Good morning !
Once we performed ridge regression for a given range of lambdas, how do we know which optimal weights we should use for our "final" model ?
Do we decide based on the minimum test dataset RMSE we obtained over this range ?
I'm mostly referring to the graph in part 3 of exercise session 3. In this example, if I want to minimize the test RMSE, I would choose \(\lambda \approx 0.1\), even if the RMSE for the training test is rising.
Is it the right approach ?
Thank you very much !
Ridge regression: the right choice for lambda
Good morning !
Once we performed ridge regression for a given range of lambdas, how do we know which optimal weights we should use for our "final" model ?
Do we decide based on the minimum test dataset RMSE we obtained over this range ?
I'm mostly referring to the graph in part 3 of exercise session 3. In this example, if I want to minimize the test RMSE, I would choose
\(\lambda \approx 0.1\), even if the RMSE for the training test is rising.
Is it the right approach ?
Thank you very much !
Philippine
1
I'm sorry, my post is useless since it's the topic of this week
Add comment