[ https://issues.apache.org/jira/browse/MAHOUT792?page=com.atlassian.jira.plugin.system.issuetabpanels:commenttabpanel&focusedCommentId=13092099#comment13092099
]
Ted Dunning commented on MAHOUT792:

{quote}
But my response has always been, if you have input thinner than projection, then why even
use projection?
{quote}
Because it is cheaper to process a thin dense matrix than a wide sparse one.
Likewise, even if B is bigger than A, it can be much cheaper to compute the SVD of B than
of A if only because the Cholesky trick works on the skinny dense case. If the Cholesky decomposition
loses some accuracy then a QR or LQ decomposition could be used the same way.
> Add new stochastic decomposition code
> 
>
> Key: MAHOUT792
> URL: https://issues.apache.org/jira/browse/MAHOUT792
> Project: Mahout
> Issue Type: New Feature
> Reporter: Ted Dunning
> Attachments: MAHOUT792.patch, MAHOUT792.patch, sd2.pdf
>
>
> I have figured out some simplification for our SSVD algorithms. This eliminates the
QR decomposition and makes life easier.
> I will produce a patch that contains the following:
>  a CholeskyDecomposition implementation that does pivoting (and thus rankrevealing)
or not. This should actually be useful for solution of large outofcore least squares problems.
>  an inmemory SSVD implementation that should work for matrices up to about 1/3 of
available memory.
>  an outofcore SSVD threaded implementation that should work for very large matrices.
It should take time about equal to the cost of reading the input matrix 4 times and will
require working disk roughly equal to the size of the input.

This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
