4. Conclusion

In this chapter, we've presented various improvements to the original GAN algorithm, first introduced in the previous chapter. WGAN proposed an algorithm to improve the stability of training by using the EMD or Wasserstein 1 loss. LSGAN argued that the original cross-entropy function of GANs is prone to vanishing gradients, unlike least squares loss. LSGAN proposed an algorithm to achieve stable training and quality outputs. ACGAN convincingly improved the quality of the conditional generation of MNIST digits by requiring the discriminator to perform a classification task on top of determining whether the input image was fake or real.

In the next chapter, we'll study how to control the attributes of generator outputs. Whilst CGAN and ACGAN are able to indicate the desired digits to produce, we have not analyzed GANs that can specify the attributes of outputs. For example, we may want to control the writing style of the MNIST digits, such as roundness, tilt...