Model selection

It’s explained but I don’t understand why choosing the Jtest function with the minimum value doesn’t work.

In the same way, I don’t understand why considering the CV set solves the problem:

please be patient with me and thank you as always.

We use the test set error as the metric because the goal is to make predictions on new data that wasn’t used in training.

We use the CV set to adjust the polynomial in order to avoid overfitting the training set.

1 Like

Hello @TMosh,
Does regularization term included in training each order of polynomial.

All the features are included in regularization.