Overfitting Gap

Problem

Return the gap between validation and training loss at the best epoch, i.e. the epoch with the lowest validation loss (ties go to the earliest).

Empty input returns 0.0.

Hint

Evaluate the generalization gap at the best validation epoch, not at the last epoch.

case 1

in: [[0.5,0.4,0.3],[0.6,0.5,0.55]]

expected: 0.09999999999999998

case 2

in: [[0.9,0.7],[0.95,0.6]]

expected: -0.09999999999999998

case 3

in: [[0.2],[0.3]]

expected: 0.09999999999999998

case 4

in: [[],[]]

expected: 0

Starter codePython

def overfitting_gap(train_losses, val_losses):
    # Your code here
    pass

4 test cases · run your code against them in the editor