Global Convergence of Shifted QR

Speaker: Nikhil Srivastava

Date: Tuesday, February 06, 2024

Time: 4:15 PM to 5:00 PM Note: all times are in the Eastern Time Zone

Public: Yes

Location: 32-G449

Event Type:

Room Description: Refreshments at 4:00pm

Host: Sam Hopkins, CSAIL MIT

Contact: Joanne Talbot Hanley, 617-253-6054, joanne@csail.mit.edu

Relevant URL:

Speaker URL: None

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu, theory-seminars@csail.mit.edu

Reminder Subject: TALK: TOC SEMINAR: Global Convergence of Shifted QR, by Nikhil Srivastava, February 6, 2024 at 4:15pm

Abstract: The Shifted QR algorithm is the most widely used algorithm since the 1960's for computing the eigenvalues and eigenvectors of a dense matrix. It is a bit like the simplex algorithm in that it is specified by a "shifting strategy" (cf. pivoting strategies for simplex). There are shifting strategies which are typically very efficient in practice and on special classes of matrices, but occasionally fail, and no strategy is known to be provably rapidly convergent on every input matrix. I will report on some recent significant progress on this question, in particular a new shifting rule which converges rapidly on a small random perturbation of every matrix. Joint work with Jorge Garza Vargas and Jess Banks.

Research Areas:

Impact Areas:

See other events that are part of the Theory of Computation (ToC) Seminar 2024.

Created by Joanne Talbot Hanley at Tuesday, January 23, 2024 at 9:33 AM.