[SPARK-37959] Fix the UT of checking norm in KMeans & BiKMeans

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 3.3.0
• Fix Version/s: 3.2.1, 3.3.0
• Component/s: ML
• Assignee: Ruifeng Zheng
• Reporter: Ruifeng Zheng
• Created: 19/Jan/22 08:20
• Updated: 19/Jan/22 17:19
• Resolved: 19/Jan/22 17:19

Description

In KMeansSuite and BisectingKMeansSuite, there are some unused lines:

 

model1.clusterCenters.forall(Vectors.norm(_, 2) == 1.0 

 

For cosine distance, the norm of centering vector should be 1, so the norm checking is meaningful;

For euclidean distance, the norm checking is meaningless;

 

Issue Links

• links to: [Github] Pull Request #35247 (zhengruifeng)