#learning-curves

Question 1: Single choice

A linear model predicting song popularity gives learning curves in which the training and validation loss level off close to each other, the validation loss never rises, and both stay well above the loss that larger networks reach. What should you try next?

It underfits, so add capacity with wider or deeper layers1
It is overfitting, so add an early-stopping callback2
It is overfitting, so remove units from the model3
It fits well, because the two learning curves stay close together4

Timed test