[
https://issues.apache.org/jira/browse/MAHOUT-796?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13094018#comment-13094018
]
Dmitriy Lyubimov commented on MAHOUT-796:
-----------------------------------------
AB' is a heavy multiplication of course.
I don't want to use standard multiplication because
-- i want to be doing more things in reducer
-- need custom grouping/sorting to ensure output is partitioned the same way as
A splits
At this point it seems that the best strategy is just to preload entire A block
into memory as a (sparse) matrix and open B' stream as a side file and hope it
is not going to generate too much flood i/o. I don't know a workaround for it
anyway since whatever blocking scheme is used, we need cartesian products from
both matrix inputs and that will cause i/o and i don't think there's any clever
collocation trick to be had there
> Modified power iterations in existing SSVD code
> -----------------------------------------------
>
> Key: MAHOUT-796
> URL: https://issues.apache.org/jira/browse/MAHOUT-796
> Project: Mahout
> Issue Type: Improvement
> Components: Math
> Affects Versions: 0.5
> Reporter: Dmitriy Lyubimov
> Assignee: Dmitriy Lyubimov
> Labels: SSVD
> Fix For: 0.6
>
>
> Nathan Halko contacted me and pointed out importance of availability of power
> iterations and their significant effect on accuracy of smaller eigenvalues
> and noise attenuation.
> Essentially, we would like to introduce yet another job parameter, q, that
> governs amount of optional power iterations. The suggestion how to modify the
> algorithm is outlined here :
> https://github.com/dlyubimov/ssvd-lsi/wiki/Power-iterations-scratchpad .
> Note that it is different from original power iterations formula in the paper
> in the sense that additional orthogonalization performed after each
> iteration. Nathan points out that that improves errors in smaller eigenvalues
> a lot (If i interpret it right).
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
