[ https://issues.apache.org/jira/browse/SPARK-5832?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14326323#comment-14326323 ]
Xiangrui Meng commented on SPARK-5832: -------------------------------------- [~viirya] Thanks for sharing the details! I'm a little worried about the `N^2` complexity. Most, if not all, algorithms in MLlib have linear dependency on the number of instances. Having N^2 complexity means that if a single machine can solve a problem of 100,000 instances, we need a 100-node cluster to solve a problem of 1,000,000 instances, not counting the communication overhead. We had some discussion about k-medoids (SPARK-5056) and DBSCAN (SPARK-5226), both of which have N^2 complexity. I'm not sure whether we should include those algorithms in MLlib. For now, it might be good if we can maintain the implementation outside Spark as a third-party package (http://spark-packages.org/). > Add Affinity Propagation clustering algorithm > --------------------------------------------- > > Key: SPARK-5832 > URL: https://issues.apache.org/jira/browse/SPARK-5832 > Project: Spark > Issue Type: New Feature > Components: MLlib > Reporter: Liang-Chi Hsieh > Assignee: Liang-Chi Hsieh > -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org