loss function name consistency in gradient boosting #3481
Description
It would be nice if the loss option in gradient boosting could be more consistent with the one in SGD. Rather than deprecating names in gradient boosting, I suggest adding aliases.