Github user mengxr commented on a diff in the pull request:
https://github.com/apache/spark/pull/7884#discussion_r38595939
--- Diff:
mllib/src/main/scala/org/apache/spark/ml/classification/LogisticRegression.scala
---
@@ -803,13 +829,17 @@ private class LogisticAggregator(
this
}
- def count: Long = totalCnt
-
- def loss: Double = lossSum / totalCnt
+ def loss: Double = {
+ require(weightSum >= 1.0, s"The effective number of samples should be
" +
--- End diff --
We shouldn't define variance that way. `weight` is not necessarily
representing how many instances. For example, the following definition would be
compatible with the case when `weight === 1.0`:
~~~
Var[X] = E[X^2] - E[X]^2
~ (\sum_i w_i x_i^2 / sum_i w_i) - (\sum_i w_i x_i / sum_i w_i)^2
~~~
Though we still need to figure out how to make it compatible with unbiased
version and implement it in a numerical stable way.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
