[ 
https://issues.apache.org/jira/browse/MAHOUT-630?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13009684#comment-13009684
 ] 

Ted Dunning commented on MAHOUT-630:
------------------------------------

Looks fine.  It uses the less accurate formula for standard deviation, but that 
isn't a big deal in most applications.

To do better would require adapting Welford's method which would require some 
thought (which this implementation does not).

There is also the question of progressively accumulating error.  Again, I don't 
think that the applications of this version
require that be addressed.


> Small bug in WeightedRunningAverage; add WeightedRunningAverageAndStdDev
> ------------------------------------------------------------------------
>
>                 Key: MAHOUT-630
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-630
>             Project: Mahout
>          Issue Type: Bug
>          Components: Collaborative Filtering
>    Affects Versions: 0.4
>            Reporter: Sean Owen
>            Assignee: Sean Owen
>            Priority: Minor
>              Labels: average, standard-deviation, weighted
>             Fix For: 0.5
>
>         Attachments: MAHOUT-630.patch
>
>
> There's a problem in the WeightedRunningAverage handling of a new weighted 
> datum. While looking at that I said, hey, we could use 
> WeightedRunningAverageAndStdDev to go with it. And some tests.

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

Reply via email to