srowen commented on issue #24470: [SPARK-27577][MLlib] Correct thresholds
downsampled in BinaryClassificationMetrics
URL: https://github.com/apache/spark/pull/24470#issuecomment-490086015
Merged to master/2.4/2.3
This is an
srowen commented on issue #24470: [SPARK-27577][MLlib] Correct thresholds
downsampled in BinaryClassificationMetrics
URL: https://github.com/apache/spark/pull/24470#issuecomment-489824722
No, I leave these open for a day or two to make sure there aren't more
comments.
srowen commented on issue #24470: [SPARK-27577][MLlib] Correct thresholds
downsampled in BinaryClassificationMetrics
URL: https://github.com/apache/spark/pull/24470#issuecomment-489433166
Ah OK I agree with you, I see the argument now. It comes from the fact that
the scores are sorted
srowen commented on issue #24470: [SPARK-27577][MLlib] Correct thresholds
downsampled in BinaryClassificationMetrics
URL: https://github.com/apache/spark/pull/24470#issuecomment-489326110
I still don't see the argument that the first or last is better. They are
simply the endpoints of the
srowen commented on issue #24470: [SPARK-27577][MLlib] Correct thresholds
downsampled in BinaryClassificationMetrics
URL: https://github.com/apache/spark/pull/24470#issuecomment-489279970
This doesn't look like a bug. I can't understand the argument in the JIRA
why the last vs first
