[
https://issues.apache.org/jira/browse/MAHOUT-1190?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13628314#comment-13628314
]
Dan Filimon commented on MAHOUT-1190:
-------------------------------------
Here's an updated prototype patch for review:
https://reviews.apache.org/r/10409/
Robin, Sequential is really two sorted arrays so performance wise it's about
the same as a map implemented with a binary search tree. Granted we don't use
those, but it's still a reasonable way of building a map IMHO. Anyway, SASV,
being implemented as arrays should also be more cache-friendly (if not abused
:).
Whether SASV should be entirely removed could should be discussed on the list.
Start a thread maybe?
In any case, I'll add a test for sure (but it's ridiculously late here now).
> SequentialAccessSparseVector function assignment is very slow
> -------------------------------------------------------------
>
> Key: MAHOUT-1190
> URL: https://issues.apache.org/jira/browse/MAHOUT-1190
> Project: Mahout
> Issue Type: Bug
> Reporter: Dan Filimon
>
> Currently when calling .assign() on a SASV with another vector and a custom
> function, it will iterate through it and assign every single entry while also
> referring it by index.
> This makes the process *hugely* expensive. (on a run of BallKMeans on the 20
> newsgroups data set, profiling reveals that 92% of the runtime was spent
> updating assigning the vectors).
> Here's a prototype patch:
> https://github.com/dfilimon/mahout/commit/63998d82bb750150a6ae09052dadf6c326c62d3d
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
