[SPARK-7134] Add regParam and featureScaling options to Logistic regression 'train' methods

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Won't Fix
• Component/s: MLlib
• Assignee: Unassigned
• Reporter: Rakesh Chalasani
• Created: 24/Apr/15 20:29
• Updated: 24/Apr/15 22:41
• Resolved: 24/Apr/15 22:41

Description

LogisticRegressionWithSGD is private to mllib and is accessible using "train" factory methods. But these train methods do not have provision to set regularization parameter or feature scaling. On the other hand, LogisticRegressionWithLBFGS is public, the user can set these parameters.

Add a more general train method that allow setting there parameters on the model before training.

Issue Links

• relates to: SPARK-6682 Deprecate static train and use builder instead for Scala/Java
• links to: [Github] Pull Request #5692 (rakeshchalasani)