You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mahout.apache.org by "Peter Wippermann (JIRA)" <ji...@apache.org> on 2009/09/02 05:18:36 UTC

[jira] Created: (MAHOUT-169) Wrong implementation of the mapper in the canopy clusterer?

Wrong implementation of the mapper in the canopy clusterer?
-----------------------------------------------------------

                 Key: MAHOUT-169
                 URL: https://issues.apache.org/jira/browse/MAHOUT-169
             Project: Mahout
          Issue Type: Question
          Components: Clustering
            Reporter: Peter Wippermann


The class
org.apache.mahout.clustering.canopy.CanopyMapper
takes use of the method "Canopy.addPointToCanopies(point, canopies)".

The documentation of the Canopy-Class says, that this function is used by the reducer - no mention of the mapper here.
It furthermore says, that the mapper would use "emitPointToNewCanopies" or "emitPointToExistingCanopies", which however he does not.

I'm just trying to figure out, how the canopy clustering in general works. But this is confusing to me. So if it is a bug, please fix it. If it is not, I'd be very happy, if you could explain me, why :-)

Furthermore I'm wondering why the syntheticcontrol clustering example will only find ONE SINGLE CLUSTER. Do I mix thinks up here???

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Resolved: (MAHOUT-169) Wrong implementation of the mapper in the canopy clusterer?

Posted by "Peter Wippermann (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/MAHOUT-169?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Peter Wippermann resolved MAHOUT-169.
-------------------------------------

    Resolution: Fixed

Maybe I viewed the wrong Mapper/Reducer.
The correct ones, where the implementation is actually correct, are in the package
org.apache.mahout.clustering.canopy;
in the mahout-core

> Wrong implementation of the mapper in the canopy clusterer?
> -----------------------------------------------------------
>
>                 Key: MAHOUT-169
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-169
>             Project: Mahout
>          Issue Type: Question
>          Components: Clustering
>            Reporter: Peter Wippermann
>   Original Estimate: 0.5h
>  Remaining Estimate: 0.5h
>
> The class
> org.apache.mahout.clustering.canopy.CanopyMapper
> takes use of the method "Canopy.addPointToCanopies(point, canopies)".
> The documentation of the Canopy-Class says, that this function is used by the reducer - no mention of the mapper here.
> It furthermore says, that the mapper would use "emitPointToNewCanopies" or "emitPointToExistingCanopies", which however he does not.
> I'm just trying to figure out, how the canopy clustering in general works. But this is confusing to me. So if it is a bug, please fix it. If it is not, I'd be very happy, if you could explain me, why :-)
> Furthermore I'm wondering why the syntheticcontrol clustering example will only find ONE SINGLE CLUSTER. Do I mix thinks up here???

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.