[
https://issues.apache.org/jira/browse/SPARK-21770?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16139364#comment-16139364
]
Weichen Xu commented on SPARK-21770:
------------------------------------
Hmm... `normalizeToProbabilitiesInPlace` is only effective in the case when we
use class instance counts as `rawPrediction`.
So, I guess the meaning is, when we have empty instance set, all the counts ==
0, in this case, how to assume the probability for each class ? The uniform
distribution is a reasonable assumption I think.
But I am not sure whether this change will cause issues in other places.
> ProbabilisticClassificationModel: Improve normalization of all-zero raw
> predictions
> -----------------------------------------------------------------------------------
>
> Key: SPARK-21770
> URL: https://issues.apache.org/jira/browse/SPARK-21770
> Project: Spark
> Issue Type: Improvement
> Components: ML
> Affects Versions: 2.3.0
> Reporter: Siddharth Murching
> Priority: Minor
>
> Given an n-element raw prediction vector of all-zeros,
> ProbabilisticClassifierModel.normalizeToProbabilitiesInPlace() should output
> a probability vector of all-equal 1/n entries
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
