[Fwd: SVD algorithm]

[Ron Boisvert <boisvert@nist.gov>]

-------- Original Message --------
Subject: 	SVD algorithm
Date: 	Fri, 14 Nov 2008 14:28:51 -0500
From: 	Keir Mierle <mierle@gmail.com>
Reply-To: 	mierle@gmail.com
To: 	boisvert@nist.gov



Which SVD algorithm does JAMA implement? For example, Golub-Reisch, etc. 
Also, is it hard to extend the algorithm to compute the full U matrix (m 
x m) rather than only the rectangular (m x n)? Many applications in 
computer vision require solving underdetermined homogeneous equations 
(eg Ax = 0, ||x|| = 1 with A mxn with m < n); for that, all of U is 
required to form a basis for the nullspace of the rowspace of A (a wide 
matrix).

Thanks,
Keir