map of Terman Engineering Center
The program will include historical presentations as well as technical ones on some topics related to Gene's research interests.
Talks will begin at 8:45am and end by 5:15pm.
The speakers, with provisional titles for their talks, are listed here:
MINI-MEMORIES:
Program Committee:
Information about parking and dining
Back to Symposium homepage, and registration form.
"Bilinear Forms and Secular Equations in Electronic Structure Calculations"
Zhaojun Bai
University of California, Davis
Bilinear forms and secular equations appear in many scientific and statistical computing problems. Gene conceptualized the idea of bilinear form and secular equation computing, and worked on its theoretical and numerical aspects over three decades. In this talk, we will discuss the continual and compelling need of large-scale bilinear form computing in modern electronic structure calculations of materials and nanostructures.
"Decay Results for Functions of Band Matrices,
the HSS Preconditioner, and Saddle Point Problems:
Some Highlights from a Collaboration"
Michele Benzi
Emory University
In this talk I will describe some of the results of my collaboration with Gene Golub. In the first part of my talk I will explain how our attempt to compute preconditioners for SPD matrices using Gene's method of matrix moments led to a theory of exponential off-diagonal decay in the entries of analytic functions of banded Hermitian matrices. This theory has recently found application in quantum information processing. Some recent extensions and further applications of the theory will be briefly mentioned. In the second part of the talk I will describe our work on the use of the Hermitian/Skew-Hermitian Splitting (introduced by Bai, Golub, and Ng) as a preconditioner for saddle point problems.
"One of Gene's Hobbies: Building a Scientific Computing Community (Netlib and
NA-Net)"
Jack Dongarra
University of Tennessee
Oak Ridge National Laboratory
University of Manchester
In addition to Gene's contributions to matrix computation he was influential in Netlib, the NA Net and the digest. These systems were created in 1984 to facilitate quick distribution of public domain software routines and community interaction. The Numerical Analysis Net (or "NA Net") began as a simple file of contact information for numerical analysts and evolving into an email forwarding service for the community. It soon evolved to support a regular electronic mail newsletter, and eventually an online directory service. This talk will look at these systems and Gene's influence on them.
"Inner-Outer Iterations and Preconditioners for Constrained Linear
Systems of Equations"
Howard Elman
Department of Computer Science
University of Maryland
The numerical solution of constrained linear systems of equations such as the Stokes and linearized Navier-Stokes equations entails "inner" iterations for certain subsidiary problems such as the discrete Poisson and convection-diffusion equations. We discuss the use of inner iterations for the subproblems, and we examine how this general methodology is affected by acceleration strategies such as Richardson and conjugate gradient iteration. We then discuss the connections between inner iteration and preconditioning and show that these connections lead to the development of efficient preconditioners for the constrained problems arising in models of incompressible flows.
"Gene Golub's Fascination with Matrices, Moments, and Quadrature"
Walter Gautschi
Department of Computer Science
Purdue University
One of the early highlights in Golub's work is the now classical Golub-Welsch characterization of Gaussian quadrature rules in terms of eigenvalues and eigenvectors of a tridiagonal matrix. Some personal recollections regarding the discovery of this and related results will be offered. It is here where moments come in, and where Golub's interest in moments and generalized moments, as well as in orthogonal polynomials and quadrature, took off. What he has done with these tools in the field of linear algebra will be briefly summarized.
"Gene Speaks for Himself: Excerpts from an Interview"
Chen Greif
Department of Computer Science
University of British Columbia
The Milestones book of Gene's selected works was completed and published just in time for the Stanford50 meeting, March 2007. One item that we editors (Raymond Chan, Dianne O'Leary and I) wanted to include was a biography of Gene. During the course of 2006 Gene and I met three times, twice at Stanford and once in Vancouver, and taped an interview several hours long. Gene talked with passion and detail about his life and his work, and the tapes are fascinating and full of interesting stories. In this talk we will hear excerpts from the interview.
"n = 8 billion. Gene Golub and the Mathematics of the World Wide Web"
Sep Kamvar
Google
From 2000-2003, Gene and I collaborated on ways to compute web search rankings fast enough to enable personalized search, which amounted to a web-scale eigenvector approximation for every user of a search engine. I will discuss the collaboration and use it as a window on Gene's far reaching impact on the information industry.
"Efficient Computations with Tensors"
Tamara G. Kolda
Sandia National Labs
In this talk I will discuss some work that came out of the Workshop on Tensor Decompositions, which was co-organized by Gene Golub. The focus will be on efficient computations with tensor. We mention the Tensor Toolbox for MATLAB, which is an object-oriented library of methods for dense, sparse, and structured tensors. We discuss the critical operations in two higher-order generalizations of the matrix SVD: the CANDECOMP/PARAFAC (CP) and Tucker decompositions. This talk includes joint work with Brett Bader (Sandia National Labs) and Jimeng Sun (IBM).
"Numerical Multilinear Algebra: From Matrices to Tensors"
Lek-Heng Lim
Gene's 2007 PhD student, jointly directed by Gunnar Carlsson:
Mathematics Department
University of California, Berkeley
Gene had often lamented that linear algebra, as taught in math departments, and CS237A, his famous course on numerical linear algebra, bore almost no relation to each other. One reason is that in math, linear algebra is regarded as a topic in algebra and is mostly about what could be deduced from the axiomatic definitions of fields and vector spaces. Notions like conditioning, least squares, norms, orthogonality, SVD, though central to numerical linear algebra, do not extend to arbitrary fields and are relegated to a secondary status. Another difference, as Gene also liked to emphasize, is the pivotal role played by matrices. Many mathematicians prefer coordinate-free objects and regard matrices with disdain. But while matrices could represent linear operators with respect to some bases, they could also represent bilinear forms, order-2 tensors, graphs, metrics, correlations, hyperlink structures, DNA microarray measurements, movie ratings by viewers -- many of these make little sense when viewed as an operator. When one realizes that a matrix is not necessarily a coordinate representation of a linear operator, and is contented with results valid only over the real and complex fields, linear algebra becomes enormously more interesting. In similar spirit, we will examine the prospects of a subject we call "numerical multilinear algebra", which is to multilinear algebra what numerical linear algebra is to linear algebra.
"50/50, 20/20, and Other Golden Ratios: Remembering a Favorite Collaboration"
Charles Van Loan
Department of Computer Science
Cornell University
Working with Gene on GVL1, GVL2, and GVL3 was a defining experience; an occasion to witness up close the breadth and beauty of the matrix computation field. During this period there was a 1:1 correspondence between my 2AM revisions and Gene's missed airplane connections! Finishing up GVL4 without Gene has been a different experience, prompting me to share book-related thoughts on collaboration, research vision, and the metaphor of 1.61803398874989...
"Serra House as a Transformation: Structure, Stability,
and Updating"
Margaret Wright
Courant Institute
New York University
Among the recurrent themes in Gene's work were (problem) structure, (numerical) stability, and (low-rank) updates. These concepts can be extended to an interpretation of Serra House as a life-changing transformation experienced for many years by a sequence of students and visitors at Stanford. I'll propose some lighthearted connections between Gene's favorite topics and an a posteriori analysis of time spent at Serra House.