[SPARK-15986] SVMWithSGD requires LabeledPoint of Regression

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: 2.0.0
• Component/s: ML, MLlib

Environment

Spark 2.0 Snapshot

• Assignee: Unassigned
• Reporter: Hayri Volkan Agun
• Created: 16/Jun/16 09:29
• Updated: 16/Jun/16 17:39
• Resolved: 16/Jun/16 17:39

Description

SVMWithSGD uses the LabelledPoint in mllib-regression however a classification routine that extends ml's Predictor class uses LabelledPoint of ml.feature. So basically any rotutine, for ex: a classifier in ml pipeline cannot uses SVMWithSGD. This consistency problem can be solved by removing LabelledPoint in regression. Two different labelled points (one uses linalg.Vector, and other ml.linalg.Vector) is also confusing.

Issue Links

• Is contained by: SPARK-15944 Make spark.ml package backward compatible with spark.mllib vectors