
processor Fujitsu AP1000, and a 16 node Viking Sparc processor AP+,
shortly to be upgraded to 32 nodes.

The level of appointment, depending on the candidate's experience and
qualifications, will lie within the following salary ranges:

Research Fellow [Level B] $43,042 - $51,113 p.a.
Research Fellow [Level C] $52,726 - $60,797 p.a.

Further information is available from Dr R Gingold, Supercomputer Facility,
phone: (06) 249 3437, Fax (06) 279 8199, and email
or Professor M Osborne, email or Professor R Brent,

Please obtain the further particulars and selection criteria before
applying, from the Executive Officer, SMS, phone (61 6) 249 2957, Fax (61
6) 249 0759, E-Mail:, or from WWW:

Closing Date: 31 May 1996 Ref: SMS 16.4.1

Email addresses and/or fax numbers should be provided for referees so that
appointment procedures can be expedited.

APPLICATIONS addressing the selection criteria should be submitted in
duplicate to The Secretary, The Australian National University, Canberra,
ACT 0200 quoting reference number and including curriculum vitae, list of
publications and names and addresses of at least three referees. Further
information is available from either of the contacts listed above. The
University has a "no smoking" policy effective in all University buildings
and vehicles.


From: Horst Simon <>
Date: Thu, 18 Apr 1996 08:36:27 -0700
Subject: Hans Lewy Postdoctoral Fellowship at LBNL

Applications are invited for the Hans Lewy Postdoctoral
Fellowship in the Computing Sciences Directorate at the Lawrence
Berkeley National Laboratory for 1996-97.

The fellowship, which was initiated in 1989,
is intended to enhance scientific research in the U.S.A. by
sponsoring a recent Ph.D. to work in areas related to scientific
computing and numerical algorithms, including the interaction of
applications with high performance architectures.
Of particular interest are those areas in which the National Energy
Research Scientific Computing (NERSC) Division,
the Center of Computational Science and Engineering (CCSE), and the
LBNL Mathematics Department are active. Among these
areas are computational mathematics, turbulence theory, material
sciences, interface methods, fluid mechanics, adaptive grid methods,
iterative methods, numerical linear algebra, and tools for parallel
processing. NERSC will provide access to a 512 processor Cray T3E, and
there will be opportunities to interact with the Mathematics faculty and
Computer Science faculty at the nearby University of California, Berkeley.
(for more information see on the WWW).

Interested applicants should send a curriculum vitae and names
of three references before May 15, 1996 to Lawrence
Berkeley Laboratory, Computing Sciences, MS 50B-4230,
University of California, Berkeley,
CA 94720 to the attention of Michelle Gachis or e-mail to
The fellowship is for one year, beginning in summer or
fall of 1996, with possibility of renewal for a second year.

The Berkeley Laboratory is an Equal Opportunity employer committed to the
development of a diverse workforce.


From: SIAM <>
Date: Mon, 15 Apr 96 12:28:57 EST
Subject: Contents, SIAM Optimization

SIAM Journal on Optimization
MAY 1996, Volume 6, Number 2

A Unified Analysis of Hoffman's Bound Via Fenchel Duality
James V. Burke and Paul Tseng

A Linesearch Procedure in Barrier Methods for Some Convex Programming Problems
A. Melman

Differentiable Piecewise Quadratic Exact Penalty Functions for Quadratic
Programs With Simple Bound Constraints
Wu Li

Convergence of an Infeasible Interior-Point Algorithm from Arbitrary Positive
Starting Points
Stephen C. Billups and Michael C. Ferris

Global Convergence Properties of Some Iterative Methods for Linear
Complementarity Problems
Christian Kanzow

An Interior-Point Method for Semidefinite Programming
Christoph Helmberg, Franz Rendl, Robert J. Vanderbei, and Henry Wolkowicz

Complete Characterizations of Global Optimality for Problems Involving the
Pointwise Minimum of Sublinear Functions
B. M. Glover, Y. Ishizuka, V. Jeyakumar, and H. D. Tuan

Projected Sequential Quadratic Programming Methods
Matthias Heinkenschloss

An Interior Trust Region Approach for Nonlinear Minimization Subject to Bounds
Thomas F. Coleman and Yuying Li

Equivalence of Complementarity Problems to Differentiable Minimization: A
Unified Approach
Paul Tseng, Nobuo Yamashita, and Masao Fukushima

An SQP Algorithm for Finely Discretized Continuous Minimax Problems and Other
Minimax Problems With Many Objective Functions
Jian L. Zhou and Andre L. Tits

A Parallel Method for Unconstrained Discrete-Time Optimal Control Problems
Daniel Ralph

A Global Search Method for Discrete Stochastic Optimization
Sigrun Andradottir

Lipschitz Stability for Stochastic Programs With Complete Recourse
Werner Romisch and Rudiger Schultz


End of NA Digest