[
https://issues.apache.org/jira/browse/MAHOUT-376?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12920267#action_12920267
]
Dmitriy Lyubimov commented on MAHOUT-376:
-----------------------------------------
My quick scan of the literature seems to indicate that best promise for full
scale QR over map/reduce is row-wise givens with top k+p rows of the blocks
being processed in combiners and reducers for the reminder of the first step of
givens QR step. this can be done in the first step of the job, but i am yet to
figure how to combine and reduce 'rho' intermediate results into final Q.
[Golub, Van Loan 3rd ed] p. $5.2.3
> 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, Modified stochastic svd algorithm for
> mapreduce.pdf, 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.
