[
https://issues.apache.org/jira/browse/SPARK-3934?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14170220#comment-14170220
]
Sean Owen commented on SPARK-3934:
----------------------------------
Yep that fixes the issue I was seeing. Thanks! I can confirm it did not affect
DecisionTree too, so it seems to match your analysis.
> RandomForest bug in sanity check in DTStatsAggregator
> -----------------------------------------------------
>
> Key: SPARK-3934
> URL: https://issues.apache.org/jira/browse/SPARK-3934
> Project: Spark
> Issue Type: Bug
> Components: MLlib
> Reporter: Joseph K. Bradley
> Assignee: Joseph K. Bradley
>
> When run with a mix of unordered categorical and continuous features, on
> multiclass classification, RandomForest fails. The bug is in the sanity
> checks in getFeatureOffset and getLeftRightFeatureOffsets, which use the
> wrong indices for checking whether features are unordered.
> Proposal: Remove the sanity checks since they are not really needed, and
> since they would require DTStatsAggregator to keep track of an extra set of
> indices (for the feature subset).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
