Description
Gene Golub, a numerical analyst and Fletcher Jones Professor of Computer Science at Stanford University, discusses his career to date. Born in Chicago in 1932, Golub attended several colleges before graduating from the University of Illinois with a bachelor’s degree in mathematics in 1953. He stayed on for graduate study, receiving a PhD in 1959 under the direction of Abe Taub. Golub discusses his involvement with the ILLIAC computer and its creators and users, including John Nash, Franz Hohn, David Wheeler and Bill Gear. He describes work on statistical applications with C.I. Rao and Charles Wrigley. In 1959, Golub left for Cambridge, England on a postdoctoral fellowship with Maurice Wilkes’ EDSAC II group, where he shared an office with William Kahan and collaborated by mail with R.S. Varga on his first major paper. After returning to the USA, Golub worked briefly at UC Berkeley’s Lawrence Radiation Laboratory and at Space Technology Laboratories, before accepting a visiting position in the Computer Science Division of Stanford University. After a one year appointment at the Courant Institute (on the invitation of Eugene Isaacson) during 1965–6 he returned to Stanford as Associate Professor and remains a faculty member there. Golub outlines the evolution of numerical analysis within the computer science department at Stanford, including its early history with George Forsythe and Jack Herriot and his later role with Joseph Keller and Joe Oliger in creating the Scientific Computing and Computational Mathematics program, which Golub directed from 1988 to 1998. He discusses his students, including Margaret Wright, Michael Overton and Eric Grosse. Golub reviews his extensive list of publications and collaborations, identifying those he considers most significant and explaining their origins and impact. Among these are his papers with Kahan, Christian Reinsch and Peter Businger on singular value decomposition, with Jim Wilkinson on the application of iterative refinement to least squares problems, Victor Pereyra on differentiation of pseudo-inverses and non-linear least squares problems, and Charles Van Loan on total least squares and the book “Matrix Computations.” Golub explains in some detail the origins of his well known work on direct methods for Poisson’s equation with Bill Buzbee and Clair Nielson, including discussion of his related work on the conjugate gradient method with Paul Concus, and of the roles of Roger Hockney and Oscar Buneman in triggering this research. Golub also summarizes his work with Carl De Boor and with DL Boley on inverse eigenvalue calculations, and with a range of collaborators on Gauss quadratrure rules. Golub has been very active within the academic community, and explains his long involvement with the Householder meeting series and participation in ACM SIGNUM, SIAM, and the national academies of sciences and of engineering. He has been most active within SIAM, where he was editor of its SIAM Classics series, the founding editor of its journals on Scientific and Statistical Computing and on Matrix Analysis and Applications, and a founder of the affiliated International Council for Industrial and Applied Mathematics (ICIAM).
Date
2005-10-22; 2005-10-23
Contributor
|
Golub, Gene, Interviewee
|
|
Haigh, Thomas, Interviewer
|
Publisher
SIAM and U.S. Department of Energy
Place of Publication
Stanford, California, United States
Extent
75 p.
Format
PDF
Copyright Holder
Computer History Museum
Category
Transcription
Subject
ILLIAC (Computer); Least-squares problem; Singular value decomposition (SVD); Poisson's equation; Domain decomposition; Differentiation of pseudo-inverses; Inverse eigenvalue problem; Conjugate gradient method; Gauss Quadrature Rule
Collection Title
Society for Industrial and Applied Mathematics (SIAM) oral history collection
Credit
Gift of SIAM and the US Department of Energy
