Re: convergence of QR-algorithm
From: hansm (mittelmann_at_asu.edu)
Date: 03/23/05
- Next message: vonfeldt: "Re: abstract algebra help please!"
- Previous message: C. Bond: "Re: QD algorithm code"
- In reply to: Jeremy Watts: "Re: convergence of QR-algorithm"
- Next in thread: Jeremy Watts: "Re: convergence of QR-algorithm"
- Reply: Jeremy Watts: "Re: convergence of QR-algorithm"
- Messages sorted by: [ date ] [ thread ]
Date: 23 Mar 2005 15:28:12 -0800
Hi,
you say you need "usually less than 10 iterations" per eigenvalue. For
general real matrices with the Francis shift you need on average 1.5 of
these double-steps and with alternative single shifts in case of
convergence to real eigenvalues you can speed it up some more etc etc.
But your goal is more simplicity of the program and robustness and not
absolute efficiency.
Hans Mittelmann
- Next message: vonfeldt: "Re: abstract algebra help please!"
- Previous message: C. Bond: "Re: QD algorithm code"
- In reply to: Jeremy Watts: "Re: convergence of QR-algorithm"
- Next in thread: Jeremy Watts: "Re: convergence of QR-algorithm"
- Reply: Jeremy Watts: "Re: convergence of QR-algorithm"
- Messages sorted by: [ date ] [ thread ]