Github user WeichenXu123 commented on the issue: https://github.com/apache/spark/pull/19565 Yes I think when dataset is large enough, using the same `miniBatchFraction`, the result RDD size of "filter before sample" and "filter after sample" will be asymptotically equal, no matter how many empty elements in dataset. (correct me if I am wrong, I am a little confused about "miniBatchFraction should be adjusted proportionally", if adjusted, then asymptotically equality is broken?) If So, does it really effect the algo ?
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org