Run KMeans with fixed seed

Question

I want to run the KMeans algorithm of MLLIB (Apache Spark), but with reproducible results.

Is it possible to run KMeans of MLLIB (Apache Spark) with fixed seed? How?

Thanks and regards,

I only used ELKI, it has an `kmeans.seed` parameter and is very fast. — Has QUIT--Anony-Mousse, Dec 25 '14 at 19:08
Thanks a lot, Anony-Mousse. However, in this case, I am constrained to use the algorithm from Apache Spark only :( — learning_spark, Dec 27 '14 at 01:22

score 0 · Answer 1 · answered Dec 23 '14 at 15:59

0

As of Spark 1.2, it appears that this is not possible. It generates a seed internally.

answered Dec 23 '14 at 15:59

Sean Owen

Thanks a lot, Sean. I was under the same impression. – learning_spark Dec 27 '14 at 01:22

score 0 · Answer 2 · answered Feb 21 '16 at 12:03

0

Set seed for KMeans is availble in Spark MLLib since version 1.4.0 See doc library here

answered Feb 21 '16 at 12:03

m.piunti

2 Answers2