Hands-On Generative Adversarial Networks with Keras

How can you kill a GAN?

We analyzed statistical measures of divergence between real data and other data and the results showed that even in simple cases, for example, the distribution of pixel intensities, the divergence between training data and fake data is high in comparison to the test data.

With respect to using specifications in order to train GANs, you could use specifications that are easy to learn from real data but are hardly differentiable, hence the hard-to-learn methods that rely on differentiation. We should bear in mind that specifications might not be satisfied by GANs.