[SPARK-7612] Use BLAS in naive Bayes training

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.4.0
• Fix Version/s: 1.4.0
• Component/s: MLlib
• Target Version/s: 1.4.0
• Assignee: Xiangrui Meng
• Reporter: Xiangrui Meng
• Created: 13/May/15 19:40
• Updated: 15/May/15 00:17
• Resolved: 14/May/15 04:28

Description

Similar to the changes we made to k-means, replacing the implementation by MLlib's BLAS wrappers gives us better control on the performance.

Issue Links

• is related to: SPARK-7652 Performance regression in naive Bayes prediction
• links to: [Github] Pull Request #6128 (mengxr)