[
https://issues.apache.org/jira/browse/SPARK-6000?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16014046#comment-16014046
]
Nick Pentreath commented on SPARK-6000:
---------------------------------------
Even though SPARK-14174 is later - it seems there is more discussion and a
related PR there. I will close this as duplicate?
> Batch K-Means clusters should support "mini-batch" updates
> ----------------------------------------------------------
>
> Key: SPARK-6000
> URL: https://issues.apache.org/jira/browse/SPARK-6000
> Project: Spark
> Issue Type: Improvement
> Components: MLlib
> Affects Versions: 1.2.1
> Reporter: Derrick Burns
> Priority: Minor
>
> One of the ways of improving the performance of the K-means clustering
> algorithm is to sample the points on each round of the Lloyd's algorithm and
> to only use those samples to update the cluster centers. (Note that this is
> similar to the update algorithm of streaming K-means.) The Spark K-Means
> clusterer should support the mini-batch algorithm for large data sets.
> The K-Means implementation at
> https://github.com/derrickburns/generalized-kmeans-clustering supports the
> mini-batch algorithm.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
