[SPARK-30380] Refactor RandomForest.findSplits

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 3.0.0
• Fix Version/s: 3.0.0
• Component/s: ML
• Assignee: Ruifeng Zheng
• Reporter: Ruifeng Zheng
• Created: 29/Dec/19 08:26
• Updated: 31/Dec/19 06:16
• Resolved: 31/Dec/19 06:16

Description

Current impl of RandomForest.findSplits uses groupByKey to collect non-zero values for each feature, so it is quite dangerous.

 

 

Issue Links

• links to: GitHub Pull Request #27040