Pseudospectral Shattering and Diagonalization in Nearly Matrix Multiplication Time

Thumbnail

Event details

Date 01.04.2021
Hour 16:1517:15
Speaker Professor Nikhil Srivastava (UC Berkeley)
Location Online
Category Conferences - Seminars

We give an algorithm which diagonalizes an n x n complex matrix up to backward error delta in O(n^{\omega+o(1)}polylog(n/delta))
bit operations  in finite arithmetic with O(polylog(n/delta)) bits of precision, substantially improving the previously known running time of O(n^10/delta^2). The two key ingredients are (1) A random matrix theory result showing that every matrix is close to one with well-conditioned eigenvalues and eigenvectors, resolving a conjecture of E.B. Davies. (2) A rigorous analysis of Roberts' Newton iteration method for computing the matrix sign function in finite arithmetic, itself an open problem in numerical analysis since the 80's.

Joint work with J. Banks, J. Garza-Vargas, A. Kulkarni, and S. Mukherjee.
 

Practical information

  • General public
  • Free

Organizer

  • Prof. Daniel Kressner
    Prof. Nicolas Boumal

Contact

  • Samantha Bettschen

Event broadcasted in

Share