[jira] [Commented] (SPARK-21061) GMM Error : Matrix is not symmetric

["Sean Owen (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-21061?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16046631#comment-16046631 ] 

Sean Owen commented on SPARK-21061:
-----------------------------------

I get a slightly different error: NotConvergedException. The problem is that some inputs to eigSym are NaN. After tracing it back a while it's because math.exp(logpdf(x)) in pdf(x) becomes Infinite, and that's likely because sigma is near-singular, so rootSigmaInv has huge values. I stopped before figuring out exactly what's wrong there, but that's pretty much the root.

For your example, I think it's because you have too many cluster centers for this data. Try k=5. It shouldn't cause an error, but, I think you'll find you don't want nearly so many clusters anyway.

> GMM Error : Matrix is not symmetric
> -----------------------------------
>
>                 Key: SPARK-21061
>                 URL: https://issues.apache.org/jira/browse/SPARK-21061
>             Project: Spark
>          Issue Type: Bug
>          Components: MLlib
>    Affects Versions: 1.6.0
>         Environment: Ubuntu 14.04, Windows
>            Reporter: Sourav Mahato
>
> I am trying to implement Gaussian Mixture Model(GMM) using Apache Spark MLlib (1.6.0). But I am getting the following error.
> https://stackoverflow.com/questions/44494599/spark-mllib-gmm-error-matrix-is-not-symmetric 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org