Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/2207#issuecomment-53938249
@dbtsai The scaladoc for `Updater` says the return value is:
```
A tuple of 2 elements. The first element is a column matrix containing
updated weights,
and the second element is the regularization value computed using updated
weights.
```
Looking at `GradientDescent` that seems like how it is intended to work,
since the loss is computed from the weights and regularization term from the
previous iteration. Wouldn't this put it out of sync by 1 iteration? You would
have theta_i weights but a regularization term computed over theta_{i-1}.
I might be overlooking something and you know the L-BFGS implementation
better but are you sure this should change for both?
---
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]
