WGAN Critic Loss

Problem

Compute the Wasserstein critic loss: mean(critic(fake)) - mean(critic(real)). Minimizing it pushes fake scores below real scores.

The signature is wgan_critic_loss(real_scores, fake_scores).

Hint

The critic outputs unbounded scores, not probabilities.

case 1

in: [[1,2],[3,4]]

expected: 2

case 2

in: [[0],[0]]

expected: 0

case 3

in: [[5],[1]]

expected: -4

case 4

in: [[-1,-3],[0,2]]

expected: 3

Starter codePython

def wgan_critic_loss(real_scores, fake_scores):
    # Your code here
    pass

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