Subject: NA Digest, V. 93, # 29 Subject: NA Digest, V. 93, # 29 NA Digest Sunday, August 9, 1993 Volume 93 : Issue 29 Today's Editor: Cleve Moler The MathWorks, Inc. moler@mathworks.com Today's Topics: Change of Address for Daniel Okunbor Numerical Analysis Ideas in Science Magazine Formation of Institute of Advanced Scientific Computation Two Preprints on the Numerical Integration of ODE's A Difficult BEVP in Astrophysics Pollution & preconditioners New Book ``How To Teach Mathematics'' Finite Element Error Analysis Postdoctoral Research at Argonne National Laboratory Scalable Parallel Libraries Conference Call For Votes: sci.op-research Postdoc Opportunity in Japan and Preprint Offer IMACS Workshop on Turbulence Contents, Linear Algebra and its Applications Submissions for NA Digest: Mail to na.digest@na-net.ornl.gov. Information about NA-NET: Mail to na.help@na-net.ornl.gov. ------------------------------------------------------- From: Daniel Okunbor Date: Mon, 9 Aug 1993 14:53:36 -0500 Subject: Change of Address for Daniel Okunbor I have taken up a teaching position at the Univesity of Missouri-Rolla effective August 3rd. My current address is Daniel Okunbor University of Missouri-Rolla Rolla, MO 65401 e-maill address: okunborkmcs213k.cs.umr.edu -Daniel. ------------------------------ From: Stephen Vavasis Date: Fri, 6 Aug 93 11:31:54 -0400 Subject: Numerical Analysis Ideas in Science Magazine This week's Science magazine (30 July 1993, pp 578-584) contains an article about pseudospectra in fluid flows. "Science" and its British counterpart "Nature" rarely publish articles about applied mathematics, so this event may be of interest to NA digest readers. The article, "Hydrodynamic Stability Without Eigenvalues", by L. N. Trefethen, A. E. Trefethen, S. C. Reddy, and T. A. Driscoll, proposes a new way to analyze the transition from stable steady-state laminar flow to unstable flow. Both Trefethens, as well as Driscoll, are at Cornell, and Reddy just recently moved from NYU to Oregon State. A key point of the article is that the Navier-Stokes equations, when linearized around the steady-state flow, have highly nonorthogonal eigenvectors. This means that the eigenvalues -- which are crucial to traditional stability analyses -- are ill-conditioned and perhaps tell the wrong story about local perturbations. The connection between nonorthogonal eigenvectors and ill-conditioned eigenvalues is well-known to numerical analysts but perhaps not to the larger scientific community. This article is a great example of spreading knowledge from within our community to the world at large. ------------------------------ From: L. M. Delves Date: Tue, 3 Aug 93 9:31:34 BST Subject: Formation of Institute of Advanced Scientific Computation Liverpool University has recently set up the Institute of Advanced Scientific Computation. This represents a merger of two previous Research Centres: Center for Mathematical Software Research; and North West Transputer Support Centre. The Institute (IASC) has interests in all areas of technical computing, but especial expertise and interest in MIMD parallelism. Current projects include: 1) Development of Distributed Data Library 2) Algorithms and software for oil reservoir simulation 3) Parallel database software for diagram retrieval 4) Temperature sensing for induction heating strip mill 5) Development of Fortran90 compiler 6) Development of HPF Fortran source-source translator 7) Parallelisation of codes for SAR processing, and Forging All of these projects are collaborative with external sites/organisations. The Institute has an active Visitor, and Postgraduate Teaching, program. Here are some requests: 1) If you would like any info, mail me at delves@liverpool.ac.uk or the institute at iasc@liverpool.ac.uk 2) If your activities overlap ours, please put us on your email list for circulation of whatever: iasc@liverpool.ac.uk We are especially interested in collaboration with European and US sites. 3) Drop in and see us some time. ------------------------------ From: Robert I McLachlan Date: Tue, 3 Aug 93 14:08:38 -0600 Subject: Two Preprints on the Numerical Integration of ODE's TWO PREPRINTS ON THE NUMERICAL INTEGRATION OF ODE's, BY ROBERT MCLACHLAN Keywords: Lie-Poisson systems, Hamiltonian systems, symplectic integrators, composition methods, operator splitting. Postscript versions of the following papers are available by anonymous ftp from newton.colorado.edu (128.138.249.1), directory pub/numerics/papers, files sine.ps.Z and composition.ps.Z. Please direct comments to the author at rxm@boulder.colorado.edu. "Explicit Lie-Poisson integration and the Euler equations" We give a wide class of Lie-Poisson systems for which explicit, Lie-Poisson integrators, preserving all Casimirs, can be constructed. The methods are extremely simple. Examples are the rigid body, a moment truncation, and a new, fast algorithm for the sine-bracket truncation of the 2D Euler equations. "On the numerical integration of ordinary differential equations by symmetric composition methods" Differential equations of the form $\dot x=X=A+B$ are considered, where the vector fields $A$ and $B$ can be integrated exactly, enabling numerical integration of $X$ by composition of the flows of $A$ and $B$. The relationships between various symmetric compositions currently in use are investigated with regard to order, complexity, and reversibility. Simple formulae are given for the number of determining equations which must be solved for a method to have a particular order. A new, more accurate way of applying the methods thus obtained to compositions of an arbitrary first-order integrator is described and tested. The determining equations are thoroughly explored, and new methods up to 100 times more accurate (at constant work) than those previously known are given. Robert McLachlan ------------------------------ From: Alfred Gautschy Date: Thu, 5 Aug 93 19:09:49 +0200 Subject: A Difficult BEVP in Astrophysics Numerical Methods for a Difficult BEVP in Astrophysics In the following I describe shortly a problem in stellar astrophysics I am working with for years now. I wonder if there is anything better we can do for its solution than what we are applying presently. I am sure the na-community would know best. In stellar stability theory we usually have to solve a system of four or six first-order linear differential equations, that, together with suitable boundary conditions, constitutes the boundary-eigenvalue problem. The domain of the independent variable extends from 0 to 1, with the system having a regular singularity at 0. Formally we can bring the system of equations into the form: d Y x --- = A * Y (1) d x where Y is a complex vector with 4 or 6 components and A is a 4X4 or 6X6 matrix with complex coefficients. The eigenvalues (normal modes of the system) enter the matrix A in two components nonlinearly. The components of the matrix A are strongly varying between x = 0 and x = 1 and the values of the components are known only at a few hundred discrete points between 0 and 1. Furthermore, matrix A has no patricular symmetry properties. We are typically interested in the lowest few dozen eigen-solutions. For particular stellar problems we have to extract high overtones, but again only a few dozen of them. For some time the BEVP was solved with finite difference methods which, however, are pretty cumbersome due to the approximate solutions which have to be guessed. Often the eigenfunctions are rapidly oscillating in space, and that is something one cannot know ahead. Another method we try for a few years now is the transformation of the BEVP into an initial-value problem of Riccati type, which is then solved by means of a shooting method and fitting the solutions by adapting the eigenvalue. Due to the singularities of the Riccati equation on the path of integration one has to transform to the inverse Riccati form rather frequently which makes the calculations rather expensive. My question is now, are there methods around that allow for the calculation of eigenfrequencies and eigenvectors of a BEVP of the type (1) that do not need any a priori guesses of the approximate solution and do (hopefully) not need supercomputers for their application. Is there anything like a generalized matrix eigenvalue method that works for systems? For any informtation concerning improvements a e-mail note would be highly appreciated. Kind regards Alfred Gautschy Max-Planck-Institut fuer Astrophysik Karl-Schwarzschild-Str.1 85748 Garching/Germany ------------------------------ From: Gustavo Montero Garcia Date: Fri, 6 Aug 93 11:54:06 +0100 Subject: Pollution & preconditioners I'll be so grateful if you send me some information about the following subjets: Modelling the pollution phenomena by finite element method in: - Atmosphera - Groundwater - Seawater Fast preconditioners for solving linear and non linear systems of equations. Thank you. Gustavo Montero Garcia Prof. Tit. U. Centro de Aplicaciones Numericas en Ingenieria CEANI Universidad de Las Palmas de Gran Canaria Campus de Tafira Baja 35017 Tafira ( Las Palmas de Gran Canaria) Islas Canarias -Espa~na- Tf.: + 34 28 451 917 FAX.: + 34 28 451 921 E-mail: gustavo@titan.ulpgc.es ------------------------------ From: Nick Higham Date: Fri, 6 Aug 93 14:44:16 +0100 (BST) Subject: New Book ``How To Teach Mathematics'' The American Mathematical Society has just published a book called ``How To Teach Mathematics: A Personal Perspective'' by Steven G. Krantz. While the book is not directly concerned with teaching numerical mathematics, I think it will be of interest to many NA-Digest readers. In 76 pages Krantz gives much valuable advice based on his twenty years experience of teaching. Among the topics he covers are - preparation of lectures - blackboard technique - how to deal with student questions - how to deal with student complaints - handling large classes - setting homework and exams, and grading - use of computers in teaching (he prefers traditional pen and paper exercises) As Krantz says, ``this is a book about the obvious'', but the obvious can be suprisingly easy to overlook and much of the advice proffered here is often learned the hard way, by making mistakes. The book contains many informative examples and entertaining anecdotes, a quote from Bereseford Parlett, and advice from Paul Halmos on how to teach the Fundamental Theorem of Algebra. Every beginning mathematics lecturer will benefit from reading this book. I recommend it to anyone who wants to improve their teaching skills. @book{kran93, author = "Steven G. Krantz", title = "How To Teach Mathematics: {A} Personal Perspective", publisher = "American Mathematical Society", address = "Providence, RI", year = 1993, isbn = "0-8218-0197-X" } Nick Higham Department of Mathematics University of Manchester ------------------------------ From: I. G. Graham Date: Fri, 6 Aug 93 20:07:40 +0100 (BST) Subject: Finite Element Error Analysis Dear Colleagues, Can anyone help me with the following technical question about finite element error analysis? Let Omega be a bounded convex polygonal domain in the plane. Triangulate Omega with a sequence of meshes which are regular in the sense of Ciarlet's book (1978), and which are parametrised by the mesh diameter h. For each h, let a_h be a continuous function on Omega which is smooth in the interior of each triangle of the mesh. Moreover assume that a_h is bounded above and below on Omega by positive constants which are independent of h. Consider (in weak form) the linear elliptic problem - div [(a_h) grad u] = f, which is to be solved for the scalar function u subject to mixed Dirichlet- Neumann boundary conditions, with the Dirichlet part of the boundary having non-trivial measure . The given function f and the given Dirichlet and Neumann data are smooth and independent of h. Let u_h be the solution of this problem by the standard finite element method. Standard arguments then show that u_h is bounded independently of h in the W^1_p norm when p = 2. I am interested in proving the same property for some p > 2 (*) (I wouldn't mind if p had to be arbitrarily close to 2.) This property of finite element approximation was proved for the Dirichlet problem for Laplace's equation by Rannacher and Scott (Math Comp 38, 1982). By the Sobolev Embedding theorem the result (*) implies the slightly weaker result that the uniform norm of the finite element solution is bounded independently of h. There are many results around along these lines e.g. Schatz and Wahlbin (Math Comp 38, 1982) or Suzuki and Fujita (Numer Math 49, 1986). Again Schatz and Wahlbin (and many other authors) considered only Laplace's equation, but Suzuki and Fujita proved their result for quite general coefficients a which could be discontinuous but which were independent of h. Both Schatz and Wahlbin and Suzuki and Fujita considered only the Dirichlet problem. These are the closest results I have found to (*). I have looked through all issues of Math Comp, Numer Math and SIAM JNA since 1985 but have not found anything else. I may of course have missed it. Can anyone help with a reference or an opinion please? Note that I do not care about convergence of finite element solutions, only boundedness independent of h. I am also interested in the analogous result but with a_h discontinuous across triangle boundaries. Many thanks in advance, Ivan Graham School of Mathematical Sciences University of Bath Bath BA2 7AY United Kingdom email: igg@maths.bath.ac.uk fax: (+44 225) 826492 ------------------------------ From: Jorge More Date: Mon, 2 Aug 93 17:18:22 CDT Subject: Postdoctoral Research at Argonne National Laboratory Postdoctoral Research Mathematics and Computer Science Division Argonne National Laboratory The Mathematics and Computer Science Division of Argonne National Laboratory invites applications for a postdoctoral research position. The successful candidate will help develop a parallel environment for solving large-scale numerical optimization problems. A strong background in numerical optimization and tool development for scientific computing applications is desirable. Nominal requirements include a Ph.D. in applied mathematics, computer science, applied science, or engineering. This project is interdisciplinary in nature, interfacing with efforts in linear algebra, computational differentiation, parallel computing tools, and large-scale simulation of physical processes. Project members have access to state-of-the art computing facilities, including the IBM SP1 and the Intel Touchstone DELTA. 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. Applications must be addressed to Walter McFall, Box mcs-postdoc, 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 further information, contact Jorge More' (708-252-7238; more@mcs.anl.gov). Argonne is an affirmative action/equal opportunity employer. ------------------------------ From: Tony Skjellum Date: Mon, 2 Aug 93 22:09:09 CDT Subject: Scalable Parallel Libraries Conference *** SECOND ANNOUNCEMENT + CALL FOR POSTERS *** Scalable Parallel Libraries Conference (including Multicomputer Toolbox Developers' & Users' Meeting) October 6-8, 1993 National Science Foundation Engineering Research Center for Computational Field Simulation, Mississippi State, Mississippi More information: Request from scl-conf@cs.msstate.edu OR anonymous ftp announcement: aurora.cs.msstate.edu directory: pub/SCL, files: SCL.2, Registration_Form Deadline for posters extended to August 15 (please see files on anonymous ftp). Speakers: * Charles L. Seitz, Caltech, [KEYNOTE ADDRESS] Title: High-Performance Workstations + High-Speed LANs >= Multicomputers * Milo R. Dorr, LLNL Title: "A Concurrent, Multigroup, Discrete Ordinates Model of Neutron Transport" * David Walker, ORNL Title: "The design of scalable linear algebra libraries for concurrent computers" * David Womble, Sandia ALBQ Title: "Out of core, out of mind: making parallel I/O practical". * Dan Quinlan, LANL Title: "Run-time Recognition of Task Parallelism Within the P++ Parallel Array Class Library * Steven Smith, LLNL Title: "High-Level Message-Passing Constructs for Zipcode 1.0: Design and Implementation" * Robert Falgout, LLNL Title: "Modeling Groundwater Flow on Massively Parallel Computers" * Anthony Skjellum, MSU/NSF ERC Title: "The Multicomputer Toolbox: Current and Future Directions," * William Gropp, Argonne National Laboratory Title: TBD * Ewing Lusk, Argonne National Laboratory Title: "The MPI Communication Library: Its Design and a Portable Implementation" * Charles H. Still, LLNL Title: "The Multicomputer Toolbox: Experiences with the Meiko CS-2". * S. Lennart Johnsson, Harvard University and Thinking Machines Corp. Title: "Scientific Libraries on Scalable Architectures" * Linda Petzold, AHPCRC & UMN Title: Solving Large-Scale Differential-Algebraic Systems via DASPK on the CM5 * Dan Reed, UIUC Title: TBD * Dan Meiron, Caltech Title: "Using Archetypes to Develop Scientific Parallel Applications" * Eric F. Van de Velde, Caltech Title: Stepwise Program Refinement in Concurrent Scientific Computing * Padma Raghavan, UIUC/NCSA Title: Parallel Solution of Linear Systems using Cholesky Factorization * Anna Tsao, SRC Title: "Performance of a Parallel Eigensolver Based on the Invariant Subspace Decomposition for Dense Symmetric Matrices." * Steve Lederman, SRC Title: "Comparison of Scalable Parallel Matrix Multiply Libraries" * Chuck Baldwin, UIUC Title: "Dense and Iterative Concurrent Linear Algebra in the Multicomputer Toolbox" * Nikos Chrisochoides, NPAC Syracuse Title: "An alternative to data-mapping for scalable iterative PDE solvers : Parallel Grid Generation" * Sanjay Ranka, NPAC Syracuse Title: "Scalable Libraries for High Performance Fortran" * Anthony Skjellum, MSU/NSF ERC Title: "Building Parallel Libraries using MPI" ------------------------------ From: John Gregory Date: Tue, 3 Aug 93 09:05:08 CDT Subject: Call For Votes: sci.op-research Perhaps most na-net readers have seen this announcement, but for those few who have not, I send this copy of what appeared recently on the Usenet group sci.math.num-analysis. CALL FOR VOTES regarding the proposed creation of the unmoderated newsgroup sci.op-research Votes must be received by 2359 GMT on Sunday 22nd August 1993. This Call for Votes will also be sent to the following mailing lists: ORCS-L (ORCS-L@OSUVM1) and msys@bass.bu.edu CHARTER ======= The main purpose of this group is to act as the umbrella group from which different operations research interest groups will branch off in the future, as envisioned by the technology committee of the ORMS board. In the interim, the newsgroup will support the RESEARCH, APPLICATION and TEACHING of operations research through the exchange of information through various activities including: (RESEARCH AND APPLICATION) -- Posting information about accepted papers -- Asking questions and posting summaries of replies -- Posting Frequently Asked Questions (FAQ) and other lists such as -Ajay Shah's list of Free C/C++ programs for numerical methods -Arthur Geoffrion's list of mail reflectors relevant to O.R. -John Gregory's FAQ on LP Those interested in a particular area could bring out regular FAQs answering questions or likely questions from those new to their area. -- Posting information about ARCHIVES (e.g. those at Rutgers, Bilkent) (TEACHING) -- Sharing teaching approaches -- Announcement of new textbooks; Discussion on existing textbooks (APPLICATION) -- New product announcements -- Users' impressions of commercial software (No advertisements.) (OTHER) -- JOB announcements in universities and industry The group will be UNMODERATED. Newsgroups line: sci.op-research Operational Research VOTING INSTRUCTIONS =================== Cut this out this ballot form, fill in your name and delete the choices that aren't applicable. Send it to iwj-vote@cam-orl.co.uk. If your software honours Reply-To lines you can just reply to this message (please trim the message down, though, if you do it by quoting; I don't need a copy of the whole CFV!). My name is: My vote on sci.op-research: yes/no/abstain End of ballot The line with your name is mandatory. Do not insert any blank lines. Votes posted to newsgroups or mailed to anyone except the vote taker will not be counted. Votes for a group with a different name to that shown above will likewise not be counted. You will receive an acknowledgement of your vote by email; A mass acknowledgement will be posted during the voting period; at the end of the voting period a complete list of votes will be published. You may only vote once and only from one address, if you have several. The end of the voting period is specified at the top of this CFV. CONTACT INFORMATION =================== This vote is being conducted by Ian Jackson , a neutral third party, with possible assistance from Ron Dippold . Please contact iwj@cam-orl.co.uk if you have any questions about the voting procedure or about particular votes. Please contact the proposer, Mohan Sodhi , if you have questions about the proposed group. Mohan Sodhi is grateful to the following, and to others, too numerous to mention, for their constructive comments and encouragement: (In alphabetical order by 1st name): Arthur Geoffrion, Chris Bullen, Craig Willits, David Lawrence, James Bean, Lester Ingber, Mark Moraes, Matthew Saltzmann, Ramesh Sharda, Ron Dippold, Scott Huddleston, Taner Bilgic. -- Ian Jackson iwj@cam-orl.co.uk ...!uknet!cam-orl!iwj acting as vote taker for sci.op-research under the auspices of UVV. These opinions are not necessarily those of Olivetti Research Ltd. ------------------------------ From: Bob Geller Date: Wed, 4 Aug 93 22:14:23 JST Subject: Postdoc Opportunity in Japan and Preprint Offer Greetings from Japan! The main purpose of this email is to inform you about the opportunities available to be a post-doc in Japan, and to ask you to call this to the attention of anyone who's interested. Naturally, I'd especially like to have inquiries from people interested in working with our group at Tokyo University. The following papers give an idea of what we're now working on. If you'd like a preprint let me know, and I'll be happy to mail one to you. The following are our recent papers that are now in the pipeline: (1) Hara, T., S. Tsuboi and R. J. Geller (1993). Inversion for laterally heterogeneous upper mantle S-wave velocity structure using iterative waveform inversion, Geophys. J. Int., in press. (2) Geller, R. J. and T. Hara (1993). Two efficient algorithms for iterative linearized inversion of seismic waveform data, Geophys. J. Int., in press. (3) Geller, R. J. and T. Ohminato (1993). Computation of synthetic seismograms and their partial derivatives for heterogeneous media with arbitrary natural boundary conditions using the Direct Solution Method, Geophys. J. Int., in press. (4) Hara, T. and R. J. Geller (1993). Anamolously large near-field Rayleigh waves excited by the 1992 Landers, California, earthquake, Bull. Seism. Soc. Am., submitted. Obviously I'd be most interested in hearing from people who'd like to work on the above topics, or topics closely related thereto. But this is by no means an absolute requirement. POSTDOC OPPORTUNITY: The Japan Society for the Promotion of Science (JSPS) Postdocs pay Y270,000 per month (tax free) with an extra housing stipend of up to Y100,000 per month. One round trip air ticket is also provided. There is a one-time settling-in allowance of Y200,000, as well as other gooodies (family allowance, health insurance, Japanese language training, research funds, etc.). The current exchange rate is Y104= $1 US (approx). Tenure of appointment is 12 months, with a possibility of renewal for up to 12 additional months. PROCEDURE: Citizens of almost any country are eligible. In many cases you can apply through your country's own agency (e.g., NSF in the case of the US) rather than through me to the JSPS. But in any case you need to have an agreement with the host researcher in Japan before proceeding further. CONTACT: For additional information contact: Bob Geller (bob@global.geoph.s.u-tokyo.ac.jp). FAX +81-3-3818-3247 TEL +81-3-5800-6973 Address: Dept. of Earth&Planetary Physics, Faculty of Science, Tokyo University, Bunkyo-ku, Tokyo 113 JAPAN Please send publication list, c.v., names of three references (letters are not necessary at this point), and reprints or preprints of three recent papers. DEADLINE: Please hurry if you're interested. The deadline may be coming up soon. (It varies from country to country). PREVIOUS JSPS FELLOW WHOM YOU CAN CONTACT: Phil Cummins was a JSPS fellow in our lab from Feb. '92 to April '93. He's kindly agreed to describe his experiences, and answer questions, for anyone who's interested. He's now at ANU in Canberra. You can contact him at: phil@rses.anu.edu.au ------------------------------ From: Karen Hahn Date: Mon, 9 Aug 93 15:16:38 EDT Subject: IMACS Workshop on Turbulence First Announcement/ Call for Participation IMACS WORKSHOP ON (theoretical and computational aspects of) TURBULENCE Rutgers University, New Brunswick, NJ, USA / February 10-11, 1994 (organized by the IMACS Technical Committee on Computational Fluid Dynamics and Rutgers University) In order for CFD to get to new levels of usefulness to industry, there is a need for better turbulence models for high Re flows. An honest appraisal by industry on needs, and a similar word by academia and researchers at national labs on what can be done now is probably necessary. The workshop will deal with the numerical analysis and CFD algorithms that could bridge the gap, and will include those aspects of turbulence theory that have relevance to modeling used in simulation. Topics to include: Modeling, Non-equilibrium Turbulence, Sprectral and High Order Methods, Adaptive Grids, Lagrangian Methods, Ocean Dynamics. It is currently planned to have proceedings published after the workshop; preparing a written paper for these proceedings is, however, optional. Organizing Committee: Prof. R. Pelz (Rutgers University) Dr. R. Agarwal (McDonnell Douglas Aerospace) CALL FOR PAPERS/CALL FOR SESSIONS: Those wishing to contribute a paper, or to organize a session are invited to make themselves known to the Workshop Committee as soon as possible. Mail to: IMACS Turbulence Workshop c/o IMACS Secretariat Department of Computer Science Rutgers University e-mail: imacs@cs.rutgers.edu New Brunswick, NJ 08903, USA fax: 908-932-0537 ------------------------------ From: Richard Brualdi Date: Wed, 4 Aug 1993 14:52:13 -0500 (CDT) Subject: Contents, Linear Algebra and its Applications LINEAR ALGEBRA AND ITS APPLICATIONS Contents Volume 191 Song Xu (Beijing, People's Republic of China) Notes on Sufficient Matrices 1 Irving S. Reed (Los Angeles, California) Generalized de Moivre's Theorem, Quaternions, and Lorentz Transformations on a Minkowski Space 15 David E. Stewart (St. Lucia, Queensland, Australia) An Index Formula for Degenerate LCPs 41 Robert E. Hartwig (Raleigh, North Carolina) The Pyramid Decomposition and Rank Minimization 53 Ali H. Sayed, Hanoch Lev-Ari, and Thomas Kailath (Stanford, California) Fast Triangular Factorization of the Sum of Quasi-Toeplitz and Quasi-Hankel Matrices 77 Bernd Fritzsche, Stefan Fuchs, and Bernd Kirstein (Leipzig, Bundesrepublik Deutschland) Schur-Sequence Parametrizations of Potapov-Normalized Full-Rank jpq-Elementary Factors 107 Han H. Cho (Seoul, Korea) Regular Matrices in the Semigroup of Hall Matrices 151 Miroslav Fiedler (Prague, Czech Republic) and Thomas L. Markham (Columbia, South Carolina) Quasidirect Addition of Matrices and Generalized Inverses 165 B. Najman (Zagreb, Bijenicka, Croatia) and Q. Ye (Winnipeg, Manitoba, Canada) A Minimax Characterization for Eigenvalues of Hermitian Pencils. II 183 Jiu Ding (Hattiesburg, Mississippi) Perturbation Analysis for the Projection of a Point to an Affine Set 199 M. T. Alcalde, C. Burgueno (Casilla, Temuco, Chile) and C. Mallol (Montpellier, France) Les Pol(n, m)-Algebres: Indentites Polynomiales Symetriques dans des Algebres 213 S. Gonzalez and C. Martinez (Zaragoza, Spain) Bernstein Algebras With Zero Derivation Algebra 235 Eugene Seneta (Sydney, Australia) Explicit Forms for Ergodicity Coefficients of Stochastic Matrices 245 Author Index 253 ------------------------------ End of NA Digest ************************** -------