Re: eigenvalue/eigenvector question again
From: Gordo (gordo20878_at_hotmail.com)
Date: 02/26/05
- Next message: John: "Re: eigenvalue/eigenvector question again"
- Previous message: Anon: "Re: Signed NaN in Kahan's CORNER(x+iy)"
- In reply to: John: "eigenvalue/eigenvector question again"
- Next in thread: John: "Re: eigenvalue/eigenvector question again"
- Reply: John: "Re: eigenvalue/eigenvector question again"
- Messages sorted by: [ date ] [ thread ]
Date: Sat, 26 Feb 2005 13:03:19 -0500
John wrote:
> I create a matrix using the following multiplication
> A = UDU'
> where U'U = I
> det U = 1
> now how do I recover U and D?
> Can U and D be exactly recovered at all?
> --j
You have left out a few pertinent details, but you are doing a coordinate
transformation to/from principal axes. In linear algebra, U usually denotes
an upper triangle, but not in this case. Your U is an orthogonal matrix
whose columns are the normalized eigenvectors of A (length=1). D is a
diagonal matrix of eigenvalues of A. The eigenvectors are the principal
axes; i.e., each column of U gives the direction cosines of a principal
axis. (I have given geometric interpretations in case you are in 2-D or
3-D.)
Gordon Everstine
- Next message: John: "Re: eigenvalue/eigenvector question again"
- Previous message: Anon: "Re: Signed NaN in Kahan's CORNER(x+iy)"
- In reply to: John: "eigenvalue/eigenvector question again"
- Next in thread: John: "Re: eigenvalue/eigenvector question again"
- Reply: John: "Re: eigenvalue/eigenvector question again"
- Messages sorted by: [ date ] [ thread ]