Real and fake reversed in AAE

[plaut]

In the adversarial autoencoder, lines 139 and 141 are:
latent_fake = self.encoder.predict(imgs)
latent_real = np.random.normal(size=(half_batch, self.encoded_dim))
Shouldn't this be reversed?

[eriklindernoren]

The encoder is trained to find an embedding of the data that mirrors a normal distribution. Samples from a normal distribution are used as real examples for the discriminator and encoded images are used a false examples. The encoder is trained to fool the discriminator into labeling the encoded images as having been sampled from a normal distribution.