[
https://issues.apache.org/jira/browse/MAHOUT-642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13012978#comment-13012978
]
Sebastian Schelter commented on MAHOUT-642:
-------------------------------------------
Looked at the formula again, it shouldn't matter for the final score, but it is
inconsistent with the documentation and Ted's blog entry, so we should change
it.
> Wrong parameter order in LoglikelihoodSimilarity and
> DistributedLoglikelihoodVectorSimilarity
> ---------------------------------------------------------------------------------------------
>
> Key: MAHOUT-642
> URL: https://issues.apache.org/jira/browse/MAHOUT-642
> Project: Mahout
> Issue Type: Bug
> Affects Versions: 0.5
> Reporter: Sebastian Schelter
> Assignee: Sebastian Schelter
>
> org.apache.mahout.math.stats.LogLikelihood.logLikelihoodRatio expects the
> following counts:
> * A and B together (k_11)
> * B without A (k_12)
> * A without B (k_21)
> * Neither A nor B (k_22)
> It seems to me that in
> org.apache.mahout.cf.taste.impl.similarity.LogLikelihoodSimilarity and
> org.apache.mahout.math.hadoop.similarity.vector.DistributedLoglikelihoodVectorSimilarity
> the counts of k_12 and k_21 are given in the wrong order (B without A should
> come before A without B)
>
> Can someone confirm that?
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
