srowen commented on a change in pull request #32734:
URL: https://github.com/apache/spark/pull/32734#discussion_r646042859
##########
File path:
mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/RowMatrix.scala
##########
@@ -421,6 +421,11 @@ class RowMatrix @Since("1.0.0") (
}
}
+ // The matrix is sparse, if all the rows has sparsity more than 0.5.
+ private def isSparseMatrix: Boolean = {
+ rows.filter(row => row.sparsity() < 0.5).isEmpty()
Review comment:
Do you have insight into why the dense computation is more accurate? it
shouldn't be in theory. I am OK with the change if it's a band-aid that
improves results, but not quite sure why this is happening.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
