SVD - Singular Value Decomposition

Hello there,

I'm fairly new to matlab and have run into a little confusion when using the SVD function. I note from the help pages that the SVD function will fail to converge after 75 iterations of the QR algorithm. Now I need to assess the suitability of the SVD for my intended application, but I don't understand what condition will cause the SVD algorithm to perform more than 75 iterations, i.e. in my matlab code I'm thinking of using [U S V] = svd(x), what conditions would need to be present in x to cause more than 75 QR iterations?

Thanks.

Answer 1

[John D'Errico]

I don't think you need to worry about that problem terribly much. I've never seen svd fail (or even heard of it failing from this circumstance), and in general the svd algorithm, while not blindingly fast for large matrices, is a robust one.