Computer Science Department Events

There will be a Colloquium on November 11 at 10:30 AM by Xiaobai Sun

Friday November 11

E&CS Building RM 2120

TIME: 10:30 AM

Title: Cauchy Transforms for Fast and Parallel Matrix Computations

Speaker: Xiaobai Sun

Abstract:

I will introduce in this talk a special kind of Cauchy transforms thatcan be used as a basic operation for accelerating and parallelizingmatrix computations that arise in existing or emerging applicationswith increasingly large data size. Cauchy transforms with interleavingnodes can be seen as a bridge, in theory and practice, between theconventional matrix computations using orthogonal transforms andcertain structured matrix computations using the celebrated fastmultipole method~(FMM). They can play the same role as the elementaryorthogonal transforms do in solving linear systems, least squaresregression problems and symmetric eigenvalue problems, which areintimately related to singular value decompositions~(SVDs). Theadvantages are in exploiting the properties of the Cauchy transformsto enable hierarchical clustering, approximation, preconditiong andparallelizing matrix computations with an analyzable andcomputationally systematic approach. I illustrate their particular usein designing parallel SVD algorithms.

Bio:

Dr. Xiaobai Sun is a Professor of Computer Science at Duke University.Her research has focused in recent years on fast matrix computationsbased on approximation, compression and stability analysis and how tointegrate them on modern computing facilities successfully forscientific simulations, high-dimensional data analysis and signal andimage processing. Before joining Duke faculty, she was a postdoctoralfellow at the Division of Mathematics and Computer Science, ArgonneNational Laboratory. She got her PhD in Computer Science, with focuson numerical computation, from University of Maryland, College Park,and she got her Master degree in Applied Mathematics in China.