Code QuizIntermediate

Overfitting Gap From Curves

Computing the train/validation gap from learning curves in the wrong direction.

Codepython

final_train_score = 0.97
final_val_score = 0.79

gap = final_val_score - final_train_score
if gap > 0.1:
    print('Overfitting: large train-val gap')
else:
    print('No significant overfitting')

Why does this never detect overfitting?