[
https://issues.apache.org/jira/browse/SPARK-9956?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Joseph K. Bradley updated SPARK-9956:
-------------------------------------
Issue Type: Bug (was: Improvement)
> Spark ML trees and ensembles fail for categorical features with 1 category
> --------------------------------------------------------------------------
>
> Key: SPARK-9956
> URL: https://issues.apache.org/jira/browse/SPARK-9956
> Project: Spark
> Issue Type: Bug
> Components: ML
> Reporter: Joseph K. Bradley
> Assignee: Joseph K. Bradley
>
> Spark ML trees and ensembles can be given metadata (e.g., from VectorIndexer)
> indicating that a certain feature is categorical with a single possible
> value. This causes learning to fail.
> Proposal: For now, fix this by making sure the algorithm still runs with such
> a feature, and remove the checks (which currently cause failure). In the
> future, we can filter out these features to improve performance when there
> are many useless features.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
