It isn't common to see GANs mentioned as unconditional since this is the default and original configuration. In this book, however, we decided to stress this characteristic of the original GAN formulation in order to make you aware of the two main GAN classifications:
- Unconditional GANs
- Conditional GANs
The generative model that we described in the previous section falls under the category of unconditional GANs. The generative model is trained to capture the training data distribution and to generate samples that have been randomly sampled from the captured distribution. The conditional configuration is a slightly modified version of the framework and is presented in the next section.
Thanks to TensorFlow 2.0's eager-by-default style, the implementation of adversarial training is straightforward. In practice, to implement the adversarial training...
