You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Xiangrui Meng (JIRA)" <ji...@apache.org> on 2015/11/19 01:22:10 UTC
[jira] [Commented] (SPARK-11543) spark.ml LogisticRegressionModel
should prefer thresholds over threshold
[ https://issues.apache.org/jira/browse/SPARK-11543?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15012450#comment-15012450 ]
Xiangrui Meng commented on SPARK-11543:
---------------------------------------
ml.LogisticRegression does not support multiclass yet. Shall we update the current implementation to ignore `thresholds`? In the next release, we can ignore `threshold` if `thresholds` are set.
> spark.ml LogisticRegressionModel should prefer thresholds over threshold
> ------------------------------------------------------------------------
>
> Key: SPARK-11543
> URL: https://issues.apache.org/jira/browse/SPARK-11543
> Project: Spark
> Issue Type: Improvement
> Components: ML
> Reporter: Joseph K. Bradley
> Assignee: Xiangrui Meng
>
> The current behavior was defined in [SPARK-9847], where the two Params must match if both are set. This causes problems for model persistence, for which it is easier (for backwards compatibility) to save a single ParamMap instead of 2 (user-set + default). This can be fixed by changing LogisticRegression to use {{thresholds}} if set (ignoring {{threshold}}), and to otherwise use {{threshold}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org