Github user dbtsai commented on the pull request:
https://github.com/apache/spark/pull/7099#issuecomment-119107982
In general, the losses returned from optimizer are for debugging purpose,
and users can see how well the optimizer works. The reason is the real
statistical metrics users care about are RSE or loglikihood which can be used
for computing other metrics don't involve the regTerm. However, the returned
loss like the following has regTerm, which is not the real RSE. Also, no
easy/cheap way to convert back to the loss which standardization. As a result,
we can have the lossHistroy for debugging purpose, and have another pass of
data to compute RSE and loglikihood.
```latex
L = 1/2n ||\sum_i w_i(x_i - \bar{x_i}) / \hat{x_i} - (y - \bar{y}) /
\hat{y}||^2 + regTerms
```
---
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]
