This preview shows page 1. Sign up to view the full content.
Unformatted text preview: Hessenberg Similarity One of the most important condensed forms in numerical linear algebra is the Hessenberg matrix. An upper Hessenberg matrix is almost upper triangular, having zeros below the first subdiagonal. Every square matrix is orthogonally similar to a Hessenberg matrix. There are methods for reducing A ∈ R n × n to H = V t AV based on GramSchmidt (the Arnoldi method ) and on Householder reflectors (a slight modification to the Householder QR method). These methods are about twice as expensive as their QR analogs (the Householder Hessenberg reduction requires about 10 3 n 3 flops), but the resulting matrix is similar to A . There are several reasons Hessenberg matrices are important, but I would suggest that the following is the most fundamental: Suppose that you wanted to solve ( A sI ) x = b for m different values of s . Gaussian elimination (or QR ) would require a new factorization for each s , resulting in a cost of O( mn 3 ) flops. But ( A sI ) x = b ⇒ V t ( A...
View
Full
Document
This note was uploaded on 12/18/2010 for the course PHYS 5073 taught by Professor Mark during the Fall '10 term at Arkansas.
 Fall '10
 MARK

Click to edit the document details