When you compute the test error for ridge regression, is it typically computed with the regularization term in it?
Asked
Active
Viewed 146 times
2
-
3It's not clear to me what you're asking. Do you mean whether the test performance is calculated using the loss function that includes the regularization penalty? Also, what does "subtle" mean here? – Acccumulation Mar 26 '21 at 18:36
-
+1 There should be a longer answer than this comment, but you should use the original loss function, not the loss function with the regularization parameter. – Dave Mar 26 '21 at 18:42
-
@Acccumulation Yes to the first question. "Subtle" should have been "subtlety" but now I am wondering why I even said that in the first place. I'm going to remove it – roulette01 Mar 26 '21 at 18:53
-
3Could you please explain what you mean by the "test error" and how you compute it? I suspect this clarification by itself might answer your question. – whuber Mar 26 '21 at 18:54