Github user srowen commented on a diff in the pull request:
https://github.com/apache/spark/pull/5033#discussion_r26448937
--- Diff: docs/mllib-optimization.md ---
@@ -203,6 +203,9 @@ regularization, as well as L2 regularizer.
recommended.
* `maxNumIterations` is the maximal number of iterations that L-BFGS can
be run.
* `regParam` is the regularization parameter when using regularization.
+* `convergenceTol` designates the end of iterations based on relative
comparison
--- End diff --
If linking to Breeze, would be better to link straight to its LBFGS.
However I don't know if we want to promise that implementation detail.
Should we also be updating the javadoc for `LBFGS.scala` and other similar
places where this param is documented?
I'm not sure this text helps someone understand what the value is or how to
set it -- what are adjusted values? You're right, it's fairly opaque beyond
being something relative (rather than something that might need to change with
the input). The thing to know is that lower values mean more iterations and the
value must be nonnegative. So how about something more like:
Controls how much relative change is still allowed when L-BFGS is
considered to converge. Must be nonnegative. Lower values are less tolerant and
therefore generally cause more iterations to be run.
---
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]
