NA Digest Sunday, March 6, 1994 Volume 94 : Issue 10

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

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to


From: Michael C. Grant <>
Date: Sat, 26 Feb 1994 16:26:53 -0800
Subject: Meta-discussion of NA Digest

> From: Tom Scavo <>
> Date: Sun, 20 Feb 94 11:53:03 -0800
> Subject: Re: A Meta-answer

> I agree with Arne. An interactive forum along the lines of the
> Usenet newsgroup sci.math.num-analysis would be more productive.

Then why not read sci.math.num-analysis? I wish more people would
actively participate in that newsgroup. I would hope that Usenet
access is not a problem for .edu sites, at least. Perhaps similar
newsgroups (or their equivalent) can be created on Compuserve, AOL,
Prodigy, etc. etc., with their contents automagically shared.

> In fact, I think NA Digest and sci.math.num-analysis would
> benefit greatly from each other. I know that some mailing lists
> are automatically gated into the appropriate newsgroup. Maybe
> that could be done in the case of NA Digest. I, for one, would
> welcome the change.

sci.math.num-analysis is small enough that an automatic
news-to-mailing-list gateway might be practical. Those people with
mail readers that can automatically extract such messages from their
inbox and move them to a separate folder for later perusal will find
this quite convenient; MH has such a facility called 'slocal' that
might work with other mail handlers (I haven't used it though).

> Just look how long it's taken to get these three related messages
> to the list. In a newsgroup, it's not uncommon for a thread to
> generate the same reponse in a matter of hours.

That's why this digest is much better suited for things such as
conference announcements, employment positions, and other general
interest announcements for which no feedback (to the Digest) is

Michael C. Grant


From: Frank Stenger <>
Date: Sun, 27 Feb 1994 11:35:19 -0700
Subject: A Function Equation

An unsolved problem:

Let f be a positive, decreasing function defined on [0,1], such that
f(0) = 1, f(1) = 0, and such that for all x on [0,1], we have

f((1 - (f(x))^2)^(1/2)) = 1-x, and f(f(x)) = x.

Can we find an explicit functional expression for f?

Frank Stenger
Department of Computer Science
University of Utah
Salt Lake City, Utah 84112
Phone: (801) 585-SINC
Fax: (801) 581-5843


From: Comon Pierre <>
Date: Mon, 28 Feb 1994 11:03:28 GMT
Subject: C-codes for GEVD and SVD

Could anyone send me C-codes for GEVD and SVD.
The library packages I have do not contain these tools.


From: Stanly Steinberg <>
Date: Mon, 28 Feb 94 09:32:04 MST
Subject: Question on Numerical Methods for ODEs

I am looking for information on discretizations of a system of
ordinary differential equations of the form

dx/dt = A x + B x

where the system with B = 0 is stiff and simple while the system
for A = 0 is not stiff and complex.

Stanly Steinberg


From: James F. Epperson <>
Date: Wed, 2 Mar 94 14:52:14 CST
Subject: Iterative Least Squares

Do there exist algorithms for solving sparse, structured
least squares problems that are analogous to, say, SOR or
Gauss-Seidel for the linear systems problem? What am looking
for is a way to solve the normal equations (A^T)Ax=(A^T)f without
forming (A^T)A, or to do an (approximate) QR decomposition of A
without all the fill-in that would ordinarily occur.

Thanks in advance

Jim Epperson
Mathematical Sciences Dept.
Univ. of Alabama-Huntsville


From: James F. Epperson <>
Date: Wed, 2 Mar 94 14:56:04 CST
Subject: Backwards Heat Equation

To save me and a colleague from the embarassment of
perhaps re-inventing the wheel, could someone direct
me to a reference for what might be considered the
"state-of-the-art" for solving the backwards heat

Thanks in advance,

Jim Epperson
Mathematical Sciences Dept.
Univ. of Alabama-Huntsville


From: Fred Kus <>
Date: Fri, 4 Mar 1994 14:40:26 -0500
Subject: Eigenvalue of Large, Banded Toepliz Matrices


A researcher here in Chemistry asked about software to solve for the
extreme eigenvalues/vectors of a symmetric, narrow banded Toeplitz matrix of
size n => 10**6. The band routines in Lapack as well as several Lanczos
codes that were tried, did not seem suitable for such a large matrix. Also,
they do not take advantage of the Toeplitz structure A(i,j) -> A(|i-j|)
to reduce the memory required.
Any information and/or software to deal with this problem would be greatly


Fred W. Kus INTERNET: fred@mcmail.CIS.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: Simon Wang <>
Date: Sat, 5 Mar 94 4:57 GMT
Subject: Stability of Recurrence

Stability of Recurrence

Please help me consider the stability problem of the following

A X_{n+1} + B Y_{n+1} = C X_n + D Y_n (1)

where A, B, C and D are known constant square matrices (nonsingular), Y and
Z are vectors.

My question is: if the following recurrences

A X_{n+1} = C X_n (2)

B Y_{n+1} = D Y_n (3)

are stable, can we draw the conclusion that the recurrence in (1) must be
stable? If so, any condition required?

Thanks for your time.
Simon Wang

Email: MEG1839@V2.QUB.AC.UK


From: Garry Tee <>
Date: Mon, 28 Feb 1994 12:35:09 +1200
Subject: A Matrix Eigenvalue Problem (Solution #1)

>On Wed, 16 Feb 94 09:25 CST, Richard Luczak <> asked
>in NA Digest, V. 94, # 8:
>> Let A be a tridiagonal matrix of the form
>> -2 2
>> 1 -2 1
>> . . .
>> 1 -2 1
>> 2 -2
>> What is the formula for all eigenvalues of matrix A?
Let A be of order n>1.
Consider the vector v[k] with elements v[k]_j = cos((k-1)(j-1)Pi/(n-1)),
j = 1,...,n. It is readily verified that v[k] is an eigenvector of A with
eigenvalue m_k = -4 sin^2((k-1)Pi/(2n-2)). (Note that v[k]_1 = 1, and
hence v[k] is non-null). For k = 1,...,n this gives n distinct eigenvalues:
0 = m_1 > m_2 > ,,, > m_n = -4.
Hence, those are all the eigenvalues of A, with eigenvectors v[1],,...v[k].
Garry J. Tee.
Department of Mathematics, University of Auckland, Auckland, New Zealand.


From: Di Benedetto <>
Date: Mon, 28 Feb 1994 17:14:57 GMT+1
Subject: A Matrix Eigenvalue Problem (Solution #2)

On Wed, 16 Feb 94 09:25 CST, Richard Luczak <> asked
in NA Digest, V. 94, # 8:

> Let A be a tridiagonal matrix of the form
> -2 2
> 1 -2 1
> . . .
> 1 -2 1
> 2 -2
> What is the formula for all eigenvalues of matrix A?

On Mon, 21 Feb 1994 12:35:37 --100, Christian Brechbuehler
<> made in NA Digest, V. 94, # 9, the following

> Let A be n by n with n > 1.
> Eigenvector j (not normalized) has the components
> cos(i j Pi/(n-1)),
> where 0 <= j < n
> 0 <= i < n.
> Arbitrarily picking the first row of A, we have the Eigenvalues
> 2(cos(j Pi/(n-1)) - 1).

The conjecture is true, here is a possible way to prove it.

Writing A = T - 2 I , it suffices to compute the eigensystem of the
tridiagonal matrix

0 2
1 0 1
1 0 1
T = ..... ,
1 0 1
1 0 1
2 0

which is Toeplitz except the first and the last row.

The eigenvalue problem for T can be treated in a similar way as in

W.F.Trench, "On the eigenvalue problem for Toeplitz band matrices"
(Lin. Alg. Appl. 64, pp.199-214)

by expressing the relation (T - l I) u = 0 as a difference equation:

u(k+2) - l u(k+1) + u(k) = 0, k = 0,...,n-3

with the boundary conditions

2 u(1) = l u(0), 2 u(n-2) = l u(n-1).

If z is a root of the characteristic equation

z^2 - l z + 1 = 0 (whence l = z + 1/z)

then u(k) has the expression

u(k) = a1 z^k + a2 / z^k,

where a1 and a2 are nonzero solutions of the linear system

(2 z - l) a1 + (2/z - l) a2 = 0

z^(n-2) (2 - l z) a1 + z^(2-n) (2 - l/z) a2 = 0,

which takes into account the boundary conditions.

Substituting l = z + 1/z in the determinant d of such linear
system and solving d = 0 with respect to z, the formulae
conjectured by Brechbuehler are easily derived.

Fabio Di Benedetto
Dip. di Matematica - Univ. di Genova
via L.B.Alberti 4
I - 16132 Genova (Italy)


From: Katalin Balla <,@HUEARN.SZTAKI.HU:H153Bal@HUELLA.BITNET>
Date: 1 Mar 94 18:31:58 +0100
Subject: A Matrix Eigenvalue Problem (Solution #3)

In Na-Digest 94, N9 , Christian Brechbuechler (
proposed a conjecture that answers the question posed
by Richard Luczak ( a week earlier. The conjecture is true.

> Let A be a tridiagonal matrix of the form
> -2 2
> 1 -2 1
> . . .
> 1 -2 1
> 2 -2
> What is the formula for all eigenvalues of matrix A?

< 2(cos(j Pi/(n-1)) - 1).
<They are in descending order for increasing j. The largest Eigenvalue
<is 0, the smallest is -4, independent of n.

One can find the problem (for -A) with all details (eigenvectors, etc.)
as Example N5 in &3.3. (pp.264-266) in the book "Linear algebra and its
applications" by Pal Rozsa written in Hungarian (Linearis algebra es
alkalmazasai. Muszaki Konyvkiado, Budapest, 1976,ISBN 963 10 1278 4).
These two pages do not require any knowledge of Hungarian. Or, one can
write directly to the Prof. P. Rozsa, e-mail: (He knows about this message.) Katalin Balla


From: Jorgen Sand <>
Date: Fri, 4 Mar 1994 13:56:07 +0100 (MET)
Subject: Interpolation on a Square Grid (Solution)

In last week's nanet-news T.R.Hopkins put forward a question concerning
interpolation on a square grid, where the values of the function and its
first order derivatives were known at the grid points. One possible solution
is to find the unique interpolating polynomium of the form:

p(x,y) = (a00+a10*x+a01*y+a11*x*y) + x^2*(a20+a30*x+a21*y+a31*x*y)
+ y^2*(a02+a12*x+a03*y+a13*x*y)

on each square. The 12 coefficients are listed below in case of the square
[0,1]x[0,1] and for the square [x1,x2]x[y1,y2] one may use

a00=f11, (* the value of the function in (x1,y1) *)
a10=fx11, (* the value of the x-derivative in (x1,y1) *)
a01=fy11, (* the value of the y-derivative in (x1,y1) *)
a20=-2*fx11+3*f21-3* f11-fx21,
a21=-2*fx12+3*f22+3*f11-fx22+fx21+2*fx11-3* f21-3*f12,
a02=-2*fy11 -fy12-3*f11+3*f12,
a03=-2 *f12+2*f11+fy12+fy11,

Jorgen Sand
Dept. of Comp.Sci.
Univ. of Copenhagen, Denmark


From: Rajni Patel <rajni@ece.concordia.CA>
Date: Sat, 26 Feb 1994 14:15:20 -0500
Subject: New Book from IEEE Press

NEW from IEEE Press

Edited by
Rajni V. Patel, Concordia University
Alan J. Laub, University of California, Santa Barbara
Paul M. Van Dooren, University of Illinois, Urbana

With an extensive 35-page introduction, and 21 pages of
bibliography, this edited collection of 47 key papers incor-
porates over 15 years of intensive research in the field of
modern numerical linear algebra and its application to com-
putational problems in systems and control. The book illus-
trates the importance and rapid growth of this area of
research, and helps to focus attention on the directions
that future research should take. The introduction and
papers presented in the book also provide a thorough back-
ground for all current computer-aided control system
analysis and design software and, in particular, address the
following issues:

* How to assess numerical stability and conditioning
most effectively in numerical analysis.

* Efficient ways to translate system and control
ideas into reliable numerical algorithms.

* Using condensed forms to check controllability,
observability, and other properties of a model.

* The effect of model uncertainties on various numer-
ical properties of a given model.

* The most reliable algorithms for the key computa-
tional problems in linear system theory.

As the only collection of papers covering the broad scope of
numerical linear algebra in systems and control, this book
will be of interest to engineers and applied mathematicians
as well as others engaged in any aspect of control system
analysis and design.

The book contains the following chapters:

1. Introduction and Survey
2. General Numerical Issues in Control
3. Controllability, Observability, and Realizations
4. ``Closeness'' Problems
5. Frequency Response, Transfer Functions, Poles and
6. Pole Assignment and Observer Design
7. Riccati, Lyapunov, and Sylvester Equations
8. Some Relevant Results from Numerical Linear Algebra
9. Bibliography


edited by
Rajni V. Patel
Alan J. Laub
Paul M. Van Dooren

IEEE Control Systems Society Member Price: $45.00
Regular IEEE Member Price: $72.00
List Price: $89.95
Hardcover, 736 pp

IEEE Order #: PC0340-0
ISBN 0-7803-0443-8

The book can be ordered by telephone or fax:

Tel: 1-800-678-IEEE (in USA) or 908-981-0060
Fax: 908-981-9667

or by mail:

Customer Service Department
445 Hoes Lane, P.O. Box 1331
Piscataway, NJ 08855-1331

The book is also available at a slightly higher price from:

IEEE Technical Activities Brussels Office
13, Avenue de L'Aquilon
B-1200 Brussels
Fax: 32.2.770.85.05
Phone: 32.2.770.22.42


From: Steve Roberts <>
Date: Mon, 28 Feb 94 14:54:25 +1100
Subject: Multigrid Workshop In Australia

Workshop on Multigrid and Multilevel Methods
School of Mathematical Sciences
Australian National University
18th - 22nd April 1994

Profs Jim Bramble and Steve McCormick will both be at the ANU
during the last two weeks of April. To take advantage of this
opportunity, we are planning to hold a workshop in which they
will be the major contributors. We plan to hold this workshop
during the week 18th - 22nd April. Jim will give
a series of talks describing some of his recent work on low
regularity results for multigrid methods and Steve is planning to
talk on algorithms and multigrid applications.
Other speakers will include

Zbigniew Leyk "Iterative methods for solving large nonsymmetric
systems of linear equations"
David Stewart "Multilevel methods for approximate inertial manifolds
(in large scale dynamical systems)"
Suely Oliveira "Parallel Multigrid for Mantle Convection"
David Singleton "Multigrid in data parallel on the Connection Machine"
Steve Roberts, "Multigrid methods applied to geometic evolution problems"

A knowledge of the Multigrid method to the level
of William Briggs' ``A Multigrid Tutorial'' will be assumed.

It is intended that other participants to the workshop will
have an opportunity to present
their work on multigrid methods, or perhaps at a more informal
level, present problems in which multilevel methods may be applied.

At present the format and content of the workshop is very fluid.
If you would like to attend the workshop, contact me (Steve Roberts,
address below) giving details of your interests and background.
We will try to arrange the main talks to suit the average background
of the audience. The format of the workshop will be as informal as
possible, with an emphasis on allowing interaction between the
participants at the workshop.

Steve Roberts
Mathematics Department, The Faculties Office: (61)(6) 249 4445
Australian National University Messages: (61)(6) 249 2908
Canberra ACT 0200 Fax: (61)(6) 249 5549


From: John Strain <>
Date: Wed, 2 Mar 1994 09:15:54 -0800
Subject: Bay Area NA Day

Bay Area Numerical Analysis Day at Berkeley
Saturday, 26 March 1994

Everyone is invited!! Bay Area Numerical Analysis Day
will be held on the UC Berkeley campus on Saturday,
26 March 1994. Young Bay Area numerical analysts will
speak on their research. It will also be an excellent
opportunity to meet colleagues.

The meeting will begin at 9:30 am and finish up about
4:00, with a Chinese banquet at 6:30 pm, costing $20.

A second announcement with details will follow shortly.

-- John Strain


From: Erling Pytte <>
Date: Fri, 4 Mar 94 10:01:25 EST
Subject: Positions at IBM Watson Research Center

The IBM T.J. Watson Research Center have several positions in the area
of Computational Science. Intellectual power and scientific excellence
are the primary criteria, but affinity for and ability to use parallel
computers will be a valued tool. Among the resources available is a
128-way SP1 (128 RISC6000 with a fast switch for serial or parallel use).
Applications with resumes, list of publications, and references should be
sent to Dr. Erling Pytte, IBM T.J. Watson Research Center, P.O. Box 218,
Yorktown Heights, NY 10598.


From: Petter Bjorstad <>
Date: Sat, 5 Mar 1994 11:16:22 +0100
Subject: Ph. D. Scholarship in Norway

A Ph.D scholarship supported by the Norwegian Research Council,
is available immediately.
The Ph.D. student will perform research in scientific computing at the
department's laboratory for parallel computing (Para//ab).
The successful candidate must document strong knowledge
in mathematics and computer science.

For more information please contact:
professor Petter E. Bj{\o}rstad,
Institutt for Informatikk,
Universitetet i Bergen,
N-5020 Bergen.
Phone: +47-55544171


From: John Pryce <>
Date: Wed, 2 Mar 94 15:40 BST
Subject: Benin Conference in Nigeria

by John Pryce and Alastair Spence

The 24th - 28th January 1994 saw the Sixth International Conference on
Scientific Computing at the University of Benin (Uniben), Nigeria, directed
as were the previous conferences by Professor Simeon Ola Fatunla, head of
the department of Mathematics and Computer Science. It was attended by
over fifty delegates from Nigeria and by eight overseas visitors.

The organizing committee had chosen Parallel Computation as a main theme of
the conference in view of its importance to the computational needs of the
Nigerian oil industry. The venue was the main Auditorium of the
university. The opening ceremony was attended by senior figures of the
university and of the state government. Unlike previous conferences no
representatives of Federal government were present - this was one sign of
the unsettled political situation in the country. The keynote addresses
included John Pryce on "Scientific computing - key to an enduring
technology?" who emphasized that while computing is essential to high
technology, teaching basic technical skills (what in the UK are generally
called Do-It-Yourself) so that they are natural everyday activities of the
ordinary man and woman, is vital to an enduring technology. Also Dr Chris
Phillips gave a lucid and "low-tech" introduction to "What parallel
computing is all about". Unfortunately Prof Ezeilo, President of the
Nigerian Mathematical Society, and billed to give a keynote address, was
unable to be present.

The overseas visitors gave mainly survey talks on their specialist areas.
Willy Govaerts (University of Gent, Belgium) spoke on Bordered Matrix
Methods in studying Bifurcations of Large Dynamical Systems; Rolf Jeltsch
(ETH, Zurich, Switzerland) on a Truly Multi-dimensional Scheme for the
Euler Equations of Gas Dynamics; Choi-Hong Lai (University of Greenwich,
UK) on Domain Decomposition Methods and Massively Parallel Processing;
Marco Marletta (University of Leicester, UK) on Numerical Solution of
Hamiltonian System Eigenvalue Problems; Chris Phillips (University of
Newcastle, UK) on Parallel Implementation of Conjugate Gradient-Type
Methods; John Pryce (RMCS, Shrivenham, UK) on Test Packages for
Mathematical Software with a Sturm-Liouville Test Set as an example; Roger
Sidje (Ivory Coast, and Irisa, Rennes, France) on Krylov Subspace Methods
for Markov Processes; Alastair Spence (University of Bath, UK) on
Eigenvalues of Block Matrices arising from Discretizations of the Navier
Stokes Equations.

The talks by local researchers were again mostly of high quality, and
perhaps the best local talk was from Prof Mrs Oni of Ibadan theoretical
physics department who tackled a difficult and important problem from the
oil industry: of devising theoretically well-founded methods for extracting
more information from seismic survey data than is done at present. It was
good to see how many of the speakers were young. There was a welcome
increase in the number of numerical methods that were backed up by
numerical experiments and performance comparisons. The 'locally produced'
talks included, as one would expect, a good proportion from Professor
Fatunla's own school on methods for ODE's but went well beyond this to
problems of control, mathematical programming, combustion theory and
queuing theory. For the first time there were also (mainly expository)
talks on mainstream computer science topics such as topological sorting and
the timetabling problem. As usual, many final-year mathematics students
came to form an enthusiastic addition to the audience, and also attended a
Matlab tutorial, at which one woman student was heard to say "Mmm, I begin
to like this package".

A new feature was a four day Software Engineering Workshop held the week
prior to the Conference. This was a fairly informal event at which Simeon
Fatunla and John Pryce presented the basics of the engineering approach to
software development, to an audience of around twenty, the majority being
from commerce and industry. On the day following the workshop, Willy
Govaerts gave a series of lectures on the theory and uses of numerical
continuation and bifurcation computations.

The need for such training had been emphasized at the 1992 conference. The
workshop was full of discussion, interaction and controversy and a whole
day was given over to delegates presenting their own experiences. Plans
for a follow-up course with more professional facilities and course
material are in hand.

The conference booklet was professionally produced (it was interesting to
see a general move from LaTeX to Wordperfect for document preparation) with
abstracts and full names and addresses of participants. For the first time
it also contained maps of Uniben, Benin City and Nigeria and also - a
valuable addition - the membership of the local organization committees who
put in many hours of work before and during the conference to make it run
smoothly within tight constraints on budget and equipment. It is invidious
to single out names but especially obvious to us visitors were the efforts
of Dr David Igbafe of the reception committee, Dr Frank Otunta who chaired
the latter, and of Dr Victor Aladesula who was on hand punctually each
morning for nearly two weeks to transport us from our NNPC accommodation to
the workshop and conference venue, as well as his wife who also gave freely
of her time.

Most of the overseas visitors (not all owing to lack of space) were
accommodated courtesy of the Nigerian National Petroleum Corporation (NNPC)
at its guesthouse, where we were very comfortable and the menu included
grasscutter (a local animal related to the porcupine) and other varieties
of "bush meat" on demand. We appreciated the unobtrusive and efficient
service of the guesthouse reception staff. The local museum and art
gallery, and the Palm Research Institute (and its palm wine) were sampled
as usual, and the visitors explored the Benin city market with the help of
local staff and their wives to negotiate for souvenirs.

Comparing this with previous conferences one can report both good news and
bad news.

The bad news is that the prevailing political uncertainty and the apparent
absence of a coherent policy for higher education continues to leave the
universities in financial decline. One talented colleague from a
neighbouring university told me how his campus had been without national
grid electricity for six months of the last academic year owing to
inability to pay its electricity bill. Obviously, in such straits, other
necessities of academic life such as purchase of up-to-date international
journals have become unattainable luxuries.

The good news is that there is clearly a core of people who are determined
to carry on come what may and who are extending their locally based
activities. The last two conference Proceedings have been edited, typeset
and printed locally. Professor Fatunla's text "Fundamentals of Fortran
Programming" (see endnote) appeared in November, again typeset and printed
locally. Marketed by the African arm of Oxford University Press it has
already achieved a wide sale in Nigeria. Mr Kayode Ojoko, an entirely
Nigeria-educated civil engineer and software writer who runs a successful
engineering consultancy in Lagos, has demonstrated his structural analysis
Software at the 1992 conference and returned this year to show the
increasing range and sophistication of his products. Necessity as always
is proving to be the mother of invention and it is from such local
technical mastery and initiatives, as much as on any outside help, that the
technological progress of Nigeria will be based.

It remains for us on behalf of all the visitors to give our thanks to all
those who gave us hospitality: to Professor Fatunla and his team for
inviting us and for their hard work on our behalf; to the Vice Chancellor
of the university and his staff for their support to the conference; to the
Nigerian National Petroleum Corporation for accommodating us; and to the
many staff and students who welcomed us so warmly.


Fundamentals of Fortran Programming by Simeon Ola Fatunla, Ada Jane
Press, Benin City, Nigeria, Nov 1993, 460 pp, price 20 UK pounds or 30 US

Proceedings of Benin Conference 1988+1990 (combined), ed. Simeon Ola
Fatunla, University of Ibadan Press plc, Jan 1992, 308 pp, price 18 UK
pounds or 27 US dollars

Proceedings of Benin Conference 1992, ed. Smeon Ola Fatunla, Ada Jane
Press, Benin City, Nigeria, Jan 1994, 230pp, price 18 UK pounds or 27 US

These are availbale from Prof Fatunla, Head of Mathematics and Computer
Science, University of Benin, P.M.B. 1154, Benin City, Nigeria. Telex:
Nigeria (905)41365 UNIBEN NG

John Pryce


From: SIAM <>
Date: Mon, 28 Feb 94 10:03:37 EST
Subject: Contents, SIAM Optimization


Line Search Procedures for the Logarithmic Barrier Function
Walter Murray and Margaret H. Wright

Superlinearly Convergent O( nL)-Iteration Interior Point Algorithms
for Linear Programming and the Monotone Linear Complementarity
Kevin McShane

Convergence Properties of a Class of Rank-two Updates
Paul T. Boggs and Jon W. Tolle

Can Parallel Branch and Bound Without Communication be Effective?
Per S. Laursen

Convergence Theory of Nonlinear Newton)Krylov Algorithms
Peter N. Brown and Youcef Saad

On the Resolution of Linearly Constrained Convex Minimization
Ana Friedlander, Jose Mario Martinez, and Sandra A. Santos

On Optimization Problems with Variational Inequality Constraints
J. V. Outrata

Triangular Decomposition Methods for Solving Reducible Nonlinear
Systems of Equations
J. E. Dennis Jr., Jose Mario Martinez, and Xiaodong Zhang

Extension of Hoffman's Error Bound to Polynomial Systems
Xiao-Dong Luo and Zhi-Quan Luo

Globally Convergent Inexact Newton Methods
Stanley C. Eisenstat and Homer F. Walker

An Interior Point Column Generation Method for Linear Programming
Using Shifted Barriers
John E. Mitchell

Predictor-Corrector Methods for a Class of Linear Complementarity
Sanjay Mehrotra and Robert A. Stubbs


From: Richard Brualdi <>
Date: Thu, 3 Mar 1994 07:07:22 -0600 (CST)
Subject: Contents, Linear Algebra and its Applications

Linear Algebra and its Applications
Contents Volume 199

Leon Jay Gleser (West Lafayette, Indiana), Michael D. Perlman
(Seattle, Washington), S. James Press (Riverside, California),
and Allan R. Sampson (Pittsburgh, Pennsylvania)
A Brief Biography and Appreciation of Ingram Olkin 1

T. Ando (Sapporo, Japan)
Majorizations and Inequalities in Matrix Theory 17

Y. L. Tong (Atlanta, Georgia)
Some Recent Developments on Majorization Inequalities in
Probability and Statistics 69

Karl Mosler (Hamburg, Germany)
Majorization in Economic Disparity Measures 91

James V. Bondar (Ottawa, Ontario, Canada)
Comments on and Complements to: Inequalities: Theory of Majorization
and Its Applications, by Albert W. Marshall and Ingram Olkin 115

Hector F. Miranda and Robert C. Thompson (Santa Barbara, California)
Group Majorization, the Convex Hulls of Sets of Matrices and
the Diagonal Element - Singular Value Inequalities 131

Khakim D. Ikramov (Waterloo, Ontario, Canada)
A Simple Proof of the Generalized Schur Inequality 143

Steen A. Andersson (Bloomington, Indiana) and Michael D. Perlman
(Seattle, Washington)
A Characterization of Matrix Groups That Act Transitively on
the Cone of Positive Definite Matrices 151

Markus Abt (Augsburg, Germany)
A Note on the Product Correlation Rule 171

Srinivasa R. Arikati and Uri N. Peled (Chicago, Illinois)
Degree Sequences and Majorization 179

Berthold Heiligers (Augsburg, Germany)
Totally Nonnegative Moment Matrices 213

J. Ferrer, Ma I. Garcia, and F. Puerta (Barcelona, Spain)
Differentiable Families of Subspaces 229

M*Ua Asuncion Beitia and Juan M. Gracia (Vitoria-Gasteiz, Spain)
Local Behavior of Sylvester Matrix Equations
Related to Block Similarity 253

R. B. Bapat and Subhash C. Kochar (New Delhi, India)
On Likelihood-Ratio Ordering of Order Statistics 281

B. Mond and J. E. Pecaric (Bundoora, Victoria, Australia)
Inequalities Involving Powers of Generalized Inverses 293

Jianming Miao and Adi Ben-Israel (New Brunswick, New Jersey)
On IP-Approximate Solutions of Linear Equations 305

S. W. Drury (Montreal, Quebec, Canada)
A Bound for the Determinant of Certain Hadamard Products
and for the Determinant of the Sum of Two Normal Matrices 329

Ingo Althofer (Bielefeld, Germany)
On Sparse Approximations to Randomized Strategies and
Convex Combinations 339

H. K. Wimmer (Wurzburg, Germany)
Roth's Theorems for Matrix Equations With Symmetry Constraints 357

M. Knott and C. S. Smith (London, England)
On a Generalization of Cyclic Monotonicity and Distances
Among Random Vectors 363

Alan Hoffman (Yorktown Heights, New York) and Uriel G. Rothblum
(Haifa, Israel)
A Proof of the Convexity of the Range of a Nonatomic Vector Measure
Using Linear Inequalities 373

Russell Merris (Hayward, California)
Degree Maximal Graphs Are Laplacian Integral 381

S. W. Drury (Montreal, Quebec, Canada)
On a Theorem of Wielandt and the Compounds of Unitary Matrices 391

Xiao-Li Meng (Chicago, Illinois) and Donald B. Rubin (Cambridge,
On the Global and Componentwise Rates of Convergence
of the EM Algorithm 413

Rafael Bru, Juana Cerdan, and Ana M. Urbano (Valencia, Spain)
An Algorithm for the Multiinput Pole Assignment Problem 427

Author Index 445

Special Issues in Progress

1. Special Issue Honoring Marvin Marcus; special editors are Bryan E. Cain,
Moshe Goldberg, Robert Grone, and Nicholas J. Higham. To appear as Volume 201,
April 1, 1994.

2. Linear Systems and Control, Third Special Issue; special editors are
A. C. Antoulas, P. A. Fuhrmann, M. L. J. Hautus, and Y. Yamamoto. Submission
deadline: November 30, 1992. To appear as Volumes 203/204/205, May/June/July
1, 1994.

3. Special Issue Honoring Chandler Davis; special editors are Rajendra
Bhatia, Shmuel Friedland, and Peter Rosenthal. Submission deadline: March 31,
1993. To appear as Volume 206, July 15, 1994.

4. Proceedings of the Third Conference of the International Linear Algebra
Society at Pensacola; special editors are Dianne P. O'Leary, Leiba Rodman, and
Helene Shapiro. Submission deadline: June 30, 1993. Details provided with the
conference announcement.

5. Proceedings of the conference ``Matrices and Graphs'' in honor of John
Maybee's 65th birthday, held at Boulder, Colorado, May 7, 8, 1993. Special
editors: C. R. Johnson and J. R. Lundgren. Submission deadline: August 31,
1993. Details provided with the conference announcement.

6. Fourth Special Issue on Linear Algebra and Statistics; special editors
are Jeffrey J. Hunter, Simo Puntanen, and George P. H. Styan. Submission
deadline: June 30, 1993. Details in Volume 177, December 1992.

7. Proceedings of the workshop ``Nonnegative Matrices, Applications and
Generalizations'' and the Eighth Haifa Matrix Theory conference held at Haifa,
Israel, May 31-June 4 and June 7-June 10, 1993, respectively. Special
editors: S. Friedland, D. Hershkowitz, and R. Loewy. Submission deadline:
September 15, 1993. Details provided with the conference announcement.

8. Special Issue Honoring Miroslav Fiedler and Vlastimil Ptak; special
editors are Wayne Barrett, Angelika Bunse-Gerstner, and Nicholas Young.
Submission deadline: August 31, 1993. Details in Volume 179.

9. Proceedings of the Fourth Conference of the International Linear
Algebra Society at Rotterdam; special editors are Harm Bart, Ludwig Elsner,
and Andre Ran. Submission deadline November 30, 1994. Details provided with
the conference announcement.

10. Special Issue Honoring J. J. Seidel: special editors are Aart Blokhuis,
Willem H. Haemers, and Alan J. Hoffman. Submission deadline: August 30, 1994.
Details in Volume 193, November 1, 1993.

Special issues are available to individuals at a reduced rate. For
further information, please contact Yusuf Guvenc, Journals Customer Service,
Elsevier Science Inc., 655 Avenue of the Americas, New York, NY
10010; Tel. 212-633-3955; Fax 212-633-3990.

Special Issues Vol. 199


End of NA Digest