[
https://issues.apache.org/jira/browse/SPARK-16653?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15386028#comment-15386028
]
Apache Spark commented on SPARK-16653:
--------------------------------------
User 'WeichenXu123' has created a pull request for this issue:
https://github.com/apache/spark/pull/14286
> Make convergence tolerance param in ANN default value consistent with other
> algorithm using LBFGS
> -------------------------------------------------------------------------------------------------
>
> Key: SPARK-16653
> URL: https://issues.apache.org/jira/browse/SPARK-16653
> Project: Spark
> Issue Type: Improvement
> Components: ML, Optimizer
> Reporter: Weichen Xu
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> The default value of convergence tolerance param in ANN is 1e-4,
> but other algorithm (such as LinearRegression, LogisticRegression, and so on)
> using LBFGS optimizers this param's default values are all 1e-6.
> I think make them the same will be better.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
