Try answering and working on the following exercises to expand the knowledge that you've gained from this chapter:
- What is the adversarial training process?
- Write the value function of the min-max game that the Discriminator and Generator are playing.
- Explain why the min-max value function formulation can saturate in the early training step of training.
- Write and explain the non-saturating value function.
- Write the rules of the adversarial training process.
- Are there any recommendations on how to feed a condition to a GAN?
- What does it mean to create a conditional GAN?
- Can only the fully connected neural networks be used to create GANs?
- Which neural network architecture works better for the image generation problem?
- Update the code of the Unconditional GAN: Log the Generator and Discriminator loss value on TensorBoard, and also log matplotlib plots.
- Unconditional...