[jira] Commented: (MAHOUT-173) Implement clustering of massive-domain attributes

["Ted Dunning (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MAHOUT-173?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12796538#action_12796538 ] 

Ted Dunning commented on MAHOUT-173:
------------------------------------


It seems that this algorithm is a combination of a hashed kernel and k-means clustering.

Would it be easiest to implement this using the vectorization algorithms described as part of MAHOUT-228 and then just using the normal k-means algorithm?



> Implement clustering of massive-domain attributes
> -------------------------------------------------
>
>                 Key: MAHOUT-173
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-173
>             Project: Mahout
>          Issue Type: New Feature
>          Components: Clustering
>    Affects Versions: 0.2
>            Reporter: Matias Bjørling
>            Priority: Trivial
>             Fix For: 0.3
>
>   Original Estimate: 30h
>  Remaining Estimate: 30h
>
> Implement the Clustering algorithm described in "A Framework for Clustering Massive-Domain Data Streams" by Chary C. Aggarwal.
> Steps: 
> 1. Implement baseline solution to compare solutions.
> 2. Figure out how to implement the loading of clustering by looking at the k-means implementation.
> 3. Implement Count-Min sketch algorithm for each cluster.
> 4. Find out how to give the user the power to choose the distance function for the input data ( Maybe already possible? )

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