Return to Colloquia & Seminar listing
Super-fast Solvers for FMM matrices
Applied Math| Speaker: | Shiv Chandrasekaran, UCSB |
| Location: | 1147 MSB |
| Start time: | Fri, Oct 20 2006, 4:10PM |
Description
The fast multi-pole method (FMM) of Greengard and Rokhlin has
revolutionized scientific computing by greatly speeding up the
calculation of certain sets of sums. The FMM can also be regarded
as a fast matrix multiplication technique for certain classes of
hierarchically structured matrices. These matrices are ubiquitous,
and they readily appear when one discretizes integral and partial
differential equations. Therefore the ability to rapidly apply the
*inverse* of these matrices to other vectors would be of great
importance. In this talk we outline some of the progress that has
been made in designing such an algorithm.
This is joint work with Professors Patrick Dewilde (TU Delft) and
Ming Gu (UC Berkeley) and our graduate students.