[SPARK-11534] MLLib uniformVectorRDD wrong data generation

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Not A Problem
• Affects Version/s: 1.5.1
• Component/s: MLlib
• Labels: mllib

Environment

Spark 1.6.0-SNAPSHOT standalone on centos 6.6
Single node VM machine, 8 cores 16 GB memory

• Assignee: Unassigned
• Reporter: Sandeep Pal
• Created: 05/Nov/15 19:19
• Updated: 05/Nov/15 19:36
• Resolved: 05/Nov/15 19:36

Description

According to the definition of uniformVectorRDD, it is supposed to generate the uniformly distributed data with mean ~0.0 and standard deviation ~1.
[Definition: Generates an RDD comprised of i.i.d. samples from the uniform distribution U(0.0, 1.0).]
But it is generating the data with mean ~0.5 and sd of ~0.28.

PS: uniformRDD is working correctly as per description with uniform distribution U(0.0, 1.0) and uniformVectorRDD should be generating data with same distribution.