Github user yanboliang commented on the issue:
https://github.com/apache/spark/pull/19020
@sethah To the issue that whether huber linear regression share codebase
with ```LinearRegression```, we have discussion at
[JIRA](https://issues.apache.org/jira/browse/SPARK-3181). At last @dbtsai and I
reached an agreement to combine them in a single class. Actually in sklearn,
there are separate
[```HuberRegressor```](http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.HuberRegressor.html)
and
[```SGDRegressor```](http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.SGDRegressor.html)
support to train linear regression with huber loss. Also, in this
[paper](http://statweb.stanford.edu/~owen/reports/hhu.pdf), huber regression is
a robust hybrid of lasso and ridge regression, so I think we can combine them
together.
@jkbradley @MLnick @WeichenXu123 @hhbyyh What's your opinion? Thanks.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
