You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@mahout.apache.org by Zhongming cai <ca...@yahoo.co.uk> on 2010/04/22 12:45:13 UTC

SVD in mahout

Hi,

Does the SVD decompsition (Lanczos algorithm) implemented in Mahout only return one eigenvectors, not both left and right eigenvectors ? 

Thanks,

Zhongming

Re: SVD in mahout

Posted by Jake Mannix <ja...@gmail.com>.

Both svd implementations in mahout return only the right singular vectors
(the vectors with dimensionality equal to the number of columns of the input
matrix).

Recovering the left singular vectors is just a matter of multiplying your
input matrix by your right singular vectors and the inverse singular values,
one more pass over the data.

-jake

On Apr 22, 2010 3:45 AM, "Zhongming cai" <ca...@yahoo.co.uk> wrote:

Hi,

Does the SVD decompsition (Lanczos algorithm) implemented in Mahout only
return one eigenvectors, not both left and right eigenvectors ?

Thanks,

Zhongming