VAE Reconstruction Loss

Problem

Compute the mean squared reconstruction error of a variational autoencoder: sum over elements of (x - x_hat)^2 divided by the number of elements.

The signature is vae_reconstruction_loss(x, x_hat).

Hint

Reconstruction terms are usually MSE or BCE; here use the mean squared error.

case 1

in: [[1,2],[1,2]]

expected: 0

case 2

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

expected: 2.5

case 3

in: [[0.5,0.5,0.5],[0,1,0]]

expected: 0.25

case 4

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

expected: 0.5

Starter codePython

def vae_reconstruction_loss(x, x_hat):
    # Your code here
    pass

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