Hands-On Generative Adversarial Networks with Keras

Tracking failure

It is important to track failure early on to speed up the training process.

Loss of the discriminator goes to 0.

When the loss of the discriminator goes to 0, as mentioned, the updates of the generator are insignificant.

Check the norm of the gradients.

When the norm of the gradients is large, it means relatively large weight updates. Although this is expected on the first iterations, the norm should decrease as training proceeds.

Check the variance over time of the Discriminator's Loss.

It is expected that the variance of the discriminator's loss decreases over time and does not have sudden spikes.