[
https://issues.apache.org/jira/browse/MAHOUT-376?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12917394#action_12917394
]
Ted Dunning commented on MAHOUT-376:
------------------------------------
I think that what you are saying is that some of the A_i blocks that make up A
may be rank deficient.
That is definitely a risk if the number of rows in A_i is small compared to the
size of Q_i. If the number of rows of A_i is much larger than the size of
Q_i, then that is very, very unlikely.
Regardless, we will still have Q*Q = I and Q_i will still span the projection
of A_i even if A_i and thus A_i \Omega is rank deficient. Thus Q will still
span A \Omega. This gives us Q Q* A \approx A as desired.
So, my feeling is that what you say is correct, but it isn't a problem.
> Implement Map-reduce version of stochastic SVD
> ----------------------------------------------
>
> Key: MAHOUT-376
> URL: https://issues.apache.org/jira/browse/MAHOUT-376
> Project: Mahout
> Issue Type: Improvement
> Components: Math
> Reporter: Ted Dunning
> Assignee: Ted Dunning
> Fix For: 0.5
>
> Attachments: MAHOUT-376.patch, sd-bib.bib, sd.pdf, sd.tex, Stochastic
> SVD using eigensolver trick.pdf
>
>
> See attached pdf for outline of proposed method.
> All comments are welcome.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
